Title: Learning Control via Probabilistic Trajectory Optimization
Date: Monday, October 31
Time: 3:00pm (EST)
Location: CCB 345
A major goal of the robotics community is to develop real-time decision-making and control algorithms for autonomous systems to operate under uncertainty. While classical optimal control provides a general theoretical framework, it relies on strong assumption of full knowledge of the system dynamics and environments. Alternatively, modern reinforcement learning (RL) offers a computational framework for controlling autonomous systems with minimal prior knowledge and user intervention. However, typical RL approaches require many interactions with the physical systems, and suffer from slow convergence. Furthermore, both optimal control and RL have the difficulty of scaling to high-dimensional state and action spaces.
In this proposal we present probabilistic trajectory optimization methods for optimal control of systems with uncertain dynamics. Our methods share two key characteristics: (1) we incorporate explicit uncertainty into modeling, prediction and decision making. (2) Our algorithms bypass the Bellman’s curse of dimensionality using local approximation and sampling. Compared to the state-of-the-art approaches, our methods offer superior combination of data efficiency and scalability. We will present experimental results and comparative analyses to demonstrate the strengths of the proposed methods. We will discuss current limitations and our future research agenda on theoretical generalizations, computational schemes and applications in spacecraft control, autonomous driving and neuromodulation.
PhD student in Robotics
School of Aerospace Engineering
Georgia Institute of Technology
Dr. Evangelos Theodorou (advisor), School of Aerospace Engineering, Georgia Institute of Technology
Dr. Byron Boots, School of Interactive Computing, Georgia Institute of Technology
Dr. Eric Johnson, School of Aerospace Engineering, Georgia Institute of Technology
Dr. Le Song, School of Computational Science and Engineering, Georgia Institute of Technology
Dr. Jonathan How, Department of Aeronautics and Astronautics, Massachusetts Institute of Technology