Nanjing University · School of Intelligent Science and Technology

SOARLAB

Flying General Intelligence Lab

Flying robots, embodied agents, real systems. Led by Hao Xu, we build embodied and collaborative intelligence for aerial robots, humanoids, and emerging flying humanoid platforms, with a focus on robotics agents, world/action models, VLA-style robot foundation models, distributed perception, SLAM, and reinforcement learning.

2027 Admissions Publications GitHub Google Scholar

Highlights

Recent Work

AUG perception-aware UAV planning
IROS 2026 · Honorable Mention

AUG: Perception-Aware UAV Graph-Based Planning

Perception-aware UAV planning for GNSS-denied, feature-sparse environments. Honorable Mention for the IROS 2026 Best Paper Award on Safety, Security, and Rescue Robotics.

Project PDF
Spatial Memory by Agents, for Agents
arXiv · Oct 2026

Spatial Memory by Agents, for Agents

Coding agents design a robot's spatial memory by writing code, running benchmarks, and iterating on feedback (Guided System Improvement). About 51% higher average unified score than the strongest baseline across five task tracks.

arXiv Project
ECHO-G co-speech humanoid motion
arXiv · Open source

ECHO-G: Embodied Co-speech Humanoid Motion

Humanoids that gesture while they talk, on the beat: joint text and speech latents, end-to-end whole-body control, running on a Jetson Orin AGX. With Mondo Robotics and the HKUST Aerial Robotics Group.

Paper Code Video

Join us

2027 Admissions

We are recruiting direct Ph.D., regular Ph.D., and master's students from the 2027 cohort who are interested in embodied intelligence, robotics, and core algorithms. The most urgent directions are WAM/VLA and AI agents, while strong students in distributed perception, SLAM, and reinforcement learning are also welcome.

Recruiting Research

PI

Principal Investigator

Hao Xu
Hao Xu Tenure-track Associate Professor Personal homepage →
Nanjing University emblem Nanjing UniversitySchool of Intelligent Science and Technology

Hao Xu received his Ph.D. from the Department of Electronic and Computer Engineering at the Hong Kong University of Science and Technology, supervised by Prof. Shaojie Shen. He has worked as a multi-sensor information fusion algorithm engineer at DJI and as an associate researcher at Beihang University, and joins Nanjing University as a tenure-track associate professor. His prior work spans spatial intelligence for aerial and legged robots, aerial swarm state estimation, distributed SLAM, and autonomous swarm systems; his future research centers on embodied and collaborative intelligence.

Dr. Xu has published in major robotics journals and conferences including IEEE TRO, IEEE RAL, ICRA, and IROS, and serves as a reviewer for these venues. The collaborative work RACER received the IEEE Transactions on Robotics King-Sun Fu Memorial Best Paper Award, the first time this award went to a team from China. His released systems include D2SLAM, Omni-Swarm, VINS-Fisheye, and FoxTracker.

Research

Research Pillars

01

Robotics Agent

Agent-oriented perception, planning/control, world models, simulation, and behavior decision-making for robots operating under real hardware constraints.

02

WAM/VLA Robot Foundation Models

World-action and vision-language-action modeling for robot foundation models validated on aerial robots, humanoids, and flying humanoid platforms.

03

Distributed Perception & Swarms

Distributed SLAM, aerial swarm state estimation, reinforcement learning, and autonomous swarm systems, building on the lab's long-standing core robotics work.

