A Method Of COA Based On Mutli-Agent Evolutionary Algorithm

In our modern world, complex situations arise that require the coordinated actions of many resources to achieve desired outcomes or effects. The first step in dealing with these complex situations is to develop and select a Course of Action that will lead to a desired outcome. The goal of COA is putting the right resource in the right place and time to perform the appropriate information gathering task on the appropriate objects of interest. The paper has done at least the following works:Firstly, COA can be defined as a multi-objective optimization problem. In this paper, a method based on multi-agent evolutionary algorithm was presented to solve COA'resource management and scheduling problems. First of all, initialize the COA population, each individual of COA population can be seen as an agent, in order to realize the local perceptivity of agents, the environment is organized as a latticelike structure. Each agent is fixed on a lattice point and it can only interact with its neighbors. In this work, constraint functions are added in competition strategy to deal with the multi-objective aspect of resource-constrained project scheduling problems. This approach avoids the use of a penalty function to deal with constraints. At the same time, the added constraint functions could make the whole algorithm evolving feasible. The simulation results demonstrated that this approach could improve searching ability of this algorithm, and the precision of this method.Secondly, a set of non-dominated solutions gained by the multi-agent evolutionary algorithm, so a problems about how to choose one most suitable COA is raised. In this paper, the weighted summation approach is used for selecting the best COA from the non-dominated solution set. In the normalized, the standard 0-1 transformation of data preprocessing in multi-attribute decision is used for values normalized. Weight is very important for the results of weighted summation. The determination of weight is very important and very difficult. In this work, three grades of weighted vector are divided into through development of fuzzy rule. DARE is used for the determination of the weight of each weighted vector, this method is practical, flexible and unlimited.Thirdly, competitive strategy of agent in multi-agent evolutionary algorithm have a defect: in the potential: offspring solution is may not dominant the previous ones, the so-called cross-out phenomenon. To some extent, it loses some population diversity. In this work, a method based on elitist multi-agent evolutionary algorithm was presented to solve COA problems. In this method, multi-agent population is divided into A and B two sub-populations, each sub-population separately has different competitive strategy. In addition, elitist population is introduction for storing outstanding individuals of each generation. Non-elitist populations are connected only with elitist one, allowing to migrate in both sides. In order to avoid the elitist population is too large, this paper also introduced a elimination mechanism to keep the elitist population size. Finally, we compare this method with MAEA and MFGA, the simulation results show that this method in the non-dominated solution set of relations of domination, as well as a broad solution set in terms of relative to other two methods has certain advantages. To a certain extent, this method makes up for the above- mentioned shortcomings.