An Improved Particle Swarm Optimization Method And Its Application On Mission Planning Of Unmanned Aerial Vehicle

Over the last few decades,with the development and revolution of unmanned technolo-gies,unmanned aerial vehicles(UAVs)have gained increasing popularity both in the military and civil fields to perform different crucial unmanned missions such as planet exploration,surveillance and landmine detection,to name but a few.In recent years,with the blooming of computer and information sciences,UAVs have become more and more autonomous and intelligent,which would be one of the inevitable trends in the future development of unmanned technologies.As one of the key technologies to achieve the high autonomy and intelligence of UAVs,mission planning has recently drawn great attention around the world.Proverbially,task allocation and path planning problems are two paramount issues in the mission planning system.As a result,both these two issues have great impacts on the autonomy and intelligence of UAVs.Therefore,this thesis mainly devotes to studying these two fundamental issues main-ly using an improved particle swarm optimization.The main contents done in this thesis and the potential contributions of this thesis can be roughly summarized as follows.1.Aiming at enhancing the performance of the conventional Particle Swarm Optimiza-tion(PSO)via overcoming its two basic deficiencies,this thesis first proposes an improved PSO algorithm,named self-adaptive evolutionary game theory based PSO(SAEGPSO),through in-tegrating the Standard Particle Swarm Optimization 2011(SPSO 2011)with the evolutionary game theory.Attempting to prevent particles in SAEGPSO plunging into stagnation,SAEG-PSO leverages the non-stagnation mechanism developed in SPSO 2011 to update velocities and positions of particles in SAEGPSO.In order to strike good trade-offs between the global and local search capabilities of particles,through using the evolutionary stable strategies of evolutionary game theory and the iteration number of SAEGPSO,this thesis proposes a novel self-adaptive strategy to fine-tune the three main control parameters of particles in SAEGP-SO.Since the three main control parameters of particles not only affect the global and local search powers of particles,but also determine the convergence of PSO,and the performance of PSO heavily depends on its convergence,this thesis analytically investigates the conver-gence of SAEGPSO through standard results of the dynamic system theory.Also,after the analytical investigation on the convergence of SAEGPSO,this thesis provides a convergence-guaranteed parameter selection principle for the proposed SAEGPSO.Moreover,the different convergence behaviors of particles in SAEGPSO are investigated.Besides,the local optimal-ity of the equilibrium point of the proposed SAEGPSO is theoretically studied in this thesis.Finally,the developed SAEGPSO is verified through 10 benchmark test functions against 5 well-established PSO algorithms.The simulation results confirm that the proposed SAEGPSO performs superior to its contenders in terms of the optimality,search reliability,as well as the average convergence speed over the majority of the selected benchmark problems.Thus,the proposed SAEGPSO can be regarded as a vital alternative for handling different optimization problems.2.Leveraging the proposed SAEGPSO algorithm,this thesis develops a SAEGPSO-based method for solving the single-UAV path planning problem.In order to release the burden of the optimization process and add more diversifications to solutions,a new constraint handling method,which enhances the self-adaptive relaxation method and the feasibility-based rule,is presented to handle the inequality and equality constraints of the single-UAV path planning problem.Moreover,in order to analytically smooth the global path searched by SAEGPSO and obtain a curvature-continuous path which satisfies the maximum curvature constraint of UAV,this thesis proposes a new path smoothing algorithm based on two specific parametric quintic Pythagorean Hodograph curves.Finally,the superiorities of the proposed methods are evaluated through numerical simulations.3.Under the background of multiple UAVs attacking multiple tasks,in an attempt to pro-duce a feasible and acceptable path for each UAV within a tractable time,using a co-evolution strategy with the proposed SAEGPSO,this paper presents a SAEGPSO-based method to solve the multiple-UAVs path planning problem in a decentralized way.Finally,the presented method is verified through different simulation scenarios,compared with two well-established multiple-UAVs path planning methods.The simulation results reveal that the proposed method dominates its peers both in the optimality and the computation time.4.For obtaining a feasible task allocation scheme for the multiple-UAVs system,utilizing the proposed SAEGPSO,a SAEGPSO-based method for the single-objective task allocation problem is designed in this thesis.Also,the designed method is tested through three different-sized simulation cases.The simulation results confirm that the designed task allocation method outperforms its competitors as far as the solution optimality is concerned.Moreover,the com-putation time of the developed method is comparable with those of its peers.5.Since there may exist some conflicting objectives in the task allocation problem and decision makers may have different preferences on different objectives in real-world task al-location applications,there is necessity to solve the task allocation problem as a multiple-objective optimization problem,so that the optimizer can simultaneously obtain multiple non-dominated solutions to diversify selections of decision makers.Under this background,on the basis of the proposed SAEGPSO algorithm,this thesis proposes a SAEGPSO-based multiple-objective optimization method.To gain a well-distributed Pareto front,the circular sorting method,which is combined with the elite-preserving method,is proposed to update non-dominated solutions searched by particles in SAEGPSO.Moreover,a Sigma method is proposed to update the global best soutions of particles in the developed SAEGPSO-based multiple-objective optimization method to fasten the convergence speed of particles toward the real Pareto front of a multiple-objective optimization problem.After the evaluation of the pro-posed SAEGPSO-based multiple-objective optimization approach through 8 benchmark test functions,this method is then implemented to solve the multiple-objective task allocation prob-lem.Finally,the feasibility of the proposed method in the multiple-objective task allocation problem is verified through a numerical simulation.