A fault diagnosis and maneuver control algorithm for autonomous vehicle

Abstract

This paper proposes the concept of autonomous vehicle failsafe system structure. It contains vehicle hardware and software structure design for autonomous vehicle failsafe system. Also, it handles the contents fail detection and emergency braking strategy in case there is no driver intervention in the fault situation. According to the 2017 ' AUTOMATED DRIVE SYSTEM 2.0 ' report released by the National Highway Traffic Safety Commission (NHTSA), it states that deployment of the crash avoidance system is essential to switch to a minimum hazardous condition in the event of a problem with the self-driving vehicle or the system cannot operate safely. Secondly, real-time monitoring of the normal operation of the self-driving system is required. Third, the monitoring of the autonomous driving system shall be such that the driver can regain adequate control of the vehicle or that the system can individually return to a minimum hazardous condition. Fallback strategies should also consider minimizing errors with driver awareness and decision making.

As mentioned above, establishing a failsafe system for autonomous driving is an essential element for autonomous driving. To achieve NHTSA stated standard, designed failsafe system. First, on the structural side, designed two actuators (PC and Autobox) to monitor each state. Fail detection part enabled

to find autonomous vehicle sensor errors. The failsafe control part proposes a sliding mode control based control strategy that enables an emergency braking control strategy in case the upper control fails. The control parts divide into longitudinal and lateral controls. The proposed algorithm suggest proper respond ways for autonomous vehicle safe driving only utilize vehicle chassis information, even without perception and decision part information. The sliding mode control based longitudinal control is suggested in this paper. The safe driving distance is calculated in real time and transmitted to the reference model. The reference model calculates reference acceleration, reference velocity, and the reference station. The sliding mode control based longitudinal acceleration control algorithm tracking the reference values in an emergency situation. For lateral control, it is suggested that the desired path transmitted to a lateral controller which using dead reckoning method.