MBSE and Safety Lifecycle of AI-enabled systems in transportation

Abstract

Recently, many autonomous driving companies  announced reducing or even closing their business. Even though billions of dollars were invested into autonomous driving in the last decade, there is still no commercially viable autonomous vehicle capable of using public roads. One of the biggest hurdles that prevent AI-enabled systems from achieving their target functionality is the lack of provable safety. This lack is rooted in the methods that are used in developing autonomous systems. Those methods are agile, in the most cases there are no clear process descriptions. As a result, no complete system description is created. This in turn makes safety analysis and creation of a structured safety argument impossible. In this article we analyze models and methods for ensuring safety of AI-enabled systems in transportation. We define the task for ensuring safety applicable to any level of driving automation according to the SAE J3016 standard. One of the goals of this publication is to characterize the changes in safety lifecycle depending on the autonomy level.