Enhancing Balance Robustness in Swarm-based Multi-Agent Reinforcement Learning Under Random Freeze Events

Riley Roberts - 101207615
Explore the docs »
Read the report »

Report Bug · Request Feature

Table of Contents

1. About The Project
   • Built With
2. Getting Started
   • Prerequisites
3. Usage

About The Project

In real world applications, individual robots or drones may occasionally lose control due to hardware malfunctions. By understanding how to design MARL systems that are robust to intermittent control losses it allows us to deploy reliable, real world multi agent systems. The insights gained from this project may inform better design of fault-tolerant algorithms in various domains such as autonomous vehicles and drone swarms.

In this project, I will attempt to adress this through aggregated swarm statistics being added the observations of agents. More information can be found in the report »

(back to top)

Built With

• Python
• VMAS
• BenchMARL

(back to top)

Getting Started

Prerequisites

This is an example of how to list things you need to use the software and how to install them.

• BenchMARL Requirements

  pip install -e requirements.txt

Usage

Simply run sim_train.py to train the base IPPO and run sim_train_modified.py to train the IPPO with swarm statistics. If you are looking to change scenario parameters ()k,n_agents) this can be done via the task.config in their respective files.

(back to top)