Functional Safety for Self-Driving Cars

Rolf Johansson and Håkan Sivencrona

Safetronic.2016 - Functional Safety in Automotive, Novmber2016.

Paper: to appear

Abstract: In recent years, the future of self-driving cars has started to seem more and more reachable. There are many advantages, and one of them is traffic safety. Today, human drivers cause a large majority of severe accidents, meaning there is a potential to increase traffic safety dramatically, given that the self-driving technology is safe in itself. This paper explains how the discipline of functional safety can be used to assure that a self-driving vehicle will not cause an unacceptable amount of severe accidents. This paper contains a methodology or approach on how the authors believe an item or function for a self-driving vehicle should be defined, i.e. what elements to include and in which traffic context and to what performance a service is offered. Given this item, we should do a hazard analysis and risk assessment (HA&RA) to cover all possible traffic scenarios inside this scope. In this paper, we elaborate on the implications on safety argumentation from the overlaying understanding that an autonomous vehicle must behave safely under all conditions. This implies a number of challenges compared to manually driven road vehicles such as:

  • Safety requirements must limit all kind of systematic design and faults, including the use of components that might not meet the system level requirements under all safety-critical conditions.
  •  Safety requirement refinement must be verified in unclear redundancy patterns like sensor fusion, even if it is hard and/or standard doesn’t tell you how to do it.
  • All safety requirements must be verified, even if it is hard and/or a standard today does not tell how to do it. .