CNN-LSTM Driving Style Classification Model Based on Driver Operation Time Series Data

Abstract

The CNN-LSTM-based driving style classification model is intended to analyse driver behaviour using operation signals and vehicle dynamics data. The model successfully recognises various driving styles by combining CNN for feature extraction and LSTM networks for temporal data processing. A dataset that includes driver operation and vehicle dynamics information improves classification performance, obtaining over 99.5% precision. This hybrid technique enhances standard driving behaviour analysis approaches by providing greater adaptability and real-time performance. The merging of accelerator and brake pedal signals improves recognition accuracy and makes the system more efficient. Compared to standalone CNNs, the CNN-LSTM model improves classification efficiency and convergence speed. The model is highly generalisable across a wide range of driving circumstances, making it ideal for intelligent vehicle applications. Its computational efficiency enables smooth integration with ADAS, which enhances safety and comfort. The system's robustness allows accurate forecasts despite changing road conditions and driver habits. The initiative helps to the evolution of vehicle control systems that are capable of learning to adapt through deep learning methods. The implementation is systematic, from data collection to model training and validation. This study emphasises the significance of personalised driver assistance in promoting safer and more efficient transportation. Future enhancements could focus on improving network design and increasing dataset diversity. The research lays the groundwork for incorporating driver behaviour analysis into autonomous driving technology, paving the path for improved road safety and intelligent mobility solutions.