Standardization of Generic Architecture for Autonomous Driving: A Reality Check

Abstract

Autonomous driving requirements regarding safety, redundancy and performance lead to an increasing number of on-board heterogeneous equipments (sensors, actuators and processing units). Systems with the highest levels of autonomy must handle a large number of real-life situations, some of them being quite challenging. Interactions between sensors, perception, planning, control and actuators are fundamental to ensure operational efficiency while providing both proactive safety and timeliness execution. To address these constraints, appropriate design patterns are required to implement complex and robust decisional architectures. As a matter of fact, current system vehicle standards do not address these design neither patterns nor their underlying complexity. Based on various experiences in defense and civilian domains, we will present a critical review of ongoing standard developments and propose a generic design pattern to qualify autonomous vehicles architectures.