Scalable Sensor Models and Simulation Methods for Seamless Transitions Within System Development: From First Digital Prototype to Final Real System

Highly automated systems are taking over an increasing number of human tasks. This includes not only the execution of certain actions but also perception and interpretation of the environment. Here, environment-perceiving sensor systems such as camera, lidar, or radar represent a key technology while posing a major challenge in terms of complex physical interactions that make validation very complicated. Typically, required system tests within dynamic environments are enabled too late, while real-world testing is neither efficient nor sufficient. Striving for a seamless transition from a first digital prototype to the final real system, we propose the use of digital twins and hybrid testbeds. These combine real and simulated components per availability or requirement to continuously enable simulation-based validation in all stages of development. To this end, we present an overall concept of scalable sensor models and simulation methods that meet the varying requirements along the development process by introducing a triple-axis scalability. Here, reusable building blocks allow for scaling across different sensor types and applications with varying level of detail, while natural interfaces and measurement-accurate ground truth annotators ensure a seamless integration into the underlying development process using methods of X-in-the-loop.