Towards Long-Horizon Robot Task Learning

Explore cutting-edge research on long-horizon robot task learning in this hour-long lecture by Zsolt Kira from Georgia Tech. Delve into the challenges of enabling robots to perform complex, multi-step tasks in home environments, and discover innovative approaches to overcome current limitations in imitation and reinforcement learning. Learn about the development of Habitat 2.0, a fast, photo-realistic simulation environment for robotics research, and examine methods for hierarchical task decomposition in object rearrangement and vision-language navigation. Compare learning-based approaches with traditional sense-plan-act methods, and gain insights into ongoing research in self-supervised learning, skill chaining, and improved inverse reinforcement learning. This captioned talk offers a comprehensive overview of the latest advancements and future directions in embodied AI and robotics.

Towards Long-Horizon Robot Task Learning (Zsolt Kira, Georgia Tech)