Research On Discussion Collective Decision Optimization Method And Its Applications

Heuristic optimization algorithms are a kind of intelligent search and optimization technique inspired by natural phenomena or laws.Due to their high efficient optimization performance and huge application potentials,heuristic optimization algorithms have been broadly applied to diversified fields,such as mechanical manufacturing,aerospace engineering,military affairs and science research etc.However,with the increasing complexity of optimization problems,it brings great challenges to the optimization algorithm.Exploring and proposing efficient search algorithms is always an open topic.Through simulating group decision-making phenomenon,a new heuristic optimization technique-discussion collective decision optimization algorithm(CDOA)is proposed and used to solve different optimization problems.The main work of this dissertation can be summarized as follows.(1)Inspired by group decision making phenomenon,this dissertation proposes discussion collective decision optimization algorithm,which simulates five significant decision courses of group decision making:experience-based phase,other decider-based phase,group thinking-based phase,leader-based phase and innovation-based phase.For different decision phase,different corresponding operators are designed and connected by multi-step position-selectable strategy.Experimental results carried out on comprehensive set of benchmark functions and neural network training model demonstrate that the proposed algorithm can achieve better results with respect to other state-of-art optimization algorithms.(2)Inspired by generalized group decision making phenomena,this dissertation proposes an extended version of discussion collective decision optimization algorithm(ECDOA).For the fixed operator connection order and redundant computation,in ECDOA,a random construction sequences strategy is introduced to provide difference search sequences for each population agent.This way not only changes the original operator connection mechanism,but also improves the search flexibility of the algorithm.In addition,some operators are further modified based on the original framework of the algorithm.Experimental results carried out on comprehensive set of benchmark functions and unmanned aerial vehicle path planning demonstrate that ECDOA obtains better results than other state-of-art optimization algorithms.(3)For solving dynamic multi-objective optimization problems effectively,this dissertation proposed a new prediction-based dynamic multi-objective optimization algorithm.The dynamic processing mechanism aims to provide high quality search population in new environment.It has three important strategies.Firstly,the non-dominated solutions prediction strategy aims to make full use of the information of the non-dominated solutions in environment to predict the feasible solutions in the new environment.Then,the sampling strategy aims to make full use of the relationship between variables to generate promising solutions,which can be used to guide the search of population.Finally,the shrinking stategy aims to search a more effective search space based on the distribution characteristics of variables for improving the quality of search population.Numerical experiments carried out on a variety of benchmark functions demonstrate that the proposed algorithm has competitive search performance with respect to other state-of-art algorithms.