Dr. Kun Li
RBE / CS Faculty Candidate
Inverse Reinforcement Learning from Sparse
High-Dimensional Motion Data in Robotic Applications
This talk presents the application of inverse reinforcement learning to evaluating human motion and teaching robot tasks. The main difficulties are the high-dimensional and insufficient data. To solve the problem, this talk introduces two simple principles in designing the algorithms, and proposes two algorithms to handle the data dimensions and data sparsity. The resultant algorithms are used to evaluate the skills of surgical robot operators, quantify the effects of therapies, and teach robot simple grasping tasks. At last, some ongoing and future works are presented.
Kun Li got his BS degree from Jilin University, China in 2010, and his PhD degree from The Chinese University of Hong Kong in 2015. Since then, he is a postdoctoral scholar in the California Institute of Technology. His main research interests are robot learning and robot vision, especially 3D visual data processing and robot imitation learning via inverse reinforcement learning.
Friday, March 16, 2018
3:00 p.m. - 4:00 p.m.
60 Gateway Park, GP 1002