DCRF-Net: A Dual-Stream Convolutional–Ensemble Paradigm for Context-Aware Driving Action Inference from Multi-Sensor Streams

Abstract

The growing evolution of intelligent transportation and autonomous driving systems has made accurate environment perception a critical component of Advanced Driver Assistance Systems (ADAS). Earlier approaches primarily depended on manual driving support and simple rule-based methods with minimal sensor fusion, which often failed to capture the complexity of real-world driving scenarios. These traditional systems lacked adaptability and struggled to interpret dynamic conditions effectively. Although machine learning techniques later improved sensor data analysis, challenges such as imbalanced datasets, high-dimensional features, and inconsistent prediction accuracy remained unresolved. This research focuses on addressing the problem of accurately classifying driving actions using diverse and heterogeneous sensor inputs in real-time environments. Conventional models often exhibit poor scalability, reduced generalization capability, and higher misclassification rates when dealing with large and complex datasets. To overcome these limitations, a comprehensive framework is proposed that integrates multiple machine learning algorithms, including Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Random Forest (RF), along with a novel hybrid model named DualStream-ConvRF (DCRF). The proposed DCRF architecture leverages the strengths of Convolutional Neural Networks (CNN) for deep feature extraction and Random Forest for robust classification. The system incorporates systematic data preprocessing, efficient feature engineering, and thorough model evaluation to enhance overall performance. Experimental results demonstrate that the DCRF model achieves an accuracy of 95.35%, significantly outperforming traditional baseline methods. This study highlights the potential of hybrid learning approaches in improving perception reliability, thereby contributing to safer and more intelligent autonomous driving systems