OptaPlanner User Guide

Teacher conflict: A teacher must not have 2 lectures in the same period.
Curriculum conflict: A curriculum must not have 2 lectures in the same period.
Room occupancy: 2 lectures must not be in the same room in the same period.
Unavailable period (specified per dataset): A specific lecture must not be assigned to a specific period.

The problem is defined by the International Timetabling Competition 2007 track 3.

Room capacity: A room's capacity should not be less than the number of students in its lecture.
Minimum working days: Lectures of the same course should be spread into a minimum number of days.
Curriculum compactness: Lectures belonging to the same curriculum should be adjacent to each other (so in consecutive periods).
Room stability: Lectures of the same course should be assigned the same room.

3.3.1.2. Problem size

comp01 has 24 teachers,  14 curricula,  30 courses, 160 lectures, 30 periods,  6 rooms and   53 unavailable period constraints with a search space of  10^360.
comp02 has 71 teachers,  70 curricula,  82 courses, 283 lectures, 25 periods, 16 rooms and  513 unavailable period constraints with a search space of  10^736.
comp03 has 61 teachers,  68 curricula,  72 courses, 251 lectures, 25 periods, 16 rooms and  382 unavailable period constraints with a search space of  10^653.
comp04 has 70 teachers,  57 curricula,  79 courses, 286 lectures, 25 periods, 18 rooms and  396 unavailable period constraints with a search space of  10^758.
comp05 has 47 teachers, 139 curricula,  54 courses, 152 lectures, 36 periods,  9 rooms and  771 unavailable period constraints with a search space of  10^381.
comp06 has 87 teachers,  70 curricula, 108 courses, 361 lectures, 25 periods, 18 rooms and  632 unavailable period constraints with a search space of  10^957.
comp07 has 99 teachers,  77 curricula, 131 courses, 434 lectures, 25 periods, 20 rooms and  667 unavailable period constraints with a search space of 10^1171.
comp08 has 76 teachers,  61 curricula,  86 courses, 324 lectures, 25 periods, 18 rooms and  478 unavailable period constraints with a search space of  10^859.
comp09 has 68 teachers,  75 curricula,  76 courses, 279 lectures, 25 periods, 18 rooms and  405 unavailable period constraints with a search space of  10^740.
comp10 has 88 teachers,  67 curricula, 115 courses, 370 lectures, 25 periods, 18 rooms and  694 unavailable period constraints with a search space of  10^981.
comp11 has 24 teachers,  13 curricula,  30 courses, 162 lectures, 45 periods,  5 rooms and   94 unavailable period constraints with a search space of  10^381.
comp12 has 74 teachers, 150 curricula,  88 courses, 218 lectures, 36 periods, 11 rooms and 1368 unavailable period constraints with a search space of  10^566.
comp13 has 77 teachers,  66 curricula,  82 courses, 308 lectures, 25 periods, 19 rooms and  468 unavailable period constraints with a search space of  10^824.
comp14 has 68 teachers,  60 curricula,  85 courses, 275 lectures, 25 periods, 17 rooms and  486 unavailable period constraints with a search space of  10^722.

3.3.1.3. Domain model

3.3.2. Machine reassignment (Google ROADEF 2012)

3.3.2.1. Problem statement

Assign each process to a machine. All processes already have an original (unoptimized) assignment. Each process requires an amount of each resource (such as CPU, RAM, ...). This is more complex version of the Cloud Balancing example.

Maximum capacity: The maximum capacity for each resource for each machine must not be exceeded.
Conflict: Processes of the same service must run on distinct machines.
Spread: Processes of the same service must be spread across locations.
Dependency: The processes of a service depending on another service must run in the neighborhood of a process of the other service.
Transient usage: Some resources are transient and count towards the maximum capacity of both the original machine as the newly assigned machine.

The problem is defined by the Google ROADEF/EURO Challenge 2012.

Load: The safety capacity for each resource for each machine should not be exceeded.
Balance: Leave room for future assignments by balancing the available resources on each machine.
Process move cost: A process has a move cost.
Service move cost: A service has a move cost.
Machine move cost: Moving a process from machine A to machine B has another A-B specific move cost.

3.3.2.2. Problem size

model_a1_1 has  2 resources,  1 neighborhoods,   4 locations,    4 machines,    79 services,   100 processes and 1 balancePenalties with a search space of     10^60.
model_a1_2 has  4 resources,  2 neighborhoods,   4 locations,  100 machines,   980 services,  1000 processes and 0 balancePenalties with a search space of   10^2000.
model_a1_3 has  3 resources,  5 neighborhoods,  25 locations,  100 machines,   216 services,  1000 processes and 0 balancePenalties with a search space of   10^2000.
model_a1_4 has  3 resources, 50 neighborhoods,  50 locations,   50 machines,   142 services,  1000 processes and 1 balancePenalties with a search space of   10^1698.
model_a1_5 has  4 resources,  2 neighborhoods,   4 locations,   12 machines,   981 services,  1000 processes and 1 balancePenalties with a search space of   10^1079.
model_a2_1 has  3 resources,  1 neighborhoods,   1 locations,  100 machines,  1000 services,  1000 processes and 0 balancePenalties with a search space of   10^2000.
model_a2_2 has 12 resources,  5 neighborhoods,  25 locations,  100 machines,   170 services,  1000 processes and 0 balancePenalties with a search space of   10^2000.
model_a2_3 has 12 resources,  5 neighborhoods,  25 locations,  100 machines,   129 services,  1000 processes and 0 balancePenalties with a search space of   10^2000.
model_a2_4 has 12 resources,  5 neighborhoods,  25 locations,   50 machines,   180 services,  1000 processes and 1 balancePenalties with a search space of   10^1698.
model_a2_5 has 12 resources,  5 neighborhoods,  25 locations,   50 machines,   153 services,  1000 processes and 0 balancePenalties with a search space of   10^1698.
model_b_1  has 12 resources,  5 neighborhoods,  10 locations,  100 machines,  2512 services,  5000 processes and 0 balancePenalties with a search space of  10^10000.
model_b_2  has 12 resources,  5 neighborhoods,  10 locations,  100 machines,  2462 services,  5000 processes and 1 balancePenalties with a search space of  10^10000.
model_b_3  has  6 resources,  5 neighborhoods,  10 locations,  100 machines, 15025 services, 20000 processes and 0 balancePenalties with a search space of  10^40000.
model_b_4  has  6 resources,  5 neighborhoods,  50 locations,  500 machines,  1732 services, 20000 processes and 1 balancePenalties with a search space of  10^53979.
model_b_5  has  6 resources,  5 neighborhoods,  10 locations,  100 machines, 35082 services, 40000 processes and 0 balancePenalties with a search space of  10^80000.
model_b_6  has  6 resources,  5 neighborhoods,  50 locations,  200 machines, 14680 services, 40000 processes and 1 balancePenalties with a search space of  10^92041.
model_b_7  has  6 resources,  5 neighborhoods,  50 locations, 4000 machines, 15050 services, 40000 processes and 1 balancePenalties with a search space of 10^144082.
model_b_8  has  3 resources,  5 neighborhoods,  10 locations,  100 machines, 45030 services, 50000 processes and 0 balancePenalties with a search space of 10^100000.
model_b_9  has  3 resources,  5 neighborhoods, 100 locations, 1000 machines,  4609 services, 50000 processes and 1 balancePenalties with a search space of 10^150000.
model_b_10 has  3 resources,  5 neighborhoods, 100 locations, 5000 machines,  4896 services, 50000 processes and 1 balancePenalties with a search space of 10^184948.

3.3.3. Vehicle routing

3.3.3.1. Problem statement

Using a fleet of vehicles, pick up the objects of each customer and bring them to the depot. Each vehicle can service multiple customers, but it has a limited capacity.

The capacitated vehicle routing problem (CRVP) is defined by the VRP web.

3.3.3.2. Problem size

A-n32-k5  has 1 depots,  5 vehicles and  31 customers with a search space of  10^46.
A-n33-k5  has 1 depots,  5 vehicles and  32 customers with a search space of  10^48.
A-n33-k6  has 1 depots,  6 vehicles and  32 customers with a search space of  10^48.
A-n34-k5  has 1 depots,  5 vehicles and  33 customers with a search space of  10^50.
A-n36-k5  has 1 depots,  5 vehicles and  35 customers with a search space of  10^54.
A-n37-k5  has 1 depots,  5 vehicles and  36 customers with a search space of  10^56.
A-n37-k6  has 1 depots,  6 vehicles and  36 customers with a search space of  10^56.
A-n38-k5  has 1 depots,  5 vehicles and  37 customers with a search space of  10^58.
A-n39-k5  has 1 depots,  5 vehicles and  38 customers with a search space of  10^60.
A-n39-k6  has 1 depots,  6 vehicles and  38 customers with a search space of  10^60.
A-n44-k7  has 1 depots,  7 vehicles and  43 customers with a search space of  10^70.
A-n45-k6  has 1 depots,  6 vehicles and  44 customers with a search space of  10^72.
A-n45-k7  has 1 depots,  7 vehicles and  44 customers with a search space of  10^72.
A-n46-k7  has 1 depots,  7 vehicles and  45 customers with a search space of  10^74.
A-n48-k7  has 1 depots,  7 vehicles and  47 customers with a search space of  10^78.
A-n53-k7  has 1 depots,  7 vehicles and  52 customers with a search space of  10^89.
A-n54-k7  has 1 depots,  7 vehicles and  53 customers with a search space of  10^91.
A-n55-k9  has 1 depots,  9 vehicles and  54 customers with a search space of  10^93.
A-n60-k9  has 1 depots,  9 vehicles and  59 customers with a search space of 10^104.
A-n61-k9  has 1 depots,  9 vehicles and  60 customers with a search space of 10^106.
A-n62-k8  has 1 depots,  8 vehicles and  61 customers with a search space of 10^108.
A-n63-k10 has 1 depots, 10 vehicles and  62 customers with a search space of 10^111.
A-n63-k9  has 1 depots,  9 vehicles and  62 customers with a search space of 10^111.
A-n64-k9  has 1 depots,  9 vehicles and  63 customers with a search space of 10^113.
A-n65-k9  has 1 depots,  9 vehicles and  64 customers with a search space of 10^115.
A-n69-k9  has 1 depots,  9 vehicles and  68 customers with a search space of 10^124.
A-n80-k10 has 1 depots, 10 vehicles and  79 customers with a search space of 10^149.
F-n135-k7 has 1 depots,  7 vehicles and 134 customers with a search space of 10^285.
F-n45-k4  has 1 depots,  4 vehicles and  44 customers with a search space of  10^72.
F-n72-k4  has 1 depots,  4 vehicles and  71 customers with a search space of 10^131.

3.3.4. Hospital bed planning (PAS - Patient admission scheduling)

3.3.4.1. Problem statement

Assign each patient (that will come to the hospital) into a bed for each night that the patient will stay in the hospital. Each bed belongs to a room and each room belongs to a department. The arrival and departure dates of the patients is fixed: only a bed needs to be assigned for each night.

2 patients must not be assigned to the same bed in the same night.
A room can have a gender limitation: only females, only males, the same gender in the same night or no gender limitation at all.
A department can have a minimum or maximum age.
A patient can require a room with specific equipment(s).

Schedule each exam into a period and into a room. Multiple exams can share the same room during the same period.

A patient can prefer a maximum room size, for example if he/she want a single room.
A patient is best assigned to a department that specializes in his/her problem.
A patient is best assigned to a room that specializes in his/her problem.
A patient can prefer a room with specific equipment(s).

The problem is defined on this webpage and the test data comes from real world hospitals.

3.3.4.2. Problem size

testdata01 has 4 specialisms, 2 equipments, 4 departments,  98 rooms, 286 beds, 14 nights,  652 patients and  652 admissions with a search space of 10^1601.
testdata02 has 6 specialisms, 2 equipments, 6 departments, 151 rooms, 465 beds, 14 nights,  755 patients and  755 admissions with a search space of 10^2013.
testdata03 has 5 specialisms, 2 equipments, 5 departments, 131 rooms, 395 beds, 14 nights,  708 patients and  708 admissions with a search space of 10^1838.
testdata04 has 6 specialisms, 2 equipments, 6 departments, 155 rooms, 471 beds, 14 nights,  746 patients and  746 admissions with a search space of 10^1994.
testdata05 has 4 specialisms, 2 equipments, 4 departments, 102 rooms, 325 beds, 14 nights,  587 patients and  587 admissions with a search space of 10^1474.
testdata06 has 4 specialisms, 2 equipments, 4 departments, 104 rooms, 313 beds, 14 nights,  685 patients and  685 admissions with a search space of 10^1709.
testdata07 has 6 specialisms, 4 equipments, 6 departments, 162 rooms, 472 beds, 14 nights,  519 patients and  519 admissions with a search space of 10^1387.
testdata08 has 6 specialisms, 4 equipments, 6 departments, 148 rooms, 441 beds, 21 nights,  895 patients and  895 admissions with a search space of 10^2366.
testdata09 has 4 specialisms, 4 equipments, 4 departments, 105 rooms, 310 beds, 28 nights, 1400 patients and 1400 admissions with a search space of 10^3487.
testdata10 has 4 specialisms, 4 equipments, 4 departments, 104 rooms, 308 beds, 56 nights, 1575 patients and 1575 admissions with a search space of 10^3919.
testdata11 has 4 specialisms, 4 equipments, 4 departments, 107 rooms, 318 beds, 91 nights, 2514 patients and 2514 admissions with a search space of 10^6291.
testdata12 has 4 specialisms, 4 equipments, 4 departments, 105 rooms, 310 beds, 84 nights, 2750 patients and 2750 admissions with a search space of 10^6851.
testdata13 has 5 specialisms, 4 equipments, 5 departments, 125 rooms, 368 beds, 28 nights,  907 patients and 1109 admissions with a search space of 10^2845.

3.4. Difficult examples

3.4.1. Exam timetabling (ITC 2007 track 1 - Examination)

3.4.1.1. Problem statement

Soft constraints (each of which has a parametrized penalty):

Exam conflict: 2 exams that share students must not occur in the same period.
Room capacity: A room's seating capacity must suffice at all times.
Period duration: A period's duration must suffice for all of its exams.
Period related hard constraints (specified per dataset):
Room related hard constraints (specified per dataset):
- Exclusive: 1 specified exam should not have to share its room with any other exam.

The same student should not have 2 exams in a row.
The same student should not have 2 exams on the same day.
Period spread: 2 exams that share students should be a number of periods apart.
Mixed durations: 2 exams that share a room should not have different durations.
Front load: Large exams should be scheduled earlier in the schedule.
Period penalty (specified per dataset): Some periods have a penalty when used.
Room penalty (specified per dataset): Some rooms have a penalty when used.

It uses large test data sets of real-life universities.

The problem is defined by the International Timetabling Competition 2007 track 1. Geoffrey De Smet finished 4th in that competition with a very early version of OptaPlanner. Many improvements have been made since then.

3.4.1.2. Problem size

exam_comp_set1 has  7883 students,  607 exams, 54 periods,  7 rooms,  12 period constraints and  0 room constraints with a search space of 10^1564.
exam_comp_set2 has 12484 students,  870 exams, 40 periods, 49 rooms,  12 period constraints and  2 room constraints with a search space of 10^2864.
exam_comp_set3 has 16365 students,  934 exams, 36 periods, 48 rooms, 168 period constraints and 15 room constraints with a search space of 10^3023.
exam_comp_set4 has  4421 students,  273 exams, 21 periods,  1 rooms,  40 period constraints and  0 room constraints with a search space of  10^360.
exam_comp_set5 has  8719 students, 1018 exams, 42 periods,  3 rooms,  27 period constraints and  0 room constraints with a search space of 10^2138.
exam_comp_set6 has  7909 students,  242 exams, 16 periods,  8 rooms,  22 period constraints and  0 room constraints with a search space of  10^509.
exam_comp_set7 has 13795 students, 1096 exams, 80 periods, 15 rooms,  28 period constraints and  0 room constraints with a search space of 10^3374.
exam_comp_set8 has  7718 students,  598 exams, 80 periods,  8 rooms,  20 period constraints and  1 room constraints with a search space of 10^1678.

3.4.1.3. Domain model

Below you can see the main examination domain classes:

Figure 3.3. Examination domain class diagram

Notice that we've split up the exam concept into an Exam class and a Topic class. The Exam instances change during solving (this is the planning entity class), when their period or room property changes. The Topic, Period and Room instances never change during solving (these are problem facts, just like some other classes).

3.4.2. Employee rostering (INRC 2010 - Nurse rostering)

3.4.2.1. Problem statement

For each shift, assign a nurse to work that shift.

The problem is defined by the International Nurse Rostering Competition 2010.

