/Multi-agent-Path-Planning-with-Reinforcement-Learning-and-Artificial-Potential-Field

Primary LanguagePythonMIT LicenseMIT

Multi-agent-Path-Planning-with-Reinforcement-Learning-and-Artificial-Potential-Field

Abstract

This program aims to solve a kind of multi-agent path planning scheme and it is one of the scenarios we constructed in our recent research. Recently, we have published one article on the Chinese Automation Conference (CAC 2021), and the program is part of the simulation of that article.

How to run?

the framework file

run pip install -r requirements.txt to install the required packages run rt_multi-agent-sarsalambda.py directly.

the supporting files

  1. maze_env_sarsa_multi_agent.py: the maze map environment for RL simulation

  2. RL_brain_sarsalambda_mul_v2.py: the states, action spaces and policy store of the UAV agents.

Cite this work

@inproceedings{liu2021solving,
  title={Solving a Multi-robot Search Problem with Bionic Sarsa Algorithm and Artificial Potential Field},
  author={Liu, Haichao and Qu, Zhenshen and Zhu, Runwen},
  booktitle={2021 China Automation Congress (CAC)},
  pages={1830--1835},
  year={2021},
  organization={IEEE}
}