Uncertainty quantification for safe robot learning in dynamic environments

Abstract: While advancing robots towards full autonomy, it is important to minimize deleterious effects on human and infrastructure. To achieve this, it is essential for robots to understand their surroundings. The surrounding can be represented as a metric map. However, if the environment is assumed to be static and deterministic, it is challenging for the decision-making algorithms to effectively account for the changing and unpredictable nature of the real-world environments. Therefore, it is essential to model uncertainties in dynamic environments. In this talk, various probabilistic modeling techniques for mapping occupancy and directions in environments that change in both space and time are discussed. Since these maps represent uncertainty, they can then be used for risk-aware decision-making.

Bio: Ransalu Senanayake is a postdoctoral scholar in the Department of Aeronautics and Astronautics at Stanford University. He focusses on safe interactions of autonomous systems with applications to autonomous driving. He is part of the Stanford Intelligent Systems Laboratory directed by Mykel Kochenderfer, Stanford Center for AI Safety, and the SAIL-Toyota Center for AI Research. Prior to joining Stanford, Ransalu obtained a PhD in Computer Science from the University of Sydney, and an MPhil in Industrial Engineering and Decision Analytics from the Hong Kong University of Science and Technology. He was also a visiting student at the Robotics and State Estimation Laboratory directed by Dieter Fox in the Department of Computer Science and Engineering at the University of Washington.