An approach to utilize FMEA for autonomous vehicles to forecast decision outcome

Abstract

Every autonomous vehicle has an analytic framework which monitors the decision making of the vehicle to keep it safe. By tweaking the FMEA (Failure Mode Effect Analysis) framework and applying this to the decision system will make significant increase in the quality of the decisions, especially in series of decision and its overall outcome. This will avoid collisions and better quality of decision. The proposed methodology uses this approach to identify the risks associated with the best alternative selected. The FMEA requires to be running at real time. It has to keep its previous experiences in hand to do quick/split time decision making. This paper considers a case study of FMEA framework applied to autonomous driving vehicles to support decision making. It shows a significant increase in the performance in the execution of FMEA over GPU. It also brings out a comparison of CUDA to TPL and sequential execution.