Distribution of Error and Performance Rates to Sensors in the Scenario-Driven Development of Autonomous Vehicles to Determine Acceptable Error Rates and Estimate the Remaining Risk

Efficient allocation of performance and error rate budgets in the development of autonomous and intelligent vehicles is paramount for ensuring both Functional Safety (ISO 26262) and Safety of the Intended Functionality (SOTIF) (ISO 21448). This paper delves into the approach of distributing performance and error rate budgets to functional scenarios, considering the stringent safety requirements outlined in these standards. First scenarios are analyzed and categorized using the functional Safety (FuSa) risk classes. In doing so, the allocation of error rate budgets to identify and mitigate potential hazards can be done effectively. As autonomous vehicles make use of a large set of sensors, it is necessary to further break down the error and performance rate requirements to specific sensors. Depending on the scenario the vehicle is encountering, the position and type of the sensor for example plays a role for the contribution to the decision making of the following behavior of the vehicle. In this paper we present a method how to break down the overall performance rates to individual sensors. We present, moreover, how a performance model for an individual sensor is created and used along with the systematic derivation of individual performance rates. The systematic approach contributes to optimizing Autonomous Driving Systems (ADS) in vehicles for correct and anticipatory perception of the environment in the relevant scenarios. We present how the functional risk and overall risk of the active ADS in a defined scenario selection can be estimated. Therefore, it facilitates that the vehicle equipped with a complex ADS will be developed, validated and released compliant with safety standards in its intended operational design domain (ODD).