Improving the performance of low-cost hardware with data-driven calibration and control
Max van Meer defended his PhD thesis at the Department of Mechanical Engineering on September 9th.
Technologies that enable global communication and drive scientific discovery often rely on high-end parts that are capable of positioning and measuring with extreme precision. Examples of these technologies are satellites that deliver high-speed internet to remote communities and microscopes that reveal structures at the scale of individual atoms. The components in these systems are expensive to manufacture, which limits their availability and slows down adoption in new fields. However, it鈥檚 challenging to achieve the same level of precision with more affordable parts. Max van Meer tackles this challenge in his PhD research by developing a set of data-driven calibration and control algorithms that make low-cost hardware perform much closer to its high-end counterpart.
The difficulty with affordable components is that they鈥檙e never perfect. Motors may turn less smoothly, sensors may give slightly inaccurate values, and each unit may behave differently. These small flaws can cause big problems in sensitive systems. Fortunately, many of these imperfections are repeatable, which means they can often be measured and corrected by smart software. Max van Meer focuses on identifying the unique quirks of each component from data and adjusting the way the system operates with calibration and control. He does this by combining advanced motion control with machine learning to automatically detect imperfections and counteract them.
Validation on industrial prototypes
Van Meer鈥檚 methods were then experimentally validated on industrially relevant prototypes in collaboration with Thermo Fisher Scientific, Sioux Technologies, and TNO. With Thermo Fisher Scientific, algorithms were developed to cancel mechanical imperfections in piezo-stepper actuators used for nanoscale positioning in electron microscopes. In addition, calibration methods were created with Sioux Technologies to significantly improve the accuracy of cost-effective magnetic position sensors. Lastly, the techniques were applied to motors and sensors for space-based optical communication systems with TNO. This technology is expected to be used on satellites that are planned to launch in the coming years.
Optical satellite communication
Especially, optical satellite communication shows how powerful Van Meer鈥檚 approach can be. For this technology, data is sent through narrow beams of light between satellites and ground stations instead of using radio waves. This allows much faster and more secure connections, but the aiming has to be extremely precise. Even the tiniest of movements can break the link. Traditional solutions rely on heavy, expensive components that are hard to launch in large numbers. By enabling lighter and more affordable parts to achieve the same precision, the developed algorithms could help make global high-speed satellite internet more widely available.
The results of this research demonstrate how precision technology can be made more affordable and accessible. Such advances can broaden the use of high-performance systems in science, communication, and industry, and enable developments in areas where such technologies are currently out of reach.
Title of PhD thesis: Supervisors: Prof. Tom Oomen and Dr. Gert Witvoet