Junwen Gu | 顾俊文

I am a Ph.D. candidate at the Key Laboratory of Cognition and Decision Intelligence for Complex Systems, Institute of Automation, Chinese Academy of Sciences (CASIA), where I have been advised by Prof. Zhengxin Wu since 2022. Previously, I received my B.Eng. degree in Automation from the University of Science and Technology Beijing (USTB).

My research primarily focuses on underwater robotics, where I strive to explore full-stack technologies and develop advanced robotic systems from a holistic perspective. I am currently exploring Embodied AI (RL and VLA) to enhance robotic autonomy and investigating how system architectures shape intelligence across broader domains, including automated floor plan design.

Junwen Gu
Email Google Scholar ResearchGate LinkedIn CV
English 中文
GitHub

Education

Institute of Automation, Chinese Academy of Sciences

Doctor of Philosophy - PhD, Robotics

Sept 2022 - June 2027 (Expected)

University of Science and Technology Beijing

Bachelor of Engineering - BE, Automation

Sept 2018 - June 2022

TOP 5% | Dean's Medal | National Scholarship (three times)

Research

* indicates equal contribution, † indicates corresponding author

First-Author Publications

TRO Paper
TRO Paper
Deformation control and thrust analysis of a flexible fishtail with muscle-like actuation
Junwen Gu*, Jian Wang*, Zhijie Liu, Min Tan, Junzhi Yu, Zhengxin Wu†
IEEE Transactions on Robotics (TRO), 41, 159-179, 2025
Paper / 中文介绍

We developed a flexible fishtail actuated by artificial muscles. Through a combination of PDE-based modeling and DRL controller, the system achieves real-time deformation control with up to a 203% propulsion enhancement.

USIM/U0 Paper
USIM/U0 Paper
USIM and U0: A Vision-Language-Action Dataset and Model for General Underwater Robots
Junwen Gu*, Zhiheng Wu*, P. Si, S. Qiu, Yukai Feng, L. Sun, L. Luo, L. Yu, Jian Wang†, Zhengxin Wu
arXiv preprint arXiv:2510.07869, 2025
Paper / Project Page

We present USIM, a underwater VLA dataset across 20 diverse tasks, and U0, a VLA model achieving an 80% success rate in underwater tasks including inspection, navigation, and tracking.

CBS 2024 Paper
CBS 2024 Paper
Learning to Turn: Deformation Control of a Novel Flexible Fishtail Actuated by Artificial Muscles
Junwen Gu, Yukai Feng, Zhengxin Wu, Jian Wang†, Junzhi Yu
2024 IEEE International Conference on Cyborg and Bionic Systems (CBS), 171-176
Paper

This work presents a flexible fishtail using SAC reinforcement learning to optimize turning agility. It enables adjustable compliance for enhanced fluid interactivity.

Co-authored Publications

TIE Paper
TIE Paper
Decentralized multirobotic fish pursuit control with attraction-enhanced reinforcement learning
Yukai Feng, Zhengxin Wu†, Jian Wang, Junwen Gu, F. Yu, Junzhi Yu, Min Tan
IEEE Transactions on Industrial Electronics (TIE), 72 (8), 8290-8300, 2025
Paper

We proposed a decentralized cooperative pursuit strategy for multi-robotic fish using curriculum learning. The algorithm enables adaptive coordination in complex underwater environments.

ICCSS 2021 Paper
ICCSS 2021 Paper
Observer-based feedback control for linear parabolic PDEs with quantized input
Xuena Zhao, Junwen Gu, Zhijie Liu, Wei He
2021 8th International Conference on Information, Cybernetics, and Computational Social Systems (ICCSS)
Paper

We developed a PDE observer and quantization feedback compensator for linear parabolic PDE systems. The framework ensures exponential stability, validated through Lyapunov-based proofs and numerical simulations.