Profile photo of Lucas Chen

Lucas Chen

I am a Masters student in the CoMMA lab at Purdue University, advised by Zachary Kingston. I am currently working on learning-based motion planning and trajectory optimization for realtime (millisecond!) planning.

I'm still sorting out my research interests, which broadly involve designing robotic systems and algorithms that can make informed adaptations to unseen and ever-changing conditions in the real world. Sometimes, this involves distilling high level semantic information into executable trajectories and motions.

In the past, I have worked on projects spanning kinodynamic motion planning for wacky situations, distributed multiagent RL, open language querying, and human-robot interaction/intent modelling. I am also interested in how we can make proof-of-concept methods from literature reliable and explainable enough to deploy to commercial robots.

Email   |   Github   |   Linkedin

Experience

Education

Publications

Differentiable Particle Optimization for Fast Sequential Manipulation
Lucas Chen, Shrutheesh R. Iyer, Zachary Kingston

ICRA 2026 (in review)
paper | github | video | project page
Publication thumbnail
Parallel Simulation of Contact and Actuation for Soft Growing Robots
Yitian Gao*, Lucas Chen*, Priyanka Bhovad, Sicheng Wang, Zachary Kingston, Laura H. Blumenschein

SoRo 2026 (in review)
paper | github
Publication thumbnail
Physics-Grounded Differentiable Simulation for Soft Growing Robots
Lucas Chen*, Yitian Gao*, Sicheng Wang, Francesco Fuentes, Laura H. Blumenschein, Zachary Kingston

RoboSoft 2025
paper | github | poster | slides
Publication thumbnail
Efficient Q-Learning over Visit Frequency Maps for Multi-Agent Exploration of Unknown Environments
Lucas Chen*, Ashvin N. Iyer*, Zixing Wang, Ahmed H. Qureshi

IROS 2023
paper | github | poster
* denotes equal contribution

Projects

movForth
Compile stack machine operations from the Forth language to LLVM IR via static graph analysis
github | talk (2021)
Project thumbnail
Learning Rearrangement from Brownian Motion
CS 592 Reinforcement Learning - Generate diverse new trajectories by diffusing in task-space, even with obstacles and self-collisions.
poster
Project thumbnail
Neural Time Fields as Metric Learning
We solve multi-query planning problems by learning a special higher dimensional manifold whose L2 distance is the same as the shortest geodesic path.
Project thumbnail