Vector has introduced Release 7 of DYNA4. The new release facilitates the adaption of DevOps workflows in modern vehicle control software development. With its ability to generate lightweight simulation packages for Linux environments, massive simulations of virtual test drives can be integrated easily into CT pipelines.
DevOps workflows and continuous software testing are essential for developing modern vehicle control software. Adopting these methods greatly satisfies the demands on quality and functional safety of automotive software. The simulation tool supports customers throughout the complete development cycle with virtual test drives – from model-based control design to hardware-in-the-loop ECU testing. The new release presents features that enable the adaption of virtual test drives in constant testing pipelines. The DYNA4 vehicle and scenario simulation models are based on freely customisable models in Simulink. These models can generate lightweight simulation packages for Windows or Linux. For instance, they may be exported to FMUs or distributed and operated as Docker containers. The generation and execution of these DYNA4 Run Packages can be automated for CT pipelines. This could be triggered by a pull request from a control function software developer or a test engineer who extended the ADAS test scenario catalogue. Consequently, scaled execution for massive simulations of millions of virtual test drive kilometres can be recognised easily, for example, using a Kubernetes cluster.
The perception of the environment is essential for realising ADAS/AD functions. Thus, it is a crucial model part of modern simulation tools for virtual test driving. Depending on the system boundaries of the ADAS/AD function under test, a suitable level of abstraction must be selected for the sensor model, varying from idealised object lists to raw sensor output. The solution closes a gap at the intermediate detection level of camera sensors: the segmented image as an output of a smart camera sensor can now be completely customised. Various distortion filters may be parameterised, and the segmentation can be either founded on single object instances or object categories. The latter can be modified to match the relevant categories and even the respective colours, which the real smart camera sensor outputs. Additional improvements involve providing ground truth data as a key asset of simulation. With R7, all visualised objects can be extracted in compliance with the ASAM OSI standard, emphasising its high integrability.
The solution presents three product editions which are tailored to serve different use cases and operation environments and, as such, ideally support modern DevOps workflows.
The 'Server Edition' offers features that foster the fully automated and scaled execution of virtual test drives in continuous testing pipelines. The 'Desktop Edition' is intended for simulation experts who focus on simulation content authoring, perform individual simulation studies, or prepare automated operations of test benches and massive simulations. The 'Test Bench Edition' is ideal for the manual and automated operation of dedicated SIL or HIL test benches.