Efficient Domain Coverage for Vehicles with Second-Order Dynamics via Multi-Agent Reinforcement Learning

TL;DR

This paper introduces a multi-agent reinforcement learning approach using MAPPO with LSTM and self-attention for efficient area coverage by vehicles with second-order dynamics, outperforming classical control methods.

Contribution

The paper presents a novel RL-based method with a specialized network architecture for multi-agent coverage, handling variable agent numbers and surpassing traditional control policies.

Findings

RL approach outperforms classical control policies in simulations

Incorporation of LSTM and self-attention improves adaptability to agent number

Method demonstrates significant efficiency in simulated coverage tasks

Abstract

Collaborative autonomous multi-agent systems covering a specified area have many potential applications, such as UAV search and rescue, forest fire fighting, and real-time high-resolution monitoring. Traditional approaches for such coverage problems involve designing a model-based control policy based on sensor data. However, designing model-based controllers is challenging, and the state-of-the-art classical control policy still exhibits a large degree of sub-optimality. In this paper, we present a reinforcement learning (RL) approach for the multi-agent efficient domain coverage problem involving agents with second-order dynamics. Our approach is based on the Multi-Agent Proximal Policy Optimization Algorithm (MAPPO). Our proposed network architecture includes the incorporation of LSTM and self-attention, which allows the trained policy to adapt to a variable number of agents. Our trained policy significantly outperforms the state-of-the-art classical control policy. We demonstrate our proposed method in a variety of simulated experiments.