news

Nov 18, 2026 Our paper “RoboStriker: Latent-Space Strategic Games for Autonomous Humanoid Boxing” has been accepted for publication in IEEE Robotics and Automation Letters (RA-L). Yang Li is a co-corresponding author.
Two humanoid robots wearing red boxing gloves spar in the RoboStriker lab demonstration.
Real-world boxing with two Unitree G1 robots. Figure 3 from the paper.
Nov 03, 2026 Our paper, “SOMA: Safe Open-ended Multi Agent Reinforcement Learning for Soft Landing of Launch Vehicles” (Co-first Author), has been accepted by IEEE Transactions on Aerospace and Electronic Systems (TAES). Many thanks to all collaborators!
Oct 24, 2026 Join us at RLChina 2026 for two events on embodied intelligence and robot learning!

November 8 (afternoon) · Forum 14: 具身智能与机器人学习 (Embodied Intelligence and Robot Learning)
Co-chairs: 张伟楠 (Weinan Zhang) and 李阳 (Yang Li), both from Shanghai Jiao Tong University.
Speakers: 郭浩宇 (上海人工智能实验室), 石野 (上海科技大学、瞬适科技), 赵乾佑 (Sharpa), 班雨彤 (上海交通大学), 李永露 (上海交通大学), and 陈立 (源策未来 / Archon Robotics).

November 11 (afternoon) · RLChina Tutorial: 具身智能关键技术 (Key Technologies in Embodied Intelligence)
Host: 李阳 (Yang Li). Three lectures cover embodied AI foundations and the frontiers and practice of Physical Intelligence Harness:
  1. 具身智能导论 — 竺正邦 (上海人工智能实验室)
  2. 物理智能Harness前沿 — 李阳 (上海交通大学)
  3. 物理智能Harness实践 — 朱觐镳 (通用决策智能研究所)
We warmly welcome researchers, students, and practitioners to join the discussions and tutorial!
Oct 15, 2026 New paper in Progress in Aerospace Sciences: “Liquid-propellant rocket engine health monitoring technologies: A survey, insights, and prospects”. Authors: Jianjun Wu, Yuwei Liu, Yuqiang Cheng, Shuming Yang, and Yang Li. From a Physical AI perspective, this survey reframes liquid-propellant rocket-engine health monitoring as a physics-grounded, closed-loop intelligence problem—linking multimodal sensing, fault diagnosis, prognostics, and autonomous decision-making to make safety-critical propulsion systems more self-aware and resilient. Progress in Aerospace Sciences is Elsevier’s invitation-only journal devoted exclusively to critical reviews across aeronautics and astronautics—including AI, machine learning, and onboard autonomy—and currently reports a CiteScore of 34.8 and an Impact Factor of 16.2.
Sep 28, 2026 Two papers have been accepted at NeurIPS 2026: “Context Value Informed In-Context Reinforcement Learning” (Corresponding Author) and “Axiomatic Reinforcement Learning for Open Multi-Agent Systems from Shapley Axioms” (Corresponding Author). Many thanks to all collaborators!
Sep 18, 2026 I am serving as an Area Chair for ICLR 2027.
Sep 18, 2026 Our paper (co-first author), “Embodied Science Agents: Making Embodied Experimental Design Learnable,” has been accepted as an oral presentation at DAI 2026.
Aug 06, 2026 New paper: “SkillTV-Bench: Benchmarking How Well Judges Perform on Skill-Augmented Agentic Execution” (Corresponding Author). SkillTV-Bench evaluates whether LLM- and agent-based judges can verify complete skill-augmented agent executions, while SkillTV-Evolve automatically refines a reusable JudgeSkill for more reliable trajectory verification and rollout selection. arXiv · Code
Aug 03, 2026 Two new papers are available. “CoNav-UAV” (Corresponding Author) models cooperative dual-altitude aerial navigation as a Stackelberg game, coupling a high-altitude VLM leader with a low-altitude control follower (arXiv). “DF³” (Corresponding Author) forecasts future states entirely within a frozen vision encoder, removing heavy decoders for efficient autonomous navigation (arXiv).
Jul 24, 2026 New preprint: “InternLab: A Governed Network of Physically Intelligent Autonomous Laboratories for Open-world Scientific Discovery” (First Author). InternLab proposes a governed network that connects physically intelligent autonomous laboratories, enabling coordinated and accountable open-world scientific discovery across distributed physical labs. DOI
Jul 06, 2026 New paper: “Multi-Robot Open Adaptive Teaming Across Unseen Environments, Partners, and Scales” (First Author). We formulate open adaptive multi-robot teaming and introduce HOLA, enabling drone and quadruped teams to coordinate with unseen partners, in unseen environments, and at varying team scales, with zero-shot transfer to physical robots. arXiv
Jun 29, 2026 New paper: “Clarus: Coordinating Autonomous Research Agents toward Web-Scale Scientific Collaboration.” Clarus turns autonomous research from a closed workflow into an open, auditable, attributable, and resource-aware collaboration network connecting agents, researchers, and physical laboratories. arXiv · Project