© iStockphoto.de/BeeBright

The Testing Aspects of Automated Driving

Steinbeis experts develop a framework to evaluate the robustness of electronic control unit software

The transition to fully autonomous vehicles, which would automatically assume total responsibility of the driving maneuvers automatically performed, is a major challenge for the automotive industry, especially when it comes to testing the safety of the new vehicle generations, at a time when they are still undergoing development. Intensive testing activities will be necessary to ensure that the functions of the electronic control unit work properly in any traffic situation. Steinbeis Interagierende Systeme GmbH has been providing support to automotive companies and their suppliers with systematic testing methods and tools for analyzing the robustness of their electronic control unit software. It is also responsible for running testing platforms worldwide.

According to the German Federal Statistical Office, despite the rising number of vehicles on the streets, the accident statistics show that there has been no rise in road traffic injuries or fatalities. One reason for this is the number of technical advancements in the passive and active safety systems now incorporated into vehicles. This is also partly due to driver assistance systems, which automatically intervene if they identify a critical traffic situation. The provided assistance can reduce the criticality of traffic situations – or at least can mitigate road accidents.

As vehicles move toward full automation, the vehicle manufacturers are taking over more and more of the responsibility for the driving maneuvers automatically performed by the vehicles over extended periods of time. This development will result in drivers being allowed to turn their attention away from the vehicle and its immediate surroundings, and to focus on something else instead – as the vehicle drives itself autonomously. As a result, in situations when the vehicle cannot cope with the traffic or is doing something wrong, automotive companies will no longer be able to rely on the immediate intervention of the driver. The areas in which these fully autonomous vehicles will need to operate are not like the previous scenarios encountered in the industry, where conveyance systems move around within defined areas. If a vehicle is being controlled automatically in the road traffic, it will have to handle a large number of completely different situations and varying conditions. Each drive will depend on a whole host of unknown external factors, all happening at the same time. These factors affect the quality of the information about the surroundings, which are provided by different sensors and are needed by the software of the electronic control unit as the basis for evaluating traffic situations. Despite all these uncertainties, the vehicle manufacturers will still have to ensure that vehicles are capable of reaching their destination, adhering to all kinds of traffic regulations. And even if they do get something wrong, they will have to be safe and not endanger passengers or other road users.

For automated driving, the various traffic situations and their temporal sequences in the road traffic represent a challenge where aspects of robustness must be taken into account. Steinbeis Interagierende Systeme GmbH has been researching new ways to automatically evaluate robustness. These have to take into account all of the different traffic scenarios and temporal sequences in order to evaluate the software of the electronic control unit in adherence to established standards of software testing. The aim of the project is to develop a tool-supported method for determining the functional limitations of a vehicle. If a vehicle reaches or exceeds its limitations, it will typically show a response deviating from the specification. The test throughput needed to do this is achieved by a resource-saving implementation, which allows the performant simulation of many thousands of test kilometers at the limits of the electronic control unit software.

For the current project being worked on by the experts at Steinbeis Interagierende Systeme, a framework called Gen4es.4L is being developed. This will make it possible to adapt the method to the project-specific requirements of the driver assistance function to be tested. This makes the framework ideal for the integration into the development and testing process. Gen4es.4L provides a stimulation of the test object on different testing levels as well as the determination and evaluation of the corresponding responses. “With Gen4es.4L, we can use evolutionary algorithms as well as random-based procedures combined with guided searches, so the driver assistance function we’re looking at can be pushed to its limits,” explains Steffen Wittel, who is managing the project. Another challenge the project team faced was how to test limitations in traffic situations that are similar to real life. To do this, Gen4es.4L enables traffic to be simulated by systematically accelerating and changing lanes. This automatically leads the generated simulation closer and closer to functional limitations. To continuously compare the expected outcomes with actual outcomes, this iterative approach requires an automatic determination of the expected vehicle response. To do this, Gen4es.eval offers different ways to analyze data and evaluate it hierarchically. It also makes it possible to influence the testing based on the response of the simulated vehicle from the viewpoint of an external observer.

Dr. Daniel Ulmer, member of the management board at Steinbeis Interagierende Systeme GmbH, sums up the value added by the new framework: “Plotting the multidimensional functional system limitations of a vehicle using Gen4es.4L allows an efficient evaluation of the electronic control unit functionality across different software versions. It also makes it possible to quickly visualize the system response, when the system is pushed to its limits. Based on this, decisions can be made about ongoing development of the functions provided by the electronic control unit software, and its impact can be analyzed.” Steffen Wittel, Dr. Daniel Ulmer, Dr. Oliver Buhler Steinbeis Interagierende Systeme GmbH (Esslingen) su1913@stw.de | www.steinbeis-ias.de Image


Steffen Wittel, Dr. Daniel Ulmer, Dr. Oliver Bühler
Steinbeis Interagierende Systeme GmbH (Esslingen)