The junior research group "Data-Driven Methods in Control Engineering" (DART) at the Heinz Nixdorf Institute at Paderborn University has spent four years researching how artificial intelligence (AI) can be used profitably in the field of control engineering. This technology is used to control and regulate systems, e.g. to automate industrial processes such as the operation of machines and systems. Data, for example in the form of measurements, plays a central role here. Accordingly, it makes sense to increasingly use data-based learning algorithms in control engineering as well.
The junior research group focussed on the development of hybrid methods that combine traditional control approaches with artificial intelligence. In doing so, they discovered that these combined techniques offer many advantages when it comes to understanding and optimising technical systems that cannot be fully described by traditional models. "At all stages of the control design process, our research showed that by using machine learning, i.e. AI, incorrect models could be improved and inaccuracies corrected. As a result, the control of the system also benefited," explains Dr. Julia Timmermann, head of the DART junior research group. The project was developed as part of the funding programme for young female AI scientists. The Federal Ministry of Education and Research (BMBF) funded the project with around 1.6 million euros. The aim of this funding line is to increase the participation of women in German research on AI and to enable them to take up academic leadership positions.
Development of practical applications
In addition to the theoretical work on hybrid methods, the employees and doctoral candidates in the DART project developed two practical applications: an autonomous golf robot and a self-balancing cube. The golf robot works with hybrid methods, for example in the field of image recognition and prediction of stroke speeds. "The hitting mechanism also offers good opportunities to evaluate the new methods under real-life conditions," says Timmermann. The self-balancing cube in turn brings with it many challenges in terms of control technology. The acceleration and strong deceleration of flywheels within the cube can cause it to move to an edge and, in future, to a corner, where it can be stabilised.
Other successes that were realised during the project directly affect the participating researchers and students at Paderborn University: three successful doctoral degrees, 24 theses written and the employment of ten student assistants. This made it possible for the students to work intensively on scientific topics in the field of machine learning. They will be able to transfer their newly acquired knowledge to companies in the region after graduation. "Even though machine learning approaches are currently very popular, their use is not appropriate in every application. This critical reflection on the application of modern research trends in the department of control engineering was also the content of the DART group," says Timmermann.
This text was translated automatically.
