tensoraerospace 0.3.13

Open source deep learning framework that focuses on aerospace objects (rockets, planes, UAVs)

🚀 TensorAeroSpace

Advanced Aerospace Control Systems & Reinforcement Learning Framework

A comprehensive Python library for aerospace simulation, control algorithms, and reinforcement learning implementations

📖 Documentation • 🚀 Quick Start • 💡 Examples • 🤝 Contributing

---

🌟 Overview

TensorAeroSpace is a cutting-edge Python framework that combines aerospace engineering with modern machine learning. It provides:

• 🎯 Control Systems: Advanced control algorithms including PID, MPC, and modern RL approaches
• ✈️ Aerospace Models: High-fidelity aircraft and spacecraft simulation models
• 🎮 OpenAI Gym Integration: Ready-to-use environments for reinforcement learning
• 🧠 RL Algorithms: State-of-the-art reinforcement learning implementations
• 🔧 Extensible Architecture: Easy to extend and customize for your specific needs

🧭 Applied Use Cases

1. Autonomous Flight Vehicle Control — stabilization, trajectory tracking, attitude control for aircraft, UAVs, and experimental vehicles.
2. Rocket & Spacecraft Systems Control — modeling and control of launch vehicles, satellites in various orbital classes, trajectory optimization.
3. Hybrid Control Systems — design and tuning of controllers combining classical and intelligent control methods.
4. Algorithm Optimization & Benchmarking — automated hyperparameter tuning, comparative analysis of control algorithms, quality metrics visualization.
5. Simulation Platform Integration — interfacing with game engines, CAD/CAE systems, model import/export between environments.
6. Reliability Analysis & Diagnostics — failure mode investigation, control system robustness assessment, training data preparation.

🚀 Quick Start

✅ Minimum Technical Requirements

Component	Minimum	Recommended
OS	Linux x86_64, Windows 10, macOS 13	Ubuntu 22.04 LTS / Windows 11
CPU	4 cores, AVX	8+ cores, AVX2/FMA
RAM	8 GB	16–32 GB for RL/Simulink
GPU	Optional	NVIDIA RTX with ≥8 GB VRAM for SAC/DSAC/PPO, CUDA 12.2 support
Python	3.10–3.13	3.11/3.12
Additional	Git, Poetry or pip, Docker (optional)	MATLAB/Simulink R2022b+ (for simulink-example), Unity 2021.3.5f1/2023.2.20f1

📦 Installation

Using Poetry (Recommended)

git clone https://github.com/tensoraerospace/tensoraerospace.git
cd tensoraerospace
poetry install

Using pip

pip install tensoraerospace

🐳 Docker

The image starts JupyterLab by default (see Dockerfile CMD).

Ubuntu / Linux (bash):

docker build -t tensoraerospace . --platform=linux/amd64
docker run --rm -it -p 8888:8888 \
  -v "$(pwd)/example:/app/example" \
  tensoraerospace

# Optional: enable NVIDIA GPU inside the container
docker run --rm -it --gpus all -p 8888:8888 \
  -v "$(pwd)/example:/app/example" \
  tensoraerospace

Windows (PowerShell):

docker build -t tensoraerospace . --platform=linux/amd64
docker run --rm -it -p 8888:8888 `
  -v "${PWD}\example:/app/example" `
  tensoraerospace

# Optional: enable NVIDIA GPU inside the container
docker run --rm -it --gpus all -p 8888:8888 `
  -v "${PWD}\example:/app/example" `
  tensoraerospace

Open the printed URL (default http://127.0.0.1:8888) and navigate to example/quickstart.ipynb to run the SAC walkthrough inside Docker.

🏃‍♂️ Quick Examples

💡 Interactive walkthrough: open the Quickstart notebook to run the SAC B747 benchmark flow end-to-end inside Jupyter/VS Code.

🚀 Pretrained SAC Agent (Boeing 747)

Run a pretrained Soft Actor-Critic agent on Boeing 747 pitch control:

Command line:

python example/reinforcement_learning/sac-b747-render.py \
    --render \
    --dt 0.1 \
    --tn 200 \
    --repo TensorAeroSpace/sac-b747 \
    --device cuda  # Optional: 'cuda', 'mps', or 'cpu' (auto-detects if not specified)

📖 See full tutorial: SAC B747 Documentation

---

🎛️ PID Controller (F-16)

import gymnasium as gym
import numpy as np

from tensoraerospace.agent.pid import PID
from tensoraerospace.utils import generate_time_period
from tensoraerospace.signals.standart import unit_step

# Simulation setup
dt = 0.01
tp = generate_time_period(tn=10, dt=dt)  # 10 seconds
N = len(tp)

# Reference signal for alpha tracking (5 deg step in radians)
reference = unit_step(
    degree=5, tp=tp, time_step=100, output_rad=True
).reshape(1, -1)

# Create F-16 longitudinal environment
env = gym.make(
    'LinearLongitudinalF16-v0',
    number_time_steps=N,
    initial_state=[[0], [0]],
    reference_signal=reference,
    use_reward=False,
)

# PID controller with tuned coefficients
pid = PID(
    env,
    kp=-14.290139135229715,
    ki=-8.240470780203491,
    kd=-1.2991634935096958,
    dt=dt
)

obs, info = env.reset()
for t in range(N - 1):
    setpoint = reference[0, t]
    alpha = float(obs[0])
    u = pid.select_action(setpoint, alpha)
    action = np.array([[float(u)]], dtype=np.float32)
    obs, reward, terminated, truncated, info = env.step(action)
    if terminated or truncated:
        break

