Research For Hybrid Cellular Genetic Algorithm And Multiayer Cellular Genetic Algorithm

A mount of problems existing in the field of real world are highly complexity, multiple targets and multiple constraints. It is very difficult to deal with these problems for the traditional search approach and the result is not satisfactory. How to find a kind of algorithm which has a strong ability of solving on problems is becoming an urgent problem. The genetic algorithm which generated first and has the biggest impact in the evolutionary algorithm is a kind of probabilistic search method. It simulates the natural evolution. There are many characteristics such as strong applicability, good robustness and parallel computing. The advantage is that it can be used to solve the highly complex optimization problems, which are implemented very difficultly by the traditional search technology. But the Standard Genetic algorithm takes no account of spatially structured environments and the interactions between local individuals in the process of biological evolution. There is no position relationship between individuals. Genetic operator is random and blind. So, it is very easy to lose the diversity of population and get into the local optimum. Cellular Automata is a discrete mathematics model. The simple interactions between a large number of cellular individuals can form overall complex behavior. CA is particularly suitable for implementing complex dynamic systems in the computer simulation. Cellular Genetic Algorithm can significantly improve the GA’s globally convergent performance. It is a kind of effective method to solve the complex problem. But it often exists some flaws:a large mount of calculation, slow speed etc. There is great development space for further improving the global convergence rate of CGA.This paper focuses on how to solve the problem of low search efficiency. The main research contents are as follows:1. A novel algorithm called hybrid particle swarm and multi-population cellular genetic algorithm (HPCGA) is proposed. First, the whole population is divided into some sub-populations, which can communicate with each other and share the evolutionary information. Division of the population appropriately reduces the selection pressure, can maintain the individual diversity more effectively. The mutation of CGA is replaced by particle swarm optimization, so as to improve the ability of local search. The above two improvements balance the tradeoff between global exploration and local exploitation. Selection pressure and individual diversity of the proposed HPCGA are studied. Optimization of six typical functions is carried out using proposed HPCGA and CGA. The experiment results show that the performance of the proposed HPCGA is obviously superior to that of CGA.2. The relevant control parameters of hybrid cellular genetic algorithm are analyzed and discussed. Optimization of four typical functions focused on different sub-populations is carried out. Selection pressure and individual diversity are also studied. The experiment results show that10sub-populations are more conducive to solution.3. This paper proposes a novel algorithm inspired by the local interactions in ecological systems, called multilayer cellular genetic algorithm (MCGA). First, the multidimensional space grid is introduced. Two new neighbor rules are defined to determine these cellular neighbors. The interaction between individuals is significant difference by changing the spatial distribution and the mutual position relationship, so as to affect the optimal performance of the algorithm. Selection pressure and diversity of the proposed MCGA are also studied. The experiment results of typical functions verity that the performance of the MCGA can obviously remain the diversity of population and improve the ability of global optimization. The more suitable neighbor structures for different types of optimization functions are given.