3.4.2.2. Problem size

There are 3 dataset types:

sprint: must be solved in seconds.
medium: must be solved in minutes.
long: must be solved in hours.

toy1          has 1 skills, 3 shiftTypes, 2 patterns, 1 contracts,  6 employees,  7 shiftDates,  35 shiftAssignments and   0 requests with a search space of   10^27.
toy2          has 1 skills, 3 shiftTypes, 3 patterns, 2 contracts, 20 employees, 28 shiftDates, 180 shiftAssignments and 140 requests with a search space of  10^234.

sprint01      has 1 skills, 4 shiftTypes, 3 patterns, 4 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint02      has 1 skills, 4 shiftTypes, 3 patterns, 4 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint03      has 1 skills, 4 shiftTypes, 3 patterns, 4 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint04      has 1 skills, 4 shiftTypes, 3 patterns, 4 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint05      has 1 skills, 4 shiftTypes, 3 patterns, 4 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint06      has 1 skills, 4 shiftTypes, 3 patterns, 4 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint07      has 1 skills, 4 shiftTypes, 3 patterns, 4 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint08      has 1 skills, 4 shiftTypes, 3 patterns, 4 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint09      has 1 skills, 4 shiftTypes, 3 patterns, 4 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint10      has 1 skills, 4 shiftTypes, 3 patterns, 4 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint_hint01 has 1 skills, 4 shiftTypes, 8 patterns, 3 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint_hint02 has 1 skills, 4 shiftTypes, 0 patterns, 3 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint_hint03 has 1 skills, 4 shiftTypes, 8 patterns, 3 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint_late01 has 1 skills, 4 shiftTypes, 8 patterns, 3 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint_late02 has 1 skills, 3 shiftTypes, 4 patterns, 3 contracts, 10 employees, 28 shiftDates, 144 shiftAssignments and 139 requests with a search space of  10^144.
sprint_late03 has 1 skills, 4 shiftTypes, 8 patterns, 3 contracts, 10 employees, 28 shiftDates, 160 shiftAssignments and 150 requests with a search space of  10^160.
sprint_late04 has 1 skills, 4 shiftTypes, 8 patterns, 3 contracts, 10 employees, 28 shiftDates, 160 shiftAssignments and 150 requests with a search space of  10^160.
sprint_late05 has 1 skills, 4 shiftTypes, 8 patterns, 3 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint_late06 has 1 skills, 4 shiftTypes, 0 patterns, 3 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint_late07 has 1 skills, 4 shiftTypes, 0 patterns, 3 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.
sprint_late08 has 1 skills, 4 shiftTypes, 0 patterns, 3 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and   0 requests with a search space of  10^152.
sprint_late09 has 1 skills, 4 shiftTypes, 0 patterns, 3 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and   0 requests with a search space of  10^152.
sprint_late10 has 1 skills, 4 shiftTypes, 0 patterns, 3 contracts, 10 employees, 28 shiftDates, 152 shiftAssignments and 150 requests with a search space of  10^152.

medium01      has 1 skills, 4 shiftTypes, 0 patterns, 4 contracts, 31 employees, 28 shiftDates, 608 shiftAssignments and 403 requests with a search space of  10^906.
medium02      has 1 skills, 4 shiftTypes, 0 patterns, 4 contracts, 31 employees, 28 shiftDates, 608 shiftAssignments and 403 requests with a search space of  10^906.
medium03      has 1 skills, 4 shiftTypes, 0 patterns, 4 contracts, 31 employees, 28 shiftDates, 608 shiftAssignments and 403 requests with a search space of  10^906.
medium04      has 1 skills, 4 shiftTypes, 0 patterns, 4 contracts, 31 employees, 28 shiftDates, 608 shiftAssignments and 403 requests with a search space of  10^906.
medium05      has 1 skills, 4 shiftTypes, 0 patterns, 4 contracts, 31 employees, 28 shiftDates, 608 shiftAssignments and 403 requests with a search space of  10^906.
medium_hint01 has 1 skills, 4 shiftTypes, 7 patterns, 4 contracts, 30 employees, 28 shiftDates, 428 shiftAssignments and 390 requests with a search space of  10^632.
medium_hint02 has 1 skills, 4 shiftTypes, 7 patterns, 3 contracts, 30 employees, 28 shiftDates, 428 shiftAssignments and 390 requests with a search space of  10^632.
medium_hint03 has 1 skills, 4 shiftTypes, 7 patterns, 4 contracts, 30 employees, 28 shiftDates, 428 shiftAssignments and 390 requests with a search space of  10^632.
medium_late01 has 1 skills, 4 shiftTypes, 7 patterns, 4 contracts, 30 employees, 28 shiftDates, 424 shiftAssignments and 390 requests with a search space of  10^626.
medium_late02 has 1 skills, 4 shiftTypes, 7 patterns, 3 contracts, 30 employees, 28 shiftDates, 428 shiftAssignments and 390 requests with a search space of  10^632.
medium_late03 has 1 skills, 4 shiftTypes, 0 patterns, 4 contracts, 30 employees, 28 shiftDates, 428 shiftAssignments and 390 requests with a search space of  10^632.
medium_late04 has 1 skills, 4 shiftTypes, 7 patterns, 3 contracts, 30 employees, 28 shiftDates, 416 shiftAssignments and 390 requests with a search space of  10^614.
medium_late05 has 2 skills, 5 shiftTypes, 7 patterns, 4 contracts, 30 employees, 28 shiftDates, 452 shiftAssignments and 390 requests with a search space of  10^667.

long01        has 2 skills, 5 shiftTypes, 3 patterns, 3 contracts, 49 employees, 28 shiftDates, 740 shiftAssignments and 735 requests with a search space of 10^1250.
long02        has 2 skills, 5 shiftTypes, 3 patterns, 3 contracts, 49 employees, 28 shiftDates, 740 shiftAssignments and 735 requests with a search space of 10^1250.
long03        has 2 skills, 5 shiftTypes, 3 patterns, 3 contracts, 49 employees, 28 shiftDates, 740 shiftAssignments and 735 requests with a search space of 10^1250.
long04        has 2 skills, 5 shiftTypes, 3 patterns, 3 contracts, 49 employees, 28 shiftDates, 740 shiftAssignments and 735 requests with a search space of 10^1250.
long05        has 2 skills, 5 shiftTypes, 3 patterns, 3 contracts, 49 employees, 28 shiftDates, 740 shiftAssignments and 735 requests with a search space of 10^1250.
long_hint01   has 2 skills, 5 shiftTypes, 9 patterns, 3 contracts, 50 employees, 28 shiftDates, 740 shiftAssignments and   0 requests with a search space of 10^1257.
long_hint02   has 2 skills, 5 shiftTypes, 7 patterns, 3 contracts, 50 employees, 28 shiftDates, 740 shiftAssignments and   0 requests with a search space of 10^1257.
long_hint03   has 2 skills, 5 shiftTypes, 7 patterns, 3 contracts, 50 employees, 28 shiftDates, 740 shiftAssignments and   0 requests with a search space of 10^1257.
long_late01   has 2 skills, 5 shiftTypes, 9 patterns, 3 contracts, 50 employees, 28 shiftDates, 752 shiftAssignments and   0 requests with a search space of 10^1277.
long_late02   has 2 skills, 5 shiftTypes, 9 patterns, 4 contracts, 50 employees, 28 shiftDates, 752 shiftAssignments and   0 requests with a search space of 10^1277.
long_late03   has 2 skills, 5 shiftTypes, 9 patterns, 3 contracts, 50 employees, 28 shiftDates, 752 shiftAssignments and   0 requests with a search space of 10^1277.
long_late04   has 2 skills, 5 shiftTypes, 9 patterns, 4 contracts, 50 employees, 28 shiftDates, 752 shiftAssignments and   0 requests with a search space of 10^1277.
long_late05   has 2 skills, 5 shiftTypes, 9 patterns, 3 contracts, 50 employees, 28 shiftDates, 740 shiftAssignments and   0 requests with a search space of 10^1257.

3.4.3. Sport scheduling (TTP - Traveling tournament problem)

3.4.3.1. Problem statement

Schedule matches between n teams.

Each team plays twice against every other team: once home and once away.
Each team has exactly 1 match on each timeslot.
No team must have more than 3 consecutive home or 3 consecutive away matches.
No repeaters: no 2 consecutive matches of the same 2 opposing teams.

The problem is defined on Michael Trick's website (which contains the world records too).

Minimize the total distance traveled by all teams.

3.4.3.2. Problem size

1-nl04     has  6 days,  4 teams and   12 matches with a search space of    10^9.
1-nl06     has 10 days,  6 teams and   30 matches with a search space of   10^30.
1-nl08     has 14 days,  8 teams and   56 matches with a search space of   10^64.
1-nl10     has 18 days, 10 teams and   90 matches with a search space of  10^112.
1-nl12     has 22 days, 12 teams and  132 matches with a search space of  10^177.
1-nl14     has 26 days, 14 teams and  182 matches with a search space of  10^257.
1-nl16     has 30 days, 16 teams and  240 matches with a search space of  10^354.
2-bra24    has 46 days, 24 teams and  552 matches with a search space of  10^917.
3-nfl16    has 30 days, 16 teams and  240 matches with a search space of  10^354.
3-nfl18    has 34 days, 18 teams and  306 matches with a search space of  10^468.
3-nfl20    has 38 days, 20 teams and  380 matches with a search space of  10^600.
3-nfl22    has 42 days, 22 teams and  462 matches with a search space of  10^749.
3-nfl24    has 46 days, 24 teams and  552 matches with a search space of  10^917.
3-nfl26    has 50 days, 26 teams and  650 matches with a search space of 10^1104.
3-nfl28    has 54 days, 28 teams and  756 matches with a search space of 10^1309.
3-nfl30    has 58 days, 30 teams and  870 matches with a search space of 10^1534.
3-nfl32    has 62 days, 32 teams and  992 matches with a search space of 10^1778.
4-super04  has  6 days,  4 teams and   12 matches with a search space of    10^9.
4-super06  has 10 days,  6 teams and   30 matches with a search space of   10^30.
4-super08  has 14 days,  8 teams and   56 matches with a search space of   10^64.
4-super10  has 18 days, 10 teams and   90 matches with a search space of  10^112.
4-super12  has 22 days, 12 teams and  132 matches with a search space of  10^177.
4-super14  has 26 days, 14 teams and  182 matches with a search space of  10^257.
5-galaxy04 has  6 days,  4 teams and   12 matches with a search space of    10^9.
5-galaxy06 has 10 days,  6 teams and   30 matches with a search space of   10^30.
5-galaxy08 has 14 days,  8 teams and   56 matches with a search space of   10^64.
5-galaxy10 has 18 days, 10 teams and   90 matches with a search space of  10^112.
5-galaxy12 has 22 days, 12 teams and  132 matches with a search space of  10^177.
5-galaxy14 has 26 days, 14 teams and  182 matches with a search space of  10^257.
5-galaxy16 has 30 days, 16 teams and  240 matches with a search space of  10^354.
5-galaxy18 has 34 days, 18 teams and  306 matches with a search space of  10^468.
5-galaxy20 has 38 days, 20 teams and  380 matches with a search space of  10^600.
5-galaxy22 has 42 days, 22 teams and  462 matches with a search space of  10^749.
5-galaxy24 has 46 days, 24 teams and  552 matches with a search space of  10^917.
5-galaxy26 has 50 days, 26 teams and  650 matches with a search space of 10^1104.
5-galaxy28 has 54 days, 28 teams and  756 matches with a search space of 10^1309.
5-galaxy30 has 58 days, 30 teams and  870 matches with a search space of 10^1534.
5-galaxy32 has 62 days, 32 teams and  992 matches with a search space of 10^1778.
5-galaxy34 has 66 days, 34 teams and 1122 matches with a search space of 10^2041.
5-galaxy36 has 70 days, 36 teams and 1260 matches with a search space of 10^2324.
5-galaxy38 has 74 days, 38 teams and 1406 matches with a search space of 10^2628.
5-galaxy40 has 78 days, 40 teams and 1560 matches with a search space of 10^2951.

Chapter 4. Planner configuration

4.1. Overview

4.2. Solver configuration

4.2.1. Solver configuration by XML file
4.2.2. Solver configuration by Java API

4.3. Model your planning problem

4.3.1. Is this class a problem fact or planning entity?
4.3.2. Problem fact
4.3.3. Planning entity and planning variables
4.3.4. Planning value and planning value ranges
4.3.5. Planning problem and planning solution

4.4. Use the Solver

4.4.1. The Solver interface
4.4.2. Solving a problem
4.4.3. Environment mode: Are there bugs in my code?
4.4.4. Logging level: What is the Solver doing?

4.1. Overview

Solving a planning problem with OptaPlanner consists out of 5 steps:

Model your planning problem as a class that implements the interface Solution, for example the class NQueens.
Configure a Solver, for example a first fit and tabu search solver for any NQueens instance.
Load a problem data set from your data layer, for example a 4 Queens instance. Set it as the planning problem on the Solver with Solver.setPlanningProblem(...).
Solve it with Solver.solve().
Get the best solution found by the Solver with Solver.getBestSolution().

4.2. Solver configuration

4.2.1. Solver configuration by XML file

You can build a Solver instance with the XmlSolverFactory. Configure it with a solver configuration XML file:

       XmlSolverFactory solverFactory = new XmlSolverFactory(

               "/org/optaplanner/examples/nqueens/solver/nqueensSolverConfig.xml");

       Solver solver = solverFactory.buildSolver();

A solver configuration file looks something like this:


<?xml version="1.0" encoding="UTF-8"?>

<solver>

  <!-- Define the model -->

  <solutionClass>org.optaplanner.examples.nqueens.domain.NQueens</solutionClass>

  <planningEntityClass>org.optaplanner.examples.nqueens.domain.Queen</planningEntityClass>



  <!-- Define the score function -->

  <scoreDirectorFactory>

    <scoreDefinitionType>SIMPLE</scoreDefinitionType>

    <scoreDrl>/org/optaplanner/examples/nqueens/solver/nQueensScoreRules.drl</scoreDrl>

  </scoreDirectorFactory>



  <!-- Configure the optimization algorithm(s) -->

  <termination>

    ...

  </termination>

  <constructionHeuristic>

    ...

  </constructionHeuristic>

  <localSearch>

    ...

  </localSearch>

</solver>

Notice the 3 parts in it:

Define the model
Define the score function
Configure the optimization algorithm(s)

We'll explain these various parts of a configuration later in this manual.

OptaPlanner makes it relatively easy to switch optimization algorithm(s) just by changing the configuration. There's even a Benchmark utility which allows you to play out different configurations against each other and report the most appropriate configuration for your problem. You could for example play out tabu search versus simulated annealing, on 4 queens and 64 queens.

4.2.2. Solver configuration by Java API

As an alternative to the XML file, a solver configuration can also be configured with the SolverConfig API:

        SolverConfig solverConfig = new SolverConfig();


        solverConfig.setSolutionClass(NQueens.class);

        solverConfig.setPlanningEntityClassSet(Collections.<Class<?>>singleton(Queen.class));


        ScoreDirectorFactoryConfig scoreDirectorFactoryConfig = new ScoreDirectorFactoryConfig();

        scoreDirectorFactoryConfig.setScoreDefinitionType(ScoreDirectorFactoryConfig.ScoreDefinitionType.SIMPLE);

        scoreDirectorFactoryConfig.setScoreDrlList(

                Arrays.asList("/org/optaplanner/examples/nqueens/solver/nQueensScoreRules.drl"));

        solverConfig.setScoreDirectorFactoryConfig(scoreDirectorFactoryConfig);


        TerminationConfig terminationConfig = new TerminationConfig();

        // ...

        solverConfig.setTerminationConfig(terminationConfig);

        List<SolverPhaseConfig> solverPhaseConfigList = new ArrayList<SolverPhaseConfig>();

        ConstructionHeuristicSolverPhaseConfig constructionHeuristicSolverPhaseConfig

                = new ConstructionHeuristicSolverPhaseConfig();

        // ...

        solverPhaseConfigList.add(constructionHeuristicSolverPhaseConfig);

        LocalSearchSolverPhaseConfig localSearchSolverPhaseConfig = new LocalSearchSolverPhaseConfig();

        // ...

        solverPhaseConfigList.add(localSearchSolverPhaseConfig);

        solverConfig.setSolverPhaseConfigList(solverPhaseConfigList);

        Solver solver = solverConfig.buildSolver();

It is highly recommended to configure by XML file instead of this API. To dynamically configure a value at runtime, use the XML file as a template and extract the SolverConfig class with getSolverConfig() to configure the dynamic value at runtime:

        XmlSolverFactory solverFactory = new XmlSolverFactory();

                "/org/optaplanner/examples/nqueens/solver/nqueensSolverConfig.xml");


        SolverConfig solverConfig = solverFactory.getSolverConfig();

        solverConfig.getTerminationConfig().setMaximumMinutesSpend(userInput);

        Solver solver = solverConfig.buildSolver();

4.3. Model your planning problem

4.3.1. Is this class a problem fact or planning entity?

Look at a dataset of your planning problem. You 'll recognize domain classes in there, each of which is one of these:

A unrelated class: not used by any of the score constraints. From a planning standpoint, this data is obsolete.
A problem fact class: used by the score constraints, but does NOT change during planning (as long as the problem stays the same). For example: Bed, Room, Shift, Employee, Topic, Period, ...
A planning entity class: used by the score constraints and changes during planning. For example: BedDesignation, ShiftAssignment, Exam, ...

Ask yourself: What class changes during planning? Which class has variables that I want the Solver to change for me? That class is a planning entity. Most use cases have only 1 planning entity class.

Note

In real-time planning, problem facts can change during planning, because the problem itself changes. However, that doesn't make them planning entities.

A good model can greatly improve the success of your planning implementation. For inspiration, take a look at how the examples modeled their domain:

When in doubt, it's usually the many side of a many to one relationship that is the planning entity. For example in employee rostering, the planning entity class is ShiftAssignment, not Employee. Vehicle routing is special, because it uses a chained planning variable.

In OptaPlanner all problems facts and planning entities are plain old JavaBeans (POJO's). You can load them from a database (JDBC/JPA/JDO), an XML file, a data repository, a noSQL cloud, ...: OptaPlanner doesn't care.

4.3.2. Problem fact

A problem fact is any JavaBean (POJO) with getters that does not change during planning. Implementing the interface Serializable is recommended (but not required). For example in n queens, the columns and rows are problem facts:

public class Column implements Serializable {


    private int index;


    // ... getters

}

public class Row implements Serializable {


    private int index;


    // ... getters

}

A problem fact can reference other problem facts of course:

public class Course implements Serializable {


    private String code;


    private Teacher teacher; // Other problem fact

    private int lectureSize;

    private int minWorkingDaySize;


    private List<Curriculum> curriculumList; // Other problem facts

    private int studentSize;


    // ... getters

}

A problem fact class does not require any Planner specific code. For example, you can reuse your domain classes, which might have JPA annotations.

Note

Generally, better designed domain classes lead to simpler and more efficient score constraints. Therefore, when dealing with a messy legacy system, it can sometimes be worth it to convert the messy domain set into a planner specific POJO set first. For example: if your domain model has 2 Teacher instances for the same teacher that teaches at 2 different departments, it's hard to write a correct score constraint that constrains a teacher's spare time.

Alternatively, you can sometimes also introduce a cached problem fact to enrich the domain model for planning only.

4.3.3. Planning entity and planning variables

4.3.3.1. Planning entity

A planning entity is a JavaBean (POJO) that changes during solving, for example a Queen that changes to another row. A planning problem has multiple planning entities, for example for a single n queens problem, each Queen is a planning entity. But there's usually only 1 planning entity class, for example the Queen class.

A planning entity class needs to be annotated with the @PlanningEntity annotation.

Each planning entity class has 1 or more planning variables. It usually also has 1 or more defining properties. For example in n queens, a Queen is defined by its Column and has a planning variable Row. This means that a Queen's column never changes during solving, while its row does change.

