Xinle Liang

About

Xinle Liang is a pioneering researcher at the intersection of autonomous driving, reinforcement learning, and federated systems. His core contributions lie in developing scalable, privacy-preserving machine learning frameworks for real-world robotic applications. Liang’s most influential work, "Federated Transfer Reinforcement Learning for Autonomous Driving" (2022), has garnered 94 citations, establishing a foundational method for pre-training RL models on simulators before securely fine-tuning them on physical vehicles. This approach dramatically reduces the data and safety risks of real-world deployment. His earlier 2019 paper on the same topic (39 citations) laid the groundwork for this sequential transfer paradigm. Liang further advanced the field with "Cross-Silo Federated Neural Architecture Search for Heterogeneous and Cooperative Systems" (2022), tackling the challenge of optimizing neural network designs across diverse, non-collaborating clients. By enabling efficient, collaborative learning without sharing raw data, Liang’s work directly addresses critical hurdles in autonomous driving and distributed robotics. His research not only accelerates the safe deployment of intelligent vehicles but also provides a blueprint for federated learning in heterogeneous, real-world systems—a vital contribution for students and engineers building the next generation of autonomous technologies.

Research Focus

Key Achievements

Top Papers

1. 1
   Federated Transfer Reinforcement Learning for Autonomous Driving

   94 citations · 2022
2. 2
   Federated Transfer Reinforcement Learning for Autonomous Driving

   39 citations · 2019
3. 3
   Cross-Silo Federated Neural Architecture Search for Heterogeneous and Cooperative Systems

   3 citations · 2022

Key Collaborators

Contact & Links