iLunch Talk (Sept 25, noon): Toward Safe Robots that Learn

What: Toward Safe Robots that Learn

When: Friday September 25, 12 noon – 1:00 pm

Where: Zoom link: https://maine.zoom.us/j/96744107270?pwd=eVNtL0tYcERZY2VNTjRwTGNsRi92dz09, Password: 155012  Meeting ID: 967 4410 7270

Who: Dr. Vikas Dhiman
Affiliation: Postdoctoral Researcher at University of California in San Diego

Abstract: As Machine Learning and Robotics move to the real world, how do we make sure that unexpected situations do not cause accidents? How much learning data is enough? This talk focuses on learning a model of system dynamics online while satisfying safety constraints. Our motivation is to avoid offline learning of system dynamics or hand-specified dynamics models and allow a system to safely and autonomously estimate and adapt its own model during online operation. Given streaming observations of the system state, we use Bayesian learning to obtain a distribution over the system dynamics. In turn, the distribution is used to optimize the system behavior and ensure safety with high probability, by specifying a chance constraint over a control barrier function.

Bio: Dr. Vikas Dhiman is a Postdoctoral Researcher, with Prof Henrik Christensen and Prof Nikolay Atanasov, at the University of California, San Diego. His works lie in the localization, mapping and control algorithms for applications in robotics. Earlier in 2019, he finished his Ph.D. under the mentorship of Prof Jason Corso at the University of Michigan, Ann Arbor.

Host: School of Computing and Information Science, University of Maine

For questions: salimeh.yasaei@maine.edu