Scalable Reinforcement Learning
Reinforcement learning (RL) is a branch of machine learning that deals with teaching machines through interaction with the environment to accomplish a task by rewarding desirable actions. With recent advances in deep learning, a promising tool known as Deep Reinforcement Learning (DRL) has emerged that has been able to solve complex-decision making tasks such as robot manipulations, playing Atari and Starcraft, and even board games such as chess and Go against world experts.
Our lab focuses on a subset of DRL problems that are applicable to robotics, especially concerning ourselves with strategies that can learn and think like people in the real world. In that aspect, we focus on the problems of imitation learning to learn expert-like behaviors and objective functions as well as to compose them into complex functions to complex tasks in a data-efficient way. We look to solve real-world problems in scaling up robot intelligence beyond traditional strategies of modeling and feedback control.
Students and Collaborators
Byron Boots (GaTech)
Adversarial Imitation via Variational Inverse Reinforcement Learning
A.H.Qureshi, B. Boots, M.C. Yip
Open-Sourced Reinforcement Learning Environments for Surgical Robotics
F. Richter, R. K. Orosco, M.C. Yip