Feiyang Wu
Papers
2
Total Citations
9
H-Index
2
About
Feiyang Wu is pioneering the frontier of legged locomotion, with a sharp focus on enabling bipedal and humanoid robots to navigate the world’s most challenging, uneven terrains. His research masterfully bridges reinforcement learning (RL) and inverse reinforcement learning (IRL), tackling the fundamental problem of translating complex, dynamic environments into robust, real-world robotic behavior. In his highly cited 2024 work, "Infer and Adapt," Wu introduced a novel IRL framework that allows bipedal robots to learn reward functions directly from human demonstrations, dramatically improving their ability to adapt to highly irregular surfaces—a breakthrough that has already garnered 6 citations for its practical elegance. Building on this, his 2025 paper, "Learn to Teach," addresses the notorious sample inefficiency of RL for humanoid locomotion. By developing a privileged learning paradigm that leverages a teacher-student architecture, Wu’s method slashes the enormous simulation samples typically required, paving the way for real-world deployment. With a cumulative impact of 9 citations on these foundational works alone, Wu is establishing himself as a rising star in robotics, pushing the boundaries of what humanoid robots can achieve outside the lab.
Research Focus
Key Achievements
Top Papers
- 1
- 2