At the RITMO centre we have developed some robots/agents for a musical robotic swarm, these include the Self-playing guitars, Dr. Squiggles, and the CAVI AI agent. They have so far mostly been studied in a physical setting, but the goal of this project is to study principles and algorithms of multi-agent systems and swarms using a simulated system. With a simulator we can easily test out new algorithms and environmental conditions, scale up the numbers of units in the swarm, and perform large-scale experiments.
A Master's project in this domain could explore one of the following topics:
- How can the robots synchronize in a decentralized fashion, given that their information is limited and noisy, and how does this work as the size of the swarm is scaled up? Previously, we have been inspired by the synchronization of fireflies in nature.
- How can the robots collaborate in a musical setting, e.g. how can they behave like a jazz band and take turns in playing solos? Previously, we have explored market-based mechanisms like auctions.
The project would typically propose some improvements or changes to our previous research on the subject and analyze the swarm dynamics using simulations. A basic simulation system using Unity has been developed in an earlier project, which can serve as a starting point. Given interest and time, the approach could also be tested out on a physical platform.