Research Topics
1. Reinforcement learning using deep neural networks:
I am interested in learning-based control strategies for multi-body systems (nonlinear
and non-smooth hybrid systems) using deep neural networks. Deep neural nets provide a rich parameterization for high dimensional control strategies and reinforcement
learning provides a generic framework for training them. My focus is on sim-to-real
transfer which utilizes a simulated environment for training control policies.
2. Rigid-body simulation:
Physics engine is perhaps the most important tool for robotics. I study necessary
theories behind physics engines and develop a fast and accurate physics engine for robotics.