A Road Friction-Aware Anti-Lock Braking System Based on Model-Structured Neural Networks

The anti-lock braking system (ABS) is a vital safety feature in modern vehicles, preventing wheel lock during emergency braking. However, the performance of conventional ABS is often limited by the lack of real-time road friction information. This paper introduces a novel road friction-aware ABS, leveraging model-structured neural networks (MS-NNs) to learn the vehicle longitudinal dynamics in different road conditions. Our framework uses a robust criterion to dynamically select from a set of pre-trained MS-NNs based on the available sensor data, enabling real-time road friction estimation and autonomous adaptation of the ABS parameters. Simulation experiments demonstrate that the proposed MS-NN-based ABS significantly improves safety and performance across varying road conditions: the braking distances are reduced by 3.0%-40.4% compared to a conventional ABS, tuned for a specific road condition. Furthermore, the MS-NN's architecture shows better accuracy, generalization and sample-efficiency compared to other neural networks in the literature, and is suitable for real-time deployment on automotive-grade hardware. Our implementation is open source and available in a public repository.