Analysis And Improvement Of The Gravitational Search Algorithm

Gravitational search algorithm (short for GSA) is a new heuristic optimization algorithm which proposed by Esmat Rashedi from Iran University in2009, it originates from the physics gravity by simulating the phenomenon. The particles in GSA are considered as a group of moving objects in the space, particles can attract by each other through gravity and move followed the rule of kinematics. A particle that has greater fitness value has larger mass quality so that all particles can move towards the particle that has the largest quality and converge to the optimal solution. GSA has a strong global search ability and fast convergence speed. With the development of GSA research work, GSA attracted much attention of scholars and is applied to many fields. However, compared to other global algorithms, GSA also has the problems such as easier to fall into local optima and lower optima precision and so on. There are many problem to be improved.This paper firstly introduces in detail the current research situation of GSA, as well as its principle, algorithm implementation. Then analyzes the parameters involved in GSA and improves the GSA algorithm based on the analysis result. Two improvement GSA algorithms were proposed to make up its weakness. The main research work of the paper is given as follow.1. First of all this paper introduces the principle and characters of Swarm Intelligent Optimal Algorithm, and then the domestic and foreign research status and application of GSA is given. The physical phenomenon, the description and the implementation of GSA is also detailed introduced. After analyzing the parameters involved in the algorithm, the relevant work for improvement is done.2. Based on the result of analyzing parameters, a fuzzy gravitational search algorithm (short for FGSA) is proposed. The principle of the fuzzy control is applied into the control of the GSA parameters. By control the value of parameter during different phase in algorithm, FGSA effectively balances the exploration and exploitation of the particles as to prevent the algorithm falling into the local optima and improve the precision of the optimal solution. Through the analysis of experimental results, FGSA has better effect.3. The gravitational search algorithm based on the composite different evolution (short for DGSA) is proposed. By introducing a variety of different evolution mutation and cross operation into the gravitational search algorithm, the particle updating strategy is diversified and the exploration of particle is improving in order to avoid the algorithm converge to the local optima, the precision of the solution is improving either. As shown by the experimental result, DGSA has better optimal ability than other improvement algorithm and the basic GSA.4. Finally, the research of this paper is summed up and prospects of next research direction are given.