Date: March 25, 2026
Speaker: Kwan-Yee, Research Fellow, EECS Department, University of Michigan
Title: Zac Manchester, Associate Professor of Aeronautics and Astronautics, MIT
Abstract: Some of the most exciting breakthroughs in robotics in the past few years have involved data-driven machine-learning techniques. However, there are many domains in which we lack good data to train these methods, and some scenarios where we likely will never get there. Meanwhile, there have also been significant advances in physics simulation and optimization-based planning and control methods. One surprising feature of these recent successes – both data-driven and model-based – is how simple they can be. In this talk, I’ll argue that these methods share more in common than one might expect at first glance, that models and data are complementary, and that we need both. I’ll highlight several recent works from my group that push the limits of how simple locomotion (and, possibly, manipulation) controllers for general-purpose robots can be from several different viewpoints, while also making connections to state-of-the-art data-driven generative methods like diffusion and flow policies.
Bio: Zac Manchester is an Associate Professor of Aeronautics and Astronautics at MIT. He holds a Ph.D. in aerospace engineering and a B.S. in applied physics from Cornell University. Zac was a postdoc in the Agile Robotics Lab at Harvard and previously worked at Carnegie Mellon, Stanford, NASA Ames Research Center and Analytical Graphics, Inc. He received a NASA Early Career Faculty Award in 2018, a Google Faculty Research Award in 2020, an NSF CAREER Award in 2025, and has led four NASA-funded satellite missions. His research interests include motion planning, control, and numerical optimization, particularly with application to robotics and space exploration.