These videos show robot teams that were evolved with multiagent HyperNEAT to learn to synchronize their movements while patrolling for the paper Multirobot Behavior Synchronization through Minimal Communication at ICIRA 2012. The robots can only see the walls, so they must synchronize through communication. The communication scheme used in these examples is called the hive brain because it directly connects the neural networks of neighboring agents to each other. Teams were trained in simulation and then transferred to Khepera III robots.
4 Robot Synchronization
These videos demonstrate the ability of teams to synchronize their patrol behavior and to resynchronize when disturbed. They were initially trained at this team size.
3 and 5 Robot Synchronization
These videos demonstrate the ability of teams to synchronize their patrol behavior on team sizes at which they were not trained, without further training.
Download .wmv
Download .wmv
Download .wmv
Download .wmv
Staggered Synchronization
This video shows a variation of the synchronization shown above in which the initial heading of every other agent is changed. The agents are still able to synchronize.