Telemetry Analysis AI

Inspiration and Learning

Inspired by the challenges motorsport teams face in analyzing vast telemetry data quickly and accurately, this project was built to streamline race data analysis with AI-driven insights. It leverages machine learning models for predictive analytics and anomaly detection, enabling race engineers and enthusiasts to gain actionable insights in real-time. Throughout development, significant learning involved data parsing, time format conversion, AI model implementation, and building a responsive cross-platform desktop app.

How It Was Built

The tool was developed using Python 3.8+ with libraries like pandas and scikit-learn for data processing and AI, PyQt6 for the user interface, and matplotlib for visualization. It processes official Toyota telemetry CSV files, converts time strings to seconds, and cleans racing data for analysis. AI components include an LSTM-based lap time predictor and an Isolation Forest anomaly detector. The app provides interactive data tables and charts for user-friendly exploration.

Challenges Faced

Key challenges included handling varied telemetry data formats and irregularities, designing an intuitive yet powerful interface for complex data, and mocking advanced AI models within environment constraints. Ensuring cross-platform compatibility and maintaining code quality with thorough testing and compliance to standards were also crucial hurdles overcome.

Mathematical Modeling

The lap time predictor uses a Long Short-Term Memory (LSTM) neural network—a recurrent model good for sequence data modeled as:
[ h_t = \text{LSTM}(x_t, h_{t-1}) ] where (x_t) includes lap features and (h_t) predicts the next lap time.

Technologies Used

• Python 3.8+
• PyQt6 (GUI)
• Pandas, NumPy (data manipulation)
• Scikit-learn (Isolation Forest)
• Matplotlib (visualization)
• PyTorch (mocked LSTM model)

Built With

• matplot
• pandas
• pyqt6
• python
• scikit-learn