a p p l i c a t i o n s

The Robot Sheepdog Project

Boid modeling techniques are just beginning to be applied in fields other than computer animation. A number of projects involving robots are ongoing, including the Robot Sheepdog Project. The goal of this project is to develop a robot that is capable of accurately interacting with animals. In this case, despite the project's name, the animals being interacted with are ducks. Apparently ducks are commonly used to train sheepdogs, as they are easier to work with. Boid-like models of animal-group behavior are used to test the robots, as simulated ducks are even easier to work with than real ducks.

Search techniques

Boid flocking and swarming models can be used to rapidly search large data sets. This article describes one method of examining data using an enhanced type of boid. The particular application under discussion is mining geographic data, or, for instance, analyzing Figure 1 to find the clusters of important information that can be seen in Figure 2. This kind of cluster analysis, although relatively easy for humans to perform, is quite difficult for computers, and no truly adequate method of analysis has yet been found.

In addition to the usual qualities of velocity, location, and orientation, the boids used to perform this cluster analysis have color properties. The color of a boid in this model influences its neighbors:

blue indicates that the boid is processing data, but hasn't yet found any appropriate information points;
green indicates that it has found an insignificant number of appropriate information points;
red indicates that the boid has found an almost significant number of appropriate information points;
yellow indicates that it has found a significant number of appropriate information points;
and black indicates that a boid has found no information (whether appropriate or not) to process, and it has died.

Thus, boids are attracted to their red and yellow neighbors, repelled from their black and blue neighbors, and ignore their green neighbors. Also, each boid uses its own color to determine the speed at which it should be travelling:

blue boids move quickly (since there is no information to examine);
red boids move slowly (so as to not miss anything important);
yellow boids don't move (they show the location of good data to other boids);
and black boids don't move (dead boids indicate areas devoid of information).

The end result is that boids travel around a set of information, examining it for points of interest. Yellow boids will rapidly group around the information sought, and black boids will gather around uninteresting areas, effectively blocking those areas off from other exploring boids.

Here is an example of this modified-boid method being used to search crime data from Baltimore for patterns.

home boid rules boid models bibliography

last updated: May 5, 2000 (chris boone, sam hillier)