Xiaolin Fang

I am a final-year Ph.D. student at MIT CSAIL, LIS Group, working with Leslie Kaelbling and Tomás Lozano-Pérez. I interned at NVIDIA Seattle Robotics Lab led by Prof. Dieter Fox during summer 2022. Previously, I obtained my bachelor's degree from Chu Kochen Honors College, Zhejiang University, where I worked with Prof. Xiaowei Zhou.

My research focuses on robot learning and planning, with an emphasis on generalization to new goals, novel environments, and long-horizon manipulation tasks. My previous work involves vision-language models (VLMs), generative models (e.g., diffusion models), and task and motion planning, aiming to develop composable robot skill models that generalize across visual changes, spatial constraints, and goal specifications. I have worked on

• A framework that composes skills at inference time through task and motion planning, leveraging estimated affordances from raw perception inputs
• Skill representation that enhances composability of learned skill models, enabling generalization to unseen constraints at testing time
• Scaling up skill learning with foundation models

All these approaches have been validated on real-world robotic platforms, including Franka (Single Arm), Rainbow Rby-1 (Mobile-base Bi-Manual), Movo (Mobile-base Bi-Manual), and BostonDynamics Spot (Mobile Single Arm).

I'm looking for full-time positions in the industry. Let's chat :)