Font Size: a A A

Research On Swarm Intelligence-based Metaheuristic Algorithm For Unconstrained Single Object Optimization Problem

Posted on:2022-12-05Degree:DoctorType:Dissertation
Country:ChinaCandidate:J H FanFull Text:PDF
GTID:1488306758479144Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
In recent years,with the rapid development of artificial intelligence,optimization problem,as one of the important research fields in applied mathematics,is widely used in the fields of automatic driving,medical diagnosis,smart city,and so on.The unconstrained single objective optimization problem is one of the most important research fields in the optimization problem.According to different types of objective functions,the unconstrained single objective optimization problem is subdivided into unconstrained single objective unimodal optimization problem,unconstrained single objective multimodal optimization problem,and unconstrained single objective composition optimization problem.Although researchers have improved the swarm intelligence-based metaheuristic algorithm from different aspects for the unconstrained single objective optimization problem,there are still three deficiencies: First,when solving the unconstrained single objective unimodal optimization problem,the swarm intelligence-based metaheuristic algorithm relies too much on artificially designed strategies and does not make full use of individual's information in the population,resulting in low optimization accuracy;Second,when solving the unconstrained single objective multimodal optimization problem,the behavior of individuals cannot be quickly adjusted according to the difficulty of the problem to be solved,resulting in poor generalization ability;Third,when solving the unconstrained single objective composition optimization problem,the global exploration ability and local exploitation ability cannot be well balanced in the optimization process,and it is easy to fall into local traps.Focusing on the problems to be solved and aiming at the shortcomings of the existing swarm intelligence-based metaheuristic algorithms,this thesis proposes a random walk strategy that can be applied to all swarm intelligence-based metaheuristic algorithms and two improved swarm intelligence-based metaheuristic algorithms.The main contents of this thesis are as follows:1.A gravity random walk strategy based on universal gravity(GRW)is proposed for the unconstrained single objective unimodal optimization problem.First,GRW assigns a mass to each individual according to its performance and position.Second,to balance the exploration ability and exploitation ability of the swarm intelligence-based metaheuristic algorithm in different optimization stages,an adaptive parameter based on an individual's position information and current iteration times is designed in GRW.Third,to improve the exploration ability and exploitation ability of the swarm intelligence-based metaheuristic algorithm,GRW computes a random disturbance based on the individual's mass and the adaptive parameter and applies it to the individual's position update.The experimental results show that the performance of GRW is not only significantly better than the commonly used LÚvy Flight but also can significantly improve the exploration ability and exploitation ability of the swarm intelligence-based metaheuristic algorithm.2.An adaptive marine predators algorithm based on random approximate opposition-based learning and chaotic map(RCAMPA)is proposed for the unconstrained single objective multimodal optimization problem.First,to improve the global exploration ability of RCAMPA and select the most appropriate chaotic map,opposition-based learning and three commonly used chaotic maps are introduced for the initialization of the population.Second,to prevent RCAMPA from falling into local traps,a random approximate opposition-based learning strategy is proposed to improve the local exploitation ability of RCAMPA.Third,to accelerate the convergence speed of RCAMPA and enhance its generalization ability,an adaptive phase switching strategy based on an individual's performance is proposed to automatically switch the behavior of marine predators.To verify the effectiveness of RCAMPA,RCAMPA is compared with the original MPA and four improved MPA algorithms proposed in recent years.Experimental results show that the performance of RCAMPA is significantly better than the comparison algorithms.3.An improved African vultures optimization algorithm based on the tent chaotic map and time-varying mechanism(TAVOA)is proposed for the unconstrained single objective composition optimization problem.First,to improve the global exploration ability of TAVOA,the tent chaotic map is introduced for population initialization.Second,to avoid TAVOA falling into the local optimal solution,the individual's historical optimal position is recorded and used for the individual's position updating.Third,to balance the exploration ability and exploitation ability of TAVOA in different optimization stages,this thesis designs a time-varying mechanism based on the current number of iterations.To verify the effectiveness and efficiency of TAVOA,this thesis compares TAVOA with five swarm intelligence-based metaheuristic algorithms.Experimental results show that the performance of TAVOA is significantly better than the other five comparison algorithms.To sum up,this thesis proposes a random walk strategy and two improved swarm intelligence-based metaheuristic algorithms to solve the unconstrained single objective unimodal optimization problem,the unconstrained single objective multimodal optimization problem,and the unconstrained single objective composition optimization problem.The experimental results show that GRW,RCAMPA,and TAVOA have achieved good optimization accuracy and stability.
Keywords/Search Tags:Metaheuristic, swarm intelligence, global exploration, local exploitation, random walk, chaotic map, opposition-based learning, time-varying mechanism
PDF Full Text Request
Related items