Chengguang (Claude) Xu

Hi, I'm a final-year Ph.D. candidate at Northeastern University in Boston, where I work closely with Professor Lawson L.S. Wong and Professor Christopher Amato. My research interests span machine learning, computer vision, and natural language processing, particularly at their intersection with robotics. My long-term goal is to equip robots with human-level multi-modality intelligence, including advanced visual perception for better scene understanding and robust natural language grounding for improved language-driven robot control.

Before joining Northeastern, I earned my Master's and Bachelor's degrees from Nankai University in China.

Resume  / Scholar  /  Linkedin

profile photo

Contact

If you are interested in my research or potential collaboration, you can reach me at: xu [dot] cheng [at] northeastern [dot] edu.

Research

Recently, I have been working on the cross-modality learning and grounding problems in vision-and-language navigation (VLN). I am also exploring the sequential decision-making problem under partial observability in different object goal navigation tasks. Please feel free to contact me if you are interested! xu dot cheng at northeastern dot edu

Robot Navigation in Unseen Environments using Coarse Maps
Chengguang Xu, Christopher Amato, Lawson L.S. Wong*
IEEE International Conference on Robotics and Automation (ICRA), 2024
video/ poster

Propose a vision-based navigation system that can localize an embodied agent on top-down 2-D coarse maps (e.g., hand-drawn maps) from photo-realistic panoramic RGB images.

Vision-and-Language Navigation in Real World using Foundation Models
Chengguang Xu*, Hieu Trung Nguyen, Christopher Amato, Lawson L.S. Wong
In the Workshop of Foundation Models for Decision Making @ NeurIPS, 2023
video / poster

Propose a navigation framework that achieves zero-shot vision-and-language navigation in real world scenarios using a Large Language Model (LLM) and a Large Visual-Language Model(VLM).

Hierarchical Robot Navigation in Novel Environments using Rough 2-D Maps
Chengguang Xu*, Christopher Amato, Lawson L.S. Wong
Conference on Robot Learning (CoRL), 2020

Propose a hierarchical navigation framework to facilitate high-level long-horizon planning and low-level goal-conditioned policy learning. To bridge the high-level map, we utilized a conditional generative model to generate RGB image goals from binary occupancy grids.

Deep Supervised Summarization: Algorithm and Application to Learning Instructions
Chengguang Xu*, Ehsan Elhamifar
Conference on Neural Information Processing Systems (NeurIPS), 2019

Propose a triple loss to learn sequential neural representations of video clips for supervised video summarization.

Design and Performance Evaluation of a Simple Semi-Physical Human-Vehicle Collaborative Driving Simulation System
Wenyu Li*, Feng Duan, Chengguang Xu
IEEE Access, 2019

We design a co-pilot driving system that uses vision controller as a vehicle assistant.

A Human-Vehicle Collaborative Simulated Driving System Based on Hybrid Brain–Computer Interfaces and Computer Vision
Wenyu Li, Feng Duan*, Shili Sheng, Chengguang Xu, Rensong Liu, Zhiwen Zhang, Xue Jiang
IEEE Transactions on Cognitive and Developmental Systems, 2018

We design a co-pilot driving system that uses vision controller as a vehicle assistant.


Feel free to steal this website's source code. Do not scrape the HTML from this page itself, as it includes analytics tags that you do not want on your own website — use the github code instead. Also, consider using Leonid Keselman's Jekyll fork of this page.