How to Simulate Adaptive AUTOSAR V-ECUs in VEOS

Version R18-10, the next specification of the AUTOSAR Adaptive Platform with a significantly extended range of functions, will be released at the end of October. For our customers, this poses the question of how an adaptive control unit can be developed and validated in combination with classic ECUs and environment models.

Even though the topic of validating adaptive ECUs is taking up speed due to the current AUTOSAR releases, dSPACE started providing a solution to this issue as early as mid-2018 (dSPACE Release 2018-A). The solution supports developers in validating Linux-based ECUs at an early stage. In the following paragraphs, I have summarized a few recurring questions on this topic:  

How can dSPACE help us develop and validate Linux-based virtual ECUs (V-ECUs)?

dSPACE VEOS enables the simulation of ECUs developed according to the Classic Platform in combination with environment models. The PC-based offline simulator can execute complex systems efficiently and reproducibly. The event-based process also makes it possible to run simulations faster than in real time or to pause them for any period of time.

Since dSPACE Release 2018-A, it has also been possible to simulate Linux-based applications in VEOS in combination with ECUs developed according to the Classic Platform and environment models, while benefiting from the advantages of event-based simulation. Especially time synchronization with the other classically developed simulation nodes is generally a difficult challenge for Linux-based applications. Here, VEOS can drastically relieve the user’s workload, even with Linux-based ECUs, because VEOS can completely take on time synchronization during the simulation.

Why should I simulate Linux-based ECUs with dSPACE software? Can I not simply run the applications in QEMU or on an evaluation board?

This depends on the use case. The crux here is whether and how you want to connect the applications to the remaining simulation components and when to synchronize them. Simulation in VEOS makes it possible to test Linux-based applications early and efficiently and synchronizes them with other simulation components. For this, various experimental and test tools are available.

This is particularly interesting for computation- or bus-intensive simulations in an adaptive environment. Since VEOS also takes over the entire time synchronization for Linux-based ECUs, the simulation can be executed faster, but also slower than real time, if required, depending on the complexity. If the simulation is paused in VEOS, the adaptive V-ECU is also stopped.

Can I simulate the AUTOSAR adaptive demonstrator examples in VEOS?

Yes, you can. I have summarized this in three steps:

• Step 1: Building adaptive demonstrator applications

In the AUTOSAR Wiki, AUTOSAR provides its members with very detailed instructions on how to build the demonstrator middleware and the corresponding examples. If you follow the instructions and choose QEMU as the target platform for the build, you will receive a Linux kernel image and a tarball with the root file system.  

• Step 2: Configuring an adaptive V-ECU in dSPACE SystemDesk

To be able to simulate an adaptive V-ECU in VEOS, you must first configure the most important simulation parameters of the adaptive V-ECU, such as the number of CPUs, memory size, and existing Ethernet controllers. In addition, you must select the binary artifacts (tarballs) to be simulated. You can conveniently do this in dSPACE SystemDesk. An additional netinit layer was inserted in each of the configured adaptive V-ECUs. This is required because the AUTOSAR demonstrator examples are designed for QEMU and assume that the Ethernet controllers are statically preconfigured at start up. Otherwise, no modifications to the demonstrator are required.   

• Step 3: Starting simulation in dSPACE VEOS

The generated simulation system can then easily be loaded to VEOS and simulated. As is usual with the dSPACE tool chain, dSPACE ControlDesk can also be used to monitor bus communication via the virtual Ethernet bus, for example.  

I have developed a Linux application on my Ubuntu system. Can I run it in VEOS?

Yes, you can. All you need is the finished executable and an image with the root file system for the dependencies of your application. The second is freely available as Ubuntu Cloud Image on the Internet. In the same manner, the binary artifacts must be configured in SystemDesk. Afterwards the simulation can be started in VEOS.

My Linux application is already running in a Docker container. Can I simply import the entire container?

Yes, I have also tried this. All I had to do was to export the Docker container. This can be easily done with the ‘docker export -o rootfs.tar <MY_CONTAINER>’ command. The tarball that is created automatically contains the executable and all required dependencies and can also be configured directly in SystemDesk.

Can I create adaptive applications using classical model-based tool chains?

Yes, because as of Release 2018-B, SystemDesk supports converting classic AUTOSAR components to adaptive applications. In this case, you can also use Basic Software Services. The TargetLink team at dSPACE is also working at full speed on the native modeling of adaptive applications.

My conclusion:

Since Release 2018-A, dSPACE has been offering an intuitive product solution to support customers in the development and early validation of Linux-based ECUs. With VEOS, Linux-based ECUs can be tested early and efficiently, and synchronized with other simulation components.

Do you have any questions? Simply contact us.