Model-based reinforcement learning, or MBRL, is a promising approach for autonomously learning to control complex real-world systems with minimal expert knowledge. Join Facebook AI Research Scientist Roberto Calandra for an overview of MRBL research in this invited talk hosted by Heni Ben Amor.
Rethinking Model-based Reinforcement Learning
Presented by Roberto Calandra, research scientist, Facebook AI Research
Tuesday, October 8, 2019
Brickyard Artisan Court (BYAC) 110, Tempe campus [map]
In this talk, Calandra will first provide an overview of recent work on MBRL, including Facebook’s state-of-the-art algorithm, PETS. He will also present a new work that identifies a critical flaw in the way MBRL is currently conceived. Identifying and addressing this issue is an important step toward better understanding and design of MBRL algorithms.
About the speaker
Roberto Calandra is a research scientist at Facebook AI Research. Previously, he was a postdoctoral scholar at the University of California, Berkeley in the Berkeley Artificial Intelligence Research Laboratory (BAIR) working with Sergey Levine.
His education includes a doctoral degree from TU Darmstadt, Germany, under the supervision of Jan Peters and Marc Deisenroth, a master’s of science in machine learning and data mining from the Aalto University, Finland, and a bachelor’s of science in computer science from the Università degli studi di Palermo, Itlay.
His interests focus at the conjunction of machine learning and robotics in what is known as robot learning. Some of the research topics that he is currently developing include deep reinforcement learning, model-based reinforcement learning, tactile sensing, dynamics modeling and Bayesian optimization.