This repository is the course project for CSE 5001 Advanced Artificial Intelligence Fall 2021 at SUSTech. In this project, we investigated the problem of genetic algorithm for multiple traveling salesman.
Multiple Travelling Salesman Problem (MTSP) is an extension of the famousTravelling Salesman Problem (TSP) that visiting each city exactly once withno sub-tours. MTSP involves assigningmsalesmen toncities, and each citymust be visited by a salesman while requiring a minimum total cost. In this project, we investigated baseline method, IPGA and VNS-GA .A novel genetic algorithm named Polar-IPGA is proposed by combing IPGA and VNS-GA. Statistical comparison is conducted on 6 datasets with varying city numbers andlayouts.The proposed method achieved superior performance on all 6 datasetscompared to baseline method within the runtime consumption limitation.
The original code implementation can be found here. To run baseline method,
$ cd ./baseline/
$ python main.py -d NAME_OF_YOUR_DATASET (mtsp51, mtsp100, mtsp150, pr76, pr152, pr226)
The original code implementation of NN-IPGA can be found here. To run NN-IPGA,
$ cd ./NN-IPGA/
$ python main.py -d NAME_OF_YOUR_DATASET (mtsp51, mtsp100, mtsp150, pr76, pr152, pr226)
To run the proposed Polar-IPGA:
$ cd ./Polar-IPGA/
$ python main.py -d NAME_OF_YOUR_DATASET (mtsp51, mtsp100, mtsp150, pr76, pr152, pr226)
The projetc is done jointly with Jieting Zhao, Zhirui Sun, Jiamin Zheng, and Mingzhe Lv. Thanks for our group's effort.