The Improvements And Extending Of Particle Swarm Optimization Based On Dynamic Search Space

With the swarm intelligence algorithms for example the particle swarm optimization algorithm are applied,mostly,in whole fields about the industry,the issue combined the premature convergence with the un-enough precise solution must be paid more attention to deal.Through the study of the traits of individual in the swarm of particle swarm optimization(PSO)algorithm,an adaptive control system is proposed with expanding and improving.There are three specific contents including:(1)For further study and optimization of particle swarm optimization(PSO)algorithm,this paper,named Particle Swarm Optimization Algorithm based on Homing(HMPSO),presents two conjoint improvements,including a kind of nonlinear wave function as learning factor and a traction strategy of particle swarm,to advance the ability of finding optimal solution at the end of calculation's process.According to the requirements of simulation experiments,the performances are also compared with different variants of PSO algorithm,using the amended benchmark functions,reported in the literature.Finally,the probability interval of issue that the solution's precision excels a threshold,built by 0-1 distribution model,is estimated exhibited in the ending.All the results clearly indicate that HMPSO performances such as robustness,steadiness and preciseness are better than other PSO algorithms reported in the literature.(2)To further ease the problems such as the slow convergence speed and the premature convergence of the particle swarm optimization algorithm,this paper proposes a new strategy named strategy of dynamic search space(DSPSO).This strategy structures and embedding system by global optimal solutions and self-adaption threshold.Reinitializing the position of particles and speed weights improve exploration capabilities of improved PSO algorithms.Simulation test of different benchmark functions clearly show that,with the same weight and learning-factors strategy,the performance of DSPSO is better than PSO.And the same things happen during the compare between the QPSO and its embedding.Studies illuminate this strategy with robustness can prevent the premature convergence and promote the capacities of exploration in the end of algorithm.(3)The task of swarm intelligence is meeting the optimized solution as precise as possible with limited time.However,the premature convergence leads to that the more precise solution need to be obtained by a more time.The layer-by-layer evolution strategy is proposed to deal with the premature convergence as a collaborator with other existing researches basing on the pre-experiments and the simple proofs.For promoting the precision of solution and eviting the premature convergence,both of the self-adaption system based on the primal algorithm,operations such as compression,selection and re-initialization using the technology of layer-by-layer,the social information including the compressed research space and the optimal solution are used.The improvements of the precision of solution can be found in results of simulation experiments with benchmark functions.All the above strategies are verified that: the final solution can be promoted by the high-level population diversity and the compressed search space and the layer-by-layer algorithm exhibitions favorable universality in most of intelligence algorithms successively.