🚁 UAV Insight Toolkit (AI-Powered)

UAV Insight Toolkit is a comprehensive, open-source telemetry analysis and visualization platform designed for PX4 and ArduPilot drones. Built with Python, Pymavlink, and PyULog, it bridges the gap between raw binary logs and actionable flight insights.

Core Value: Provides engineers with an interactive dashboard to diagnose vibration issues, PID performance, and 3D flight trajectories instantly—now powered by GLM-4.5 AI for automated diagnostics.

📸 Dashboard Preview

✨ Key Features

📡 Dual-Stack Support (ArduPilot & PX4)

• Auto-Detection: Seamlessly handles ArduPilot .bin and PX4 .ulg logs. The system automatically selects the appropriate parser based on file extension.
• Unified Interface: Standardizes data from different firmwares (MavLink vs. ULog) into a unified analysis pipeline.

🤖 AI Smart Diagnostics (AI Copilot)

• Powered by GLM-4.5: Integrated with ZhipuAI's latest Large Language Model via zai-sdk for real-time inference.
• Time-Series Analysis: Automatically downsamples flight logs to extract vibration peaks, altitude shifts, and attitude errors, generating a context-aware diagnostic report.
• One-Click Checkup: Automatically identifies issues like "Frame Resonance," "PID Oscillation," and "Sensor Clipping."

👁️ Optical Flow & Indoor Navigation

• GPS-Denied Trajectory: Automatically visualizes Local Position (NED) trajectories when GPS is unavailable, perfect for indoor flight analysis.
• Flow Analytics: Dedicated charts for Optical Flow quality and integrated flow (rad), helping to diagnose "toilet bowl effect" or position drift issues.

🔧 PID Tuning Analysis (Rate Controller)

• Desired vs. Actual: Overlays Desired Rate (Red) and Actual Rate (Blue) curves for Roll, Pitch, and Yaw axes.
• Quantified Metrics: Automatically calculates Mean Absolute Error (MAE) to quantify tracking performance.

📊 Multi-Dimensional Visualization

• Interactive 3D Replay: Visualizes flight paths with relative altitude encoding using Plotly 3D engines.
• Vibration & Clipping: * ArduPilot: Visualizes VIBE message data.
  • PX4: Automatically calculates vibration levels (Rolling Std Dev) from high-frequency IMU data and detects accel clipping events.

⚡ Modern Engineering Experience

• Drag & Drop Upload: Supports direct file uploads via the web UI or local scanning of the data/ directory (EXE compatible).
• Secure Configuration: Built-in API Key validation with secure storage (session-based or local file), ensuring keys are never exposed.

🛠️ Tech Stack

• Core Parsing: * pymavlink (for ArduPilot .bin)
  • pyulog (for PX4 .ulg)
• AI Engine: ZhipuAI GLM-4.5-Flash (via zai-sdk)
• Data Processing: Pandas (Time-series alignment, merge_asof & forward-fill algorithms)
• Visualization: Streamlit (Web UI), Plotly (Interactive Charts)

🚀 Quick Start

1. Installation

# Clone the repository
git clone [https://github.com/Tartistbz/UAV-Insight-Toolkit.git](https://github.com/Tartistbz/UAV-Insight-Toolkit.git)
cd UAV-Insight-Toolkit

# Create virtual environment (Recommended)
conda create -n uav-env python=3.9
conda activate uav-env

# Install dependencies
pip install -r requirements.txt

2. AI Configuration (Optional)

To use the AI Diagnostics feature, you need a ZhipuAI API Key.

1. Get your key from the ZhipuAI Open Platform.
2. Method A (GUI): Enter the key in the sidebar and click "Verify".
3. Method B (File): Create a file named apikey.txt in the root directory and paste your key there (this file is Git-ignored).

3. Usage

Run the dashboard:

Bash

streamlit run src/app.py

The app will launch at http://localhost:8501. You can then drag and drop your .bin or .ulg log files to start analysis.

📂 Project Structure

Plaintext

UAV-Insight-Toolkit/
├── data/               # Log storage (Git-ignored)
├── src/
│   ├── analyzer/       # Core Parsing Logic
│   │   ├── parser_base.py   # Abstract Base Class
│   │   ├── ardu_parser.py   # ArduPilot Implementation
│   │   └── px4_parser.py    # PX4 Implementation (New)
│   └── app.py          # Streamlit Dashboard Entry
├── requirements.txt    # Dependency Management
└── README.md           # Documentation

📝 Roadmap

• Basic .bin Log Parsing
• Attitude & 3D Trajectory Visualization
• Vibration Level Diagnostics & Clipping Detection
• PID Analysis (Rate Roll/Pitch/Yaw)
• AI Smart Diagnostics (GLM-4.5 Integration)
• File Upload & EXE Path Compatibility
• PX4 .ulg Support (with Optical Flow & Local Pos)
• Export to CSV/MAT format

🤝 Contribution

Contributions are welcome! Please fork the repository and submit a Pull Request.

📄 License

Distributed under the GNU General Public License v3.0. See LICENSE for more information.