🤖 Supported Algorithms

Algorithm	Type	Save/Load	HuggingFace Hub	Status
SAC	Soft Actor-Critic	✅	✅ from_pretrained / publish_to_hub	✅
PPO	Proximal Policy Optimization	✅	✅ from_pretrained / publish_to_hub	✅
DDPG	Deep Deterministic Policy Gradient	✅	✅ from_pretrained / publish_to_hub	✅
DSAC	Distributional Soft Actor-Critic	✅	✅ from_pretrained / publish_to_hub	✅
DQN	Deep Q-Learning	✅	✅ from_pretrained / publish_to_hub	✅
A2C	Advantage Actor-Critic	✅	✅ from_pretrained / publish_to_hub	✅
A2C-NARX	A2C with NARX Critic	✅	✅ from_pretrained / publish_to_hub	✅
A3C	Asynchronous Advantage Actor-Critic	✅	✅ from_pretrained / publish_to_hub	✅
GAIL	Imitation Learning (Adversarial)	✅	✅ from_pretrained / publish_to_hub	✅
MPC	Model Predictive Control	✅	✅ from_pretrained / publish_to_hub	✅
ADP	Adaptive Dynamic Programming	✅	✅ from_pretrained / publish_to_hub	✅
ADHDP	Action-Dependent HDP	✅	✅ from_pretrained / publish_to_hub	✅
HDP	Heuristic Dynamic Programming	✅	✅ from_pretrained / publish_to_hub	✅
PID	Proportional-Integral-Derivative	✅	✅ from_pretrained / publish_to_hub	✅
IHDP	Incremental Heuristic Dynamic Programming	✅	❌	✅
NARX	Nonlinear Autoregressive Network	✅	✅ from_pretrained / publish_to_hub	✅

✈️ Aircraft & Spacecraft Models

🛩️ Fixed-Wing Aircraft

• General Dynamics F-16 Fighting Falcon - High-fidelity fighter jet model
• Boeing 747 - Commercial airliner dynamics
• McDonnell Douglas F-4C Phantom II - Military aircraft model
• North American X-15 - Hypersonic research aircraft

🚁 UAVs & Drones

• LAPAN Surveillance Aircraft (LSU)-05 - Indonesian surveillance UAV
• Ultrastick-25e - RC aircraft model
• Generic UAV State Space - Configurable UAV dynamics

🚀 Rockets & Satellites

• ELV (Expendable Launch Vehicle) - Launch vehicle dynamics
• Generic Rocket Model - Customizable rocket simulation
• Geostationary Satellite - Orbital mechanics simulation
• Communication Satellite - ComSat dynamics and control

🎮 Simulation Environments

🎯 Unity ML-Agents Integration

TensorAeroSpace seamlessly integrates with Unity ML-Agents for immersive 3D simulations:

• 🎮 3D Visualization: Real-time 3D aircraft simulation
• 🔄 Real-time Training: Train agents in realistic environments
• 📊 Rich Sensors: Camera, LiDAR, and physics-based sensors
• 🌍 Custom Environments: Build your own aerospace scenarios

📁 Example Environment: UnityAirplaneEnvironment

🔧 MATLAB Simulink Support

• 📐 Model Import: Convert Simulink models to Python
• ⚡ High Performance: Compiled C++ integration
• 🔄 Bidirectional: MATLAB ↔ Python workflow
• 📊 Validation: Cross-platform model validation

📊 State Space Matrices

Mathematical foundation for control system design:

• 🧮 Linear Models: State-space representation
• 🎛️ Control Design: Modern control theory implementation
• 📈 Analysis Tools: Stability, controllability, observability
• 🔄 Linearization: Nonlinear model linearization

📚 Examples & Tutorials

Explore our comprehensive example collection in the ./example directory:

Category	Description	Notebooks
🚀 Quick Start	Basic usage and concepts	quickstart.ipynb
🤖 Reinforcement Learning	RL algorithm implementations	reinforcement_learning/
🎛️ Control Systems	PID, MPC controllers	pid_controllers/, mpc_controllers/
✈️ Aircraft Models	Environment examples	environments/
🔧 Optimization	Hyperparameter tuning	optimization/

🛠️ Development & Contributing

We welcome contributions! Please see our Contributing Guide for details.

🏗️ Development Setup

git clone https://github.com/tensoraerospace/tensoraerospace.git
cd tensoraerospace
poetry install --with dev
poetry run pytest  # Run tests

🧪 Testing

# Run all tests
poetry run pytest

# Run specific test category
poetry run pytest tests/envs/
poetry run pytest tests/agents/

📖 Documentation

• 📚 Full Documentation: tensoraerospace.readthedocs.io
• 🚀 API Reference: Detailed API documentation
• 📝 Tutorials: Step-by-step guides
• 💡 Examples: Practical use cases

🤝 Community & Support

• 💬 Discussions: GitHub Discussions
• 🐛 Issues: Bug Reports
• 📧 Contact: Email Support

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• OpenAI Gym team for the excellent RL framework
• Unity ML-Agents team for 3D simulation capabilities
• The aerospace engineering community for domain expertise
• All contributors who make this project possible

---

⭐ Star us on GitHub if you find TensorAeroSpace useful! ⭐

Made with ❤️ by the TensorAeroSpace team