TimeTableScheduler

Solve time table scheduling using `Constraint Satisfaction Problem`

6
0
Java

GitHub repo: https://github.com/KajanM/TimeTableScheduler

CS3612 - Intelligent Systems

Assignment: Time_Tabling_With_CSP

To run the program

  1. navigate to the executable directory which has scheduler.jar
  2. ensure the input, output csv files are in the same directory
  3. open terminal from that directory
  4. run java -jar scheduler.jar <input_file_name> <output_file_name>
    (do not include .csv extension)

Sample output of the program

[kajan@kajan executable]$ java -jar scheduler.jar test-input-2 test-output
Debug enabled: false

Found Linux OS, setting path separator to /
Input file contents
# 6 subjects, 6 available slots

S1, o, M1, M2, M3, T1
S2, o, M1, M2, M3, T2
S3, o, M1, M2, M3, T3
S4, o, M3
S5, c, M2
S6, o, M1
R1

Parsing input csv file completed

Assigning priority to subjects based on `Minimum Remaining Value` and `Degree` heuristics.

Recursive scheduling job started at Mon Sep 03 06:46:30 IST 2018

Scheduling job completed at Mon Sep 03 06:46:30 IST 2018
Scheduling process took 15 milli seconds

Results
S4 | compulsory: false | R1 | M3
S6 | compulsory: false | R1 | M1
S5 | compulsory: true | R1 | M2
S1 | compulsory: false | R1 | T1
S2 | compulsory: false | R1 | T2
S3 | compulsory: false | R1 | T3

Running test cases to verify all constraints are met...
Passed: No subjects are assigned to same room and time
Passed: No other subjects assigned at a compulsory subject time
Passed: Assigned time is one of given applicable times
[kajan@kajan executable]$ 

Modeling the problem

Variables

  1. Subject object representing a subject that need to be scheduled with the following fields

    • name
    • compulsory or not
    • scheduled room and time (represented by a single object RoomAndTime)
    • priority (to apply heuristics when selecting which subject to schedule next/first)
    • set of applicable slots
  2. RoomAndTime object to hold combination of room and time. Since both room and time need to be assigned for a subject they are combined to form a single object to apply OOP concepts.

  3. subjectsToSchedule - a Set(no duplicates) of Subjects to be scheduled.

  4. subjectsScheduled - Ordered set of Subjects that are scheduled so far.

  5. availableRoomsAndTimes - Set of RoomAndTime objects that are available.

  6. assignedRoomsAndTimes- Set of RoomAndTime objects that are scheduled.

Domain

Domain for each variable is extracted from the input csv file.
Input file should be of the following format for correct processing.

#comments are ignored from parsing

Subject_1, c, M1, M3, Tu2
Subject_2, o, Tu1, W1, Th2
Subject_3, c, M1, M3, W1
.
.
.
Subject_n, o, M3, Th2
R1, R2, R3

Initializing

  1. While parsing the input csv file, create RoomAndTime and Subject objects.
  2. add all Subjects to subjectsToSchedule
  3. add all RoomAndTime to availableRoomsAndTimes
  4. initialize subjectsScheduled as empty set
  5. initialize assignedRoomsAndTimes as empty set

Assigning a time slot to a subject

This is a recursive operation, i.e this operation calls itself until there is no Subject to assign or infinite loop is detected.

  1. get a Subject from subjectsToSchedule
  2. iterate through each of the applicable timeslot for the selected Subject
  3. check if the applicable slot is consistent to current assignment
  4. forward check to see if the assignment will fail any of the remaining Subject to schedule
  5. once a consistent slot is assigned to the Subject move it to the subjectsScheduled set
  6. update availableRoomsAndTimes, assignedRoomsAndTimes accordingly
  7. if no consistent value is found then backtrack

Checking consistency

When selecting a RoomAndTime to assign to a Subject following constraints are checked.

  1. A given subjects can be assigned only to one of the possible time slots given for that subject.
  2. No other subject is assigned to a time that is already assigned to a compulsory subject regardless of room.
  3. Two subjects cannot be assigned to the same room if they are assigned to the same time slot.

Test cases are also written to ensure the end result satisfies above constraints. They are availabe at TestCases.java. Below is a snippet of the code.

public static void onlyApplicableTimeIsAssigned(Set<Subject> subjectsScheduled) {
	for(Subject subject : subjectsScheduled) {
		if(!subject.getApplicableSlots().contains(subject.getRoomAndTime())) {
			System.out.println("Failed: Assigned time is not applicable for the subject " + subject);
			return;
		}
	}
	System.out.println("Passed: Assigned time is one of given applicable times");
}

Forward checking

Before assigning a consistent RoomAndTime value to a Subject, the logic ensures that remaining Subjects to be scheduled has atleast one applicable RoomAndTime which is also availabe.

Backtracking

When the logic finds there is no consistent RoomAndTime to assign to a Subject, it removes the cause of assignment from subjectsScheduled and put into subjectsToSchedule and update other relevant parameters(availableRoomsAndTimes, ``assignedRoomsAndTimes`)

Heuristics

Priority is assigned to each Subject based on following heuristics. Subject with smaller priority value is considered to have higer priority.

Minimum Remaining Values

Subject with the fewest number of applicable slots is chosen for next assignment.

Least-Constraining Value

Priority values of compulsory Subjects are increased by one to give higer priority to optional Subjects.