@PlanningEntity

public class Queen {


    private Column column;


    // Planning variables: changes during planning, between score calculations.

    private Row row;


    // ... getters and setters

}

A planning entity class can have multiple planning variables. For example, a Lecture is defined by its Course and its index in that course (because 1 course has multiple lectures). Each Lecture needs to be scheduled into a Period and a Room so it has 2 planning variables (period and room). For example: the course Mathematics has 8 lectures per week, of which the first lecture is Monday morning at 08:00 in room 212.

@PlanningEntity

public class Lecture {


    private Course course;

    private int lectureIndexInCourse;


    // Planning variables: changes during planning, between score calculations.

    private Period period;

    private Room room;


    // ...

}

The solver configuration also needs to be made aware of each planning entity class:

<solver>

  ...

  <planningEntityClass>org.optaplanner.examples.nqueens.domain.Queen</planningEntityClass>

  ...

</solver>

Some uses cases have multiple planning entity classes. For example: route freight and trains into railway network arcs, where each freight can use multiple trains over its journey and each train can carry multiple freights per arc. Having multiple planning entity classes directly raises the implementation complexity of your use case.

Note

Do not create unnecessary planning entity classes. This leads to difficult Move implementations and slower score calculation.

For example, do not create a planning entity class to hold the total free time of a teacher, which needs to be kept up to date as the Lecture planning entities change. Instead, calculate the free time in the score constraints and put the result per teacher into a logically inserted score object.

If historic data needs to be considered too, then create problem fact to hold the historic data up to, but not including, the planning window (so it doesn't change when a planning entity changes) and let the score constraints take it into account.

4.3.3.2. Planning entity difficulty

Some optimization algorithms work more efficiently if they have an estimation of which planning entities are more difficult to plan. For example: in bin packing bigger items are harder to fit, in course scheduling lectures with more students are more difficult to schedule and in n queens the middle queens are more difficult to fit on the board.

Therefore, you can set a difficultyComparatorClass to the @PlanningEntity annotation:

@PlanningEntity(difficultyComparatorClass = CloudProcessDifficultyComparator.class)

public class CloudProcess {

    // ...

}

public class CloudProcessDifficultyComparator implements Comparator<CloudProcess> {


    public int compare(CloudProcess a, CloudProcess b) {

        return new CompareToBuilder()

                .append(a.getRequiredMultiplicand(), b.getRequiredMultiplicand())

                .append(a.getId(), b.getId())

                .toComparison();

    }


}

Alternatively, you can also set a difficultyWeightFactoryClass to the @PlanningEntity annotation, so you have access to the rest of the problem facts from the Solution too:

@PlanningEntity(difficultyWeightFactoryClass = QueenDifficultyWeightFactory.class)

public class Queen {

    // ...

}

See Sorted Selection for more information.

Important

Difficulty should be implemented ascending: easy entities are lower, difficult entities are higher. For example in bin packing: small item < medium item < big item.

Even though some algorithms start with the more difficult entities first, they just reverse the ordering.

None of the current planning variable state should be used to compare planning entity difficult. During construction heuristics, those variables are likely to be null anyway. For example, a Queen's row variable should not be used.

4.3.3.3. Planning variable

A planning variable is a property (including getter and setter) on a planning entity. It points to a planning value, which changes during planning. For example, a Queen's row property is a planning variable. Note that even though a Queen's row property changes to another Row during planning, no Row instance itself is changed.

A planning variable getter needs to be annotated with the @PlanningVariable annotation. Furthermore, it needs a @ValueRange annotation too.

@PlanningEntity

public class Queen {


    private Row row;


    // ...


    @PlanningVariable

    @ValueRange(type = ValueRangeType.FROM_SOLUTION_PROPERTY, solutionProperty = rowList")

    public Row getRow() {

        return row;

    }


    public void setRow(Row row) {

        this.row = row;

    }


}

4.3.3.4. Nullable planning variable

By default, an initialized planning variable cannot be null, so an initialized solution will never use null for any of its planning variables. In over-constrained use case, this can be contra productive. For example: in task assignment with too many tasks for the workforce, we would rather leave low priority tasks unassigned instead of assigning them to an overloaded worker.

To allow an initialized planning variable to be null, set nullable to true:

    @PlanningVariable(nullable = true)

    @ValueRange(...)

    public Worker getWorker() {

        return worker;

    }

Important

Planner will automatically add the value null to the value range. There is no need to add null in a collection used by a ValueRange.

Note

Using a nullable planning variable implies that your score calculation is responsible for punishing (or even rewarding) variables with a null value.

Repeated planning (especially real-time planning) does not mix well with a nullable planning variable: every time the Solver starts or a problem fact change is made, the construction heuristics will try to initialize all the null variables again, which can be a huge waste of time. One way to deal with this, is to change when a planning entity should be reinitialized with an reinitializeVariableEntityFilter:

    @PlanningVariable(nullable = true, reinitializeVariableEntityFilter = ReinitializeTaskFilter.class)

    @ValueRange(...)

    public Worker getWorker() {

        return worker;

    }

4.3.3.5. When is a planning variable considered initialized?

A planning variable is considered initialized if its value is not null or if the variable is nullable. So a nullable variable is always considered initialized, even when a custom reinitializeVariableEntityFilter triggers a reinitialization.

A planning entity is initialized if all of its planning variables are initialized.

A Solution is initialized if all of its planning entities are initialized.

4.3.4. Planning value and planning value ranges

4.3.4.1. Planning value

A planning value is a possible value for a planning variable. Usually, a planning value is a problem fact, but it can also be any object, for example a double. It can even be another planning entity or even a interface implemented by a planning entity and a problem fact.

A planning value range is the set of possible planning values for a planning variable. This set can be a discrete (for example row 1, 2, 3 or 4) or continuous (for example any double between 0.0 and 1.0). Continuous planning variables are currently undersupported and require the use of custom moves.

There are several ways to define the value range of a planning variable with the @ValueRange annotation.

4.3.4.2. Planning value range

4.3.4.2.1. `ValueRange` from `Solution` property

All instances of the same planning entity class share the same set of possible planning values for that planning variable. This is the most common way to configure a value range.

The Solution implementation has property which returns a Collection. Any value from that Collection is a possible planning value for this planning variable.

    @PlanningVariable

    @ValueRange(type = ValueRangeType.FROM_SOLUTION_PROPERTY, solutionProperty = "rowList")

    public Row getRow() {

        return row;

    }

public class NQueens implements Solution<SimpleScore> {


    // ...


    public List<Row> getRowList() {

        return rowList;

    }


}

That Collection must not contain the value null, even for a nullable planning variable.

4.3.4.2.2. `ValueRange` from planning entity

Each planning entity has its own set of possible planning values for a planning variable. For example, if a teacher can never teach in a room that does not belong to his department, lectures of that teacher can limit their room value range to the rooms of his department.

    @PlanningVariable

    @ValueRange(type = ValueRangeType.FROM_PLANNING_ENTITY_PROPERTY, planningEntityProperty = "possibleRoomList")

    public Room getRoom() {

        return room;

    }


    public List<Room> getPossibleRoomList() {

        return getCourse().getTeacher().getPossibleRoomList();

    }

Never use this to enforce a soft constraint (or even a hard constraint when the problem might not have a feasible solution). For example: Unless there is no other way, a teacher can not teach in a room that does not belong to his department. In this case, the teacher should not be limited in his room value range (because sometimes there is no other way).

Note

By limiting the value range specifically of 1 planning entity, you are effectively making a build-in hard constraint. This can be a very good thing, as the number of possible solutions is severely lowered. But this can also be a bad thing because it takes away the freedom of the optimization algorithms to temporarily break that constraint in order to escape a local optima.

A planning entity should not use other planning entities to determinate its value range. That would only try to make it solve the planning problem itself and interfere with the optimization algorithms.

This value range is not compatible with a chained variable.

4.3.4.2.3. `ValueRange` undefined

Leaves the value range undefined. Most optimization algorithms do not support this value range.

    @PlanningVariable

    @ValueRange(type = ValueRangeType.UNDEFINED)

    public Row getRow() {

        return row;

    }

4.3.4.2.4. Combining ValueRanges

Value ranges can be combined, for example:

    @PlanningVariable(...)

    @ValueRanges({

            @ValueRange(type = ValueRangeType.FROM_SOLUTION_PROPERTY, solutionProperty = "companyCarList"),

            @ValueRange(type = ValueRangeType.FROM_PLANNING_ENTITY_PROPERTY, planningEntityProperty = "personalCarList"})

    public Car getCar() {

        return car;

    }

4.3.4.2.5. A `ValueRange` which includes other planning entities

In some cases (such as in chaining), the planning value itself is sometimes another planning entity. In such cases, it's often required that a planning entity is only eligible as a planning value if it's initialized:

    @PlanningVariable

    @ValueRange(type = ValueRangeType.FROM_SOLUTION_PROPERTY, solutionProperty = "copList", excludeUninitializedPlanningEntity = true)

    public Cop getPartner() {

        return partner;

    }

TODO: this is likely to change in the future (jira), as it should support specific planning variable initialization too.

4.3.4.3. Chained planning variable

Some use cases, such as TSP and Vehicle Routing, require chaining. This means the planning entities point to each other and form a chain.

A planning variable that is chained either:

Directly points to a planning fact, which is called an anchor.
Points to another planning entity with the same planning variable, which recursively points to an anchor.

Here are some example of valid and invalid chains:

Every initialized planning entity is part of an open-ended chain that begins from an anchor. A valid model means that:

A chain is never a loop. The tail is always open.
Every chain always has exactly 1 anchor. The anchor is a problem fact, never a planning entity.
A chain is never a tree, it is always a line. Every anchor or planning entity has at most 1 trailing planning entity.
Every initialized planning entity is part of a chain.
An anchor with no planning entities pointing to it, is also considered a chain.

Warning

A planning problem instance given to the Solver must be valid.

Note

If your constraints dictate a closed chain, model it as an open-ended chain (which is easier to persist in a database) and implement a score constraint for the last entity back to the anchor.

The optimization algorithms and build-in MoveFactory's do chain correction to guarantee that the model stays valid:

Warning

A custom Move implementation must leave the model in a valid state.

For example, in TSP the anchor is a Domicile (in vehicle routing it is the vehicle):

public class Domicile ... implements Appearance {


    ...


    public City getCity() {...}


}

The anchor (which is a problem fact) and the planning entity implement a common interface, for example TSP's Appearance:

public interface Appearance {


    City getCity();


}

That interface is the return type of the planning variable. Furthermore, the planning variable is chained. For example TSP's Visit (in vehicle routing it is the customer):

@PlanningEntity

public class Visit ... implements Appearance {


    ...


    public City getCity() {...}


    @PlanningVariable(chained = true)

    @ValueRanges({

            @ValueRange(type = ValueRangeType.FROM_SOLUTION_PROPERTY, solutionProperty = "domicileList"),

            @ValueRange(type = ValueRangeType.FROM_SOLUTION_PROPERTY, solutionProperty = "visitList",

                    excludeUninitializedPlanningEntity = true)})

    public Appearance getPreviousAppearance() {

        return previousAppearance;

    }


    public void setPreviousAppearance(Appearance previousAppearance) {

        this.previousAppearance = previousAppearance;

    }


}

Notice how 2 value ranges need to be combined:

The value range which holds the anchors, for example domicileList.
The value range which holds the initialized planning entities, for example visitList. This always requires an enabled excludeUninitializedPlanningEntity, because an initialized entity should never point to an uninitialized entity: that would break the principle that every chain must have an anchor.

4.3.4.4. Planning value strength

Some optimization algorithms work more efficiently if they have an estimation of which planning values are stronger, which means they are more likely to satisfy a planning entity. For example: in bin packing bigger containers are more likely to fit an item and in course scheduling bigger rooms are less likely to break the student capacity constraint.

Therefore, you can set a strengthComparatorClass to the @PlanningVariable annotation:

    @PlanningVariable(strengthComparatorClass = CloudComputerStrengthComparator.class)

    // ...

    public CloudComputer getComputer() {

        // ...

    }

public class CloudComputerStrengthComparator implements Comparator<CloudComputer> {


    public int compare(CloudComputer a, CloudComputer b) {

        return new CompareToBuilder()

                .append(a.getMultiplicand(), b.getMultiplicand())

                .append(b.getCost(), a.getCost()) // Descending (but this is debatable)

                .append(a.getId(), b.getId())

                .toComparison();

    }


}

Note

If you have multiple planning value classes in the same value range, the strengthComparatorClass needs to implement a Comparator of a common superclass (for example Comparator<Object>) and be able to handle comparing instances of those different classes.

Alternatively, you can also set a strengthWeightFactoryClass to the @PlanningVariable annotation, so you have access to the rest of the problem facts from the solution too:

    @PlanningVariable(strengthWeightFactoryClass = RowStrengthWeightFactory.class)

    // ...

    public Row getRow() {

        // ...

    }

See Sorted Selection for more information.

Important

Strength should be implemented ascending: weaker values are lower, stronger values are higher. For example in bin packing: small container < medium container < big container.

None of the current planning variable state in any of the planning entities should be used to compare planning values. During construction heuristics, those variables are likely to be null anyway. For example, none of the row variables of any Queen may be used to determine the strength of a Row.

4.3.5. Planning problem and planning solution

4.3.5.1. Planning problem instance

A dataset for a planning problem needs to be wrapped in a class for the Solver to solve. You must implement this class. For example in n queens, this in the NQueens class which contains a Column list, a Row list and a Queen list.

A planning problem is actually a unsolved planning solution or - stated differently - an uninitialized Solution. Therefor, that wrapping class must implement the Solution interface. For example in n queens, that NQueens class implements Solution, yet every Queen in a fresh NQueens class is not yet assigned to a Row (their row property is null). So it's not a feasible solution. It's not even a possible solution. It's an uninitialized solution.

4.3.5.2. The `Solution` interface

You need to present the problem as a Solution instance to the Solver. So you need to have a class that implements the Solution interface:

public interface Solution<S extends Score> {


    S getScore();

    void setScore(S score);


    Collection<? extends Object> getProblemFacts();


}

For example, an NQueens instance holds a list of all columns, all rows and all Queen instances:

public class NQueens implements Solution<SimpleScore> {


    private int n;


    // Problem facts

    private List<Column> columnList;

    private List<Row> rowList;


    // Planning entities

    private List<Queen> queenList;


    // ...

}

4.3.5.3. The `getScore()` and `setScore()` methods

A Solution requires a score property. The score property is null if the Solution is uninitialized or if the score has not yet been (re)calculated. The score property is usually typed to the specific Score implementation you use. For example, NQueens uses a SimpleScore:

public class NQueens implements Solution<SimpleScore> {


    private SimpleScore score;


    public SimpleScore getScore() {

        return score;

    }


    public void setScore(SimpleScore score) {

        this.score = score;

    }


    // ...

}

Most use cases use a HardSoftScore instead:

public class CourseSchedule implements Solution<HardSoftScore> {


    private HardSoftScore score;


    public HardSoftScore getScore() {

        return score;

    }


    public void setScore(HardSoftScore score) {

        this.score = score;

    }


    // ...

}

See the Score calculation section for more information on the Score implementations.

4.3.5.4. The `getProblemFacts()` method

The method is only used if Drools is used for score calculation. Other score directors do not use it.

All objects returned by the getProblemFacts() method will be asserted into the Drools working memory, so the score rules can access them. For example, NQueens just returns all Column and Row instances.

    public Collection<? extends Object> getProblemFacts() {

        List<Object> facts = new ArrayList<Object>();

        facts.addAll(columnList);

        facts.addAll(rowList);

        // Do not add the planning entity's (queenList) because that will be done automatically

        return facts;

    }

All planning entities are automatically inserted into the Drools working memory. Do not add them in the method getProblemFacts().

Note

A common mistake is to use facts.add(...) instead of fact.addAll(...) for a Collection, which leads to score rules failing to match because the elements of that Collection aren't in the Drools working memory.

The method getProblemFacts() is not called much: at most only once per solver phase per solver thread.

4.3.5.5. Cached problem fact

A cached problem fact is a problem fact that doesn't exist in the real domain model, but is calculated before the Solver really starts solving. The method getProblemFacts() has the chance to enrich the domain model with such cached problem facts, which can lead to simpler and faster score constraints.

For example in examination, a cached problem fact TopicConflict is created for every 2 Topic's which share at least 1 Student.

    public Collection<? extends Object> getProblemFacts() {

        List<Object> facts = new ArrayList<Object>();

        // ...

        facts.addAll(calculateTopicConflictList());

        // ...

        return facts;

    }


    private List<TopicConflict> calculateTopicConflictList() {

        List<TopicConflict> topicConflictList = new ArrayList<TopicConflict>();

        for (Topic leftTopic : topicList) {

            for (Topic rightTopic : topicList) {

                if (leftTopic.getId() < rightTopic.getId()) {

                    int studentSize = 0;

                    for (Student student : leftTopic.getStudentList()) {

                        if (rightTopic.getStudentList().contains(student)) {

                            studentSize++;

                        }

                    }

                    if (studentSize > 0) {

                        topicConflictList.add(new TopicConflict(leftTopic, rightTopic, studentSize));

                    }

                }

            }

        }

        return topicConflictList;

    }

Any score constraint that needs to check if no 2 exams have a topic which share a student are being scheduled close together (depending on the constraint: at the same time, in a row or in the same day), can simply use the TopicConflict instance as a problem fact, instead of having to combine every 2 Student instances.

4.3.5.6. Cloning a `Solution`

Most (if not all) optimization algorithms clone the solution each time they encounter a new best solution (so they can recall it later) or to work with multiple solutions in parallel.

Note

There are many ways to clone, such as a shallow clone, deep clone, ... This context focuses on a planning clone.

A planning clone of a Solution must fulfill these requirements:

The clone must represent the same planning problem. Usually it reuses the same instances of the problem facts and problem fact collections as the original.
The clone must use different, cloned instances of the entities and entity collections. Changes to an original Solution's entity's variables must not effect its clone.

Implementing a planning clone method is hard, therefore you don't need to implement it.

4.3.5.6.1. `FieldAccessingSolutionCloner`

This SolutionCloner is used by default. It works for the majority of use cases.

