Scaling issues with robot search and tagging

One of the authors is involved with the Robot Simulator project at Kansas State University. While running experiments to compare scaling issues between cooperative algorithms and swarm algorithms using a simple search and tag problem, it was noticed that in some configurations, the number of targets found decreased when the number of targets was increased [Gu04]. Increasing the number of robots searching also decreased the success rate. The authors have been investigating this phenomenon to attempt to understand the effect of interference on scaling issues. We have abstracted the problem into a cellular automata (CA) simulation. Specifically, we modeled a robot as a wanderer [Ma95], capable of avoiding collision with obstacles (including other robots, walls and tagged-targets), with a non-directional heat sensor for detecting targets and a sonar sensor for determining direction when a heat source is detected. The sonar is incapable of differentiating between targets and obstacles. Intuitively, one might think that increasing robot numbers, or sensor strength would be beneficial. However, experience suggests that path and sensor interference caused by an increased number of robots and sensor range is often harmful. The following investigation used several models to elucidate the issues of robot scaling and sensor noise. Our initial cellular automaton was one dimensional with a width of 20 cells. The robots were either left moving or right moving, with two targets placed at end points. When a robot is next to a target, the target is tagged and becomes an obstacle. The robot’s sonar detect objects, either targets or other robots, within a specified distance, S. The robot’s heat sensor detects heat sources within a distance, H. Experiments were run for all combinations of 1 to 5 robots with varying sonar (S) and heat (H) ranges from 1 to 3 for 1000 steps. The results showed a clear effect of scaling from 1 to 5 robots. From one to 5