Tong Xu

Tong Xu

PhD student for Robotics

George Mason University

RobotiXX Lab

Hello there!

I am a third-year PhD student in RobotiXX Lab at George Mason University, advised by Prof. Xuesu Xiao. I received my Master’s degree from University of Southern California and Bachelor’s degree from Nanjing University of Information Science & Technology.

My primary research interests include robotics, reinforcement learning and foundation models. My current work on fast adaptation for cross-embediments reflects my dedication to advance real-world general-purpose applications in robotics through large-scale machine learning techniques.

Interests

  • Robotics
  • Motion Planning
  • Reinforcement Learning
  • Foundation Models

Education

  • PhD in Computer Science, 2023

    George Mason University

  • M.S. in Computer Science, 2021

    University of Southern California

  • B.E. in Network Engineering, 2017

    Nanjing University of Information Science & Technology

News

Publications

Preprint

Adaptive Dynamics Planning for Robot Navigation

Adaptive Dynamics Planning for Robot Navigation

Traverse the Non-Traversable: Estimating Traversability for Wheeled Mobility on Vertically Challenging Terrain

Traverse the Non-Traversable: Estimating Traversability for Wheeled Mobility on Vertically Challenging Terrain

Journal

Autonomous Ground Navigation in Highly Constrained Spaces: Lessons Learned from The Forth BARN Challenge at ICRA 2025

Autonomous Ground Navigation in Highly Constrained Spaces: Lessons Learned from The Forth BARN Challenge at ICRA 2025

CARoL: Context-aware Adaptation for Robot Learning

CARoL: Context-aware Adaptation for Robot Learning

Pietra: Physics-informed evidential learning for traversing out-of-distribution terrain

Pietra: Physics-informed evidential learning for traversing out-of-distribution terrain

Conference

Verti-Arena: A Controllable and Standardized Indoor Testbed for Multi-Terrain Off-Road Autonomy

Verti-Arena: A Controllable and Standardized Indoor Testbed for Multi-Terrain Off-Road Autonomy

Decremental Dynamics Planning for Robot Navigation

Decremental Dynamics Planning for Robot Navigation

Reward Training Wheels: Adaptive Auxiliary Rewards for Robotics Reinforcement Learning

Reward Training Wheels: Adaptive Auxiliary Rewards for Robotics Reinforcement Learning

VertiSelector: Automatic Curriculum Learning for Wheeled Mobility on Vertically Challenging Terrain

VertiSelector: Automatic Curriculum Learning for Wheeled Mobility on Vertically Challenging Terrain

Verti-Bench: A General and Scalable Off-Road Mobility Benchmark for Vertically Challenging Terrain

Verti-Bench: A General and Scalable Off-Road Mobility Benchmark for Vertically Challenging Terrain

Reinforcement learning for wheeled mobility on vertically challenging terrain

Reinforcement learning for wheeled mobility on vertically challenging terrain

Experience

 
 
 
 
 
RobotiXX Logo

Graduate Research Assistant

RobotiXX

May 2024 – Present Fairfax
  • Research in off-road navigation, deep reinforcement learning, humaniods, foundation models
  • Leading the Verti-Bench for fast adaptation cross different types of vehicles in off-road navigation scenario
 
 
 
 
 
George Mason University Logo

Graduate Teaching Assistant

George Mason University

Aug 2023 – May 2025 Fairfax
  • Designed student lab contents involving data structure, led weekly lab recitations and office hours
  • Created grading scripts and managed a team of 8 undergraduate teaching assistants
 
 
 
 
 
H2X Lab Logo

Research Intern

H2X Lab

May 2022 – Aug 2022 Boston
  • DeepVO - Visual Odometry with Deep Learning
  • OpenGuide - A Scalable Human-Like Guidance System for Travelers with Visual Impairment

Projects

*

DeepVO

Reproduced Deep Visual Odometry architecture by integrating a pretrained FlowNetSimple model with LSTM. Achieved 14.8% performance improvement through epoch optimization and 8.4% improvement through hyperparameter tuning in the loss function, evaluated on KITTI and nuScenes datasets using translation and rotation RMSE metrics.

Cone Detection

Implemented Faster R-CNN for cone detection by integrating Region Proposal Network with Fast R-CNN. Achieved 37.4% higher recall rate compared to YOLOv3 on a cone-annotated dataset, demonstrating superior detection performance.

OpenGuide

A Scalable Human-Like Guidance System for Travelers with Visual Impairment.

Contact

  • txu25@gmu.edu
  • 4400 University Dr, Fairfax, VA 22030
  • RobotiXX Lab