Data-Driven Fault Detection for Vehicle Lateral Dynamics

In this paper, we investigate a data-driven fault detection (FD) problem and its application to vehicle lateral dynamics for the improvement of vehicle lateral safety, reliability and feasibility. A more practical situation in this work is the generalized (non-affine) models and real-time implementation. In particular, a linear parameter-varying (LPV) model is established for vehicle lateral dynamics with varying longitudinal velocity, and the least square support vector machines (LS-SVM) aided datadriven approach is proposed to design the parity vector based residual generation and evaluation framework directly for fault detection. The proposed data-driven FD approach is illustrated via a benchmark simulation of vehicle model and environment from veDYNA.