Computational Aspects of Overlapping Coordination Methodology for Linear Hierarchical Systems

Large hierarchical systems may have more than one potential decomposition, based on the interests of the many affected groups which interact to form the system. The technique of over-lapping coordination¿which optimizes such a system by decomposing it in two ways¿has been proposed in the literature. This paper presents computational aspects of this technique as related to linear hierarchical structures with selected problems used as examples. The technique is shown to work well on the example problems. Complete convergence to the overall optimum solution, without decomposition, is obtained in the first two problems (which have a minimum level of subsystem couplings), and a close-to-complete level of convergence is achieved in the third problem. The conditions that guarantee convergence to the overall optimum solution impose limits on the class ofoverlapped problems to which the technique may be applied. The method, however, is useful in both assessing a system from different viewpoints and generating information which may lead to more acute decisionmaking and management.