Shang-Hsuan Yang

About

Shang-Hsuan Yang is a researcher whose work lies at the intersection of reinforcement learning and constrained optimization, with a particular focus on solving real-world control problems. Their most notable contribution, "Escaping from Zero Gradient: Revisiting Action-Constrained Reinforcement Learning via Frank-Wolfe Policy Optimization" (2021), tackles a critical limitation in action-constrained RL: the vanishing gradient problem that plagues projection-based methods. By introducing Frank-Wolfe policy optimization, Yang provides an elegant alternative that avoids zero-gradient pitfalls while maintaining constraint satisfaction, making it especially valuable for applications like resource-constrained scheduling and kinematically-limited robot control. This work has garnered 3 citations, establishing Yang as an emerging voice in bridging theoretical RL advances with practical deployment challenges. Their research addresses a fundamental tension in constrained RL—how to learn effectively when the action space is bounded—offering a pathway toward more reliable autonomous systems in safety-critical domains.

Research Focus

Key Achievements

Top Papers

1. 1
   Escaping from Zero Gradient: Revisiting Action-Constrained Reinforcement Learning via Frank-Wolfe Policy Optimization

   3 citations · 2021

Key Collaborators

Contact & Links