An efficient distributed optimization and coordination protocol: Application to the emergency vehicle management

Abstract

In this paper we deal with the distributed optimization of problems where the objective/constraints are related to the whole variables of the system. In this kind of problems we need to bring into play all the distributed agents of the system simultaneously to guarantee that the solution is feasible and of a good quality. We propose an efficient agents coordination protocol which avoids centralizing control and computation to one agent. It overcomes the shortcoming of tree search methods related to the agents order and accelerates the search process. The proposed protocol is applied to the distributed emergency vehicle management problem that consists in taking, in a distributed way, the decisions about the locations of a set of emergency vehicles in order to solve the dispatching and covering issues simultaneously. The protocol is compared to the Synchronous Limited Discrepancy Search method. It is also compared to other distributed and centralized methods. The obtained results showed the efficiency of the proposed protocol in finding good solutions in short times and its capacity to lessen the effect of the agents ordering on the solution quality.