Yanbing Mao | Ph.D.
 |
Assistant Professor |
Research Interests
Embodied AI
Robot Learning
Dynamics and Control of Social Information
Application Domains
Self-Driving Vehicles
Safety-Critical Robots
Social Cybersecurity
Work Experience
Assistant Professor, Wayne State University, Aug. 2022 – Present
Postdoctoral Research Associate, University of Illinois at Urbana-Champaign, Oct. 2019 – Jul. 2022
Education
Ph.D. in Electrical and Computer Engineering, State University of New York at Binghamton, Sept. 2015 – Sept. 2019
M.E. in Circuits & Systems, University of Electronic Science and Technology of China, Sept. 2010 – Jul. 2013
News
[10-15-2025] We open the "Real-DRL’’, including source codes and demo videos.
[10-01-2025] Our Paper "Real-DRL: Teach and Learn in Reality’’ got accepted at NeurIPS 2025. Real-DRL enables safe runtime learning of deep reinforcement learning agents to develop safe and high-performance action policies in safety-critical autonomous systems. At a high level, Real-DRL features Real Data + Good Data + Novel Learning Architecture. Congratulations to my student Yihao Cai!
[06-15-2025] Our Paper "Runtime Learning Machine’’ got accepted at ACM Transactions on Cyber Physical Systems (Special Issue on Embodied Artificial Intelligence in Cyber-Physical Systems: Algorithms, Computing Systems, Applications, and Trustworthiness). Congratulations to my student Yihao Cai!
[03-20-2025] Our project, Runtime Learning Machine, got funded by Nvidia Academic Grant. With Nvidia’s hardware support, we will upgrade runtime learning to physical edge learning for robots in the wild.
[03-15-2025] The "Demo video: Year 2024’’ summarized our research progress in the Year 2024.
[02-20-2025] We open-sourced my student Yihao Cai's leading work on "Runtime Learning of Quadruped Robots in Wild Environments’’.
[03-15-2024] We open the "Phy-DRL’’, including source codes and demo videos.
[01-16-2024] Our Physics-AI work "Phy-DRL: Physics-Regulated Deep Reinforcement Learning: Invariant Embeddings’’ got accepted at ICLR 2024 as a spotlight! Phy-DRL features 1) automatic construction of safety-embedded reward, given many safety conditions or regulations, 2) mathematically-provable safety guarantee, and 3) fast training towards safety guarantee. The theoretical claims were demonstrated by training a quadruped robot and a cart-pole system.