Virtual Stochastic Testing of Advanced Driver Assistance Systems

Abstract

With Advanced Driver Assistance Systems becoming increasingly complex, testing methods must keep up to efficiently test and validate these systems. This paper focuses on a method of testing vision-based Advanced Driver Assistance Systems on a state-of-the-art hardware-in-the-loop test bench. Virtual driving scenarios are being used for functional testing. This paper suggests a framework where the driving scenarios are constructed using a stochastical approach. This allows the testing of the parameter combinations that might otherwise be forgotten or disregarded by a human creating the scenarios. The first step of this framework, a road generator, is introduced. Generic courses of roads are created using the Markov Chain and Markov Chain Monte Carlo methods reconstructing real-life scenarios by analyzing map data.