Bingqi Huang.

Robotics Research · Tsinghua University

Robot learning.
Real-world systems.

I’m Bingqi Huang, a robotics researcher and full-stack engineer. I study how robots learn, reason, and act—and build the hardware and software to put those ideas to work.

Explore my research GitHub ↗
My focusVision-language-action policiesWorld action modelsTrustworthy robot agentsMobile manipulation

About

I am a Ph.D. student in the School of Aerospace Engineering at Tsinghua University, advised by Prof. Zhaokui Wang. I lead robotics research and development at the Distributed and Intelligent Space System Laboratory (DSSL), spanning learning algorithms, robot platforms, and real-world deployment.

Previously, I studied robotics at TU Delft as an exchange student.

01 / Selected research

Learning, action & autonomy.

From camera-robust policies to accountable robot decision-making.

01
Robot learningFirst author · 2026

CVAC Cross-View Action Consistency

Learning camera-robust vision-language-action policies by aligning flow-matching action predictions across observations of the same robot state. Built on π0.5 / OpenPI, with single-view inference and no camera calibration input.

87.2%LIBERO-Plus success
74.4%Real-robot held-out views
Read paper
02
World action modelsFirst author · 2026

SCVC Selective Cross-View Consistency

Adapting NVIDIA Cosmos Policy with selective consistency for view-invariant action, proprioceptive-state, and value predictions, while preserving view-dependent future video predictions.

+12.2 ppHeld-out orbital-view success
NVIDIA Cosmos Policy · 2B parametersRead paper
03
Agentic roboticsManuscript in preparation

Thoth A Harness for Lunar Robot Agents

A Codex-inspired agent harness for adapting mission structure at runtime. Typed robot capabilities, authorization boundaries, independent effect verification, and auditable execution traces make autonomous decisions inspectable and accountable.

NVIDIA Isaac Sim / Space Robotics Bench · ROS 2Ongoing research

02 / Robotic systems

Built from the ground up.

Mechanical design, electronics, control, and learning—connected in real hardware.

LunarBot mobile manipulator with a RealMan arm and independently steered wheelsLunarBot / Research platform
Principal engineer2024–Present

LunarBot

A holonomic mobile manipulation platform for robotics research.

I led its end-to-end development, from mechanical and electrical design to assembly, ROS software, system integration, and debugging.

  • Hardware4-wheel independent steering and driving; RealMan RM75 manipulator.
  • ControlQP-based whole-body control for coordinated mobile manipulation.
  • LearningLeRobot-format data collection, π0.5 training and adaptation, and real-robot deployment.
Explore DSSL Robotics ↗
LunarBot operating in a university corridor
From platform bring-up to real-world operation.
LunarBot manipulator, gripper, and camera integration
Manipulator, sensing, and system integration.
Design rendering of LunarBot-Mini, a dual-arm mobile robot with a sensor mast
Design rendering · Platform in development
In development2026–Present

LunarBot-Mini

A dual-arm mobile robot built around a Ranger Mini base and two PiPER arms.

I lead platform design, hardware integration, ROS development, and supplier coordination.

03 / Experience & education

Research with engineering depth.

Experience

Current

DSSL · Tsinghua University

Robotics Research Lead

Lead robotics projects, mentor undergraduate researchers, and maintain original ROS packages and adapted robot-learning systems. Coordinate engineering partners and own robot and GPU infrastructure.

Education

2025–Present

Tsinghua University

Ph.D. Student

Aerospace Science and Technology

2021–2025

Tsinghua University

B.S. & B.Eng.

Theoretical and Applied Mechanics / Aerospace Engineering

2023–2024

Delft University of Technology

Exchange Student

Robotics

Selected honors

2026Tsinghua Graduate Comprehensive Scholarship, Second Class

2024Future Technology Taihu Scholarship

2022–23Xingjian College Scholarships

04 / Publications

Selected papers.

  1. 2026

    Cross-View Action Consistency for Camera-Robust Vision-Language-Action Policies

    B. Huang et al.

    arXiv:2608.06965 · Preprint

  2. 2026

    Selective Cross-View Consistency for World Action Models

    B. Huang et al.

    arXiv:2608.21402 · Preprint

  3. 2026

    Toward Self-Evolving Lunar Robotic Autonomy Through Contract-Governed Skill Registration

    B. Huang et al.

    Astronautics 1(3), 15

  4. 2026

    Whole-Body Control of a Holonomic Wheeled Mobile Manipulator for Coverage-Continuous Surface Interaction on Lunar Infrastructure

    B. Huang et al.

    77th International Astronautical Congress · IAC-26,A3,IP,51,x109136

  5. 2026

    Gesture-Driven Action Generation for Lunar Robotic Systems Based on Multimodal Large Language Models

    C. Wei, B. Huang, et al.

    77th International Astronautical Congress · IAC-26,B3,6-A5.3,2,x109721

  6. 2026

    MOAL: Mutual Odometry and Localization for Lunar Robot Cluster

    C. Wei, B. Huang, et al.

    77th International Astronautical Congress · IAC-26,B2,IPB,14,x109225

Contact

Let’s talk robotics.

bingqih0311@gmail.com
GitHub ↗DSSL Robotics ↗