The objective of this research is to find a method of producing cooperative agents using co-evolution (each individual evolving separately) that maximizes accuracy while minimizing computational cost. Robots that cooperate can often achieve much more than the sum of what they could do individually. Learning heterogeneous behaviors for robots to cooperate in the performance of a task is a difficult problem. In order to promote the greatest level of specialization, team members should be evolved in separate populations. The greatest complication in the evolution of separate populations is finding suitable partners for evaluation at trial time. If too few combinations are tested, the genetic algorithm loses its ability to recognize possible solutions and if too many combinations are tested the algorithm becomes too computationally expensive.
Using the concept behind punctuated anytime learning with sampling to
reduce the computational complexity of co-evolving individuals can
significantly improve our ability to evolve coordinated teams.
These concepts being developed will be applicable to all forms of
evolutionary computation which are used in co-evolving team coordination.