Automation is an inseparable part of the modern world, which is why the accuracy of robots is becoming an increasingly important factor in the practice in various fields, anything from industrial production to medicine. This fall, VILNIUS TECH alumnus Marius Sumanas, who earned his PhD, focused his doctoral research on reducing robot positioning errors using innovative machine learning methods.

This work not only reveals the solutions proposed by science but also opens up opportunities to improve older systems.

Accuracy – the foundation of modern automation

Dr. Sumanas chose the topic "Compensating robot positioning errors using deep Q-learning algorithm" due to its novelty and practical relevance. Moreover, the researcher believes his studies will contribute to significant changes in the field of automation.

In his dissertation, Dr. Sumanas examined the problem of robot accuracy. Mechanical wear, external forces, and environmental factors often lead to positioning errors, reducing the reliability of robot movements. Such inaccuracies can cause problems in manufacturing or medicine, where micron-level precision is a critical criterion.

"We, along with my team, developed a method that uses a combination of machine learning algorithms to help correct robot movements so that they are as precise and reliable as possible," says the young scientist.

The results of the dissertation promise not only practical benefits but also a significant contribution to scientific progress. The developed method is universal and can be applied in various fields that require extremely high precision.

"This research is a great example of how innovative technologies and artificial intelligence can solve real problems, improve processes, and contribute to building a sustainable future."

Errors are as unique as the robots themselves

Each robot and its operating conditions are unique, so it is essential to accurately identify and compensate for kinematic errors. As explained by the VILNIUS TECH alumnus, the dissertation aimed to create a system that would not only identify errors but also correct them effectively.

"During the experiments, we used a robot with special experimental equipment and inertial sensors to collect data on the robot's movements and their accuracy. The data collected was then processed using the deep Q-learning algorithm, which provided corrective coordinates enabling a significant improvement of the robot's accuracy," explains Dr. Sumanas.

The results confirmed that the deep Q-learning algorithm is effective in solving positioning issues. During the study, the error rate was significantly reduced, which opens up the possibility for broader application of this method in the future.

The researcher states that the application of such a method could not only improve the accuracy of new robots but also extend the lifespan of older systems. This would undoubtedly contribute to the development of automation and the more efficient use of older equipment.