*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Title: Machine Learning for Safe and Effective Robotic Control
Date: Monday, July 26, 2021
Time: 10AM - 12PM EST
Location (Virtual): https://bluejeans.com/707943714/4137
Nolan Wagener
Robotics PhD Student
School of Interactive Computing
Georgia Institute of Technology
Committee:
Dr. Byron Boots (Advisor) - School of Computer Science and Engineering, University of Washington
Dr. Panagiotis Tsiotras (Co-Advisor) - School of Aerospace Engineering, Georgia Institute of Technology
Dr. Sehoon Ha - School of Interactive Computing, Georgia Institute of Technology
Dr. Seth Hutchinson - School of Interactive Computing, Georgia Institute of Technology
Dr. Andreas Krause - Department of Computer Science, ETH Zurich
Abstract:
For robotic systems to take greater roles in industrial and public settings, there is a great need for them to operate in unstructured or dynamic environments. With this increase in task and environment complexity, it is necessary for robotic systems to learn from interactions and data coming from the environment.
This research studies several ways that learning approaches can be incorporated for control tasks: system identification, model predictive control (MPC), and safe reinforcement learning. More specifically, a neural network approach to modeling rally car dynamics will be presented which, along with a sampling-based MPC algorithm, results in state-of-the-art aggressive off-road driving. Then, the MPC framework will be re-examined from an online learning perspective. Finally, a safe reinforcement learning algorithm based on interventions will be presented, and proposed work will focus on an application of the algorithm to a quadrupedal robot for locomotion.
