Software testing with simulations

Autonomous Reply helps companies develop more secure systems while saving time with new methods in software testing.

Software as part of the autonomous ecosystem

In level 3 of autonomous driving, the backend, that is the cloud, becomes an integrated part of the ecosystem. This means that the complexity in the interaction of software modules and algorithms for functions increases greatly, which makes the development of safe systems more challenging.

Saving time with new test methods

The process of testing the security of systems is still one of the most important stages in development. With the new complexity, however, classic software testing methods that take all variables into account are reaching their time and economic limits. In addition, it is becoming increasingly apparent that machine learning (ML) algorithms are displacing the previously hand-written software codes, especially for complex and non-linear problems. Autonomous Reply uses virtual testing methods so that companies can already carry out scenario tests during the training of machine learning models and thus save time.

Automated simulations

Autonomous Reply uses virtual testing tools for example for the black box testing of ML models. Scenarios that best represent the use case are derived from the software or system requirements. After this definition of the scenario, a large number of simulation runs can be created through automated variations of the virtual environment and further parameterisation. These simulations are then run in parallel side by side in a cloud infrastructure to automate a large number of simulation iterations in a relatively short time.

Meticulus alignment of system boundaries

In each time step of the simulation, the system under test is compared with and evaluated against the system limits defined in the requirement. In this process, abnormalities and bugs are reported back to the tester and can be taken into account for the renewed training of the software. As a result, errors are already corrected in the ML training process. In software development for complex systems with increased use of machine learning algorithms, Autonomous Reply thus minimises the time required for testing the components. The use of virtual simulation processes also makes the components safer altogether, because more bugs are found that can be corrected before the release.

Autonomous Reply

Within the Reply Group, Autonomous Reply is the specialised company for the software and system integration of autonomous things. The experts advise companies in the industrial, automotive and new mobility sectors from the sensor to the infrastructure. The portfolio includes holistic solutions across the entire value chain – from strategy definition and advice on application possibilities to design and implementation. The offer includes edge computing, embedded software, cloud services and integration into different eco-systems. State-of-the-art technologies and methods from the fields of deep learning, machine learning and computer vision are used.