A Method for Slip Angle Estimation of Intelligent Tire System Based on Hierarchical Temporal Fusion Network

Accurately estimating the slip angle under dynamic conditions is critical for enhancing vehicle stability and safety. Intelligent tire systems and data-driven deep learning methods have been adopted to slip angle estimation. However, existing approaches often suffer from inadequate temporal feature extraction and limited model–data compatibility. To address these limitations, we perform a temporal dependency analysis of tire acceleration data, revealing strong local and global correlations. Based on this, we propose a novel feature extraction algorithm and a Hierarchical Temporal Fusion Network to fully capture these multiscale temporal correlations. Additionally, an Intelligent Linkage mechanism is designed to enhance model convergence speed and generalization capabilities. Linear probing and principal component analysis further confirm that the network learns highly discriminative and physically continuous representations. Experiments demonstrate that the proposed method achieves 99.38% prediction accuracy with only 11,842 parameters and processes more than 96,000 samples/s, outperforming baseline networks in both accuracy and efficiency. On-road validation confirms its applicability, reaching over 92% agreement on both dry and wet pavements against a kinematic reference.