Decentralized Multi-Agent Coverage Path Planning with Greedy Entropy Maximization
Kale Champagnie, Farshad Arvin, Junyan Hu · 2024
In this paper, we present GEM, a novel approach to online coverage path planning in which a swarm of homogeneous agents act to maximize the entropy of pheromone deposited within their environment. We show that entropy maximization (EM) coincides with many conventional goals in offline coverage path planning, while also generalizing to online settings. We first propose the concept of uniformity, which is a generalised metric that allows offline and online CPP approaches to be viewed through a unified lens. We then evaluate our approach by mea-suring the rate at which entropy is maximized within a variety of static and dynamic environments. Our experimental results demonstrate that GEM achieves state-of-the-art performance in online coverage, competitive with offline methods, despite requiring no direct communication among agents.