Cellular Genetic Algorithm Base On Predator And Prey System

In nature, the mainly relationship between species is predators-prey, and the foraging theory of this relationship between species, as the one of ecology, plays a pivotal and fundamental role. Some researchers have brought forward that combined predation theory with artificial intelligence, make predation theory not only be confined to theory study phase, but also can successfully applied to practical problem, meanwhile, accelerating the progress of artificial intelligence. Informed Predatory Search Algorithm only to simulates animal predatory behavior, lack of features, therefore, it has great significance for predatory and the algorithm of predatory study. CGA (Cellular Genetic Algorithm) could increase the diversity of the population by its own Space overlapping structure conquers the problem of CGA in local optimum easily, but the conquers individual will die whether its fitness value sick or well, may lead to the loss of the excellent solution, this is not consistent with the Natural evolution process, consequently, it is urgent to how to simulates Natural evolution process more real.CGA and predator-prey system are two valuable fields in practical applications, they have been widely used in many fields, but the above combination of the two is a challenging and fire-new subject. This thesis process from the combination of the two, making a deep research to the predator-prey system by using deceptive problem, numerical optimization and other one. Papers main contents include the following respects:(1) The predator-prey problem is an ideal method of studying multi-agent coordination and cooperation of distributed system, although it couldn’t describe the complex problems of the real world, it could make lots of idea concretize and therefore make simplify complex issues. A novel cellular genetic algorithm with predator and prey mechanism (PPCGA) is proposed in this paper, for the purpose of improving the ability of escaping from premature trap. To mimic the predator-prey model from natural ecology, the evolution rule of CGA is replaced by predator and prey mechanism. If the individual is survival is decided not only by the fitness of predator and prey, but also by the density of predator and prey in neighborhood. The population size of predator and prey individuals is maintained in the reasonable range by a certain population size control strategy. The predator and prey mechanism balances the tradeoff between exploration and exploitation. In the experiment of optimization of several typical complicated functions, the proposed algorithm shows better performance of avoiding premature trap and can obtain higher convergence rate of global optimum.(2) In evolutionary algorithm, selection pressure has been defined as a selected probability ratio of the best to the worst, it could increase the chances of survival. If selection pressure becomes too big, the algorithm may not converges globally; In turn, may lead the algorithm into local optimum. Thus, keeping appropriate selection pressure will be a key problem that improving property of the algorithm has to be concerned. Based on the CGA with predator and prey mechanism, the selection pressure of algorithm is studied. Reach the goal of selection pressure adjustment by made a few adjustments to new algorithm. For different parameter, algorithm will have different growth curves, observing growth curves changes by comparing growth curves.(3) Using growth curves variation rule to achieve selection pressure adjustment. The end of this article, from the study of selection pressure—we propose a new algorithm. Through the adjustment of the adaptive algorithm parameters to adjust the selection pressure, which can search for global exploration and local searching balance, make the algorithm get optimum effect. In the experiments, three functions are used to validate the performance. The results show that whether the dimensions are high or two, it still can find high quality solutions at a low computational cost. The simulation experiment has prove that the algorithm can maintain diversity efficiently, solving problems quickly with good stability, preventing premature convergence, improving the global searching performance, this method is effective and feasible.