In this project, we investigate a Vehicle Routing Problem (VRP). This is an NP-hard problem, which means that it is problematic to find the optimal solution for large test cases, therefore we use the two metaheuristics Simulated Annealing (SA) and Genetic Algorithm (GA) to find an approximate solution.
The VRP was formulated as an optimization problem and the optimal solution was found for some small test cases using exhaustive search. SA and GA was benchmarked against these optimal solutions. For larger test cases, we benchmarked SA and GA against each other without looking the actual optimal solution.
Our results were that SA performed better in all our experiments, both in terms of time and average achieved score.
Our results are included in the results/
directory, the plots can be viewed
in the plot.ipynb
notebook.
The test suite can be run by executing make test
in the src/
directory.