Limei Cheng

About

Limei Cheng is a researcher advancing the theoretical foundations of reinforcement learning, with a focus on continuous control systems for robotics and autonomous driving. Her most cited work, "Incremental Reinforcement Learning — a New Continuous Reinforcement Learning Frame Based on Stochastic Differential Equation Methods" (2019, 5 citations), identifies critical weaknesses in widely-used algorithms like DDPG and A3C. Specifically, she demonstrates that DDPG fails to manage noise in control processes, while A3C violates continuity conditions under Gaussian policies. To address these gaps, Cheng introduces a novel framework that leverages stochastic differential equations to provide more robust and theoretically sound learning dynamics. This contribution is particularly significant for applications requiring precise, real-time decision-making in uncertain environments. By bridging the gap between stochastic calculus and reinforcement learning, Cheng’s work offers a principled path forward for developing safer and more reliable autonomous systems. Her research underscores the importance of rigorous mathematical foundations in AI, making her a notable voice in the ongoing evolution of continuous reinforcement learning.

Research Focus

Key Achievements

Top Papers

1. 1
   Incremental Reinforcement Learning --- a New Continuous Reinforcement Learning Frame Based on Stochastic Differential Equation methods

   5 citations · 2019

Key Collaborators

Contact & Links