An alternative approach to the beacon navigation idea:
Imagine the robots as photophobes with light sensors pointing downwards (e.g. one in the front and one in the back). A gradient on the floor will guide them to the center and keep them from escaping, making an arena fence unnecessary.
Yesterday we gave a short speech in the Scope Seesions event. To let the audience know our respective backgrounds, Mey showed her Lumibots and I the robotic Quad Play. Then we exposed the ongoing project.
There was a nice feedback from the audience and a debate started about the most convinient way towards “artificial intelligence”: either a set of rules performed by a swarm, whose emergent behavior would the appear to us as “intelligently organised”; or to set up a neural networks and let the robots learn to reach their goals.