Automate Software Tests Faster with Artificial Intelligence

Software tests are complex and are therefore increasingly automated. An AI-based solution from IAV accelerates this process: It creates test scripts from the manual test descriptions and achieves a high degree of accuracy in the selection of automation components. As part of a modular toolkit of tools, it is designed to make the work of test experts much more efficient in the future.

Test automation is becoming increasingly important: Many companies rely on machine assistance for software quality assurance because it increases efficiency and allows them to run tests independently overnight, for example. Compared to manual processing, this saves valuable time and reduces costs, especially for repeated tests. The automation of software tests is not a trivial task, however: “As a rule, you need a great deal of expert knowledge and sound programming skills to do this,” explains Dr. Remo Lachmann, Team Manager Software Test Automation & Data Quality at IAV.

What’s more, there is always a need to make adjustments to the tests as the project progresses. For example, the automated test sequences can be disrupted by changes in the software. Particularly in black-box tests, where the tester does not know the source code of the software, such problems further increase the effort required for automation.

«We have therefore set ourselves the goal of noticeably accelerating the implementation of test automation with the help of artificial intelligence.»

Dr. Remo Lachmann — Team Manager Software Test Automation & Data Quality at IAV

The IAV solution learns from known test case descriptions and the corresponding test scripts how it can later select the appropriate components for automatic testing from a text it does not know. For example, a simple test execution can consist of starting a web application, entering the user name and password and then pressing the login button. The individual steps “Start application,” “Enter user name,” “Enter password” and “Press login button” correspond to predefined components of an automation script. The IAV tool displays the automatically predicted components in an easy-to-use editor.

High hit rate through Machine Learning

Until now, a human expert had to manually select these components from a library of dozens of components and assemble them correctly. In initial tests, the new IAV solution has managed to determine the required script components correctly and independently for around half of all test cases through supervised learning. For the other half, the AI was already close to the desired result, so only minimal manual rework by the experts was necessary.

“This is a very good result because we are dealing with a particularly difficult task for Machine Learning when it comes to the automatic generation of test automation,” explains Lachmann. “On the one hand, different testers use different terms for the same steps, and typing errors make life difficult for us. Natural Language Processing – i.e. the automatic preparation of texts – can only help us to a limited extent with such difficulties. On the other hand, one step in testing can correspond to any number of script components – so we are not dealing with a 1:1 relationship, but with a multi-label problem.”

The intelligent test case editor accelerates test script creation through Machine Learning.

Customer projects sought for further development of the tool

The new solution promises to significantly speed up the testing process in the future. “Employees with little experience in particular save a lot of time when searching for the right components and can therefore work quickly and very effectively,” says Lachmann. “In the next step, we want to automatically generate test cases directly from the customer’s written requirements, which will accelerate the test creation process even more.”

In order to further advance test automation, Lachmann and his colleagues are currently looking for customers who are interested in joint pilot projects. “Everyone has different requirements and boundary conditions,” the IAV expert explains. “The more practical examples we can take into account when developing our new tool, the more precisely we will be able to improve the quality of our solution.”

«We don't want to replace the human expert, but we do want to significantly increase their efficiency so that the expert can concentrate on more important things.»

Dr. Remo Lachmann — Team Manager Software Test Automation & Data Quality at IAV

The aim of the IAV team is to create a modular kit of solutions that includes not only AI-based test automation but also the IAV Scout tool for the automatic prioritization of software tests. “We don’t want to replace the human expert, but we do want to significantly increase their efficiency so that the expert can concentrate on more important things,” summarizes Lachmann. “Our approach has already aroused a great deal of interest. It is suitable not only for automotive applications, but for software testing in all areas.”

The article was published in automotion 03/2020, the automotive engineering magazine of IAV. Here you can order the automotion free of charge.

Stay up to date

Subscribe to the newsletter