Research

My research interests lie in developing both general-purpose algorithms and specialized autonomous systems for service robots, including planning, human-robot interaction, and service robot design. I aim to bridge the gap between robotics research and its real-world applications in socially aware scenarios.

My long-term research goal is to develop deployable service robotic systems capable of operating in open-world environments. I define the open world as one characterized by (1) natural and unconstrained human behavior, (2) incomplete environmental knowledge at both the semantic and geometric levels, and (3) unknown environmental structures populated with novel objects.

To tackle the challenges of open-world operation, I adopt the system design grounded in compositionality and modularity. I explore structured representations that link semantic concepts with their grounding in the physical world. At the low level, I develop robust navigation and manipulation systems that passively tolerate unforeseen objects and environments in a passive manner. At the high level, I design systems and algorithms that actively expand the robot's representations and models. I also incorporate human presence to interact with both high-level and low-level representations.

Socially-Aware Navigation/Exploration Multimodal Domain Induction for TAMP Mobile Manipulation in the Changing Environment Planning with Partial World Knowledge Multimodal Instructions Grounding Planning and Control for Physical HRI Learning from Human Videos Open-world Interaction Enabling seamless human-robot interaction in daily assistive tasks by developing algorithms that reduce human effort. Open-world Reasoning Enabling robots to reason and generate plans with limited knowledge about the unknown environment. Open-world Cability Empowering robots the ability to perform generalizable navigation/manipulation skills in unknown environments.

Open-world Interaction


GSON: A Group-based Social Navigation Framework with Large Multimodal Model

IEEE Robotics and Automation Letters (RA-L) 2025.
Shangyi Luo, Peng Sun, Ji Zhu, Yuhong Deng, Cunjun Yu, Anxing Xiao (Mentor), Xueqian Wang

TL;DR: Group-based social navigation framework using large multimodal models for socially-aware robot navigation.
Paper Video Code

Robi Butler: Multimodal Remote Interactions with a Household Robot Assistant

IEEE International Conference on Robotics and Automation (ICRA) 2025.
Anxing Xiao, Nuwan Janaka, Tianrun Hu, Anshul Gupta, Kaixin Li, Cunjun Yu, David Hsu

TL;DR: The robot should be able to understand and execute multimodal instructions that combine language and gesture.
Paper Video Website

Quadruped Guidance Robot for the Visually Impaired: A Comfort-Based Approach

IEEE International Conference on Robotics and Automation (ICRA) 2023.
Yanbo Chen, Zhengzhe Xu, Zhuozhu Jian, Gengpan Tang, Yunong Yangli, Anxing Xiao (Mentor), Xueqian Wang, Bin Liang

TL;DR: Comfort-based guidance robot system that safely guides visually impaired users through complex environments.
Paper Video

Robotic Guide Dog: Leading a Human with Leash-Guided Hybrid Physical Interactions

IEEE International Conference on Robotics and Automation (ICRA) 2021.
Anxing Xiao*, Wenzhe Tong*, Lizhi Yang*, Jun Zeng, Zhongyu Li, Koushil Sreenath
This paper was the ICRA Best Paper Award Finalist for Service Robotics.
TL;DR: The world's first robotic guide dog that can lead visually impaired to safely travel in a confined space.
Paper Video


Open-world Reasoning

Robot Operation of Home Appliances by Reading User Manuals

Conference on Robot Learning (CoRL) 2025.
Jian Zhang, Hanbo Zhang, Anxing Xiao, David Hsu

TL;DR: Enabling robots to operate home appliances by understanding and following user manuals.
Paper Video Code

Octopi: Object Property Reasoning with Large Tactile-Language Models

Robotics: Science and Systems (RSS) 2024.
Samson Yu, Kelvin Lin, Anxing Xiao, Jiafei Duan, and Harold Soh

TL;DR: Combining tactile perception with language enables robots to understand physical properties through interaction.
Paper Website Code

LLM-State: Expandable State Representation for Long-horizon Task Planning in the Open World

Preprint 2024.
Siwei Chen, Anxing Xiao, David Hsu

TL;DR: Expandable state representation that continuously updates object attributes for long-horizon task planning.
Paper Video

© Anxing Xiao (01/2024)
Designed by BootstrapMade