My goal is to create intelligent robots that spontaneously help humans. Technically, I am interested in adaptivity, safety, and spontaneity (behavior emergence) of the agents. My research interests include reinforcement learning, control, and adaptive algorithms.
R3. RL for the Systems Described by Stochastic Differential Equations
- Proposed a reinforcement learning technique working in reproducing kernel Hilbert spaces for systems described by stochastic differential equations.
- Showed that the continuous-time formulation is methodologically more desirable over the discrete-time formulation in terms of affinity for the control theory and susceptibility to numerical errors (Used Python).
- Showed that the proposed framework also reproduces some of the existing kernel-based discrete-time RL techniques.
R2. Barrier-Certified Adaptive RL -- Applications to Brushbot Navigation at Robotarium
- Developed a safe learning framework that guarantees Lyapunov stability of the safe set under nonstationary dynamics
- Developed an algorithm for adaptively identifying dynamical structure (e.g. control affine) via sparse kernel adaptive filter.
- Proposed a novel kernel-based model-free adaptive reinforcement learning.
- Proposed a discrete-time control barrier certificate that defines convex constraints.
- Efficiently combined all of the above.
R1. Adaptive Nonlinear Estimation in the L2 Space - Fixed Point Theory
- Began with the question: Should we always select reproducing kernel Hilbert spaces for efficiently conducting adaptive nonlinear estimation?
- Proposed a novel nonlinear estimation technique working in the L2 space to speed-up learning.
- Showed importance of properly selecting metric for nonlinear estimation tasks.
- Compared with a bunch of machine learning batch methods and the Extended Kalman Filter.