Warning

When the FieldAccessingSolutionCloner clones your entity collection, it might not recognize the implementation and replace it with ArrayList, LinkedHashSet or TreeSet (whichever is more applicable). It recognizes most of the common JDK Collection implementations.

4.3.5.6.2. Custom cloning: Make `Solution` implement `PlanningCloneable`

If your Solution implements PlanningCloneable, Planner will automatically choose to clone it by calling the method planningClone().

public interface PlanningCloneable<T> {


    T planningClone();


}

For example: If NQueens implements PlanningCloneable, it would only deep clone all Queen instances. When the original solution is changed during planning, by changing a Queen, the clone stays the same.

public class NQueens implements Solution<...>, PlanningCloneable<NQueens> {

    ...


    /**

     * Clone will only deep copy the {@link #queenList}.

     */

    public NQueens planningClone() {

        NQueens clone = new NQueens();

        clone.id = id;

        clone.n = n;

        clone.columnList = columnList;

        clone.rowList = rowList;

        List<Queen> clonedQueenList = new ArrayList<Queen>(queenList.size());

        for (Queen queen : queenList) {

            clonedQueenList.add(queen.planningClone());

        }

        clone.queenList = clonedQueenList;

        clone.score = score;

        return clone;

    }

}

The planningClone() method should only deep clone the planning entities. Notice that the problem facts, such as Column and Row are normally not cloned: even their List instances are not cloned. If you were to clone the problem facts too, then you'd have to make sure that the new planning entity clones also refer to the new problem facts clones used by the solution. For example, if you would clone all Row instances, then each Queen clone and the NQueens clone itself should refer to those new Row clones.

Warning

Cloning an entity with a chained variable is devious: a variable of an entity A might point to another entity B. If A is cloned, then it's variable must point to the clone of B, not the original B.

4.3.5.7. Build an uninitialized solution

Build a Solution instance to represent your planning problem, so you can set it on the Solver as the planning problem to solve. For example in n queens, an NQueens instance is created with the required Column and Row instances and every Queen set to a different column and every row set to null.

    private NQueens createNQueens(int n) {

        NQueens nQueens = new NQueens();

        nQueens.setId(0L);

        nQueens.setN(n);

        nQueens.setColumnList(createColumnList(nQueens));

        nQueens.setRowList(createRowList(nQueens));

        nQueens.setQueenList(createQueenList(nQueens));

        return nQueens;

    }


    private List<Queen> createQueenList(NQueens nQueens) {

        int n = nQueens.getN();

        List<Queen> queenList = new ArrayList<Queen>(n);

        long id = 0;

        for (Column column : nQueens.getColumnList()) {

            Queen queen = new Queen();

            queen.setId(id);

            id++;

            queen.setColumn(column);

            // Notice that we leave the PlanningVariable properties on null

            queenList.add(queen);

        }

        return queenList;

    }

Figure 4.1. Uninitialized solution for the 4 queens puzzle

Usually, most of this data comes from your data layer, and your Solution implementation just aggregates that data and creates the uninitialized planning entity instances to plan:

        private void createLectureList(CourseSchedule schedule) {

            List<Course> courseList = schedule.getCourseList();

            List<Lecture> lectureList = new ArrayList<Lecture>(courseList.size());

            for (Course course : courseList) {

                for (int i = 0; i < course.getLectureSize(); i++) {

                    Lecture lecture = new Lecture();

                    lecture.setCourse(course);

                    lecture.setLectureIndexInCourse(i);

                    // Notice that we leave the PlanningVariable properties (period and room) on null

                    lectureList.add(lecture);

                }

            }

            schedule.setLectureList(lectureList);

        }

4.4. Use the `Solver`

4.4.1. The Solver interface

A Solver implementation will solve your planning problem.

public interface Solver {


    void setPlanningProblem(Solution planningProblem);

    

    void solve();


    Solution getBestSolution();


    // ...


}

A Solver can only solve 1 planning problem instance at a time. A Solver should only be accessed from a single thread, except for the methods that are specifically javadocced as being thread-safe. It's build with a SolverFactory, do not implement or build it yourself.

4.4.2. Solving a problem

Solving a problem is quite easy once you have:

A Solver build from a solver configuration
A Solution that represents the planning problem instance

Just set the planning problem, solve it and extract the best solution:

    solver.setPlanningProblem(planningProblem);

    solver.solve();

    Solution bestSolution = solver.getBestSolution();

For example in n queens, the method getBestSolution() will return an NQueens instance with every Queen assigned to a Row.

Figure 4.2. Best solution for the 4 queens puzzle in 8 ms (also an optimal solution)

The solve() method can take a long time (depending on the problem size and the solver configuration). The Solver will remember (actually clone) the best solution it encounters during its solving. Depending on a number factors (including problem size, how much time the Solver has, the solver configuration, ...), that best solution will be a feasible or even an optimal solution.

Note

The Solution instance given to the method setPlanningProblem() will be changed by the Solver, but it do not mistake it for the best solution.

The Solution instance returned by the method getBestSolution() will most likely be a clone of the instance given to the method setPlanningProblem(), which means it's a different instance.

Note

The Solution instance given to the method setPlanningProblem() does not need to be uninitialized. It can be partially or fully initialized, which is likely to be the case in repeated planning.

4.4.3. Environment mode: Are there bugs in my code?

The environment mode allows you to detect common bugs in your implementation. It does not affect the logging level.

You can set the environment mode in the solver configuration XML file:


<solver>

  <environmentMode>FAST_ASSERT</environmentMode>

  ...

</solver>

A solver has a single Random instance. Some solver configurations use the Random instance a lot more than others. For example simulated annealing depends highly on random numbers, while tabu search only depends on it to deal with score ties. The environment mode influences the seed of that Random instance.

There are 4 environment modes:

Note

Despite the reproducible mode, your application might still not be fully reproducible because of:

Use of HashSet (or another Collection which has an inconsistent order between JVM runs) for collections of planning entities or planning values (but not normal problem facts), especially in the Solution implementation. Replace it with LinkedHashSet.
Combining a time gradient dependent algorithms (most notably simulated annealing) together with time spend termination. A sufficiently large difference in allocated CPU time will influence the time gradient values. Replace simulated annealing with late acceptance. Or instead, replace time spend termination with step count termination.

The reproducible mode is not much slower than the production mode. If your production environment requires reproducibility, use it in production too.

In practice, this mode uses the default random seed, and it also disables certain concurrency optimizations (such as work stealing).

4.4.3.4. PRODUCTION

The production mode is the fastest and the most robust, but not reproducible. It is recommended for a production environment.

The random seed is different on every run, which makes it more robust against an unlucky random seed. An unlucky random seed gives a bad result on a certain data set with a certain solver configuration. Note that in most use cases the impact of the random seed is relatively low on the result (even with simulated annealing). An occasional bad result is far more likely to be caused by another issue (such as a score trap).

4.4.4. Logging level: What is the `Solver` doing?

The best way to illuminate the black box that is a Solver, is to play with the logging level:

error: Log errors, except those that are thrown to the calling code as a RuntimeException.
Note
If an error happens, Planner normally fails fast: it throws a subclass of RuntimeException with a detailed message to the calling code. It does not log it as an error itself to avoid duplicate log messages. Except if the calling code explicitly catches and eats that RuntimeException, a Thread's default ExceptionHandler will log it as an error anyway. Meanwhile, the code is disrupted from doing further harm or obfuscating the error.
warn: Log suspicious circumstances.
info: Log every phase and the solver itself. See scope overview.
debug: Log every step of every phase. See scope overview.
trace: Log every move of every step of every phase. See scope overview.
Note
Turning on trace logging, will slow down performance considerably: it's often 4 times slower. However, it's invaluable during development to discover a bottleneck.

For example, set it to debug logging, to see when the phases end and how fast steps are taken:

INFO  Solving started: time spend (0), score (null), new best score (null), random seed (0).
DEBUG     Step index (0), time spend (1), score (0), initialized planning entity (col2@row0).
DEBUG     Step index (1), time spend (3), score (0), initialized planning entity (col1@row2).
DEBUG     Step index (2), time spend (4), score (0), initialized planning entity (col3@row3).
DEBUG     Step index (3), time spend (5), score (-1), initialized planning entity (col0@row1).
INFO  Phase (0) constructionHeuristic ended: step total (4), time spend (6), best score (-1).
DEBUG     Step index (0), time spend (10), score (-1),     best score (-1), accepted/selected move count (12/12) for picked step (col1@row2 => row3).
DEBUG     Step index (1), time spend (12), score (0), new best score (0), accepted/selected move count (12/12) for picked step (col3@row3 => row2).
INFO  Phase (1) localSearch ended: step total (2), time spend (13), best score (0).
INFO  Solving ended: time spend (13), best score (0), average calculate count per second (4846).

All time spends are in milliseconds.

Everything is logged to SLF4J, which is a simple logging facade which delegates every log message to Logback, Apache Commons Logging, Log4j or java.util.logging. Add a dependency to the logging adaptor for your logging framework of choice.

If you're not using any logging framework yet, use Logback by adding this Maven dependency (there is no need to add an extra bridge dependency):


    <dependency>

      <groupId>ch.qos.logback</groupId>

      <artifactId>logback-classic</artifactId>

      <version>1.x</version>

    </dependency>

Configure the logging level on the package org.optaplanner in your logback.xml file:


<configuration>



  <logger name="org.optaplanner" level="debug"/>



  ...



<configuration>

If instead, you're still using Log4J (and you don't want to switch to its faster successor, Logback), add the bridge dependency:


    <dependency>

      <groupId>org.slf4j</groupId>

      <artifactId>slf4j-log4j12</artifactId>

      <version>1.x</version>

    </dependency>

And configure the logging level on the package org.optaplanner in your log4j.xml file:


<log4j:configuration xmlns:log4j="http://jakarta.apache.org/log4j/">



  <category name="org.optaplanner">

    <priority value="debug" />

  </category>



  ...



</log4j:configuration>

Chapter 5. Score calculation

5.1. Score terminology

5.1.1. What is a score?
5.1.2. Score constraint signum (positive or negative)
5.1.3. Score constraint weight
5.1.4. Score level
5.1.5. Pareto scoring (AKA multi-objective optimization scoring)
5.1.6. Combining score techniques
5.1.7. The Score interface

5.2. Choose a Score definition

5.2.1. SimpleScore
5.2.2. HardSoftScore (recommended)
5.2.3. HardMediumSoftScore
5.2.4. BendableScore
5.2.5. Implementing a custom Score

5.3. Calculate the Score

5.3.1. Score calculation types
5.3.2. Simple Java score calculation
5.3.3. Incremental Java score calculation
5.3.4. Drools score calculation
5.3.5. Detecting invalid scores

5.4. Score calculation performance tricks

5.4.1. Overview
5.4.2. Average calculation count per second
5.4.3. Incremental score calculation (with delta's)
5.4.4. Avoid calling remote services during score calculation
5.4.5. Unused constraint
5.4.6. Build-in hard constraint
5.4.7. Other performance tricks
5.4.8. Score trap
5.4.9. stepLimit benchmark

5.5. Reusing the score calculation outside the Solver

5.1. Score terminology

5.1.1. What is a score?

Every initialized Solution has a score. That score is an objective way to compare 2 solutions: the solution with the higher score is better. The Solver aims to find the Solution with the highest Score of all possible solutions. The best solution is the Solution with the highest Score that Solver has encountered during solving, which might be the optimal solution.

Planner cannot automatically know which Solution is best for your business, so you need to tell it how to calculate the score of a given Solution according to your business needs. There are multiple score techniques that you can use and combine:

Maximize or minimize a constraint: score constraint signum (positive or negative)
Put a cost/profit on constraints: score constraint weight
Prioritize constraints: score level
Pareto scoring

5.1.2. Score constraint signum (positive or negative)

All score techniques are based on constraints. Such a constraint can be a simple pattern (such as Maximize the apple harvest in the solution) or a more complex pattern. A positive constraint is a constraint you're trying to maximize. A negative constraint is a constraint you're trying to minimize.

Notice in the image above, that the optimal solution always has the highest score, regardless if the constraints are positive or negative.

Most planning problems have only negative constraints and therefore have a negative score. In that case, the score is usually the sum of the weight of the negative constraints being broken, with a perfect score of 0. This explains why the score of a solution of 4 queens is the negative (and not the positive!) of the number of queen couples which can attack each other.

Negative and positive constraints can be combined, even in the same score level.

Note

Don't presume your business knows all its score constraints in advance. Expect score constraints to be added or changed after the first releases.

When a constraint activates (because the negative constraint is broken or the positive constraint is fulfilled) on a certain planning entity set, it is called a constraint match.

5.1.3. Score constraint weight

Not all score constraints are equally important. If breaking one constraint is equally bad as breaking another constraint x times, then those 2 constraints have a different weight (but they are in the same score level). For example in vehicle routing, you can make 2 "unhappy driver" constraint matches count as much as 1 "fuel tank usage" constraint match:

Score weighting is often used in use cases where you can put a price tag on everything. In that case, the positive constraints maximize revenue and the negative constraints minimize expenses: together they maximize profit. Alternatively, score weighting is also often used to create social fairness. For example: nurses that request a free day on New Year's eve pay a higher weight than on a normal day.

The weight of a constraint match can be dynamically based on the planning entities involved. For example in cloud balance: the weight of the soft constraint match for an active Computer is the cost of that Computer.

5.1.4. Score level

Sometimes a score constraint outranks another score constraint, no matter how many times the other is broken. In that case, those score constraints are in different levels. For example: a nurse cannot do 2 shifts at the same time (due to the constraints of physical reality), this outranks all nurse happiness constraints.

Most use cases have only 2 score levels: hard and soft. When comparing 2 scores, they are compared lexicographically: the first score level gets compared first. If those differ, the others score levels are ignored. For example: a score that breaks 0 hard constraints and 1000000 soft constraints is better than a score that breaks 1 hard constraint and 0 soft constraints.

Score levels often employ score weighting per level. In such case, the hard constraint level usually makes the solution feasible and the soft constraint level maximizes profit by weighting the constraints on price.

Don't use a big constraint weight when your business actually wants different score levels. That hack, known as score folding, is broken:

Note

Your business will probably tell you that your hard constraints all have the same weight, because they cannot be broken (so their weight does not matter). This is not true and it could create a score trap. For example in cloud balance: if a Computer has 7 CPU too little for its Processes, then it must be weighted 7 times as much as if it had only 1 CPU too little. This way, there is an incentive to move a Process with 6 CPU or less away from that Computer.

3 or more score levels is supported. For example: a company might decide that profit outranks employee satisfaction (or visa versa), while both are outranked by the constraints of physical reality.

5.1.5. Pareto scoring (AKA multi-objective optimization scoring)

Far less common is the use case of pareto optimization, which is also known under the more confusing term multi-objective optimization. In pareto scoring, score constraints are in the same score level, yet they are not weighted against each other. When 2 scores are compared, each of the score constraints are compared individually and the score with the most dominating score constraints wins. Pareto scoring can even be combined with score levels and score constraint weighting.

Consider this example with positive constraints, where we want to get the most apples and oranges. Since it's impossible to compare apples and oranges, we can't weight them against each other. Yet, despite that we can't compare them, we can state that 2 apples are better then 1 apple. Similarly, we can state that 2 apples and 1 orange are better than just 1 orange. So despite our inability to compare some Scores conclusively (at which point we declare them equal), we can find a set of optimal scores. Those are called pareto optimal.

Scores are considered equal far more often. It's left up to a human to choose the better out of a set of best solutions (with equal scores) found by Planner. In the example above, the user must choose between solution A (3 apples and 1 orange) and solution B (1 apples and 6 oranges). It's guaranteed that Planner has not found another solution which has more apples or more oranges or even a better combination of both (such as 2 apples and 3 oranges).

To implement pareto scoring in Planner, implement a custom ScoreDefinition and Score. Future versions will provide out-of-the-box support.

Note

A pareto Score's method compareTo is not transitive because it does a pareto comparison. For example: 2 apples is greater than 1 apple. 1 apples is equal to 1 orange. Yet, 2 apples are not greater than 1 orange (but actually equal). Pareto comparison violates the contract of the interface java.lang.Comparable's method compareTo, but Planner's systems are pareto comparison safe, unless explicitly stated otherwise in this documentation.

5.1.6. Combining score techniques

All the score techniques mentioned above, can be combined seamlessly:

5.1.7. The `Score` interface

A score is represented by the Score interface, which naturally extends Comparable:

public interface Score<...> extends Comparable<...> {

    ...

}

The Score implementation to use depends on your use case. Your score might not efficiently fit in a single double value. Planner has several build-in Score implementations, but you can implement a custom Score too. Most use cases will just use the build-in HardSoftScore.

The Score implementation (for example HardSoftScore) must be the same throughout a Solver runtime. The Score implementation is configured in the solver configuration as a ScoreDefinition:


  <scoreDirectorFactory>

    <scoreDefinitionType>HARD_SOFT</scoreDefinitionType>

    ...

  </scoreDirectorFactory>

As for all calculations with a computer, choose the correct type. Don't use a double for financial data.

Based on your score constraints and score level requirements, you'll choose a certain ScoreDefinition:

5.2. Choose a Score definition

Each Score implementation also has a ScoreDefinition implementation. For example: SimpleScore is definied by SimpleScoreDefinition.

5.2.1. SimpleScore

A SimpleScore has a single int value, for example -123. It has a single score level.


  <scoreDirectorFactory>

    <scoreDefinitionType>SIMPLE</scoreDefinitionType>

    ...

  </scoreDirectorFactory>

Variants of this scoreDefinitionType:

SIMPLE_LONG: Uses SimpleLongScore which has a long value instead of an int value.
SIMPLE_DOUBLE: Uses SimpleDoubleScore which has a double value instead of an int value.
SIMPLE_BIG_DECIMAL: Uses SimpleBigDecimalScore which has a BigDecimal value instead of an int value.

5.2.2. HardSoftScore (recommended)

A HardSoftScore has a hard int value and a soft int value, for example -123hard/-456soft. It has 2 score levels (hard and soft).


  <scoreDirectorFactory>

    <scoreDefinitionType>HARD_SOFT</scoreDefinitionType>

    ...

