Multilevel Concept for Verification of Automated Driving Systems by Using Model in the Loop Simulations in Early Development Phases 2022-01-0102

Abstract

Due to the increasing complexity of Automated Driving Systems (ADS) in combination with the multitude of possible traffic scenarios, the question arises how the functionality of ADS should be verified in an early development phase. Since there is currently no comprehensive industrially feasible solution published, this paper addresses the verification of ADS by means of Model in the Loop (MiL) simulations in advance of real prototype testing. The vehicle dynamic behavior is represented by high fidelity multi-body simulation complete vehicle models and the considered driving functions are integrated into the simulation framework via standardized interfaces and are assessed by specific safety and comfort Key Performance Indicators. The developed solution can be subdivided into four phases. (1) First, the simulation environment is used to verify the driving functions based on standardized test scenarios. (2) Next, based on one-dimensional and multi-dimensional experimental designs, the robustness of the driving function is analyzed in the entire Operational Design Domain. (3) The third phase includes system behavior studies for specific scenarios. (4) In contrast to the scenario-based function verification considered so far, the MiL simulation is integrated into a validated closed-loop stochastic traffic flow simulation to perform an additional system verification within random traffic scenarios. For the explanation of the individual phases, exemplary simulation campaigns are used. The presented verification concept allows to verify an ADS before the first test drives are started and possible improvements can be objectively shown in advance.