litneet64 / inf-295-csp

Simple solver for the Car Sequencing Problem on C++ using either an Evolutive Algorithm or Backtracking.

Simple Solver for the CSP

Solver for the Car Sequencing Problem on C++ using either Backtracking or an Evolutive Algorithm. Made for the Artificial Intelligence course (INF-295) at UTFSM, 2021-1.

Features a cross-over operator using Large Neighborhood Search (LNS) and a hybrid mutation operator.

Usage

To compile just run on your shell where the Makefile is:

$ make install

Then, to use the solver:

$ ./bin/solver <path to instance file> <algorithm to use [evol/bt]> <n gens>

An already made file with the expected data and format can be found at instances/sample.txt.

For cleaning objects and binary:

$ make clean

Explanation

Heuristics

Based on Siala et al. (2015), in order to reliably/fastly get to a result for both the Evolutive Alg and the Backtracking Alg, variable instantiation was made following the most constrained vehicle class using the difficulty and utilities found on Gottlieb et al. (2003) on a "first-to-last" order.

Initial Solutions

Created randomly such that they already satisfy the demand constraints.

Evaluation Function

Uses the improved "sliding windows" function from Zhang et al.(2017), that can be expressed as "the number of required extra workers for an overloaded station". If this yields 0 then it means no (ratio) constraint was violated.

Crossover Operator

Mutation Operator