Task 007: Distributed Learning and Inference

ABSTRACT: Linear Temporal Logic (LTL) and its variants are a useful tool to describe and specify objectives in Learning-based environments. Prior work has demonstrated the need for distributing approaches used to satisfy a given specification within a Multi-agent Reinforcement Learning (MARL) System. We notice a deficiency in these methods to handle safety-related objectives in a scalable manner for these multi-agent systems and propose that Lyapunov functions may be a way to allow scaling to complex compositions of these objectives specific to many multi-agent problems. By utilizing goal-conditioned reinforcement learning, our method further improves generalizability within a range of complex specifications. Preliminary results show the benefits of these approaches in a set of navigation tasks.

 

BIO: Joe Kurian Eappen is a graduate student in the School of Electrical and Computer Engineering at Purdue University working under the supervision of Prof. Suresh Jagannathan. He completed his Dual Degree (B.Tech & M.Tech) in Electrical Engineering at the Indian Institute of Technology (IIT) Madras, India in 2018. His current research lies at the intersection of Safety in Reinforcement Learning and Multi-Agent systems.