Deep Reinforcement Learning for Decentralized Multi-Robot Exploration With Macro Actions

Cooperative multi-robot teams need to be able explore cluttered and unstructured environments while dealing with communication dropouts that prevent them from exchanging local information maintain team coordination. Therefore, robots consider high-level teammate intentions during action selection. In this letter, we present the first Macro Action Decentralized Exploration Network (MADE-Net) using multi-agent deep reinforcement learning (DRL) address challenges of exploration in unseen, unstructured, environments. Simulated robot experiments were conducted compared against classical DRL methods where MADE-Net outperformed all benchmark terms computation time, total travel distance, number interactions between robots, rate across various degrees dropouts. A scalability study 3D showed a decrease time increasing environment sizes. The presented highlight effectiveness robustness our method.