Xiuwei Xu

Papers

Total Citations

H-Index

About

Xiuwei Xu is a rising researcher at the forefront of embodied AI and robotic manipulation, with a focus on bridging large language models and 3D scene understanding for real-world robotics. Xu’s key contributions lie in enabling robots to perform complex, human-like tasks through commonsense reasoning and adaptive perception. Their highly cited 2023 work on “Embodied Task Planning with Large Language Models” (18 citations) demonstrates how LLMs can generate task plans for robots, overcoming their lack of spatial awareness. Building on this, Xu introduced “MoManipVLA” (2025, 7 citations), a vision-language-action model that generalizes mobile manipulation across diverse environments, addressing the scalability bottleneck in robotics. In “Memory-based Adapters for Online 3D Scene Perception” (2024, 6 citations), Xu pioneered a framework for streaming RGB-D data, enabling real-time 3D perception critical for dynamic robotic applications. Notably, Xu also applied AI to healthcare with a “Visual Feedback System for Traditional Chinese Medical Massage Robot” (2019, 5 citations), blending robotics with traditional medicine. With a growing citation footprint and work spanning from foundational perception to task planning, Xu is shaping the next generation of generalist robots that can understand, plan, and act in the physical world.