Research On Two Kinds Of Many-objective Optimization Algorithms Based On Evolutionary Computing

The multi-objective optimization problem(MOP)and many-objective optimization problem(MaOP)are widely exist in the field of practical production and engineering.Due to the high complexity,it is difficult to be solved by traditional methods.Heuristic algorithm has certain advantages in solving such problems,and it has been successfully applied in the field of MOP and MaOP.Among them,the evolutionary algorithm(EA)is one of most widely used algorithm in heuristic algorithms.The EA provides a relatively novel method in solving MaOP by referring to the mechanism of biogenetics and natural selection.Therefore,EA have attracted the attention of researchers in many fields.In this paper,many-objective evolutionary algorithm(MaOEA)is studied from the view of optimization problem.Many MaOEAs are proposed for unconstrained and constrained MaOP.In this thesis,the background,the motivation,and the related works for MaOP are first introduced.Then,the unconstrained and constrained MaOP are analyzed independently.Finally,the many-objective evolutionary algorithm based on two-round selection strategy,and the constrained many-objective algorithm based on two-population strategy are proposed.The main work of this thesis are as follow:1)For the unconstrained MaOPs,a two-round environmental selection strategy is proposed to pursue a good trade-off of population diversity and convergence for MaOEAs.In the first round,the solutions with small neighborhood density are picked out to form a candidate pool,where the neighborhood density of a solution is calculated based a novel adaptive position transformation strategy.In the second round,the best solution in terms of convergence is selected from the candidate pool and inserted into the next generation.The procedure is repeated until a new population is generated.2)For the constrained MaOPs,a constrained many-objective algorithm based on two-population is proposed in this thesis.This algorithm maintains two collaborative population simultaneously: one represents the main population PC,which is the main driving force for the evolution of the Pareto front to the feasible region;the other is the assistant population PN,the main purpose of PN is to provide more feasible information for the PC.Specifically,in order to obtain more useful information,the PN is intended to detect areas of the PC that are underutilized,including infeasible areas,so that the PN can explore throughout the objective space.In addition,a novel two-population parent selection mechanism is also proposed to enhance communication between the two populations.