Papers
25
Total Citations
633
H-Index
14
About
Yezhou Yang is a prominent researcher at the intersection of computer vision, robotics, and artificial intelligence, with a particular focus on enabling robots to perceive, understand, and interact with the world in human-like ways. His most celebrated contribution — a system that teaches robots manipulation action plans by "watching" unconstrained web videos (180 citations) — demonstrated a groundbreaking approach to machine learning for robotics, bypassing the need for laboriously curated training data. Building on this, Yang has pioneered work in manipulation action recognition, developing invariant representations that allow robots to interpret human activities from visual input alone. His research extends deeply into cognitive robotics and autonomous navigation: he has developed recognition-guided policies for mobile robots to actively search for objects in indoor environments, and explored functional scene understanding that moves robots beyond simple object recognition toward genuine environmental reasoning. His work on human-robot collaboration includes vision-based hand movement prediction for collision-free trajectory planning, improving safety in shared workspaces. More recently, Yang has advanced hierarchical reinforcement learning frameworks with relational graph structures for goal-driven navigation tasks. Collectively accumulating over 500 citations, his body of work represents a sustained and impactful effort to bridge perception, language, and action in intelligent robotic systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2Detection of Manipulation Action Consequences (MAC)60 citations · 2013
- 3
- 4
- 5
- 6Towards a Watson that sees: Language-guided action recognition for robots26 citations · 2012
- 7
- 8Learning hand movements from markerless demonstrations for humanoid tasks24 citations · 2014
- 9A survey on semantic-based methods for the understanding of human movements22 citations · 2019
- 10