Data Efficient Reinforcement learning for Autonomous Robots with Simulated and Off-policy Data

June 17, 2019
Josiah Hanna | University of Texas at Austin

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, traffic management, or healthcare.

My research focuses on giving agents the ability to predict how their actions influence their ability to solve a given task. In this talk, I will describe my research in this area and how efficient prediction connects to efficient 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 it’s 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 real world. Finally, I will describe directions for future work that will lead to an increased applicability of reinforcement learning to real-world problems.

[SLIDES]

Speaker Details

Josiah Hanna is a PhD candidate in the computer science department at the University of Texas at Austin advised by Professor Peter Stone. He is joining the faculty of the University of Wisconsin-Madison in fall of 2021. Prior to attending UT Austin, he completed his bachelors degree in computer science and mathematics at the University of Kentucky advised by Professor Judy Goldsmith. During the summer of 2017, he completed a research internship at Google. Josiah is an NSF Graduate Research Fellow and an IBM PhD Fellow.