The challenge for OEMs:
software and hardware must be tested faster and better.

Software experts have long known that the test coverage of new electronic features is still a weak point today. The interplay of hardware and software, versions, country variants and equipment details with increasingly complex features in software-defined vehicles inevitably leads to a large number of edge cases in testing. Under these conditions the safety of the function is not immediately apparent. In practice, this problem is being tackled with additional testing effort: new test vehicles are being purchased, more extensive test hardware is being installed and the testing team is being expanded. In addition, the ever-increasing volumes of test data (several terabytes per day) are causing headaches for many OEM and suppliers – and not just in terms of the mobile network operator’s bill. However, there is a better and cheaper alternative.

Reducing costs with edge computing

Following the basic principle of edge computing, in which a smart end-device also has local computing power, the IT experts at Cognizant Mobility have developed a mobile data analysis device that significantly accelerates the testing of new, intelligent features. Instead of recording the data in the vehicle, sending it via the 5G network at great expense or reading it out manually after the test drive, the raw data from Cognizant Mobility’s data analysis device generally does not leave the vehicle. It is already analyzed on board. Only the desired algorithms for data analysis are sent to the analysis device via the 5G interface and the analysis results are transferred to the cloud interface.

With the combination of onboard computing power, real-time capability and modern measurement technology, the experts provide a solution to the complexity and cost problem in the development of smart, connected products.

The results speak for themselves

The results are impressive. “Measured against the average testing scenarios of a vehicle OEM or system supplier, the Cognizant Mobility data analysis device can complete over 80% of the test tasks for just 1% of the cost of the previously used test hardware,” says Cognizant Mobility test and validation expert Joachim Rudolph. The cost savings become even clearer when you look at the previous data transfer during the test stages: Instead of collecting the data in the vehicle and transmitting it cost-intensively via the mobile network operator to the testing team in the backend, search and analysis jobs are already completed fully automatically on board the vehicle. This eliminates up to 99% of the usual over-the-air volume. Instead of the previous offline analysis in the backend, the developers have direct real-time access to the analysis results while driving. The error detection and bug fixing loops that used to take several days are now reduced to just a few hours.

Tracking down even rare faults with A.I.

If faults are not only difficult to detect, but also rare, the search using classic search algorithms and their triggers reaches its limits. What in the most harmless case for the end-user is merely a failed pairing with the mobile phone can, in the worst case, also be a fault in the safety shutdown of the high-voltage system – and lead to a total loss of the vehicle.

A.I.-based anomaly detection is required to detect such faults. The data analysis device enables the use of self-learning systems with the help of its own specially developed runtime environment. This container environment allows A.I. models and their neural networks to be executed directly on the device.

On the road to the virtualization of embedded development

Cognizant Mobility’s data analysis device is also breaking new ground in the field of virtualization: From a development cost perspective, it makes sense to first simulate future features before they are partitioned onto an ECU or executed in a central computer architecture on an HPC. This allows a new product feature to be validated in live operation at a very early stage. This accelerates all development steps up to the final rollout. The basic principle here: As an active member of the vehicle-bus, the data analysis device also has a writing access on the overall system and can execute code locally.

Win-win for the development:

not only does the data analysis device from Cognizant Mobility ensure faster test results and clear added value in terms of virtualization, but it also sets cost standards that are currently urgently needed in the automotive industry.