🦅 OpenBird (open-bird-robot-mujoco-ppo) AI Models & Motion Datasets

Official Physical AI simulation platform, pre-trained policy weights, and 15-stage trajectory datasets for the OpenBird-artnfull Bipedal VTOL Robot Platform.
🎮 Playable Interactive Simulator: Download from GitHub or Hugging Face and directly control the robot bird via keyboard (WASD)!

---

🌟 Key Features & Kinematics

• 🦿 Digitigrade Bipedal Locomotion: Avian joint kinematics with dynamic reverse-knee walking and counter-balancing.
• 🪽 180° Deployable Duct Wings: Stowed vertically behind the back during walking; deployed horizontally (180°) for high-thrust VTOL flight.
• 🦅 Look-Down Eye-Gaze & Beak Pitch Control: Independent 60° downward ground scan without disturbing body equilibrium.
• 📷 Multi-Modal Vision Perception: Stereo Eye Cameras + Cockpit FPV view.

---

🎮 Interactive Keyboard Control Guide

You can download and directly control the robot bird locally in real-time 3D simulation!

Key	Action	Description
A / D	Yaw Turn	Turn left / right (Walking turn / Flight yaw)
↑ / ↓ (W/S)	Forward / Backward	Walk forward/backward (Ground) or Fly forward/backward (Air)
← / →	Strafe	Sideways translation in air
R / F	Altitude	Altitude Climb (+15cm) / Descend (-15cm)
SPACE	Stop / Hover	Stand still (Ground) / Precision Hover (Air)
1 $\rightarrow$ 2 $\rightarrow$ 3	Flight Sequence	Deploy Wings $\rightarrow$ VTOL Takeoff & Hover $\rightarrow$ Soft Landing & Fold
C	Camera Switch	3rd-Person Orbit $\rightarrow$ Head FPV $\rightarrow$ Left Eye $\rightarrow$ Right Eye
V / B	Eye-Gaze Pitch	Look down 60° (V) / Look forward (B)
G / Y / K / P	Gestures	Peck (G), Crouch (Y), Knockdown (K), Auto-Right (P)
Q / ESC	Exit	Close simulator

---

🚀 How to Download & Run Locally

Option A: From GitHub (Recommended)

# 1. Clone repository
git clone https://github.com/artnfull-bot/OpenBird-artnfull.git
cd OpenBird-artnfull

# 2. Install dependencies (MuJoCo, NumPy, OpenCV)
pip install -r requirements.txt

# 3. Launch Interactive Flight (Manual Control)
run_manual.bat   # or python run_robot_bird.py

# 4. Launch Autonomous 15-Stage Cinematic AI
run_ai_auto.bat  # or python run_ai_robot_bird.py

Option B: Download via Hugging Face Hub (Python)

from huggingface_hub import snapshot_download

# Download complete model assets, datasets, and simulator scripts
repo_path = snapshot_download(repo_id="artnfull/open-bird-robot-mujoco-ppo", repo_type="model")
print(f"Downloaded OpenBird assets to: {repo_path}")

---

📂 Repository Contents

OpenBird-artnfull/
├── policies/               # Trained neural network weights (.pt, .onnx)
│   └── baseline_flight/    # Baseline VTOL takeoff, hover, and landing models
├── datasets/               # MuJoCo / LeRobot compatible motion trajectories
│   └── demonstration/      # Teleoperation & autonomous behavioral data
├── outputs/eval/           # Official evaluation metrics & trajectory logs
│   ├── eval_info.json      # Rollout evaluation statistics
│   └── videos/             # High-definition evaluation rollouts
├── robot_bird/             # MuJoCo 3D kinematics, FSM controller & XML models
├── run_robot_bird.py       # Interactive Keyboard Flight simulator
├── run_ai_robot_bird.py    # 15-Stage Autonomous Cinematic Showcase
├── run_manual.bat          # Windows 1-Click Interactive launcher
├── run_ai_auto.bat         # Windows 1-Click Autonomous launcher
├── requirements.txt        # Dependency list (mujoco, numpy, opencv-python)
└── README.md               # Model & dataset card