This project uses DRL to address waypoint following task allocation and planning for multi-robot systems.

This software relies on Python 3.9.

Please see the requirements.txt file for the details of dependencies. To install the dependencies, run the following command:

pip install -r requirements.txt

Otherwise, you can use the Dockerfile provided to build the environment. To build the docker image, run the following command:

docker build -t rlwaypointmrta:latest .

To run the docker image, run the following command:

docker run --rm -it  rlwaypointmrta:latest

The TSPLIB data files are from: http://elib.zib.de/pub/mp-testdata/tsp/tsplib/tsplib.html

The Pre-trained RL models for solving TSP problems are from: https://github.com/wouterkool/attention-learn-to-route

python main.py --train

python main.py --eval --eval_dir trained_sessions/moe_mlp/rand_100-3/trained_model/batch31200.pt

Arguments for training and evaluation that can be specified are defined in arg_parser.py. The docmentation of each argument is avaible by running the following command:

python main.py --help

Pre-trained models are available in the trained_sessions folder. The trained_model folder contains the trained model. Some relavent information about the training session is accessible using tesorboard. To run tensorboard, run the following command:

tensorboard --logdir=trained_sessions/$MODEL/$SESSION