  </scoreDirectorFactory>

Variants of this scoreDefinitionType:

HARD_SOFT_LONG: Uses HardSoftLongScore which has long values instead of int values.
HARD_SOFT_DOUBLE: Uses HardSoftDoubleScore which has double values instead of int values.
HARD_SOFT_BIG_DECIMAL: Uses HardSoftBigDecimalScore which has BigDecimal values instead of int values..

5.2.3. HardMediumSoftScore

A HardMediumSoftScore which has a hard int value, a medium int value and a soft int value, for example -123hard/-456medium/-789soft. It has 3 score levels (hard, medium and soft).


  <scoreDirectorFactory>

    <scoreDefinitionType>HARD_MEDIUM_SOFT</scoreDefinitionType>

    ...

  </scoreDirectorFactory>

5.2.4. BendableScore

A BendableScore has a configurable number of score levels. It has an array of hard int values and an array of soft int value, for example 2 hard levels and 3 soft levels for a score -123/-456/-789/-012/-345. The number of hard and soft score levels needs to be set at configuration time, it's not flexible to change during solving.


  <scoreDirectorFactory>

    <scoreDefinitionType>BENDABLE</scoreDefinitionType>

    <bendableHardLevelCount>2</bendableHardLevelCount>

    <bendableSoftLevelCount>3</bendableSoftLevelCount>

    ...

  </scoreDirectorFactory>

5.2.5. Implementing a custom Score

The ScoreDefinition interface defines the score representation.

To implement a custom Score, you'll also need to implement a custom ScoreDefinition. Extend AbstractScoreDefinition (preferable by copy pasting HardSoftScoreDefinition or SimpleScoreDefinition) and start from there.

Then hook your custom ScoreDefinition in your SolverConfig.xml:


  <scoreDirectorFactory>

    <scoreDefinitionClass>...MyScoreDefinition</scoreDefinitionClass>

    ...

  </scoreDirectorFactory>

5.3. Calculate the `Score`

5.3.1. Score calculation types

There are several ways to calculate the Score of a Solution:

Simple Java score calculation: implement a single Java method
Incremental Java score calculation: implement multiple Java methods
Drools score calculation: implement score rules

Every score calculation type can use any Score definition. For example, simple Java score calculation can output a HardSoftScore.

All score calculation types are Object Orientated and can reuse existing Java code.

5.3.2. Simple Java score calculation

A simple way to implement your score calculation in Java.

Advantages:
- Plain old Java: no learning curve
- Opportunity to delegate score calculation to an existing code base or legacy system
Disadvantages:
- Slower and less scalable
  - Because there is no incremental score calculation

Just implement one method of the interface SimpleScoreCalculator:

public interface SimpleScoreCalculator<Sol extends Solution> {


    Score calculateScore(Sol solution);

   

}

For example in n queens:

public class NQueensSimpleScoreCalculator implements SimpleScoreCalculator<NQueens> {


    public SimpleScore calculateScore(NQueens nQueens) {

        int n = nQueens.getN();

        List<Queen> queenList = nQueens.getQueenList();

        

        int score = 0;

        for (int i = 0; i < n; i++) {

            for (int j = i + 1; j < n; j++) {

                Queen leftQueen = queenList.get(i);

                Queen rightQueen = queenList.get(j);

                if (leftQueen.getRow() != null && rightQueen.getRow() != null) {

                    if (leftQueen.getRowIndex() == rightQueen.getRowIndex()) {

                        score--;

                    }

                    if (leftQueen.getAscendingDiagonalIndex() == rightQueen.getAscendingDiagonalIndex()) {

                        score--;

                    }

                    if (leftQueen.getDescendingDiagonalIndex() == rightQueen.getDescendingDiagonalIndex()) {

                        score--;

                    }

                }

            }

        }

        return SimpleScore.valueOf(score);

    }


}

Configure it in your solver configuration:


  <scoreDirectorFactory>

    <scoreDefinitionType>...</scoreDefinitionType>

    <simpleScoreCalculatorClass>org.optaplanner.examples.nqueens.solver.score.NQueensSimpleScoreCalculator</simpleScoreCalculatorClass>

  </scoreDirectorFactory>

Alternatively, build a SimpleScoreCalculator instance at runtime and set it with the programmatic API:

    solverFactory.getSolverConfig().getScoreDirectorFactoryConfig.setSimpleScoreCalculator(simpleScoreCalculator);

5.3.3. Incremental Java score calculation

A way to implement your score calculation incrementally in Java.

Advantages:
- Very fast and scalable
  - Currently the fastest if implemented correctly
Disadvantages:
- Hard to write
  - A scalable implementation heavily uses maps, indexes, ... (things the Drools rule engine can do for you)
  - You have to learn, design, write and improve all these performance optimizations yourself
- Hard to read
  - Regular score constraint changes can lead to a high maintenance cost

Implement all the methods of the interface IncrementalScoreCalculator and extend the class AbstractIncrementalScoreCalculator:

public interface IncrementalScoreCalculator<Sol extends Solution> {


    void resetWorkingSolution(Sol workingSolution);


    void beforeEntityAdded(Object entity);


    void afterEntityAdded(Object entity);


    void beforeAllVariablesChanged(Object entity);


    void afterAllVariablesChanged(Object entity);


    void beforeVariableChanged(Object entity, String variableName);


    void afterVariableChanged(Object entity, String variableName);


    void beforeEntityRemoved(Object entity);


    void afterEntityRemoved(Object entity);


    Score calculateScore();

    

}

For example in n queens:

public class NQueensAdvancedIncrementalScoreCalculator extends AbstractIncrementalScoreCalculator<NQueens> {


    private Map<Integer, List<Queen>> rowIndexMap;

    private Map<Integer, List<Queen>> ascendingDiagonalIndexMap;

    private Map<Integer, List<Queen>> descendingDiagonalIndexMap;


    private int score;


    public void resetWorkingSolution(NQueens nQueens) {

        int n = nQueens.getN();

        rowIndexMap = new HashMap<Integer, List<Queen>>(n);

        ascendingDiagonalIndexMap = new HashMap<Integer, List<Queen>>(n * 2);

        descendingDiagonalIndexMap = new HashMap<Integer, List<Queen>>(n * 2);

        for (int i = 0; i < n; i++) {

            rowIndexMap.put(i, new ArrayList<Queen>(n));

            ascendingDiagonalIndexMap.put(i, new ArrayList<Queen>(n));

            descendingDiagonalIndexMap.put(i, new ArrayList<Queen>(n));

            if (i != 0) {

                ascendingDiagonalIndexMap.put(n - 1 + i, new ArrayList<Queen>(n));

                descendingDiagonalIndexMap.put((-i), new ArrayList<Queen>(n));

            }

        }

        score = 0;

        for (Queen queen : nQueens.getQueenList()) {

            insert(queen);

        }

    }


    public void beforeEntityAdded(Object entity) {

        // Do nothing

    }


    public void afterEntityAdded(Object entity) {

        insert((Queen) entity);

    }


    public void beforeAllVariablesChanged(Object entity) {

        retract((Queen) entity);

    }


    public void afterAllVariablesChanged(Object entity) {

        insert((Queen) entity);

    }


    public void beforeVariableChanged(Object entity, String variableName) {

        retract((Queen) entity);

    }


    public void afterVariableChanged(Object entity, String variableName) {

        insert((Queen) entity);

    }


    public void beforeEntityRemoved(Object entity) {

        retract((Queen) entity);

    }


    public void afterEntityRemoved(Object entity) {

        // Do nothing

    }


    private void insert(Queen queen) {

        Row row = queen.getRow();

        if (row != null) {

            int rowIndex = queen.getRowIndex();

            List<Queen> rowIndexList = rowIndexMap.get(rowIndex);

            score -= rowIndexList.size();

            rowIndexList.add(queen);

            List<Queen> ascendingDiagonalIndexList = ascendingDiagonalIndexMap.get(queen.getAscendingDiagonalIndex());

            score -= ascendingDiagonalIndexList.size();

            ascendingDiagonalIndexList.add(queen);

            List<Queen> descendingDiagonalIndexList = descendingDiagonalIndexMap.get(queen.getDescendingDiagonalIndex());

            score -= descendingDiagonalIndexList.size();

            descendingDiagonalIndexList.add(queen);

        }

    }


    private void retract(Queen queen) {

        Row row = queen.getRow();

        if (row != null) {

            List<Queen> rowIndexList = rowIndexMap.get(queen.getRowIndex());

            rowIndexList.remove(queen);

            score += rowIndexList.size();

            List<Queen> ascendingDiagonalIndexList = ascendingDiagonalIndexMap.get(queen.getAscendingDiagonalIndex());

            ascendingDiagonalIndexList.remove(queen);

            score += ascendingDiagonalIndexList.size();

            List<Queen> descendingDiagonalIndexList = descendingDiagonalIndexMap.get(queen.getDescendingDiagonalIndex());

            descendingDiagonalIndexList.remove(queen);

            score += descendingDiagonalIndexList.size();

        }

    }


    public SimpleScore calculateScore() {

        return SimpleScore.valueOf(score);

    }


}

Configure it in your solver configuration:


  <scoreDirectorFactory>

    <scoreDefinitionType>...</scoreDefinitionType>

    <incrementalScoreCalculatorClass>org.optaplanner.examples.nqueens.solver.score.NQueensAdvancedIncrementalScoreCalculator</incrementalScoreCalculatorClass>

  </scoreDirectorFactory>

Optionally, to get better output when the IncrementalScoreCalculator is corrupted in environmentMode FAST_ASSERT or FULL_ASSERT, you can overwrite the method buildScoreCorruptionAnalysis from AbstractIncrementalScoreCalculator.

5.3.4. Drools score calculation

5.3.4.1. Overview

Implement your score calculation using the Drools rule engine. Every score constraint is written as one or more score rules.

Advantages:
Disadvantages:
- DRL learning curve
- Usage of DRL
  - Polyglot fear can prohibit the use of a new language such as DRL in some organizations

5.3.4.2. Drools score rules configuration

There are several ways to define where your score rules live.

5.3.4.2.1. A scoreDrl resource on the classpath

This is the easy way: the score rule live in a DRL file which is a resource on the classpath. Just add your score rules *.drl file in the solver configuration:


  <scoreDirectorFactory>

    <scoreDefinitionType>...</scoreDefinitionType>

    <scoreDrl>/org/optaplanner/examples/nqueens/solver/nQueensScoreRules.drl</scoreDrl>

  </scoreDirectorFactory>

You can add multiple <scoreDrl> entries if needed, but normally you'll define all your score rules in 1 file.

Optionally, you can also set drools configuration properties but beware of backwards compatibility issues:


  <scoreDirectorFactory>

    ...

    <scoreDrl>/org/optaplanner/examples/nqueens/solver/nQueensScoreRules.drl</scoreDrl>

    <kieBaseConfigurationProperties>

      <drools.equalityBehavior>...</drools.equalityBehavior>

    </kieBaseConfigurationProperties>

  </scoreDirectorFactory>

5.3.4.2.2. A KieBase (possibly defined by Guvnor)

If you prefer to build the KieBase yourself or if you're combining Planner with Guvnor, you can set the KieBase on the SolverFactory before building the Solver:

    solverFactory.getSolverConfig().getScoreDirectorFactoryConfig.setKieBase(kieBase);

5.3.4.3. Implementing a score rule

Here's an example of a score constraint implemented as a score rule in a DRL file:

rule "multipleQueensHorizontal"
    when
        $q1 : Queen($id : id, $y : y);
        $q2 : Queen(id > $id, y == $y);
    then
        scoreHolder.addConstraintMatch(kcontext, -1);
end

This score rule will fire once for every 2 queens with the same y. The (id > $id) condition is needed to assure that for 2 queens A and B, it can only fire for (A, B) and not for (B, A), (A, A) or (B, B). Let's take a closer look at this score rule on this solution of 4 queens:

In this solution the multipleQueensHorizontal score rule will fire for 6 queen couples: (A, B), (A, C), (A, D), (B, C), (B, D) and (C, D). Because none of the queens are on the same vertical or diagonal line, this solution will have a score of -6. An optimal solution of 4 queens has a score of 0.

Note

Notice that every score rule will relate to at least 1 planning entity class (directly or indirectly though a logically inserted fact).

This is normal: it would be a waste of time to write a score rule that only relates to problem facts, as the consequence will never change during planning, no matter what the possible solution.

Note

The variable kcontext is a magic variable in Drools Expert. The scoreHolder's method uses it to do incremental score calculation correctly and to create a ConstraintMatch instance.

5.3.4.4. Weighing score rules

A ScoreHolder instance is asserted into the KieSession as a global called scoreHolder. Your score rules need to (directly or indirectly) update that instance.

global SimpleScoreHolder scoreHolder;

rule "multipleQueensHorizontal"
    when
        $q1 : Queen($id : id, $y : y);
        $q2 : Queen(id > $id, y == $y);
    then
        scoreHolder.addConstraintMatch(kcontext, -1);
end

// multipleQueensVertical is obsolete because it is always 0

rule "multipleQueensAscendingDiagonal"
    when
        $q1 : Queen($id : id, $ascendingD : ascendingD);
        $q2 : Queen(id > $id, ascendingD == $ascendingD);
    then
        scoreHolder.addConstraintMatch(kcontext, -1);
end

rule "multipleQueensDescendingDiagonal"
    when
        $q1 : Queen($id : id, $descendingD : descendingD);
        $q2 : Queen(id > $id, descendingD == $descendingD);
    then
        scoreHolder.addConstraintMatch(kcontext, -1);
end

Most use cases will also weigh their constraint types or even their matches differently, by using a specific weight for each constraint match.

Here's an example from CurriculumCourse, where assigning a Lecture to a Room which is missing 2 seats is weighted equally bad as having 1 isolated Lecture in a Curriculum:

global HardSoftScoreHolder scoreHolder;

// RoomCapacity: For each lecture, the number of students that attend the course must be less or equal
// than the number of seats of all the rooms that host its lectures.
// Each student above the capacity counts as 1 point of penalty.
rule "roomCapacity"
    when
        $room : Room($capacity : capacity)
        $lecture : Lecture(room == $room, studentSize > $capacity, $studentSize : studentSize)
    then
        scoreHolder.addSoftConstraintMatch(kcontext, ($capacity - $studentSize));
end

// CurriculumCompactness: Lectures belonging to a curriculum should be adjacent
// to each other (i.e., in consecutive periods).
// For a given curriculum we account for a violation every time there is one lecture not adjacent
// to any other lecture within the same day.
// Each isolated lecture in a curriculum counts as 2 points of penalty.
rule "curriculumCompactness"
    when
        ...
    then
        scoreHolder.addSoftConstraintMatch(kcontext, -2);
end

5.3.5. Detecting invalid scores

Put the environmentMode in FULL_ASSERT (or FAST_ASSERT) to detect corruption in the incremental score calculation. For more information, see the section about environmentMode. However, that will not verify that your score calculator implements your score constraints as your business actually desires.

A piece of incremental score calculator code can be difficult to write and to review. Assert its correctness by using a different implementation (for example a SimpleScoreCalculator) to do the assertions triggered by the environmentMode. Just configure the different implementation as a assertionScoreDirectorFactory:


  <environmentMode>FAST_ASSERT</environmentMode>

  ...

  <scoreDirectorFactory>

    <scoreDefinitionType>...</scoreDefinitionType>

    <scoreDrl>/org/optaplanner/examples/nqueens/solver/nQueensScoreRules.drl</scoreDrl>

    <assertionScoreDirectorFactory>

      <simpleScoreCalculatorClass>org.optaplanner.examples.nqueens.solver.score.NQueensSimpleScoreCalculator</simpleScoreCalculatorClass>

    </assertionScoreDirectorFactory>

  </scoreDirectorFactory>

This way, the scoreDrl will be validated by the SimpleScoreCalculator.

5.4. Score calculation performance tricks

5.4.1. Overview

The Solver will normally spend most of its execution time running the score calculation (which is called in its deepest loops). Faster score calculation will return the same solution in less time with the same algorithm, which normally means a better solution in equal time.

5.4.2. Average calculation count per second

After solving a problem, the Solver will log the average calculation count per second. This is a good measurement of Score calculation performance, despite that it is affected by non score calculation execution time. It depends on the problem scale of the problem dataset. Normally, even for high scale problems, it is higher than 1000, except when you're using SimpleScoreCalculator.

Important

When improving your score calculation, focus on maximizing the average calculation count per second, instead of maximizing the best score. A big improvement in score calculation can sometimes yield little or no best score improvement, for example when the algorithm is stuck in a local or global optima. If you're watching the calculation count instead, score calculation improvements are far more visible.

Furthermore, watching the calculation count, allows you to remove or add score constraints, and still compare it with the original calculation count. Comparing the best score with the original would be wrong, because it's comparing apples and oranges.

5.4.3. Incremental score calculation (with delta's)

When a Solution changes, incremental score calculation (AKA delta based score calculation), will calculate the delta with the previous state to find the new Score, instead of recalculating the entire score on every solution evaluation.

For example, if a single queen A moves from row 1 to 2, it won't bother to check if queen B and C can attack each other, since neither of them changed.

Figure 5.1. Incremental score calculation for the 4 queens puzzle

This is a huge performance and scalability gain. Drools score calculation gives you this huge scalability gain without forcing you to write a complicated incremental score calculation algorithm. Just let the Drools rule engine do the hard work.

Notice that the speedup is relative to the size of your planning problem (your n), making incremental score calculation far more scalable.

5.4.4. Avoid calling remote services during score calculation

Do not call remote services in your score calculation (except if you're bridging SimpleScoreCalculator to a legacy system). The network latency will kill your score calculation performance. Cache the results of those remote services if possible.

If some parts of a constraint can be calculated once, when the Solver starts, and never change during solving, then turn them into cached problem facts.

5.4.5. Unused constraint

If you know a certain constraint can never be broken, don't bother writing a score constraint for it. For example in n queens, the score calculation doesn't check if multiple queens occupy the same column, because a Queen's column never changes and every Solution starts with each Queen on a different column.

Note

Don't go overboard with this. If some datasets don't use a specific constraint but others do, just return out of the constraint as soon as you can. There is no need to dynamically change your score calculation based on the dataset.

5.4.6. Build-in hard constraint

Instead of implementing a hard constraint, you can sometimes make it build-in too. For example: If Course A should never be assigned to Room X, but it uses ValueRange from Solution, the Solver will often try to assign it to Room X too (only to find out that it breaks a hard constraint). Use filtered selection to define that Course A should only be assigned a Room other than X.

