Data Driven Sensor Model Validation

Abstract

A common method to increase the simulation accuracy of scenario-based testing for ADAS/AD systems is the integration of sensor models to simulate the behaviour of sensors in the real world. In order to use the sensor model for validation of the ADAS/AD system, all simulation components as well as the sensor model must be validated. In this presentation we present a data-driven approach to validate the behaviour of sensor models. At the core of this methodology, we use many data sources to replicate a real driving scenario on a public highway and record the real sensor data. Using this data, we replicate the real scenario in a simulation and compare and validate the sensor model output. We also present first results of an open source sensor model.