FYP_NUSRI_RL_BOTSP

This work is in partial fulfilment of the requirements for the Degree of Bachelor of Engineering National University of Singapore (Suzhou)

This code is the model with four-dimension input (Euclidean-type) based on reproduction of https://github.com/kevin031060/RL_TSP_4static

"ubuntu_18.04_pytorch-gpu" is used to configure the environment in Ubuntu 18.04 in a sophisticated way.

"win10 or linux_pytorch-gpu_fast_version" is employed for environmental configuration in Win10 or Linux in a convenient and fast method.

FYP_DRL is used for training and testing bi-objective TSP by DRL

1. To test the model, use the load_all_rewards.py

2. To train the model, run train_motsp_transfer.py

Tips. three constraints are added in FYP_DRL/model.py/class DRL4TSP.../def forward...

FYP_MATLAB is used to visualize and compare with classical methods

1. tour_route_line is used to draw the travel lines of both objectives in different subproblems

2. compared_with_classical_methods is used to compare with evolutionary algorithms (MOEA/D & NSGAII)

(run compare.m)

3. add_constrains is used to draw the travel lines after adding constrains

0. MATLAB_results stores the graph results of 1.2.3.