Papers
7
Total Citations
155
H-Index
6
About
Bohan Wu is a robotics researcher whose work sits at the intersection of deep reinforcement learning, computer vision, and dexterous manipulation. Her primary contributions focus on enabling multi-fingered robotic hands to grasp objects in cluttered, real-world environments—a notoriously difficult problem requiring both precise control and robust perception. Wu pioneered the “GenerAL” framework for generative attention learning, which achieved high-performance grasping in clutter (52 citations), and introduced the Pixel-Attentive Policy Gradient method, an on-policy RL approach that learns to attend to task-relevant visual features for dexterous grasping (39 citations). She also developed MAT (Multi-Fingered Adaptive Tactile Grasping), a system that combines tactile feedback with deep RL to recover from grasp failures caused by calibration errors (31 citations). Beyond grasping, Wu has advanced long-horizon manipulation with SQUIRL, a method for learning from video demonstrations that reduces the need for extensive real-world robot experience (16 citations), and M-EMBER, which tackles mobile manipulation through factorized domain transfer from simulation to reality (6 citations). Her work consistently pushes toward more sample-efficient, robust, and transferable robotic learning systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5
- 6
- 7