Haoyang He

Visiting Researcher @ FieldAI

prof_pic.jpg

Hi! My name is 何昊洋 (hé hào yáng). I’m a researcher at FieldAI working on Humanoid Whole-Body Loco-Manipulation and Field Foundation Models. I’m applying for PhD opportunities with 2027 start, please reach out if you’re interested!

I recently completed my master’s at the Carnegie Mellon University Robotics Institute (MRSD). At the CMU AirLab, I worked on VLM-based planning and spatial understanding. I was also a research intern at FieldAI, and worked on grounded world models. I’m very lucky to have worked under Sebastian Scherer and Shayegan Omidshafiei and to have been mentored by Cherie Ho, Jay Patrikar, and Dong-Ki Kim.

Previously, I obtained my Integrated M.S./B.S. in Electrical and Computer Engineering at CMU, during which I interned at Apple and Bilibili.

My research interests include robot learning, world modeling, and foundation models for robotics. I am particularly interested in leveraging the spatial and embodied reasoning from large-scale foundation models in deployable robotic systems with reinforcement learning and grounded alignment.

Recent Work

  1. MuscleMan: Humanoid Whole-Body Control for Heavy-Object Prehensile Interaction
    CoRL 2026 Workshop on Contact-Rich Loco-Manipulation

    Olympic-style clean-and-jerk with 45lbs+ (20.4kg) on G1 with Dex3 hands rated for 500g each.

  2. Retargeting Through Reasoning for Humanoid–Object Interaction
    Project Blog

    An agentic approach to retargeting built around reasoning about how to perform the interaction and querying the environment for evidence to guide refinement.

  3. ProprioTrack3R: 3D Point Tracking with Humanoid Proprioception
    Coming Soon

    Tracking 3D object motion through humanoid proprioception, even when objects are occluded or out of view.

  4. FieldAI Humanoid Project
    Demos and Deployments

    Training humanoid whole-body loco-manipulation for real-world use cases, and building safe policy deployment in uncertain environments.

  5. GrndCtrl: Grounding World Models via Self-Supervised Reward Alignment
    IROS 2026

    Bridge generative pretraining of World Models and grounded behavior with verifiable rewards, improving counterfactual generation.

  6. AnyWill: A Human-Interactive Autonomous Wheelchair for Outdoor Urban Navigation
    MRSD Capstone Project 2025

    User-interactive autonomous wheelchair designed for end-to-end mobility on urban sidewalks while respecting individual preferences.

  7. RayFronts: Open-Set Semantic Ray Frontiers for Online Scene Understanding and Exploration
    IROS 2025

    Real-time semantic mapping system that enables fine-grained scene understanding within and beyond the depth perception range.