Experimental Correlation of Virtual and Physical Testing for Regulatory ADAS Features: A Roadmap towards Fully Virtual ADAS Validation”

Validation of Advanced Driver Assistance Systems (ADAS), under global regulations is primarily anchored to vehicle tests. While possessing accuracy, validation through physical means is costly, slow, and not scalable in the variety of situations. The development of high-fidelity simulation environments and advanced sensor modelling presents the possibility of shifting regulatory testing for ADAS to virtual platforms. This paper describes an experimental analysis conducted with the aim of full virtual homologation by correlating real world scenarios, test scenarios and virtual vehicle simulated results for the most important ADAS regulatory features. A calibrated vehicle simulator with embedded sensor and actuator models was used to create standard ADAS test scenarios, which were then validated against data collected from a proving ground. Comparison benchmarks included response time for brake application, inter-vehicle spacing, lateral travel deviation, activation of the system, and accuracy of targeting endeavour precision. A number of scenarios were identified for which efficacy was achieved along with a high degree of interchangeability with respect to simulated verification, proving the hypothesis of substituting physical for virtual validation for selected features. The study proposes a correlation threshold framework and a scenario classification model to propose a road map to try to implement regulatory approval of virtual testing. The outcome provides a procedure which can be adopted to lay the groundwork for future regulatory adoption of fully virtual homologation protocols.