Bowen Jiang

I am a PhD student at UT Austin, advised by Prof. Roberto Martín-Martín. My current research focuses on dexterous manipulation.

From 2022 to 2024, I was a Master's student at the Robotics Institute (RI) at Carnegie Mellon University, advised by Prof. David Held and mentored by Wenxuan Zhou. My research focused on reinforcement learning and manipulation.

Prior to CMU, I earned my B.S. in Engineering at Harvey Mudd College with distinction in 2022, with a concentration in Visual Arts.

Email  /  GitHub  /  LinkedIn

Research

I'm interested in anything with arms and/or legs.

CoDex: Learning Compositional Dexterous Functional Manipulation without Demonstrations

CoDex learns dexterous functional object manipulation without human demonstrations by translating vision-language-model guidance into semantic constraints for optimization and reinforcement learning.

Learning Generalizable Tool-use Skills through Trajectory Generation

ToolGen learns generalizable tool-use skills for deformable object manipulation by generating tool point-cloud trajectories and aligning novel tools to them for execution.

HACMan++: Spatially-Grounded Motion Primitives for Manipulation

We present HACMan++, a reinforcement learning framework using a novel action space of spatially-grounded parameterized motion primitives for manipulation tasks.

HACMan: Learning Hybrid Actor-Critic Maps for 6D Non-Prehensile Manipulation

We propose a spatially-grounded and temporally-abstracted action representation with a hybrid discrete-continuous reinforcement learning framework.

POIS: Policy-Oriented Instance Segmentation for Ambidextrous Robot Picking

We propose policy-oriented Instance Segmentation for Ambidextrous Robot Picking, which predicts a pair of target masks allowing ambidextrous robots to pick objects in cluttered scenes without mutual interference.

Design and source code from Jon Barron's website