Latest

News

All news →
  1. Paper
    New preprint: Spatial Memory by Agents, for Agents. Coding agents design a robot’s spatial memory by writing code, running benchmarks, and iterating on feedback. We call this Guided System Improvement (GSI). Across five task tracks, the best configuration improves the average unified score by about 51% over the strongest all-task baseline. Joint work with Mondo Robotics. arXiv · Project page
  2. Award
    Our collaborative paper AUG: Perception-Aware UAV Graph-Based Planning for GNSS-Denied Navigation in Feature-Sparse Environments received an Honorable Mention for the IROS 2026 Best Paper Award on Safety, Security, and Rescue Robotics.
  3. Paper
    ECHO-G is out: humanoid robots that gesture while they talk, landing each move on the beat. ECHO-G combines text and speech latents for semantically rich full-body gestures with end-to-end whole-body control, and runs on an embedded Jetson Orin AGX. Joint work with Mondo Robotics and the HKUST Aerial Robotics Group. arXiv · Code · Video
  4. Lab
    Early members will help build the lab’s infrastructure, including compute servers, experimental space, and UAV and robot validation platforms.
  5. Admissions
    Our most urgent directions are WAM/VLA robot foundation models and AI agents. Students with strong interests in distributed perception, SLAM, reinforcement learning, aerial swarms, and real robotic systems are also welcome. See the recruiting page.

Research

Selected Publications

All publications →
  1. arXiv
    Spatial Memory by Agents, for Agents
    Qianwei Wang, Ce Zhang, Zhanpeng Luo, and 3 more authors
    arXiv preprint arXiv:2610.22736, 2026
    arXiv Website
  2. arXiv
    echo-g.jpg
    ECHO-G: Embodied Co-speech Humanoid mOtion Generation
    Yizhao Li*, Pusen Gao*, Ming Wang, and 3 more authors
    arXiv preprint arXiv:2609.39575, 2026
    arXiv Video Code Website
  3. TRO
    fint.jpg
    Autonomous Flights Inside Narrow Tunnels
    Luqi Wang, Yan Ning, Hongming Chen, and 5 more authors
    IEEE Transactions on Robotics, 2025
    DOI Cited by 4 arXiv IEEE Xplore Code
  4. IROS
    omninxt.jpg
    OmniNxt: A Fully Open-Source and Compact Aerial Robot with Omnidirectional Visual Perception
    Peize Liu, Chen Feng, Yang Xu, and 3 more authors
    In IEEE/RSJ International Conference on Intelligent Robots and Systems, 2024
    DOI Cited by 23 arXiv IEEE Xplore Code Website
  5. TRO
    d2slam.jpg
    D2SLAM: Decentralized and Distributed Collaborative Visual-Inertial SLAM System for Aerial Swarm
    Hao Xu, Peize Liu, Xinyi Chen, and 1 more author
    IEEE Transactions on Robotics, 2024
    DOI Cited by 100 arXiv IEEE Xplore Code
  6. TRO
    racer.jpg
    RACER: Rapid Collaborative Exploration With a Decentralized Multi-UAV System
    Boyu Zhou, Hao Xu, and Shaojie Shen
    IEEE Transactions on Robotics, 2023
    TRO Best Paper DOI Cited by 264 arXiv IEEE Xplore Code
  7. TRO
    omni-swarm.jpg
    Omni-Swarm: A Decentralized Omnidirectional Visual-Inertial-UWB State Estimation System for Aerial Swarms
    Hao Xu, Yichen Zhang, Boyu Zhou, and 4 more authors
    IEEE Transactions on Robotics, 2022
    DOI Cited by 206 arXiv IEEE Xplore Code
  8. IROS
    aug.jpg
    AUG: Perception-Aware UAV Graph-Based Planning for GNSS-Denied Navigation in Feature-Sparse Environments
    Enguang Feng, Haohuan Yuan, Hong Zhang, and 4 more authors
    In IEEE/RSJ International Conference on Intelligent Robots and Systems, 2026
    IROS SSRR Best Paper · Honorable Mention HTML PDF Website
  9. RAL
    narrow-tunnel-ral.jpg
    Neither Fast nor Slow: How to Fly Through Narrow Tunnels
    Luqi Wang, Hao Xu, Yichen Zhang, and 1 more author
    IEEE Robotics and Automation Letters, 2022
    DOI Cited by 20 arXiv IEEE Xplore
  10. ICRA
    icra-uwb.jpg
    Decentralized Visual-Inertial-UWB Fusion for Relative State Estimation of Aerial Swarm
    Hao Xu, Luqi Wang, Yichen Zhang, and 2 more authors
    In IEEE International Conference on Robotics and Automation, 2020
    DOI Cited by 191 arXiv IEEE Xplore