A Deep Learning Method for Fault Detection of Autonomous Vehicles

Abstract

Fault detection is a crucial step for the safe operation of autonomous vehicles. Failure to detect faults can result in component failure leading to the breakdown of the car or even catastrophic accidents. In this paper, we propose a general fault detection method using deep learning techniques to learn patterns of faults reflected in the dynamic model of an autonomous vehicle. We have applied the proposed method to a remotely operated scaled multi-wheeled combat vehicle and evaluated the algorithm using normal and defective signals. The results show that the proposed deep learning method can accurately identify faults that are caused by mechanical problems or changes in system parameter which are reflected in the dynamic models. This general deep learning technique can be tailored to detect many defects or faults in the manufacturing and/or operation of autonomous vehicles.