Researching and testing multi-agent systems requires an environment where distributed structures can not only be simulated, but also touched, observed and jointly operated. Existing simulation environments capture the logic of distributed systems, but tend to remain abstract and hard to communicate, both in our own research and when presenting results to students, partners and the public.
The MAS project addresses this gap by building a purpose-built experimentation table that combines interaction, flexible presentation and distributed computing power in one physical setup.
Setting up and networking four 55-inch touch displays so all four screens function as one continuous surface, including recognition of physical objects placed on it.
Building a table frame that can be reconfigured between a horizontal experimentation surface and a vertical presentation surface.
Setting up and configuring the network for a high-performance computer (CPU/GPU) and a cluster of 128 Raspberry Pi 4 units, which can be configured into different topologies via variable VLANs.
This creates a test environment in which control and coordination algorithms for multi-agent systems can be tested at scale, while the results can be demonstrated in a directly tangible way.
The project's overarching goal is to make multi-agent systems not just simulated on screen, but directly experienceable and testable at a realistic scale. Specifically, the project pursues the following goals:
Demonstrate human-machine interaction with multi-agent systems directly at the table, rather than explaining it abstractly on a screen.
Use the same setup both as an experimentation surface for ongoing projects and as a presentation surface for teaching, partners and the public.
Test control and coordination algorithms for large vehicle fleets on the actually distributed cluster, rather than only replicating them in software.
