Research On Bionic Swarm Optimization Algorithm And Its Application

The bionic swarm optimization algorithms(BSOA)mimic the behaviors of living beiings.They are simple,but can be used to solve complex problems.Particle swarm optimization algorithm(PSO),firefly algorithm(FA)and cuckoo search algorithm(CS)are three kinds of BSOA.These three algorithms have obvious advantages but they also have some defects in their global search ability,convergence speed and convergence precision.In order to solve these problems,this paper introduces the chaos model,simulated annealing mechanism,bacterial quorum sensing mechanism,master-slave structure,hierarchical evolution and opposition-based learning strategy.In the last part,some improved algorithms are used to solve the robot inverse kinematics problem.In this paper,several aspects are studied which are as follows.(1)PSO improvementsThis paper has proposed three improved PSO which are chaotic simulated annealing PSO(CSAPSO),adaptive dynamic learning factor PSO(ADCPSO)and adaptive learning factor chaotic master-slave PSO(ACCMSPSO).CSAPSO,in the evaluation stage,introduces the simulated annealing mechanism and chaos model to select and generate the individual optimal value and the global optimal value.In ADCPSO,the learning factors change along with the particle's own optimum fitness and the global optimum fitness automatically.Each particle has different evolution strategy.ACCMSPSO is developed based on ADCPSO.There are two particle swarm,master swarm and slave swarm.Once the master swarm has evolved some generations,a slave swarm will be produced which initial particles are generated from the optimal particle of the master swarm in a chaos way.After some generations,some best particles in the slave swarm are selected to replace the same number of particles not good enough in the master swarm.(2)FA improvementsFA has the defects of slow convergence and poor global search ability.LNAFA improves the step factor ?.The ? decreases nonlinearly in LNAFA.The performance of LNAFA is better than that of FA,but its convergence and global search ability are still not good enough.In order to improve FA,we propose three improved FA.The first one is adaptive minimum attractiveness min? LNAFA(ABLNAFA).In ABLNAFA,min? can change adaptively.The global searching ability,convergence speed and convergence precision are improved in different degrees.The second one is adative beta & alpha FA(ABAFA).In ABAFA,both attractiveness and step can change adaptively.The improvements promote the overall performance.The third one is simulated annealing master-slave FA(SAMSFA).There are four improvement measures on the basis of LNAFA.The first measure is to bring an adaptive min?.The second measure is to bring a global optimum into the position update formula.The third measure is to bring in simulated annealing mechanism,which is used to select a global optimum for calculation.The fourth measure is to introduce the concept of stratification.(3)CS improvementsCS is is inspired by the way that the cuckoos feed their chicks in a parasitic way.The Lévy flight rule is brought in the position update formula.We propose two improved CS,opposition-based learning CS(OBLCS)and local search capability enhanced OBLCS(LOBLCS).OBLCS will generate an opposition-based swarm on the basis of original swarm in the selection stage.LOBLCS introduces enhanced local search to OBLCS.At the end of each generation,LOBLCS will search for potential better solution around the current optimum solution in the evolution direction.(4)Bionic hybrid optimization algorithm researchAdaptability of single BSOA is limited.In order to improve the adaptability,two or more bionic algorithms can be mixed.Combined with the characteristics of PSO,FA and CS,we propose CS and PSO paralleled algorithm(CSPSOPA),CS and PSO mixed algorithm(CSPSOMA),and FA and PSO mixed algorithm(FAPSOMA).These three algorithms greatly improve the global search ability,convergence speed and adaptability.(5)Robot inverse kinematics solutionACCMSPSO,SAMSFA and CSPSOMA are introduced to solve the solution of robot inverse kinematics in this paper.ACCMSPSO improves the similar method.SAMSFA and CSPSOMA are introduced to this field for the first time.Compared with the analytical solution,the BSOA method overcomes the robot structure limitation.Finally,the last part summarizes the research work,and points out the further research area.