About

I am a master’s student in Control Science and Engineering at Shanghai University of Engineering Science. My research focuses on robot learning, embodied AI, LLM agents, mobile agents, and human-feedback decision systems. I am interested in building agents that connect perception, task-level reasoning, external knowledge, and low-level control in deployable autonomous systems.

Email / Google Scholar / GitHub

News

  • May 2026: Our human-in-the-loop multi-agent ventilator decision support work is available on arXiv.
  • Mar. 2026: Our hierarchical quadruped navigation work is available on arXiv.
  • Jan. 2026: Our mobile-agent knowledge retrieval work is available on arXiv and appears in WWW 2026.
  • 2025: HeStIa was accepted to IEEE/RSJ IROS 2025.

Selected Research

Curiosity-driven knowledge retrieval framework
We introduce a curiosity-driven knowledge retrieval framework for mobile agents.

Curiosity Driven Knowledge Retrieval for Mobile Agents

Sijia Li, Xiaoyu Tan, Shahir Ali, Niels Schmidt, Gengchen Ma, Xihe Qiu

ACM Web Conference 2026 (WWW'26), Top 20%, 2026.

paper / project

The method estimates execution uncertainty, retrieves task-relevant external knowledge, and organizes it as AppCards that encode app functions, UI mappings, parameters, and interaction patterns.

Hierarchical policy framework for quadruped navigation
We study hierarchical policy learning for quadruped navigation.

Task-Level Decisions to Gait-Level Control

Sijia Li, Haoyu Wang, Shenghai Yuan, Yizhuo Yang, Thien-Minh Nguyen

IROS 2026 under review / arXiv preprint, 2026.

paper

The framework connects sparse task-level semantic and geometric cues with gait-conditioned low-level reinforcement-learning control, improving interpretability and policy debugging on mixed terrains.

Human-in-the-loop multi-agent decision framework
We develop a human-in-the-loop multi-agent decision support framework for ventilator setting recommendation.

Human-in-the-Loop Multi-Agent Ventilator Decision Support

Sijia Li, Xiaoyu Tan, Qixing Wang, Weiyi Zhao, Chen Zhan, Teqi Hao, Xuemin Wang, Lei Gu, Roland Eils, Xihe Qiu

MICCAI 2026, Top 8%, 2026.

paper

The system uses contextual bandit preference learning to incorporate expert feedback and update decision policies in high-risk clinical scenarios.

HeStIa: Asynchronous Embodied Dynamic Locomotion Learning

Xiaoyu Tan, Haoyu Wang, Sijia Li, Yinghui Xu, Xihe Qiu

IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2025.

paper

HeStIa explores asynchronous embodied dynamic locomotion learning for walking robots through multimodal large language models, connecting high-level multimodal reasoning with dynamic locomotion behavior.

Research Interests

  • Robot learning and embodied AI: hierarchical policies, reinforcement learning, quadruped navigation, sim-to-real transfer, and deployable autonomous systems.
  • LLM and mobile agents: curiosity-driven retrieval, tool use, GUI automation, AppCards, and long-horizon task execution.
  • Human feedback and decision learning: contextual bandits, preference modeling, human-in-the-loop evaluation, and reliable policy updates.
  • Multimodal perception and control: MLLM perception, semantic meta-commands, and reinforcement-learning control for embodied systems.

Selected Awards

  • National Third Prize, 10th China Graduate Smart City Technology and Creative Design Competition, 2024.
  • Shanghai Regional Third Prize, China Robotics and Artificial Intelligence Competition, Quadruped Bionic Robot Track, 2024.
  • Shanghai Regional First Prize, 12th Shanghai College Student Engineering Practice and Innovation Ability Competition, 2022.
  • National Third Prize, 24th China Robotics and Artificial Intelligence Competition National Final, 2022.
  • Shanghai Regional First Prize, 2022 Shanghai TI Cup Undergraduate Electronic Design Contest, 2022.

Links