The aim of the project ‘Automation of an electromechanical pinball machine by reinforcement learning’ is to investigate different strategies for controlling the pinball machine in terms of game duration and point gain. This should be done by the method of reinforcement learning. The central challenge is the large tolerance in the dynamic behavior of the old electromechanical system.
For the implementation of the project a motion capture system of the company VICON is used. Reflective markers are observed simultaneously by several infrared cameras. By triangulation the position of the marker in space is determined exactly. Thus, the position of the pinball is captured in real time and further processed by appropriate algorithms.
In recent years, the application of artificial intelligence in production processes has become increasingly important. Originally focused on the process of quality management, artificial intelligence is increasingly used in production control processes and to control other complex dynamic systems. In addition to the learning speed, the robustness of the resulting algorithm is an important quality criterion here. Due to the high tolerances, an electromechanical pinball machine is therefore suitable as a demonstrator for machine learning.
The project is motivated by members of our institute and is not supported by external research funds. If you would like to be part of this exciting development process or learn more about this project, feel free to contact Mirco Alpen
Hands-on research
Almost 2000 people visited us at the “Sommer des Wissens 2019”, an event at which Hamburg’s research institutions presented themselves to the public over 5 days. Many of them got involved in the “man versus machine” duel and some even managed to win it. Next time we will be better 😉
Letzte Änderung: 15. April 2021