This tends to give a good performance gain, not just because the score calculation is faster, but mainly because most optimization algorithms will spend less time evaluating unfeasible solutions.

Note

Don't go overboard with this. Many optimization algorithms rely on the freedom to break hard constraints when changing planning entities, to get out of local optima. There is a real risk of trading short term benefits for long term harm.

5.4.7. Other performance tricks

Verify that your score calculation happens in the correct Number type. If you're making the sum of int values, don't let Drools sum it in a double which takes longer.
For optimal performance, always use server mode (java -server). We have seen performance increases of 50% by turning on server mode.
For optimal performance, use the latest Java version. For example, in the past we have seen performance increases of 30% by switching from java 1.5 to 1.6.
Always remember that premature optimization is the root of all evil. Make sure your design is flexible enough to allow configuration based tweaking.

5.4.8. Score trap

Make sure that none of your score constraints cause a score trap. A trapped score constraint uses the same weight for different constraint matches, when it could just as easily use a different weight. It effectively lumps its constraint matches together, which creates a flatlined score function for that constraint. This can cause a solution state in which several moves need to be done to resolve or lower the weight of that single constraint. Some examples of score traps:

If you need 2 doctors at each table, but you're only moving 1 doctor at a time. So the solver has no incentive to move a doctor to a table with no doctors. Punish a table with no doctors more then a table with only 1 doctor in that score constraint in the score function.
2 exams needs to be conducted at the same time, but you're only move 1 exam at a time. So the solver has a disincentive move one of those exams to another timeslot without moving the other in the same move. Add a course-grained move that moves both exams at the same time.

For example, consider this score trap. If the blue item moves from an overloaded computer to an empty computer, the hard score should improve. The trapped score implementation fails to do that:

The Solver should eventually get out of this trap, but it will take a lot of effort (especially if there are even more processes on the overloaded computer). Before they do that, they might actually start moving more processes into that overloaded computer, as there is no penalty for doing so.

Note

Avoiding score traps does not mean that your score function should be smart enough to avoid local optima. Leave it to the optimization algorithms to deal with the local optima.

Avoiding score traps means to avoid - for each score constraint individually - a flatlined score function.

Important

Always specify the degree of infeasibility. The business will often say: "if the solution is infeasible, it doesn't matter how infeasible it." While that's true for the business, it's not true for score calculation: it benefits from knowing how infeasible it is. In practice, soft constraints usually do this naturally and it's just a matter of doing it for the hard constraints too.

There are several ways to deal with a score trap:

Improve the score constraint to make a distinction in the score weight. For example: penalize -1hard for every missing CPU, instead of just -1hard if any CPU is missing.
If changing the score constraint is not allowed from the business perspective, add a lower score level with a score constraint that makes such a distinction. For example: penalize -1subsoft for every missing CPU, on top of -1hard if any CPU is missing. The business ignores the subsoft score level.
Add course-grained moves and union select them with the existing fine-grained moves. A course-grained move effectively does multiple moves to directly get out of a score trap with a single move. For example: move multiple items from the same container to another container.

5.4.9. stepLimit benchmark

Not all score constraints have the same performance cost. Sometimes 1 score constraint can kill the score calculation performance outright. Use the benchmarker to do a 1 minute run and check what happens to the average calculation count per second if you comment out all but 1 of the score constraints.

5.5. Reusing the score calculation outside the Solver

Other parts of your application, for example your webUI, might need to calculate the score too. Do that by reusing the ScoreDirectorFactory of the Solver to build a separate ScoreDirector for that webUI:

ScoreDirectorFactory scoreDirectorFactory = solver.getScoreDirectorFactory();

ScoreDirector guiScoreDirector = scoreDirectorFactory.buildScoreDirector();

Then use it when you need to calculate the Score of a Solution:

guiScoreDirector.setWorkingSolution(solution);

Score score = guiScoreDirector.calculateScore();

Tto explain in the GUI what entities are causing which part of the Score, get the ConstraintMatch objects from the ScoreDirector.

Chapter 6. Optimization algorithms

6.1. Search space size in the real world

6.2. Does Planner find the optimal solution?

6.3. Architecture overview

6.4. Optimization algorithms overview

6.5. Which optimization algorithms should I use?

6.6. SolverPhase

6.7. Scope overview

6.8. Termination

6.8.1. TimeMillisSpendTermination
6.8.2. ScoreAttainedTermination
6.8.3. StepCountTermination
6.8.4. UnimprovedStepCountTermination
6.8.5. Combining multiple Terminations
6.8.6. Asynchronous termination from another thread

6.9. SolverEventListener

6.10. Custom SolverPhase

6.1. Search space size in the real world

The number of possible solutions for a planning problem can be mind blowing. For example:

4 queens has 256 possible solutions (4 ^ 4) and 2 optimal solutions.
5 queens has 3125 possible solutions (5 ^ 5) and 1 optimal solution.
8 queens has 16777216 possible solutions (8 ^ 8) and 92 optimal solutions.
64 queens has more than 10^115 possible solutions (64 ^ 64).
Most real-life planning problems have an incredible number of possible solutions and only 1 or a few optimal solutions.

For comparison: the minimal number of atoms in the known universe (10^80). As a planning problem gets bigger, the search space tends to blow up really fast. Adding only 1 extra planning entity or planning value can heavily multiply the running time of some algorithms.

An algorithm that checks every possible solution (even with pruning) can easily run for billions of years on a single real-life planning problem. What we really want is to find the best solution in the limited time at our disposal. Planning competitions (such as the International Timetabling Competition) show that local search variations (tabu search, simulated annealing, ...) usually perform best for real-world problems given real-world time limitations.

6.2. Does Planner find the optimal solution?

The business wants the optimal solution, but they also have other requirements:

Scale out: Large production datasets must not crash and have good results too.
Optimize the right problem: The constraints must match the actual business needs.
Available time: The solution must be found in time, before it becomes useless to execute.
Reliability: Every dataset must have at least a decent result (better than a human planner).

Given these requirements, and despite the promises of some salesmen, it's usually impossible for anyone or anything to find the optimal solution. Therefore, Planner focuses on finding the best solution in available time. In realistic, independent competitions, Planner often comes out as the best reusable software.

The nature of NP-complete problems make scaling a prime concern. The result quality of a small dataset guarantees nothing about the result quality of a large dataset. Scaling problems cannot be mitigated by hardware purchases. Start testing with a production sized dataset as soon as possible. Don't asses quality on small datasets (unless production encounters such datasets). Instead, solve a production sized dataset and compare with the results of longer execution, different algorithms and - if available - the human planner.

6.3. Architecture overview

OptaPlanner is the first framework to combine optimization algorithms (metaheuristics, ...) with score calculation by a rule engine such as Drools Expert. This combination turns out to be a very efficient, because:

A rule engine such as Drools Expert is great for calculating the score of a solution of a planning problem. It makes it easy and scalable to add additional soft or hard constraints such as "a teacher shouldn't teach more then 7 hours a day". It does delta based score calculation without any extra code. However it tends to be not suitable to actually find new solutions.
An optimization algorithm is great at finding new improving solutions for a planning problem, without necessarily brute-forcing every possibility. However it needs to know the score of a solution and offers no support in calculating that score efficiently.

6.4. Optimization algorithms overview

Table 6.1. Optimization algorithms overview

Algorithm	Scalable?	Optimal?	Easy to use?	Tweakable?	Requires CH?
Exact algorithms
Brute force	0/5	5/5	5/5	0/5	No
Branch and bound	0/5	5/5	4/5	1/5	No
Construction heuristics (CH)
First Fit	5/5	1/5	5/5	1/5	No
First Fit Decreasing	5/5	2/5	4/5	2/5	No
Best Fit	5/5	2/5	4/5	2/5	No
Best Fit Decreasing	5/5	2/5	4/5	2/5	No
Cheapest Insertion	3/5	2/5	5/5	2/5	No
Metaheuristics (MH)
Local search
Hill-climbing	5/5	2/5	4/5	3/5	Yes
Tabu search	5/5	4/5	3/5	5/5	Yes
Simulated annealing	5/5	4/5	2/5	5/5	Yes
Late acceptance	5/5	4/5	3/5	5/5	Yes
Evolutionary algorithms
Evolutionary strategies	4/5	3/5	?/5	?/5	Yes
Genetic algorithms	4/5	3/5	?/5	?/5	Yes

If you want to learn more about metaheuristics, read the free book Essentials of Metaheuristics or Clever Algorithms.

6.5. Which optimization algorithms should I use?

The best optimization algorithms configuration for your use case depends heavily on your use case. Nevertheless, this vanilla recipe will get you into the game with a pretty good configuration, probably much better than what you're used to.

Start with a quick configuration that involves little or no configuration and optimization code:

First Fit

Next, implement planning entity difficulty comparison and turn it into:

First Fit Decreasing

Next, implement moves and add tabu search behind it:

First Fit Decreasing
Tabu search (use planning entity tabu)

At this point the free lunch is over. The return on invested time lowers. The result is probably already more than good enough.

But you can do even better, at a lower return on invested time. Use the Benchmarker and try a couple of simulated annealing configurations:

First Fit Decreasing
Simulated annealing (try several starting temperatures)

And combine them with tabu search:

First Fit Decreasing
Simulated annealing (relatively long time)
Tabu search (relatively short time)

If you have time, continue experimenting even further. Blog about your experiments!

6.6. `SolverPhase`

A Solver can use multiple optimization algorithms in sequence. Each optimization algorithm is represented by a SolverPhase. There is never more than 1 SolverPhase solving at the same time.

Note

Some SolverPhase implementations can combine techniques from multiple optimization algorithms, but they are still just 1 SolverPhase. For example: a local search SolverPhase can do simulated annealing with planning entity tabu.

Here's a configuration that runs 3 phases in sequence:


<solver>

  ...

  <constructionHeuristic>

    ... <!-- First phase: First Fit decreasing -->

  </constructionHeuristic>

  <localSearch>

    ... <!-- Second phase: Simulated annealing -->

  </localSearch>

  <localSearch>

    ... <!-- Third phase: Tabu search -->

  </localSearch>

</solver>

The solver phases are run in the order defined by solver configuration. When the first phase terminates, the second phase starts, and so on. When the last phase terminates, the Solver terminates. Usually, a solver will first run a construction heuristic and then run 1 or multiple metaheuristics:

Some phases (especially construction heuristics) will terminate automatically. Other phases (especially metaheuristics) will only terminate if the phase is configured to terminate:


<solver>

  ...

  <termination><!-- Solver termination -->

    <maximumSecondsSpend>90</maximumSecondsSpend>

  </termination>

  <localSearch>

    <termination><!-- Phase termination -->

      <maximumSecondsSpend>60</maximumSecondsSpend><!-- Give the next phase a chance to run too, before the Solver terminates -->

    </termination>

    ...

  </localSearch>

  <localSearch>

    ...

  </localSearch>

</solver>

If the Solver terminates (before the last phase terminates itself), the current phase is terminated and all subsequent phases won't run.

6.7. Scope overview

A solver will iteratively run phases. Each phase will usually iteratively run steps. Each step, in turn, usually iteratively runs moves. These form 4 nested scopes: solver, phase, step and move.

Configure logging to display the log messages of each scope.

6.8. Termination

Not all phases terminate automatically and sometimes you don't want to wait that long anyway. A Solver can be terminated synchronously by up-front configuration or asynchronously from another thread.

Especially metaheuristic phases will need to be told when to stop solving. This can be because of a number of reasons: the time is up, the perfect score has been reached, ... The only thing you can't depend on, is on finding the optimal solution (unless you know the optimal score), because a metaheuristic algorithm generally doesn't know it when it finds the optimal solution. For real-life problems this doesn't turn out to be much of a problem, because finding the optimal solution could take billions of years, so you 'll want to terminate sooner anyway. The only thing that matters is finding the best solution in the available time.

For synchronous termination, configure a Termination on a Solver or a SolverPhase when it needs to stop. You can implement your own Termination, but the build-in implementations should suffice for most needs. Every Termination can calculate a time gradient (needed for some optimization algorithms), which is a ratio between the time already spend solving and the estimated entire solving time of the Solver or SolverPhase.

6.8.1. TimeMillisSpendTermination

Terminates when an amount of time has been reached:


    <termination>

        <maximumTimeMillisSpend>500</maximumTimeMillisSpend>

    </termination>


    <termination>

        <maximumSecondsSpend>10</maximumSecondsSpend>

    </termination>


    <termination>

        <maximumMinutesSpend>5</maximumMinutesSpend>

    </termination>


    <termination>

        <maximumHoursSpend>1</maximumHoursSpend>

    </termination>

Note

If you use this Termination, you will most likely sacrifice perfect reproducibility (even with environmentMode REPRODUCIBLE) because of an available CPU time difference:

The available CPU time influences the number of steps that can be taken, which might be a few more or less.
The Termination might produce slightly different time gradient values, which will send time gradient based algorithms (such as simulated annealing) on a radically different path.

6.8.2. ScoreAttainedTermination

Terminates when a certain score has been reached. You can use this Termination if you know the perfect score, for example for 4 queens:


    <termination>

        <scoreAttained>0</scoreAttained>

    </termination>

For a planning problem with hard and soft constraints, it could look like this:


    <termination>

        <scoreAttained>0hard/-5000soft</scoreAttained>

    </termination>

You can use this Termination to terminate once it reaches a feasible solution.

6.8.3. StepCountTermination

Terminates when an amount of steps has been reached:


    <termination>

        <maximumStepCount>100</maximumStepCount>

    </termination>

This Termination can only be used for a SolverPhase, not for the Solver itself.

6.8.4. UnimprovedStepCountTermination

Terminates when the best score hasn't improved in a number of steps:


    <termination>

        <maximumUnimprovedStepCount>100</maximumUnimprovedStepCount>

    </termination>

If the score hasn't improved recently, it's probably not going to improve soon anyway and it's not worth the effort to continue. We have observed that once a new best solution is found (even after a long time of no improvement on the best solution), the next few steps tend to improve the best solution too.

This Termination can only be used for a SolverPhase, not for the Solver itself.

6.8.5. Combining multiple Terminations

Terminations can be combined, for example: terminate after 100 steps or if a score of 0 has been reached:


    <termination>

        <terminationCompositionStyle>OR</terminationCompositionStyle>

        <maximumStepCount>100</maximumStepCount>

        <scoreAttained>0</scoreAttained>

    </termination>

Alternatively you can use AND, for example: terminate after reaching a feasible score of at least -100 and no improvements in 5 steps:


    <termination>

        <terminationCompositionStyle>AND</terminationCompositionStyle>

        <maximumUnimprovedStepCount>5</maximumUnimprovedStepCount>

        <scoreAttained>-100</scoreAttained>

    </termination>

This example ensures it doesn't just terminate after finding a feasible solution, but also completes any obvious improvements on that solution before terminating.

6.8.6. Asynchronous termination from another thread

Sometimes you'll want to terminate a Solver early from another thread, for example because a user action or a server restart. This cannot be configured by a Termination as it's impossible to predict when and if it will occur. Therefore the Solver interface has these 2 thread-safe methods:

public interface Solver {


    // ...


    boolean terminateEarly();

    boolean isTerminateEarly();


}

If you call the terminateEarly() method from another thread, the Solver will terminate at its earliest convenience and the solve() method will return in the original Solver thread.

6.9. SolverEventListener

Each time a new best solution is found, the Solver fires a BestSolutionChangedEvent.

To listen to such events, add a SolverEventListener to the Solver:

public interface Solver {


    // ...


    void addEventListener(SolverEventListener eventListener);

    void removeEventListener(SolverEventListener eventListener);


}

6.10. Custom SolverPhase

Between phases or before the first phase, you might want to execute a custom action on the Solution to get a better score. Yet you'll still want to reuse the score calculation. For example, to implement a custom construction heuristic without implementing an entire SolverPhase.

Note

Most of the time, a custom construction heuristic is not worth the hassle. The supported constructions heuristics are configurable (so you can tweak them with the Benchmarker), Termination aware and support partially initialized solutions too.

Implement the CustomSolverPhaseCommand interface:

public interface CustomSolverPhaseCommand {


    void changeWorkingSolution(ScoreDirector scoreDirector);


}

For example:

public class ExaminationSolutionInitializer implements CustomSolverPhaseCommand {


    public void changeWorkingSolution(ScoreDirector scoreDirector) {

        Examination examination = (Examination) scoreDirector.getWorkingSolution();

        for (Exam exam : examination.getExamList()) {

            Score unscheduledScore = scoreDirector.calculateScore();

            ...

            for (Period period : examination.getPeriodList()) {

                scoreDirector.beforeVariableChanged(exam, "period");

                exam.setPeriod(period)

                scoreDirector.afterVariableChanged(exam, "period");

                Score score = scoreDirector.calculateScore();

                ...

            }

            ...

        }

    }


}

Warning

Any change on the planning entities in a CustomSolverPhaseCommand must be notified to the ScoreDirector.

Warning

Do not change any of the planning facts in a CustomSolverPhaseCommand. That will corrupt the Solver because any previous score or solution was for a different problem. If you want to do that, see repeated planning and real-time planning instead.

And configure it like this:


<solver>

  ...

  <customSolverPhase>

    <customSolverPhaseCommandClass>org.optaplanner.examples.examination.solver.solution.initializer.ExaminationSolutionInitializer</customSolverPhaseCommandClass>

  </customSolverPhase>

  ... <!-- Other phases -->

</solver>

It's possible to configure multiple customSolverPhaseCommandClass instances, which will be run in sequence.

Important

If the changes of a CustomSolverPhaseCommand don't result in a better score, the best solution won't be changed (so effectively nothing will have changed for the next SolverPhase or CustomSolverPhaseCommand). To force such changes anyway, use forceUpdateBestSolution:


  <customSolverPhase>

    <customSolverPhaseCommandClass>...MyUninitializer</customSolverPhaseCommandClass>

    <forceUpdateBestSolution>true</forceUpdateBestSolution>

  </customSolverPhase>

Note

If the Solver or SolverPhase wants to terminate while a CustomSolverPhaseCommand is still running, it will wait to terminate until the CustomSolverPhaseCommand is done, however long that takes.

