Hi, I'm Yufei Jia (贾宇飞)

Embodied AI Systems | High-Performance Simulation & Robot Learning Infrastructure

PhD Student, Department of Electronic Engineering, Tsinghua University
Building scalable simulation, sensing, and sim-to-real systems for robot learning.

       

---

I design and build high-throughput infrastructure for embodied intelligence: RL training frameworks, photorealistic robot simulators, GPU-accelerated sensor pipelines, real-sim-real evaluation workflows, and whole-body control systems that move research from algorithm prototypes toward deployable robot stacks.

I am a PhD student at Tsinghua University, advised in embodied AI and robot learning. I received my B.E. from the Department of Automation, Tsinghua University in 2023.

Infrastructure Focus

• High-throughput robot simulation: scalable simulators for vision-informed robot learning, high-fidelity environments, and large-scale policy training.
• Heterogeneous RL training systems: decoupled CPU simulation, GPU learning, and runtime buffering for cross-platform embodied reinforcement learning.
• GPU-accelerated perception infrastructure: fast LiDAR and 3D Gaussian Splatting based sensing pipelines for simulation and evaluation.
• Sim-to-real system design: reusable bridges across simulators, robot assets, controllers, ROS2 interfaces, and deployment workflows.
• Whole-body mobile manipulation: loco-manipulation systems that combine control, perception, and reinforcement learning under real robot constraints.

Research

First / Co-first Author

• [CoRL 2026] UniLab: A Heterogeneous Architecture for Robot RL Beyond GPU-Dominant Paradigms
  Webpage / arXiv / Repo

• MuJoCoUni: Persistent Batched Runtime Primitives for MuJoCo
  Webpage / arXiv / Repo

• [RSS 2026] GS-Playground: A High-Throughput Photorealistic Simulator for Vision-Informed Robot Learning
  Webpage / arXiv / Repo

• [IROS 2026] FGGS-LiDAR: Ultra-Fast, GPU-Accelerated Simulation from General 3DGS Models to LiDAR
  arXiv / Repo

• [IROS 2026] FILIC: Dual-Loop Force-Guided Imitation Learning with Impedance Torque Control for Contact-Rich Manipulation Tasks
  arXiv / Repo

• [RAL 2026] ManiVID-3D: Generalizable View-Invariant Reinforcement Learning for Robotic Manipulation via Disentangled 3D Representations
  Webpage / arXiv

• [IROS 2025 Oral] DISCOVERSE: Efficient Robot Simulation in Complex High-Fidelity Environments
  Webpage / arXiv / Repo

• [IROS 2024 Oral] Arm-Constrained Curriculum Learning for Loco-Manipulation of a Wheel-Legged Robot
  Webpage / arXiv / Repo

• [ICRA 2023 Poster] Designing the Whole-Body Controller for a Wheel-Legged Mobile Manipulator
  Webpage

Co-author

• [IROS 2026] OmniDP: Beyond-FOV Large-Workspace Humanoid Manipulation with Omnidirectional 3D Perception arXiv

• [RSS 2026] HiWET: Hierarchical World-Frame End-Effector Tracking for Long-Horizon Humanoid Loco-Manipulation
  arXiv

• [CoRL 2025 Oral] Omni-Perception: Omnidirectional Collision Avoidance for Legged Locomotion in Dynamic Environments
  arXiv / Repo

• [RAL 2025] Imit Diff: Semantics Guided Diffusion Transformer with Dual Resolution Fusion for Imitation Learning
  arXiv

• A Real-Sim-Real Loop Framework for Generalizable Robotic Policy Transfer with Differentiable Simulation
  arXiv

Open-source Projects

• MuJoCo-LiDAR: A high-performance LiDAR simulation tool for MuJoCo.

• urdf2mjcf: One-click generation of simulation-ready MJCF files from URDF.

• ROBOCON2026-MuJoCo: A MuJoCo simulation environment for ROBOCON 2026 with motion-control strategies, multi-sensor support, and ROS2 interfaces.

Academic Service

• ICRA 2025 Sim2Real Challenge (S2R2025)
  Organizer & Technical Support. A premier ICRA 2025 event focused on high-fidelity simulation and real-world robot deployment.

• RoboMaster 2024 University Sim2Real Challenge (RMUS 2024)
  Organizer & Technical Support. Part of ICRA 2024, advancing sim-to-real robotics through autonomous ore-exchange tasks.