Research And Application Of Multi-strategy Ensemble Particle Swarm Optimization Algorithm

Particle Swarm Optimization(PSO)is an evolutionary algorithm proposed by Kennedy and Eberhart in 1995,which is inspired by simulating the survival behavior of natural organisms,and has been widely used in practical complex optimization problems such as flow shop scheduling,trajectory planning,carpooling services,and wastewater treatment since proposed.Compared with other evolutionary algorithms proposed in recent years,the theoretical analysis and strategy improvement of PSO are more perfect,the algorithm structure is simple and clear,there is no complex parameter setting,the implementation is relatively simple,and the convergence speed of the algorithm is fast.Therefore,PSO is competitive in the field of evolutionary computation and is currently one of the most popular population-based stochastic optimization paradigms.According to the no free lunch theorem,no algorithm can outperform other algorithms on all problems,and PSO also has problems such as lack of diversity,precocious convergence,and stagnation at local optimum.In view of the inherent defects of PSO,this paper conducts in-depth research on the algorithm,optimizes and improves it in many aspects,and applies the improved algorithm to practical problems.The specific work indicated as below:(1)An adaptive multistrategy ensemble particle swarm optimization(AMSEPSO)based on signal-to-noise ratio distance metric is proposed,which aims to solve the problem that PSO is prone to premature convergence and single learning mode.In AMSEPSO,an evolutionary state estimation strategy selection framework based on signal-to-noise ratio distance metric is proposed.The framework can select appropriate learning strategies more flexibly and efficiently.In addition,a nonlinear acceleration coefficient based on Singer mapping is proposed,which can better balance diversity and convergence by scaling the step size,and effectively reduce the probability of particles falling into the local optimum.Finally,the global best perturbation mechanism is used to help the population escape the local optimal.A large number of experimental results based on the CEC2017 benchmarks,show that the proposed AMSEPSO is significantly superior to other advanced PSO variants and meta-heuristic evolutionary algorithms.(2)A top-level dual exploitation particle swarm optimization(TLDEPSO)is proposed,which aims to better use the evolutionary information between particles and enhance the convergence performance of the algorithm.In TLDEPSO,particles are layered for good or bad fitness,and each iteration is also executed in two phases.A particle modification method based on gene editing technology and a top-level neighborhood exploration mechanism are proposed,the former is used to adjust the search direction of the population,and the latter is used to accelerate the convergence of the algorithm.In addition,an adaptive dynamic adjustment mechanism for the parameters of acceleration coefficient,inertia coefficient and the number of top-level particles is proposed to balance better the global exploration and local exploitation capabilities of the algorithm.In the latest CEC2022 benchmarks,the effectiveness of the proposed TLDPSO in solving different fitness landscape problems is verified.(3)Research on path planning problems of unmanned aerial vehicle(UAV).Based on the comprehensive consideration of environmental constraints and the characteristics of UAV itself,a UAV path planning model affected by various factors is constructed.At the same time,cubic B-spline curves are introduced to smooth the flight path to ensure the flyability of the obtained optimal path.Experimental results in multiple complex flight scenarios show that the proposed AMSEPSO and TLDEPSO can quickly and effectively obtain the optimal flight path of UAV.