However, it is often the case that the players (or agents) are not uniform out of circumstance or design. For example one might imagine two kinds of robots used to clear mines. The first is a small flying bot with ground penetrating sensors that detects and marks the mine but being a flying device has no mine removal capability. This bot can drop miniature disposable transmitters on the mine. The second is a slower ground traversing bot with armor plating and deactivation hardware. It finds mines by looking for the transmitters but would be too slow to search the whole area.

We have developed a simple problem space that represents this scenario. Our questions are basic and address issues of practicality, but also deeper questions of the mechanics evolving cooperation such as: Are the evolutionary techniques for evolving software for these bots efficient or even applicable in this situation? How can their performance be improved? What is the relation between individual performance and team performance? What is the mechanics of the relationship? What does that tell us about what is being learned in this heterogeneous environment?

This is the work of Dr. Terry Soule and myself both of the University of Idaho. This work is made possible by NIH Grant # P20 RR16448 from the COBRE Program of the National Center for Research Resources and by NSF grant #0535130.