Research On Dynamic Multi-Objective Evolutionary Response Mechanism And Algorithm

In real-world applications,most optimization problems are dynamisms in nature.For example,in express delivery,researchers seek the optimal scheme with the optimization objectives of maximum customer satisfaction,high distribution efficiency,and short distribution time by comprehensively considering the distance between cities,express volume,express size,express quantity,distribution vehicles,and other factors.Meanwhile,the factors or objectives change with time.Thus,the obtained optimal scheme needs dynamic adjustment when a change occurs.Optimal problems whose objectives and/or parameters change are named dynamic multi-objective optimization problems.Although researchers have made considerable progress in dealing with dynamic multi-objective optimization problems,there is lots of room for improvement in solving complex dynamic multi-objective optimization problems.The response mechanism is challenging to accelerate the population to adapt to the new environment when the historical information is insufficient or the feedback is unreasonable.In addition,it is difficult to ensure that the population diversity is well restored in the dynamic optimization.In order to effectively deal with the open problems,this paper proposes the research of dynamic multi-objective evolutionary optimization response mechanisms and algorithms.The major contributions and innovations of this paper include the followings.1)To make the population adapt to the new environment quickly,a multi-direction search strategy is presented to find the Pareto-optimal set as quickly and accurately as possible before the next environmental change occurs.The proposed method mainly includes an improved local search and a global search.The first part uses individuals from the current population to produce solutions along each decision variable’s direction within a certain range and updates the population using the generated solutions.As a result,the first strategy enhances the convergence of the population.In part two,individuals are generated in a specific random method along with every dimension’s orientation in the decision variable space to achieve good diversity and guarantee the avoidance of locally optimal solutions.Experimental results show that this algorithm is very competitive in dealing with dynamic multi-objective optimization problems.2)Aiming at the different effects of dynamic environments on decision variables,an evolutionary algorithm based on the intensity of environmental change(IEC)is proposed,which can effectively track the moving Pareto-optimal set(POS)in dynamic optimization.The IEC divides each individual into two parts according to the evolutionary information feedback from the POS in the current and former evolutionary environment when an environmental change occurs.Two parts are the micro-changing decision and macro-changing decision,which are implemented in different situations of decisions to build an efficient information exchange among dynamic environments.In addition,if a new evolutionary environment is similar to its historical evolutionary environment,the history information will be used for reference to guide the search toward promising decision regions.Experimental results show that the proposed IEC is promising.3)An evolutionary algorithm based on layered prediction(LP)and subspace-based diversity maintenance(SDM)is proposed for handling dynamic multi-objective optimization(DMO)environments.The LP strategy takes into account different levels of progress by different individuals in evolution and historical information to predict the population in the event of environmental changes for a prompt change response.The SDM strategy identifies gaps in population distribution and employs a gap-filling technique to increase population diversity.SDM further guides rational population reproduction with a subspace-based probability model to balance population diversity and convergence in every generation of evolution regardless of environmental changes.The proposed algorithm has been studied by comparing five state-of-the-art algorithms,demonstrating its effectiveness.4)In real-world scenarios,the decision-maker(DM)may be only interested in a portion of the corresponding POF(i.e.,the region of interest)for different instances,rather than the whole POF.Consequently,the DMOEA based decomposition and preference(DACP)is proposed,which incorporates the preference of DM into the dynamic search process and tracks a subset of Pareto-optimal set(POS)approximation with respect to the region of interest(ROI).Due to the presence of dynamics,the ROI,which is defined in which DM gives both the preference point and the neighborhood size,may change with time-varying DMOPs.Consequently,our algorithm moves the well-distributed reference points,which are located in the neighborhood range,to around the preference point to lead the evolution of the whole population.When a change occurs,a novel strategy is performed for responding to the current change.Particularly,the population will be reinitialized according to a promising direction obtained by letting a few solutions evolve independently for a short time.Comprehensive experiments show that this approach is very competitive.The dynamic multi-objective evolutionary response mechanism and algorithm realize real-time dynamic optimization through the research of the above methods,and provide theoretical basis and method reference for solving DMOPs in dynamic multi-objective optimization.