Automation, artificial intelligence and sensor-enabled applications are part of everyday life comprising technical systems that utilize environment-perceiving sensors. Here, rough environments often require robust sensor technologies such as radar, coupled with application-specific software evaluating the measured data. The rising complexity of underlying sensor systems (hardand software) and complex physical interactions with dynamic environments require development-related validations that use virtual methods primarily focusing on close-to-reality sensor simulation. To allow for a wide range of applications, we propose the use of reusable building blocks that generically model and mimic the physical structure from the real world. This includes the whole path between sensor and 3D environment with respect to sensor-intrinsic characteristics as well as environmental influences such as rain. At the same time, computational scalability of each building block allows physical simulations generating reliable sensor data even in real-time applications. As a result, Digital Twins of relevant components, systems and environments are no longer acting on their own but interacting within Virtual Testbeds allowing for virtual validation of radar-based applications with respect to the operational environment.
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.