Fault-tolerant environmental perception architecture for robust automated driving

Abstract

Autonomous vehicles gain more and more attention. Moving towards highly automated vehicles requires the implementation of fault-tolerant systems. In this paper we propose an architecture for a fault-tolerant environmental perception, where either one fault in the hardware or one in the software can be detected. The hardware fault detection relies on a Landmark (LM) tracking approach. The software fault detection is based on comparing the outputs of redundant programs. The faulty module is then excluded in the data fusion algorithm by a fault masking. The functionality of the proposed approach is tested in simulation via injecting one hardware and one software fault.