GRASP on Robotics: Russ Tedrake, Massachusetts Institute of Technology

ABSTRACT
Motivated by the challenging problems in planning and control through contact, our group has been working to connect more deeply the motion planning formulations we often use in robotics with a different lineage of algorithms and results in combinatorial optimization. What we’ve found surprised me: I feel that we’ve been formulating our mixed-integer motion planning problems incorrectly, and that there are better formulations that can lead to dramatically faster solve times and tighter convex relaxations. I’d like to tell you that story, through the lens of optimization-based collision-free motion planning. This is joint work with Tobia Marcucci, Jack Umenberger, and Pablo Parrilo. I’ll also briefly describe a few other ongoing projects in robotic manipulation, and promise to show some robot videos.