A Fault Diagnosis Framework for Autonomous Vehicles Based on Hybrid Data Analysis Methods Combined with Fuzzy PID Control

This paper presents a fault diagnosis framework for autonomous vehicles on the basis of several hybrid data analysis approaches and fuzzy Proportional Integral Derivative (PID) control method. The framework consists of sensor monitor cluster, novel anomaly detector and actuator fault testing cluster. The Discrete Wavelet Transform (DWT) are used for denoising and feature extracting when constructing the sensor monitor. The extreme learning machine based autoencoder (ELM_AE) are applied for novel anomaly detection. Further, system approximation using neural networks and actuator fault testing via fuzzy PID control are presented. Contributions are as follow: 1) An algorithm using DWT with slide window is proposed for fatal sensor fault detection, which considers the sequential arrival characteristic of the sensor data; 2) Combining the neural network and fuzzy PID control for actuator fault testing, which solves the problem of fault location from the perspective of control. Experiments on the real autonomous vehicle platform 'Xinda' and related simulations validate the effectiveness of the proposed approaches in this fault diagnosis framework.