Research On Multi-agent Genetic Algorithm

The agent can simulate human behavior,be familiar with the surrounding environment,autonomous operation,and can have a certain impact on the environment.Genetic algorithm is a kind of adaptive global optimization algorithm which simulates the survival,reproduction,genetic and evolutionary process of biology.According to the characteristics of the agent system,the individual in the genetic algorithm as a sensing environment,competing agent collaboration and self learning,research combining the optimization algorithm and its application question makes sense.In this paper,the agent and genetic algorithm are studied.The main tasks are as follows:First,the basic theory,research status,research significance and algorithm flow of the agent and genetic algorithm are introduced.Based on the mechanism of the interaction between the particles,a multi-agent genetic algorithm based on the particle model is proposed,and the energy of the ageent is constantly improved through the characteristics of the gravitational interaction between the particles.The numerical results show that the algorithm has high optimization efficiency and is especially suitable for the optimization of higher-dimensional functions.Then,a combinatorial optimization multi-agent genetic algorithm is proposed.In the experiment,the algorithm was tested by using four small cheat functions(subfunctions)to form a large-scale deception problem.By comparison,the computation required by this algorithm is far less than that of other literatures,which shows good performance and fully shows the great potential of combination of the agent and genetic algorithm.Finally,in the solution of Kepler's transcendental equation,sensitive to the gradient calculation and initial point asked higher difficulties,this paper proposes a multi-agent based genetic algorithm calculation method of the satellite orbit,and on the basis of calculating the trajectory under a variety of satellite star point.Numerical experiments show that the above algorithms have strong universality and global convergence.