Chapter 7. `Move` and neighborhood selection

7.1. Move and neighborhood introduction

7.1.1. What is a Move?
7.1.2. What is a MoveSelector?
7.1.3. Subselecting of entities, values and other moves

7.2. General Selector features

7.2.1. CacheType: Create moves ahead of time or Just In Time
7.2.2. SelectionOrder: original, sorted, random, shuffled or probabilistic
7.2.3. Recommended combinations of CacheType and SelectionOrder
7.2.4. Filtered selection
7.2.5. Sorted selection
7.2.6. Probabilistic selection

7.3. Generic MoveSelectors

7.3.1. changeMoveSelector
7.3.2. swapMoveSelector
7.3.3. pillarSwapMoveSelector
7.3.4. subChainChangeMoveSelector
7.3.5. subChainSwapMoveSelector

7.4. Combining multiple MoveSelectors

7.4.1. unionMoveSelector
7.4.2. cartesianProductMoveSelector

7.5. EntitySelector

7.6. ValueSelector

7.7. Custom moves

7.7.1. Which move types might be missing in my implementation?
7.7.2. Custom moves introduction
7.7.3. The interface Move
7.7.4. MoveListFactory: the easy way to generate custom moves
7.7.5. MoveIteratorFactory: generate custom moves just in time
7.7.6. Move generation through DRL

7.1. `Move` and neighborhood introduction

7.1.1. What is a `Move`?

A Move is a change (or set of changes) from a solution A to a solution B. For example, the move below changes queen C from row 0 to row 2:

The new solution is called a neighbor of the original solution, because it can be reached in a single Move. Although a single move can change multiple queens, the neighbors of a solution should always be a very small subset of all possible solutions. For example, on that original solution, these are all possible changeMove's:

If we ignore the 4 changeMove's that have not impact and are therefore not doable, we can see that number of moves is n * (n - 1) = 12. This is far less than the number of possible solutions, which is n ^ n = 256. As the problem scales out, the number of possible moves increases far less than the number of possible solutions.

Yet, in 4 changeMove's or less we can reach any solution. For example we can reach a very different solution in 3 changeMove's:

Note

There are many other types of moves besides changeMove's. Many move types are included out-of-the-box, but you can also implement custom moves.

A Move can affect multiple entities or even create/delete entities. But it must not change the problem facts.

All optimization algorithms use Move's to transition from one solution to a neighbor solution. Therefor, all the optimization algorithms are confronted with Move selection: the craft of creating and iterating moves efficiently and the art of finding the most promising subset of random moves to evaluate first.

7.1.2. What is a `MoveSelector`?

A MoveSelector's main function is to create Iterator<Move> when needed. An optimization algorithm will iterate through a subset of those moves.

Here's an example how to configure a changeMoveSelector for the optimization algorithm Local Search:


  <localSearch>

    <changeMoveSelector/>

    ...

  </localSearch>

Out of the box, this works and all properties of the changeMoveSelector are defaulted sensibly (unless that fails fast due to ambiguity). On the other hand, the configuration can be customized significantly for specific use cases. For example: you want want to configure a filter to discard pointless moves.

7.1.3. Subselecting of entities, values and other moves

To create a Move, we need to select 1 or more planning entities and/or planning values to move. Just like MoveSelectors, EntitySelectors and ValueSelectors need to support a similar feature set (such as scalable just-in-time selection). Therefor, they implement a common interface Selector and they are configured similarly.

A MoveSelector is often composed out of EntitySelectors, ValueSelectors or even other MoveSelectors, which can be configured individually if desired:


    <unionMoveSelector>

      <changeMoveSelector>

        <entitySelector>

          ...

        </entitySelector>

        <valueSelector>

          ...

        </valueSelector>

        ...

      </changeMoveSelector>

      ...

    </unionMoveSelector>

Together, this structure forms a Selector tree:

The root of this tree is a MoveSelector which is injected into the optimization algorithm implementation to be (partially) iterated in every step.

7.2. General `Selector` features

7.2.1. `CacheType`: Create moves ahead of time or Just In Time

A Selector's cacheType determines when a selection (such as a Move, an entity, a value, ...) is created and how long it lives.

Almost every Selector supports setting a cacheType:


    <changeMoveSelector>

      <cacheType>PHASE</cacheType>

      ...

    </changeMoveSelector>

The following cacheTypes are supported:

JUST_IN_TIME (default): Not cached. Construct each selection (Move, ...) just before it's used. This scales up well in memory footprint.
STEP: Cached. Create each selection (Move, ...) at the beginning of a step and cache them in a list for the remainder of the step. This scales up badly in memory footprint.
PHASE: Cached. Create each selection (Move, ...) at the beginning of a SolverPhase and cache them in a list for the remainder of the SolverPhase. Some selections cannot be phase cached because the list changes every step. This scales up badly in memory footprint, but has a slight performance gain.
SOLVER: Cached. Create each selection (Move, ...) at the beginning of a Solver and cache them in a list for the remainder of the Solver. Some selections cannot be solver cached because the list changes every step. This scales up badly in memory footprint, but has a slight performance gain.

A cacheType can be set on composite selectors too:


    <unionMoveSelector>

      <cacheType>PHASE</cacheType>

      <changeMoveSelector/>

      <swapMoveSelector/>

      ...

    </unionMoveSelector>

Nested selectors of a cached selector cannot be configured to be cached themselves, unless it's a higher cacheType. For example: a STEP cached unionMoveSelector can hold a PHASE cached changeMoveSelector, but not a STEP cached changeMoveSelector.

7.2.2. SelectionOrder: original, sorted, random, shuffled or probabilistic

A Selector's selectionOrder determines the order in which the selections (such as Moves, entities, values, ...) are iterated. An optimization algorithm will usually only iterate through a subset of its MoveSelector's selections, starting from the start, so the selectionOrder is critical to decide which Moves are actually evaluated.

Almost every Selector supports setting a selectionOrder:


    <changeMoveSelector>

      ...

      <selectionOrder>RANDOM</selectionOrder>

      ...

    </changeMoveSelector>

The following selectionOrders are supported:

ORIGINAL: Select the selections (Moves, entities, values, ...) in default order. Each selection will be selected only once.
- For example: A0, A1, A2, A3, ..., B0, B1, B2, B3, ..., C0, C1, C2, C3, ...
SORTED: Select the selections (Moves, entities, values, ...) in sorted order. Each selection will be selected only once. Requires cacheType >= STEP. Mostly used on an entitySelector or valueSelector for construction heuristics. See sorted selection.
- For example: A0, B0, C0, ..., A2, B2, C2, ..., A1, B1, C1, ...
RANDOM (default): Select the selections (Moves, entities, values, ...) in non-shuffled random order. A selection might be selected multiple times. This scales up well in performance because it does not require caching.
- For example: C2, A3, B1, C2, A0, C0, ...
SHUFFLED: Select the selections (Moves, entities, values, ...) in shuffled random order. Each selection will be selected only once. Requires cacheType >= STEP. This scales up badly in performance, not just because it requires caching, but also because a random number is generated for each element, even if it's not selected (which is the grand majority when scaling up).
- For example: C2, A3, B1, A0, C0, ...
PROBABILISTIC: Select the selections (Moves, entities, values, ...) in random order, based on the selection probability of each element. A selection with a higher probability has a higher chance to be selected than elements with a lower probability. A selection might be selected multiple times. Requires cacheType >= STEP. Mostly used on an entitySelector or valueSelector. See probabilistic selection.
- For example: B1, B1, A1, B2, B1, C2, B1, B1, ...

A selectionOrder can be set on composite selectors too.

Note

When a Selector is cached, all of its nested Selectors will naturally default to selectionOrder ORIGINAL. Avoid overwriting the selectionOrder of those nested Selectors.

7.2.3. Recommended combinations of `CacheType` and `SelectionOrder`

7.2.3.1. Just in time random selection (default)

This combination is great for big use cases (10 000 entities or more), as it scales up well in memory footprint and performance. Other combinations are often not even viable on such sizes. It works for smaller use cases too, so it's a good way to start out. It's the default, so this explicit configuration of cacheType and selectionOrder is actually obsolete:


    <unionMoveSelector>

      <cacheType>JUST_IN_TIME</cacheType>

      <selectionOrder>RANDOM</selectionOrder>



      <changeMoveSelector/>

      <swapMoveSelector/>

    </unionMoveSelector>

Here's how it works. When Iterator<Move>.next() is called, a child MoveSelector is randomly selected (1), which creates a random Move is created (2, 3, 4) and is then returned (5):

Notice that it never creates a list of Moves and it generates random numbers only for Moves that are actually selected.

7.2.3.2. Cached shuffled selection

This combination often wins for small and medium use cases (5000 entities or less). Beyond that size, it scales up badly in memory footprint and performance.


    <unionMoveSelector>

      <cacheType>PHASE</cacheType>

      <selectionOrder>SHUFFLED</selectionOrder>



      <changeMoveSelector/>

      <swapMoveSelector/>

    </unionMoveSelector>

Here's how it works: At the start of the phase (or step depending on the cacheType), all moves are created (1) and cached (2). When MoveSelector.iterator() is called, the moves are shuffled (3). When Iterator<Move>.next() is called, the next element in the shuffled list is returned (4):

Notice that each Move will only be selected once, even though they are selected in random order.

Use cacheType PHASE if none of the (possibly nested) Selectors require STEP. Otherwise, do something like this:


    <unionMoveSelector>

      <cacheType>STEP</cacheType>

      <selectionOrder>SHUFFLED</selectionOrder>



      <changeMoveSelector>

        <cacheType>PHASE</cacheType>

      </changeMoveSelector>

      <swapMoveSelector/>

        <cacheType>PHASE</cacheType>

      </swapMoveSelector>

      <pillarSwapMoveSelector/><!-- Does not support cacheType PHASE -->

    </unionMoveSelector>

7.2.3.3. Cached random selection

This combination is often a worthy competitor for medium use cases, especially with fast stepping optimization algorithms (such as simulated annealing). Unlike cached shuffled selection, it doesn't waste time shuffling the move list at the beginning of every step.


    <unionMoveSelector>

      <cacheType>PHASE</cacheType>

      <selectionOrder>RANDOM</selectionOrder>



      <changeMoveSelector/>

      <swapMoveSelector/>

    </unionMoveSelector>

7.2.4. Filtered selection

There are certain moves that you don't want to select, because:

The move is pointless and would only waste CPU time. For example, swapping 2 lectures of the same course will result in the same score and the same schedule because all lectures of 1 course are interchangeable (same teacher, same students, same topic).
Doing the move would break a build-in hard constraint, so the solution would be infeasible but the score function doesn't check build-in hard constraints (for performance gain). For example, don't change a gym lecture to a room which is not a gym room.
- Note that any build-in hard constraint must usually be filtered on every move type. For example, don't swap the room of a gym lecture with another lecture if the other lecture's original room isn't a gym room.

Filtered selection can happen on any Selector in the selector tree, including any MoveSelector, EntitySelector or ValueSelector. It works with any kind of cacheType and selectionOrder.

Filtering uses the interface SelectionFilter:

public interface SelectionFilter<T> {


    boolean accept(ScoreDirector scoreDirector, T selection);


}

Implement the method accept to return false on a discarded selection. Unaccepted moves will not be selected and will therefore never have their method doMove called.

public class DifferentCourseSwapMoveFilter implements SelectionFilter<SwapMove> {


    public boolean accept(ScoreDirector scoreDirector, SwapMove move) {

        Lecture leftLecture = (Lecture) move.getLeftEntity();

        Lecture rightLecture = (Lecture) move.getRightEntity();

        return !leftLecture.getCourse().equals(rightLecture.getCourse());

    }


}

Apply the filter on the lowest level possible. In most cases, you 'll need to know both the entity and the value involved and you'll have to apply a filterClass on the moveSelector:


    <swapMoveSelector>

      <filterClass>org.optaplanner.examples.curriculumcourse.solver.move.DifferentCourseSwapMoveFilter</filterClass>

    </swapMoveSelector>

But if possible, apply it on a lower levels, such as a filterClass on the entitySelector or valueSelector:


    <changeMoveSelector>

      <entitySelector>

        <filterClass>...EntityFilter</filterClass>

      </entitySelector>

    </changeMoveSelector>

You can configure multiple filterClass elements on a single selector.

7.2.5. Sorted selection

Sorted selection can happen on any Selector in the selector tree, including any MoveSelector, EntitySelector or ValueSelector. It does not work with cacheType JUST_IN_TIME and it only works with selectionOrder SORTED.

It's mostly used in construction heuristics.

7.2.5.1. Sorted selection by `Comparator`

The easiest way to sort a Selector is with a plain old Comparator:

public class CloudProcessDifficultyComparator implements Comparator<CloudProcess> {


    public int compare(CloudProcess a, CloudProcess b) {

        return new CompareToBuilder()

                .append(a.getRequiredMultiplicand(), b.getRequiredMultiplicand())

                .append(a.getId(), b.getId())

                .toComparison();

    }


}

You 'll also need to configure it (unless it's annotated applied for this optimization algorithm):


    <entitySelector>

      <cacheType>PHASE</cacheType>

      <selectionOrder>SORTED</selectionOrder>

      <sorterComparatorClass>...CloudProcessDifficultyComparator</sorterComparatorClass>

      <sorterOrder>DESCENDING</sorterOrder>

    </entitySelector>

7.2.5.2. Sorted selection by `SelectionSorterWeightFactory`

If you need the entire Solution to sort a Selector, use a SelectionSorterWeightFactory instead:

public interface SelectionSorterWeightFactory<Sol extends Solution, T> {


    Comparable createSorterWeight(Sol solution, T selection);


}

public class QueenDifficultyWeightFactory implements SelectionSorterWeightFactory<NQueens, Queen> {


    public Comparable createSorterWeight(NQueens nQueens, Queen queen) {

        int distanceFromMiddle = calculateDistanceFromMiddle(nQueens.getN(), queen.getColumnIndex());

        return new QueenDifficultyWeight(queen, distanceFromMiddle);

    }


    // ...


    public static class QueenDifficultyWeight implements Comparable<QueenDifficultyWeight> {


        private final Queen queen;

        private final int distanceFromMiddle;


        public QueenDifficultyWeight(Queen queen, int distanceFromMiddle) {

            this.queen = queen;

            this.distanceFromMiddle = distanceFromMiddle;

        }


        public int compareTo(QueenDifficultyWeight other) {

            return new CompareToBuilder()

                    // The more difficult queens have a lower distance to the middle

                    .append(other.distanceFromMiddle, distanceFromMiddle) // Decreasing

                    // Tie breaker

                    .append(queen.getColumnIndex(), other.queen.getColumnIndex())

                    .toComparison();

        }


    }


}

You 'll also need to configure it (unless it's annotated and automatically applied for this optimization algorithm):


    <entitySelector>

      <cacheType>PHASE</cacheType>

      <selectionOrder>SORTED</selectionOrder>

      <sorterWeightFactoryClass>...QueenDifficultyWeightFactory</sorterWeightFactoryClass>

      <sorterOrder>DESCENDING</sorterOrder>

    </entitySelector>

7.2.5.3. Sorted selection by `SelectionSorter`

Alternatively, you can also use the interface SelectionSorter directly:

public interface SelectionSorter<T> {


    void sort(ScoreDirector scoreDirector, List<T> selectionList);


}


    <entitySelector>

      <cacheType>PHASE</cacheType>

      <selectionOrder>SORTED</selectionOrder>

      <sorterClass>...MyEntitySorter</sorterClass>

    </entitySelector>

7.2.6. Probabilistic selection

Probabilistic selection can happen on any Selector in the selector tree, including any MoveSelector, EntitySelector or ValueSelector. It does not work with cacheType JUST_IN_TIME and it only works with selectionOrder PROBABILISTIC.

Each selection has a probabilityWeight, which determines the chance that's that selection will be selected:

public interface SelectionProbabilityWeightFactory<T> {


    double createProbabilityWeight(ScoreDirector scoreDirector, T selection);


}


    <entitySelector>

      <cacheType>PHASE</cacheType>

      <selectionOrder>PROBABILISTIC</selectionOrder>

      <probabilityWeightFactoryClass>...MyEntityProbabilityWeightFactoryClass</probabilityWeightFactoryClass>

    </entitySelector>

For example, if there are 3 entities: process A (probabilityWeight 2.0), process B (probabilityWeight 0.5) and process C (probabilityWeight 0.5), then process A will be selected 4 times more than B and C.

7.3. Generic MoveSelectors

7.3.1. `changeMoveSelector`

For 1 planning variable, the ChangeMove selects 1 planning entity and 1 planning value and assigns the entity's variable to that value.


    <changeMoveSelector/>

A ChangeMove is the finest grained move.


    <changeMoveSelector>

      ... <!-- Normal selector properties -->

      <entitySelector>

        <entityClass>...Lecture</entityClass>

        ...

      </entitySelector>

      <valueSelector>

        <variableName>room</variableName>

        ...

      </valueSelector>

    </changeMoveSelector>

Important

Almost every moveSelector configuration injected into a metaheuristic algorithm should include a changeMoveSelector or a custom implementation. This guarantees that every possible Solution can be reached through applying a number of moves in sequence (not taking score traps into account). Of course, normally it is unioned with other, more course grained move selectors.

7.3.2. swapMoveSelector

The SwapMove selects 2 different planning entities and swaps the planning values of all their planning variables.

Although a SwapMove on a single variable is essentially just 2 ChangeMoves, it's often the winning step where the first of the 2 ChangeMoves would not be the winning step because it leave the solution in a state with broken hard constraints. For example: swapping the room of 2 lectures doesn't bring the solution in a intermediate state where both lectures are in the same room which breaks a hard constraint.


    <swapMoveSelector/>

The secondaryEntitySelector is rarely needed: if it is not specified, entities from the same entitySelector are swapped.


    <swapMoveSelector>

      ... <!-- Normal selector properties -->

      <entitySelector>

        <entityClass>...Lecture</entityClass>

        ...

      </entitySelector>

      <secondaryEntitySelector>

        ...

      </secondaryEntitySelector>

      <variableNameInclude>room</variableNameInclude>

      <variableNameInclude>...</variableNameInclude>

    </swapMoveSelector>

If one or more variableNameInclude properties are specified, not all planning variables will be swapped, but only those specified. For example for course scheduling, specifying only variableNameInclude room will make it only swap room, not period.

