Adaptive parallelization of strategies in agent based systems

We present an approach for adaptive parallelization of strategies in agent oriented systems. Our work is based on algorithmic skeletons (M. Cole, 1989) which provide the user with parallel programming templates supporting the modeling of agent strategies and ensuring their efficient parallelization. Information provided by these skeletons supports the decision for an efficient partitioning. We describe agent based modeling and a sample example taken from manufacturing. Possible types of strategies are mentioned and the idea of algorithmic skeletons is introduced. Additionally we present the concept of PARTITIONER, a meta agent which computes a parallelization of strategies based on the knowledge provided by the skeletons. First implementation results are discussed and future research is motivated.