I am currently a postoctoral fellow in the School of Engineering and Applied Sciences at Harvard University, working with Professor Na Li. I received my B.S. degree in Eletronic Engineering from Tsinghua University in 2013. I received my Ph.D. degree in Electrical Enginnering from the California Institute of Technology in 2019, under the supervision of Professor Steven Low.
Lack of model information. It is common in real-world applications that decision makers may not have access to a model describing the mechanism of the physical system. To deal with this issue, we have exploited tools from zeroth-order optimization and developed novel distributed optimization and reinforcement learning algorithms, with rigorous theoretical analysis on their performance.
Partial observability. In many practical situations, the decision-making agents will only receive partial observation of the system’s state. So far, our understanding of partially observable systems has been limited. I am interested in developing theories and algorithms that can push forward our understanding on model-free reinforcement learning for partially observable systems.
Nonstationary and time-varying components. For many cyber-physical networks, the physical layer can be subject to the influence of exogenous nonstationary and time-varying components that are hard to predict. We have developed theories and algorithms for time-varying nonconvex optimization problems, with applications in smart grids that lead to real-time optimal power flow algorithms.
My goal is to build an interdisciplinary research program to develop advanced optimization, control and learning methods algorithms for cyber-physical networks that are scalable, adaptive to uncertainties, robust to disturbances, and with guaranteed optimality, by integrating both theoretical tools and engineering insights.