Modification And Application Of Gravitational Search Algorithm

Gravitational search algorithm(GSA)is a new heuristic optimization algorithm which obeys the laws of gravity and motion.The algorithm simulates the process of particles moving in the gravitational field.The basic principle of GSA is to take the position of the particle as a solution to the problem,and the mass of each particle is determined by the fitness value.Particle can attract by each other through gravity,and a particle that has less mass will move towards the larger one due to the larger accelerated velocity.Finally,the algorithm can converge to the optimal solution.The paper firstly introduces the current research situation of GSA,as well as the principle and implementation of GSA.Then we make several groups of experiments and discuss the effect of parameters on the performance of algorithm.What's more,this paper proposes two kinds of self-adaptive strategy to improve the algorithm.Finally,the modified gravitational search algorithm is applied in the identification of T-S fuzzy model and dynamic biological systems.The main contents and contributions of this paper are as follow:(1)We have analyzed three control parameters0 G,? and Rp in GSA deeply,especially the parameter Rp which is easy to be ignored.In addition,we discuss the effect of control parameters on the performance of the algorithm,to deepen the understanding of GSA algorithm.(2)A self-adaption gravitational search algorithm is proposed in this work.In the proposed algorithm,two kinds of adaptive strategies are used to improve the exploration and exploitation ability of the algorithm.The first strategy that based on the distance of the population can strengthen the exchange of information between the particles,and it can also improve the convergence ability of GSA.The second strategy that based on the gravitational constant can improve the global search ability of the algorithm.And it can prevent the algorithm fall into a local optimum.SGSA not only can guarantee the diversity of population,but also to ensure rapid convergence of the algorithm.(3)In the part of T-S model identification,structures and parameters of T-S fuzzy model are all encoded into a particle vector.Because the new method strengthens the relationship of interdependence,it can obtain a highly accurate model than traditional two stage identification process methods.(4)In the part of biological network model identification,this paper proposes a new T-S model which is proposed for biological network.We can use SGSA to approximate an unknown system based on given input-output data.Simulation results demonstrate the effectiveness of this method.