Jinchen Ruan

I'm an M.S. student in Mechanical Engineering (Robotics and Control) at Columbia University, where I am a research assistant in the Creative Machines Lab. My research focuses on robotics, control, and machine learning. Previously I completed my B.Eng. in Robotics Engineering at Beijing University of Technology and worked on robotic control and perception in both academia and industry.

Email  /  CV  /  Scholar  /  LinkedIn  /  Homepage  /  Github  /  BiliBili

Research

My recent work includes generative models for articulated mechanisms, self-modeling of tendon-driven continuum robots, and motion planning and control for robots in complex environments. In the long term, I'm excited about embodied world models that can be updated online so robots can reuse learned skills across homes, factories, and farms.

ArticFlow is a two-stage flow-matching generative model for articulated mechanisms that couples a latent flow and a point flow to synthesize action-conditioned 3D point clouds of articulated objects, significantly reducing Earth Mover's Distance compared to baseline methods.

We build new tendon-driven continuum robot hardware and a MuJoCo simulation with hybrid model-based and learning-based control, and train a flow-matching self-model that maps motor states to external 3D point clouds. This self-model achieves accurate reconstruction in both simulation and real-world experiments; more details, videos, and code are coming soon. (* Equal contribution with Jiong Lin.)

Template adapted from Jon Barron.
Content and modifications © Jinchen Ruan.