Method Development of Automotive Ultrasound Simulations

Abstract

Within the automotive industry, autonomous driving is a promising avenue with potentially major advantages in terms of safety, fuel economy, and convenience. In order to secure the engineering of autonomous vehicles, active safety systems will need to be combined to provide a 360 degree view around the vehicle at all times. Risks arise when the different sensors are contaminated or yield conflicting information, especially when determining a proper response in a critical situation. One area of special interest in active safety systems is geared towards parking applications, such as parking assistance or self-parking systems. These systems are primarily based on ultrasonic sensors that are used in a sonar-like fashion to detect potential obstacles in the proximity of the vehicle. In order to gain a better understanding of the strengths and limitations of ultrasonic sensors in applications like these, computer modeling can be used. In this report a proof-of-concept framework for ultrasonic simulations has been designed and implemented. High performance computational aspects have been considered and domain decomposition techniques have been investigated for large-scale simulations on computer clusters. The framework has a high level interface for building virtual parking simulation scenarios, where the following aspects can be controlled and/or specified: the three-dimensional simulation domain, various environment and control parameters, placement of cars and other potential obstacles, the exportation of data.