Colloquium: Data Efficient Reinforcement Learning for Autonomous Robots with Simulated Off-policy Data
Learning from interaction with the environment - trying untested actions, observing successes and failures, and tying effects back to causes - is one of the first capabilities thought of when considering intelligent agents. Reinforcement learning is the area of artificial intelligence research that has the goal of allowing autonomous agents to learn in this way. Despite many recent empirical successes, most modern reinforcement learning algorithms are still limited by the large amounts of experience required before useful skills are learned. Making reinforcement learning more data efficient would allow computers to autonomously solve complex tasks in dynamic environments such as those found in robotics or healthcare.
My research focuses on enhancing the data-efficiency of an agent learning to predict how its actions influence the ability to solve a given task. In this talk, I will describe my research into more effective prediction for reinforcement learning. In the first part of the talk, I will introduce an algorithm that allows an agent to find informative exploratory behaviors for learning how its actions influence task performance. In the second part of the talk, I will introduce an algorithm that allows robot skills learned in simulated environments to transfer to the read world. Finally, I will describe directions for future work that will leard to an increased applicability learning to real world problems.