Coordinated Multi-Robot Exploration using Reinforcement Learning

Exploring an unknown environment by multiple autonomous robots is a long-studied problem in robotics. The agents need to coordinate the exploration to minimize the overlapping region and avoid interference with each other. This is particularly challenging in decentralized execution, where no central system guides the exploration. In such scenarios, agents need to incorporate temporal planning and the intentions of other agents into the decision-making process. In this work, we focus on several challenges involved in multi-UAV exploration in unseen, unstructured, and cluttered environments. Consequently, we propose a Multi-Agent Reinforcement Learning (MARL) based framework wherein agents learn the effective strategy to allocate and explore the environment. We evaluate the performance of our proposed framework in terms of average distance traveled, percentage of redundant exploration, and the rate of exploration against a classical approach.