Nelson Chen

Ph.D. Candidate · Computer Science (AI & Robotics)

Rutgers University Co-advised by Prof. Mridul Aanjaneya & Prof. Kostas Bekris

I'm a computer science Ph.D. candidate specializing in AI & Robotics at Rutgers University. My research focuses on differentiable simulation and learning-based control for tensegrity robots that are compliant, cable-driven structures with rich contact dynamics. I build graph neural network methods for learning robot dynamics and integrate them with model predictive controllers for real-world navigation. Previously, I worked as a Senior Data Scientist building NLP and document-understanding systems, and completed my M.S. in mechanical engineering at UC Berkeley and B.S in mechanical engineering at Northwestern.

Email CV LinkedIn GitHub Google Scholar
Nelson Chen

News

Feb 2026 CableRobotGraphSim accepted as an Oral at L4DC 2026 — preprint on arXiv.
2026 New paper on MPC of Tensegrity Robots via Contact-Aware GNN Dynamics submitted to IROS 2026.
2026 Tensegrity navigation paper accepted at IEEE Robotics and Automation Letters.
Sep 2025 Completed research internship at Dyna Robotics, working on VLA models and offline RL for robotic manipulation.
CoRL 2024 Paper on differentiable tensegrity dynamics with GNNs published at CoRL 2024 — arXiv:2410.12216.

Selected Publications

MPPI Navigation Tasks
Model Predictive Control of Tensegrity Robots via Contact-Aware Graph Neural Dynamics Model
N. Chen, P. Meng, C. Tang, A. Degay, Z. Brei, R. Kramer-Bottiglio, K. Bekris, & M. Aanjaneya
IROS 2026 Under Review
Project Page Video
CableRobotGraphSim Pipeline
CableRobotGraphSim: A Graph Neural Network for Modeling Partially Observable Cable-Driven Robot Dynamics
N. Chen, W. R. Johnson, R. Kramer-Bottiglio, K. Bekris, & M. Aanjaneya
L4DC 2026 Oral
Project Page arXiv PDF Code
Open-Source Tensegrity for Navigation
An Open-Source, Reproducible Tensegrity Robot that can Navigate Among Obstacles
W. R. Johnson*, P. Meng*, N. Chen, M. Aanjaneya, R. Kramer-Bottiglio, & K. Bekris
IEEE RA-L 2026
Project Page arXiv PDF
GNN for Tensegrity
Learning Differentiable Tensegrity Dynamics with Graph Neural Networks
N. Chen, K. Wang, W. R. Johnson, R. Kramer-Bottiglio, K. Bekris, & M. Aanjaneya
CoRL 2024
Project Page arXiv PDF Code Video

Experience

Research Intern — Dyna Robotics
May – Sep 2025
Built simulation infrastructure and training pipelines for Vision-Language-Action (VLA) benchmark models on LIBERO and Robomimic datasets using the internal DYNA-1 architecture. Developed dense learning-based reward functions and Q-functions for offline RL applied to towel-flattening manipulation.
Senior Data Scientist — Argo Group International Holdings
Mar 2020 – Sep 2022
Built Named-Entity-Recognition ensemble pipelines using Faster-RCNN, BERT, and LayoutLM to extract structured data chunks from insurance documents.
Senior Data Scientist — WorkFusion
Mar 2017 – Mar 2020
Led R&D to expand WorkFusion's AutoML framework; trained NLP-based data extraction models and a handwritten signature detector using object detection methods.

Education

Ph.D. in Computer Science (AI & Robotics)
Rutgers University, New Brunswick
2022 – 2027 (expected)
M.S. in Mechanical Engineering (Computational Fluid Dynamics)
University of California, Berkeley
2014 – 2016
B.S. in Mechanical Engineering
Northwestern University, Evanston
2010 – 2014