Matthew Giamou on Convex Relaxations for Geometric Problems in Robotics | Toronto AIR Seminar

Abstract:
Many tasks in robotics and computer vision can be formulated as nonlinear optimization problems, enabling the use of powerful general-purpose solution methods. While convenient, most truly challenging problems (e.g., localization, mapping, and optimal motion planning) are nonconvex, and therefore admit multiple suboptimal minima. As a result, much research activity has been devoted to developing clever convex relaxations which allow a user to efficiently obtain and certify a globally optimal solution to these difficult problems. In this talk, I will summarize my recent work on applying convex semidefinite relaxations and related ideas to several important geometric problems in robotics: extrinsic sensor calibration, inverse kinematics, and deep rotation learning. Expressing the geometric constraints of robotic systems with quadratic polynomials allows us to leverage globally optimal solution methods, and reveals and exploits problem structure like the inherent sparseness of robot geometry. Experimental results demonstrate that this problem-solving paradigm leads to fast and scalable algorithms with elegant theoretical properties.


Paper:
https://arxiv.org/abs/2108.13720 (accepted to T-RO)
https://arxiv.org/abs/2109.03374 (accepted to RA-L)
http://www.roboticsproceedings.org/rs...


Bio:
Matthew Giamou is a PhD student in the Space and Terrestrial Autonomous Robotic Systems Laboratory at the University of Toronto Institute for Aerospace Studies. His research applies the theory of convex optimization to estimation, learning, and planning tasks which are critical to safe and effective autonomous robotic systems.
https://mattgiamou.ca/


Toronto AIR Seminar:
The Toronto AI Robotics Seminar Series is a set of events featuring young robotics and AI experts. The talks are given by local as well as global speakers and organized by the Faculty and Students at University of Toronto’s Department of Computer Science. We welcome students, researchers and robotics enthusiasts from around the world to join us and interact with the Toronto Robotics Community.
Find out more at: https://robotics.cs.toronto.edu/toron...