A Data Based Method Road Surface Parameters Estimation for Anti-Lock Braking System

Abstract

Accurate road surface parameter identification is considered essential for selecting the appropriate controlling threshold in the Anti-lock Braking System (ABS) utilized in modern vehicles. This paper presents a data-based method for road surface parameter estimation. The proposed method utilizes a pattern recognition technique that works to estimate the road type during braking. A detailed analysis and related comparison is provided for several pattern recognition techniques such as Support Vector Machine (SVM), K-Nearest Neighbor (KNN), and Decision Tree (DT), which were chosen among previously studied pattern recognition techniques. A model for the ABS system is implemented with MATLAB Simulink, and the required data is extracted to be utilized to train each model individually. After training is complete, a test has been applied in order to obtain the performance of each trained model. In particular, accuracy and sensitivity are utilized to compare the effectiveness of these models, with 96% for the SVM, 95.2% for the DT model, and 94% for the KNN model. Although the SVM classifier accuracy was better than both the KNN and DT classifiers, all classifiers presented a high performance accuracy that proves the possibility of utilizing a data-based method for road surface parameter identification that increases the reliability of safety systems like the ABS.

Keywords

• Anti-lock Braking System
• Burkhardt tire model
• k-nearest neighbor
• Support Vector Machine
• Decision Tree