Department of EEE, IIT GUWAHATI
Talk title: Reinforcement learning algorithms for robotic applications
Abstract:Determining decision in uncertain and unknown environments for robotic vehicles is difficult. In this talk, I present two such algorithms for aerial and surface vehicles. The first problem is on persistent area coverage using a set of anchor and auxiliary agents that have limited sensing, communication and localization constraints. We use deep reinforcement learning to solve this challenging problem. The second problem is about bathymetry mapping for autonomous surface vehicles is static lakes. In India, lake bathymetry is rarely available. We develop a low-cost surface vehicle for deployment, and develop novel adaptive sampling algorithms based on a Gaussian process assumption. We demonstrate the planning in the field through real-world experiments.