7.3.3. pillarSwapMoveSelector

A pillar is a set of planning entities which have the same planning value(s) for each of their planning variables. The pillarSwapMove selects 2 different entity pillars and swaps the values of all their variables for all their entities.


    <pillarSwapMoveSelector/>

The secondaryPillarSelector is rarely needed: if it is not specified, entities from the same pillarSelector are swapped.


    <pillarSwapMoveSelector>

      ... <!-- Normal selector properties -->

      <pillarSelector>

        <entitySelector>

          <entityClass>...Lecture</entityClass>

          ...

        </entitySelector>

      </pillarSelector>

      <secondaryPillarSelector>

        <entitySelector>

          ...

        </entitySelector>

      </secondaryPillarSelector>

      <variableNameInclude>room</variableNameInclude>

      <variableNameInclude>...</variableNameInclude>

    </pillarSwapMoveSelector>

The other properties are explained in swapMoveSelector.

7.3.4. subChainChangeMoveSelector

A subChain is a set of planning entities with a chained planning variable which form part of a chain. The subChainChangeMove selects a subChain and moves it to another place in a different or the same anchor chain.


    <subChainChangeMoveSelector/>

The subChainSelector selects a number of entities, no less than minimumSubChainSize (defaults to 2) and no more than maximumSubChainSize (defaults to infinity).


    <subChainChangeMoveSelector>

      ... <!-- Normal selector properties -->

      <subChainSelector>

        <entitySelector>

          <entityClass>...VrpCustomer</entityClass>

          ...

        </entitySelector>

        <minimumSubChainSize>2</minimumSubChainSize>

        <maximumSubChainSize>40</maximumSubChainSize>

      </subChainSelector>

      <valueSelector>

        <variableName>previousAppearance</variableName>

        ...

      </valueSelector>

      <selectReversingMoveToo>true</selectReversingMoveToo>

    </subChainChangeMoveSelector>

The property selectReversingMoveToo (defaults to true) enabled selecting the reverse of every subchain too.

7.3.5. subChainSwapMoveSelector

The subChainSwapMove selects 2 different subChains and moves it to another place in a different or the same anchor chain.


    <subChainSwapMoveSelector/>

The secondarySubChainSelector is rarely needed: if it is not specified, entities from the same subChainSelector are swapped.


    <subChainSwapMoveSelector>

      ... <!-- Normal selector properties -->

      <subChainSelector>

        <entitySelector>

          <entityClass>...VrpCustomer</entityClass>

          ...

        </entitySelector>

        <minimumSubChainSize>2</minimumSubChainSize>

        <maximumSubChainSize>40</maximumSubChainSize>

      </subChainSelector>

      <secondarySubChainSelector>

        <entitySelector>

          <entityClass>...VrpCustomer</entityClass>

          ...

        </entitySelector>

        <minimumSubChainSize>2</minimumSubChainSize>

        <maximumSubChainSize>40</maximumSubChainSize>

      </secondarySubChainSelector>

      <selectReversingMoveToo>true</selectReversingMoveToo>

    </subChainSwapMoveSelector>

The other properties are explained in subChainChangeMoveSelector.

7.4. Combining multiple `MoveSelector`s

7.4.1. unionMoveSelector

A unionMoveSelector selects a Move by selecting 1 of its child MoveSelectors to supply the next Move.


    <unionMoveSelector>

      <...MoveSelector/>

      <...MoveSelector/>

      <...MoveSelector/>

      ...

    </unionMoveSelector>

The selectorProbabilityWeightFactory determines in selectionOrder RANDOM how often a child MoveSelector is selected to supply the next Move. By default, each child MoveSelector has the same chance of being selected. Change the fixedProbabilityWeight of such a child to select it more often. For example, the unionMoveSelector can return a SwapMove twice as often as a ChangeMove:


    <unionMoveSelector>

      ... <!-- Normal selector properties -->

      <selectorProbabilityWeightFactoryClass>...ProbabilityWeightFactory</selectorProbabilityWeightFactoryClass>

      <changeMoveSelector>

        <fixedProbabilityWeight>...</fixedProbabilityWeight>

        ...

      </changeMoveSelector>

      <swapMoveSelector>

        <fixedProbabilityWeight>...</fixedProbabilityWeight>

        ...

      </swapMoveSelector>

      <...MoveSelector>

        <fixedProbabilityWeight>...</fixedProbabilityWeight>

        ...

      </...MoveSelector>

      ...

    </unionMoveSelector>


    <unionMoveSelector>

      <changeMoveSelector>

        <fixedProbabilityWeight>1.0</fixedProbabilityWeight>

        ...

      </changeMoveSelector>

      <swapMoveSelector>

        <fixedProbabilityWeight>2.0</fixedProbabilityWeight>

        ...

      </swapMoveSelector>

    </unionMoveSelector>

The number of possible ChangeMoves is very different from the number of possible SwapMoves and furthermore it's problem dependent. To give each individual Move the same selection chance (as opposed to each MoveSelector), use the FairSelectorProbabilityWeightFactory:


    <unionMoveSelector>

      <selectorProbabilityWeightFactoryClass>org.optaplanner.core.impl.heuristic.selector.common.decorator.FairSelectorProbabilityWeightFactory</selectorProbabilityWeightFactoryClass>

      <changeMoveSelector/>

      <swapMoveSelector/>

    </unionMoveSelector>

7.4.2. cartesianProductMoveSelector

A cartesianProductMoveSelector selects a new CompositeMove. It builds that CompositeMove by selecting 1 Move per child MoveSelector and adding it to the CompositeMove.


    <cartesianProductMoveSelector>

      <...MoveSelector/>

      <...MoveSelector/>

      <...MoveSelector/>

      ...

    </cartesianProductMoveSelector>


    <cartesianProductMoveSelector>

      ... <!-- Normal selector properties -->

      <changeMoveSelector>

        ...

      </changeMoveSelector>

      <swapMoveSelector>

        ...

      </swapMoveSelector>

      <...MoveSelector>

        ...

      </...MoveSelector>

      ...

    </cartesianProductMoveSelector>

7.5. EntitySelector


      <entitySelector/>

The entityClass property is only required if it cannot be deduced automatically because there are multiple entity classes.


      <entitySelector>

        ... <!-- Normal selector properties -->

        <entityClass>org.optaplanner.examples.curriculumcourse.domain.Lecture</entityClass>

      </entitySelector>

7.6. ValueSelector


      <valueSelector/>

The variableName property is only required if it cannot be deduced automatically because there are multiple variables (for the related entity class).


      <valueSelector>

        ... <!-- Normal selector properties -->

        <variableName>room</variableName>

      </valueSelector>

7.7. Custom moves

7.7.1. Which move types might be missing in my implementation?

To determine which move types might be missing in your implementation, run a benchmarker for a short amount of time and configure it to write the best solutions to disk. Take a look at such a best solution: it will likely be a local optima. Try to figure out if there's a move that could get out of that local optima faster.

If you find one, implement that course-grained move, mix it with the existing moves and benchmark it against the previous configurations to see if you want to keep it.

7.7.2. Custom moves introduction

Instead of reusing the generic Moves (such as ChangeMove) you can also implement your own Moves. Generic and custom MoveSelectors can be combined as wanted.

A custom Move can be tailored to work to the advantage of your constraints. For example, in examination scheduling, changing the period of an exam A also changes te period of all the exams that need to coincide with exam A.

A custom Move is also slightly faster than a generic Move. However, it's far more work to implement and much harder to avoid bugs. After implementing a custom Move, make sure to turn on environmentMode FULL_ASSERT to check for score corruptions.

7.7.3. The interface `Move`

Your custom moves must implement the Move interface:

public interface Move {


    boolean isMoveDoable(ScoreDirector scoreDirector);


    Move createUndoMove(ScoreDirector scoreDirector);

    void doMove(ScoreDirector scoreDirector);


    Collection<? extends Object> getPlanningEntities();

    Collection<? extends Object> getPlanningValues();


}

Let's take a look at the Move implementation for 4 queens which moves a queen to a different row:

public class RowChangeMove implements Move {


    private Queen queen;

    private Row toRow;


    public RowChangeMove(Queen queen, Row toRow) {

        this.queen = queen;

        this.toRow = toRow;

    }


    // ... see below


}

An instance of RowChangeMove moves a queen from its current row to a different row.

Planner calls the doMove(ScoreDirector) method to do a move. The Move implementation must notify the ScoreDirector of any changes it make to planning entity's variables:

    public void doMove(ScoreDirector scoreDirector) {

        scoreDirector.beforeVariableChanged(queen, "row"); // before changes are made to the queen.row

        queen.setRow(toRow);

        scoreDirector.afterVariableChanged(queen, "row"); // after changes are made to the queen.row

    }

You need to call the methods scoreDirector.beforeVariableChanged(Object, String) and scoreDirector.afterVariableChanged(Object, String) directly before and after modifying the entity. Alternatively, you can also call the methods scoreDirector.beforeAllVariablesChanged(Object) and scoreDirector.afterAllVariablesChanged(Object).

Note

You can alter multiple entities in a single move and effectively create a big move (also known as a coarse-grained move).

Warning

A Move can only change/add/remove planning entities, it must not change any of the problem facts.

Planner automatically filters out non doable moves by calling the isDoable(ScoreDirector) method on a move. A non doable move is:

A move that changes nothing on the current solution. For example, moving queen B0 to row 0 is not doable, because it is already there.
A move that is impossible to do on the current solution. For example, moving queen B0 to row 10 is not doable because it would move it outside the board limits.

In the n queens example, a move which moves the queen from its current row to the same row isn't doable:

    public boolean isMoveDoable(ScoreDirector scoreDirector) {

        return !ObjectUtils.equals(queen.getRow(), toRow);

    }

Because we won't generate a move which can move a queen outside the board limits, we don't need to check it. A move that is currently not doable could become doable on the working Solution of a later step.

Each move has an undo move: a move (normally of the same type) which does the exact opposite. In the example above the undo move of C0 to C2 would be the move C2 to C0. An undo move is created from a Move, before the Move has been done on the current solution.

    public Move createUndoMove(ScoreDirector scoreDirector) {

        return new RowChangeMove(queen, queen.getRow());

    }

Notice that if C0 would have already been moved to C2, the undo move would create the move C2 to C2, instead of the move C2 to C0.

A solver phase might do and undo the same Move more than once. In fact, many solver phases will iteratively do an undo a number of moves to evaluate them, before selecting one of those and doing that move again (without undoing it this time).

A Move must implement the getPlanningEntities() and getPlanningValues() methods. They are used by entity tabu and value tabu respectively. When they are called, the Move has already been done.

    public List<? extends Object> getPlanningEntities() {

        return Collections.singletonList(queen);

    }


    public Collection<? extends Object> getPlanningValues() {

        return Collections.singletonList(toRow);

    }

If your Move changes multiple planning entities, return all of them in getPlanningEntities() and return all their values (to which they are changing) in getPlanningValues().

    public Collection<? extends Object> getPlanningEntities() {

        return Arrays.asList(leftCloudProcess, rightCloudProcess);

    }


    public Collection<? extends Object> getPlanningValues() {

        return Arrays.asList(leftCloudProcess.getComputer(), rightCloudProcess.getComputer());

    }

A Move must implement the equals() and hashCode() methods. 2 moves which make the same change on a solution, should be equal.

    public boolean equals(Object o) {

        if (this == o) {

            return true;

        } else if (o instanceof RowChangeMove) {

            RowChangeMove other = (RowChangeMove) o;

            return new EqualsBuilder()

                    .append(queen, other.queen)

                    .append(toRow, other.toRow)

                    .isEquals();

        } else {

            return false;

        }

    }


    public int hashCode() {

        return new HashCodeBuilder()

                .append(queen)

                .append(toRow)

                .toHashCode();

    }

Notice that it checks if the other move is an instance of the same move type. This instanceof check is important because a move will be compared to a move with another move type if you're using more then 1 move type.

It's also recommended to implement the toString() method as it allows you to read Planner's logging more easily:

    public String toString() {

        return queen + " => " + toRow;

    }

Now that we can implement a single custom Move, let's take a look at generating such custom moves.

7.7.4. `MoveListFactory`: the easy way to generate custom moves

The easiest way to generate custom moves is by implementing the interface MoveListFactory:

public interface MoveListFactory {


    List<Move> createMoveList(Solution solution);


}

For example:

public class RowChangeMoveFactory implements MoveListFactory {


    public List<Move> createMoveList(Solution solution) {

        NQueens nQueens = (NQueens) solution;

        List<Move> moveList = new ArrayList<Move>();

        for (Queen queen : nQueens.getQueenList()) {

            for (Row toRow : nQueens.getRowList()) {

                moveList.add(new RowChangeMove(queen, toRow));

            }

        }

        return moveList;

    }


}

Simple configuration (which can be nested in a unionMoveSelector just like any other MoveSelector):


    <moveListFactory>

      <moveListFactoryClass>org.optaplanner.examples.nqueens.solver.move.factory.RowChangeMoveFactory</moveListFactoryClass>

    </moveListFactory>

Because the MoveListFactory generates all moves at once in a List<Move>, it does not support cacheType JUST_IN_TIME. Therefore, moveListFactory uses cacheType STEP by default and it scales badly in memory footprint.


    <moveListFactory>

      ... <!-- Normal moveSelector properties -->

      <moveListFactoryClass>org.optaplanner.examples.nqueens.solver.move.factory.RowChangeMoveFactory</moveListFactoryClass>

    </moveListFactory>

7.7.5. `MoveIteratorFactory`: generate custom moves just in time

Use this advanced form to generate custom moves by implementing the interface MoveIteratorFactory:

public interface MoveIteratorFactory {


    long getSize(ScoreDirector scoreDirector);


    Iterator<Move> createOriginalMoveIterator(ScoreDirector scoreDirector);


    Iterator<Move> createRandomMoveIterator(ScoreDirector scoreDirector, Random workingRandom);


}

The method getSize() must give an estimation of the size. It doesn't need to be correct. The method createOriginalMoveIterator is called if the selectionOrder is ORIGINAL or if it is cached. The method createRandomMoveIterator is called for selectionOrder RANDOM combined with cacheType JUST_IN_TIME.

Important

Don't create a collection (list, array, map, set) of Moves when creating the Iterator<Move>: the whole purpose of MoveIteratorFactory over MoveListFactory is giving you the ability to create a Move just in time in the Iterator's method next().

Simple configuration (which can be nested in a unionMoveSelector just like any other MoveSelector):


    <moveIteratorFactory>

      <moveIteratorFactoryClass>...</moveIteratorFactoryClass>

    </moveIteratorFactory>

A construction heuristic builds a pretty good initial solution in a finite length of time. Its solution isn't always feasible, but it finds it fast and metaheuristics can finish the job.


    <moveIteratorFactory>

      ... <!-- Normal moveSelector properties -->

      <moveIteratorFactoryClass>...</moveIteratorFactoryClass>

    </moveIteratorFactory>

7.7.6. Move generation through DRL

Not yet supported.

Chapter 8. Construction heuristics

8.1. Overview

Construction heuristics terminate automatically, so there's usually no need to configure a Termination on the construction heuristic phase specifically.

8.2. First Fit

8.2.1. Algorithm description

The First Fit algorithm cycles through all the planning entities (in default order), initializing 1 planning entity at a time. It assigns the planning entity to the best available planning value, taking the already initialized planning entities into account. It terminates when all planning entities have been initialized. It never changes a planning entity after it has been assigned.

Notice that it starts with putting Queen A into row 0 (and never moving it later), which makes it impossible to reach the optimal solution. Suffixing this construction heuristic with metaheurstics can remedy that.

8.2.2. Configuration

The constructionHeuristicPickEarlyType of FIRST_LAST_STEP_SCORE_EQUAL_OR_IMPROVING is a big speedup, which should be applied when initializing a planning entity which can only make the score lower or equal. So if:


  <constructionHeuristic>

    <constructionHeuristicType>FIRST_FIT</constructionHeuristicType>

    <!-- Speedup that can be applied to most, but not all use cases: -->

    <!-- <constructionHeuristicPickEarlyType>FIRST_LAST_STEP_SCORE_EQUAL_OR_IMPROVING</constructionHeuristicPickEarlyType> -->

  </constructionHeuristic>

Note

There are no positive constraints.
There is no negative constraint that can stop being broken by adding a planning entity (except if another negative constraint gets broken which outweighs the first negative constraint).

If that is not the case, then it can still be good to apply it in some cases, but not in most cases. Use the Benchmarker to decide.

8.3. First Fit Decreasing

8.3.1. Algorithm description

Like First Fit, but assigns the more difficult planning entities first, because they are less likely to fit in the leftovers. So it sorts the planning entities on decreasing difficulty.

Requires the model to support planning entity difficulty comparison.

Note

One would expect that this algorithm always performs better than First Fit. That's not always the case, but usually is.

8.3.2. Configuration

Like First Fit, but uses the weaker planning values first, because the strong planning values are more likely to be able to accommodate later planning entities. So it sorts the planning values on increasing strength.


  <constructionHeuristic>

    <constructionHeuristicType>FIRST_FIT_DECREASING</constructionHeuristicType>

    <!-- Speedup that can be applied to most, but not all use cases: -->

    <!-- <constructionHeuristicPickEarlyType>FIRST_LAST_STEP_SCORE_EQUAL_OR_IMPROVING</constructionHeuristicPickEarlyType> -->

  </constructionHeuristic>

8.4. Best Fit

8.4.1. Algorithm description

Requires the model to support planning value strength comparison.

Note

One would expect that this algorithm always performs better than First Fit. That's not always the case.

8.4.2. Configuration

Combines First Fit Decreasing and Best Fit. So it sorts the planning entities on decreasing difficulty and the planning values on increasing strength.


  <constructionHeuristic>

    <constructionHeuristicType>BEST_FIT</constructionHeuristicType>

    <!-- Speedup that can be applied to most, but not all use cases: -->

    <!-- <constructionHeuristicPickEarlyType>FIRST_LAST_STEP_SCORE_EQUAL_OR_IMPROVING</constructionHeuristicPickEarlyType> -->

  </constructionHeuristic>

8.5. Best Fit Decreasing

8.5.1. Algorithm description

Requires the model to support planning entity difficulty comparison and planning value strength comparison.

Note

One would expect that this algorithm always performs better than First Fit, First Fit Decreasing and Best Fit. That's not always the case.

8.5.2. Configuration