Skip to content
View Functionhx's full-sized avatar
🎯
Focusing
🎯
Focusing

Highlights

  • Pro

Block or report Functionhx

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Functionhx/README.md

Header

English · 简体中文

Typing SVG

Followers Stars Profile Views

Website Email


About Me

I am an undergraduate Robotics Engineering student at Beijing Institute of Technology, working at the intersection of robotics, autonomous driving, embodied AI, 3D scene representation, and AI systems engineering.

I enjoy turning research prototypes into real systems—from perception, state estimation, and navigation to inference deployment, communication architecture, testing, and fault recovery. Beyond getting a demo to run, I care about reliability, reproducibility, and graceful recovery when things go wrong.

  • Currently building — 3DGS visual localization, instance navigation, and editable road-scene generation
  • Exploring — ROS 2, Nav2, Habitat, VLA/VLN, world models, and AI agents
  • Engineering — reliable real-robot systems, AI inference deployment, autonomous-driving infrastructure, and open-source software
  • Let's talk about — embodied AI, 3D scene intelligence, robotics middleware, AI infrastructure, and research automation
  • Highlights — 28 merged upstream PRs; Kaggle Bronze Medal (solo, top 9.6%)

Featured Projects

Stars

A batch-wise extension of Point-LIO featuring within-window motion undistortion, batched EKF updates, and parallel computation for high-bandwidth LiDAR-inertial odometry.

C++ LiDAR-Inertial Odometry EKF ROS

Stars

A 3D annotation and AI-assisted pre-labeling platform for Formula Student Driverless point clouds, with quality gates, OpenPCDet export, and a TensorRT inference backend.

3D Vision TensorRT Vue FastAPI

Stars

An open simulation and verification framework for reusable-launch-vehicle terminal recovery — return flight, GNC, tower and arrestor-cable catch dynamics, and a SIL → HIL → scaled-experiment path built for explainable, reproducible sim-to-real.

Rocket Recovery GNC Sim-to-Real C++20

Stars

A three-backend (Isaac Gym / Isaac Lab / MuJoCo) reinforcement-learning classroom for robust locomotion — train a policy in one simulator, then make it prove itself under unfamiliar dynamics, sustained pushes, stairs, and a different physics engine.

Reinforcement Learning Sim2Sim Robot Control Isaac Lab


Open-Source Contributions

Contributed 28 merged pull requests, grouped below by domain — focused on robotics middleware, embodied AI & RL, AI inference infrastructure, and 3D perception, plus broader engineering work.

Project Pull Request Focus
vLLM #48153 Mistral Large 3 → AutoWeightsLoader
DeepSpeed #8154 Pipeline gradient-scaling fix
ros2_control #3454 update_rate modulo-by-zero UB guard
CycloneDDS #2425 Signal-mask restore on pthread_create
OpenRLHF #1261 Qwen3.5 ZeRO-3 frozen-weight detection
Hugging Face TRL #6439 GRPO loss normalization
OpenCV #29487 Calibration & projection-matrix fixes
Gymnasium #1618 default_camera_config render fix
View all 28 merged PRs (by domain)

🤖 Robotics & autonomous-systems middleware

🧠 Embodied AI & reinforcement learning

⚡ AI systems & inference infrastructure

👁️ 3D & computer-vision perception

🛠️ Other engineering contributions


Technologies & Tools

C++ Python Linux CMake

ROS 2 Nav2 CycloneDDS Docker

PyTorch ONNX TensorRT OpenCV

Open3D 3DGS Habitat MuJoCo


GitHub Stats

Snake animation


If you are exploring robotics, embodied AI, 3D scene representation, or reliable AI systems, let's connect

GitHub Website Email

Footer

Pinned Loading

  1. BITFSD-Annotator BITFSD-Annotator Public

    An industrial-grade 3D point cloud annotator tailored for Formula Student Driverless (FSD/FSAC). | BITFSD (北理工) 开源的无人方程式点云标注平台。专为赛道场景打造,解决远距离稀疏点云痛点。支持锥桶局部质心智能吸附、三视图精准微调,一键无缝导出 OpenPCDet 训练数据集。

    Vue

  2. Diff-MPC Diff-MPC Public

    A PyTorch-powered Differentiable MPC solver that unifiedly handles system identification and optimal trajectory planning via backpropagation.

    Python 2