Fizzy & Education
Fizzy moves in a way that’s both fascinating and unpredictable. Because of its under-actuated system, the ball can’t move directly in every direction. However, with smart control and learning algorithms, Fizzy’s movements can be guided as desired. As an open-source hardware project, the ball is ready for exploration and serves as a hands-on platform for learning about control and the locomotion of spherical objects.
Summer school fizzy
In 2024 the first summer school: artificial intelligence for robotics, was held in Aachen at the IRT department. The students learned about AI and building controllers for the ball. At the end of the week a demonstration was given to show the different control models, from classical control to reinforcement learning algorithms.
Fizzy with AI
In a collaboration between Tilburg University and the IRT department at RWTH Aachen, the possibilities of AI and Fizzy are explored.Â
Fizzy can potentially enhance cognitive functioning by grabbing the attention and focus with its unpredictable movements or being equipped with games encouraging activities such as turn-taking and counting¹.Â
Needs for realizing this potential
- Decreasing the gab between simulation and reality
- Hardware in the ball capable of running larger models
- Effects of the off center mass to the turning behavior of the ball
Publications & Research Activities
¹ Giacomo Spigler, Simone Silenzi, Bart Horstman, and Heike Vallery. 2024. Intuitive Engineering as a Proxy Facilitates the Evolutionary Optimization of an Underactuated Robot Ball. In Workshop on Embodiment-Aware Robot Learning@ RSS 24 (June 2024).
² Bart Horstman. 2023. The Design of an Impact-Resistant Drivetrain for a Robotic Ball. Master’s thesis, Delft University of Technology, Delft, Netherlands. https://repository.tudelft.nl/record/uuid:dff15bcf-9240-43a4-81e2-42f7120840fe
³Â