市場原理を用いた分散合意形成における全体最適化

抄録

In optimization of a multi-agent system, when the constraints among agents are given in a simple form, onecan use a distributed method by carrying out the Lagrangian relaxation method. This paper considers the distributedoptimization method that uses the Walrasian auction. However, convergence of such a distributed method is not generallyguaranteed. Moreover, even if it converges, global optimality of the solution obtained is not guaranteed. This paperintroduces several methods to modify objective function and to enhance convergency; this paper shows conditions for thesolution obtained to become the optimal solution. The effect of the technique is confirmed by computer experiment.