UNIVERSITY OF HERTFORDSHIRE
             COMPUTER SCIENCE RESEARCH COLLOQUIUM

                            presents

     
	 "Monte Carlo Methods and Gaussian Processes for Optimal Control"
	

                        Joe Watson
            (Technische Universitaet Darmstadt, Germany)

                     28 October 2022 (Friday)
                        13:00 -14:00 

                  Everyone is Welcome to Attend

                             C400

Abstract:

Monte Carlo methods have become increasingly relevant for control of non-differentiable systems, approximate dynamics models and learning from data. These methods scale to high-dimensional spaces and are effective at the non-convex optimizations often seen in robot learning. This talk will present recent work looking at sample-based methods from the perspective of inference-based control, specifically posterior policy iteration. From this perspective, we highlight how Gaussian noise priors produce rough control actions that are unsuitable for physical robot deployment. Considering smoother Gaussian process priors, as used in episodic reinforcement learning and motion planning, we demonstrate how smoother model predictive control can be achieved using online sequential inference.


Short Bio:

Joe Watson is a fourth year PhD student with the Intelligent Autonomous Systems group, TU Darmstadt, supervised by Prof. Jan Peters. Joe's research involves developing statistical methods for robot learning, for settings such as trajectory optimization, system identification and model-free reinforcement learning. Joe received in MEng in Information Engineering from Peterhouse, University of Cambridge, where he was the Charles Babbage senior scholar. Prior to starting his PhD, Joe worked for a surgical robotics startup developing an assistive robotic system for laparoscopic surgery.


---------------------------------------------------
Hertfordshire Computer Science Research Colloquium
http://cs-colloq.cs.herts.ac.uk