In these experiments we see the effect of executing the communication-preserving (CP) behavior on a team of robots. We have two simple tasks that involve 4 robots going from their start location to a goal location(s) with 15% obstacle coverage. We run each experiment with twenty different worlds, one using robots executing the CP behavior and another using control robots which do not execute the behavior. We experimented with two different scenarios one where there is only one goal, and another where there is 4 goals which each robot must visit in sequence.

We collect four different statistics. The graphs below show the mean of twenty runs and their 95% confidence intervals.

The communication preserving behavior is composed of two parts. The first part calculates a 'communications center of mass' in which the robots query the network for the locations of any robots they currently can communicate with and calculate the center of communications by averaging the positions of the robots. The positions are weighted by the number of hops taken to reach a particular robot which in effect moves the center of communications towards the more communicationally distant robot. The center of communications is then treated as a comfort zone in which the robots, based on several parameters, are attracted to. The second part of the behavior is a modified obstacle avoidance behavior. This behavior takes the center of communications and modifies the avoidance vector such that the robots have a tendency to avoid the obstacles in a way which will preserve communications links.

Time Until Completion


One Network


B

Full Network


Loners