Motion planning and decision making for autonomous systems
When
Mobile robots are becoming increasingly ubiquitous due to technological advances in sensing, actuation, and computation. However, devising control strategies to ensure that robots perform their assigned tasks in unpredictable real-world environments remains a challenge. An approach that has proved useful in practice is to encode tasks as attractors of dynamical systems, e.g., design vector fields that steer a vehicle to a desired location. To encode richer behaviors, one can develop methods to switch among low-level point-attractor controllers; this is often done by developing an automaton that discretely switches between available controllers, yielding a hybrid dynamical system. In this work, we develop an alternative approach based on a dynamical systems mechanism for making decisions, i.e., choosing controllers. The system associates a value with each possible controller and then selects the controller with the highest value. We derive conditions under which the system exhibits a stable limit cycle which corresponds to persistently carrying out the desired tasks. We conclude the talk by discussing some more recent results integrating stochastic external stimuli into the dynamical system.
Bio: Paul Reverdy received the B.S. degree in Engineering Physics and the B.A. degree in Applied Mathematics from the University of California, Berkeley in 2007, and the M.A. and Ph.D. degrees in Mechanical and Aerospace Engineering from Princeton University in 2011 and 2014, respectively. He is currently an Assistant Professor in Aerospace and Mechanical Engineering at the University of Arizona.
From 2007 to 2009, he worked as a research assistant at the Federal Reserve Board of Governors, Washington, DC. From 2014 to 2017, he was a postdoctoral fellow in Electrical and Systems Engineering at the University of Pennsylvania, where he was affiliated with the GRASP laboratory. His research interests lie at the intersection of human and machine decision making and control, with applications in robotics, machine learning, and engineering design optimization. Dr. Reverdy's awards include a National Defense Science and Engineering Graduate (NDSEG) Fellowship for graduate study and the best student paper award from the 2014 European Control Conference.