The Research And Application Of Multimodal Multi-objective Evolutionary Algorithms

There are some multi-objective optimization problems in practical application which have more than one Pareto optimal set corresponding to the same Pareto frontier,called multimodal multi-objective optimization problems.The conventional multi-objective optimization algorithm usually tends to improve the distribution of Pareto frontier in objective space,which often leads to poor diversity of Pareto optimal set,so those algorithms cannot solve multimodal multi-objective optimization problems very well.Therefore,the performance of the traditional multi-objective optimization algorithm needs to be improved to enhance the distribution of the Pareto optimal set in the decision space.Flower pollination algorithm,bat algorithm and harmony search algorithm can balance the global exploration ability and local exploitation ability,and have good convergence speed and search efficiency,so it is widely applied to multi-objective optimization problems.To sum up,three multimodal multi-objective optimization algorithms are proposed to solve multimodal multi-objective optimization problems.Multimodal multi-objective flower pollination algorithm combines special crowding distance mechanism to ensure the diversity of the population.The performance enhanced niching multi-objective bat algorithm uses the concept of dynamic ring topology to enhance the local exploitation ability of the algorithm.The niching multi-objective harmony search algorithm can improve the convergence speed of the algorithm by using the dynamic harmony memory.Experimental results show that the proposed algorithms are an effective and feasible optimizer for multimodal multi-objective optimization algorithms,and perform significantly better than the state-of-the-art six multimodal multi-objective optimizers.The main contents of this thesis are as follows:First of all,the research background and significance of multimodal andmulti-objective optimization are given,and the current state of research of flower pollination algorithm,bat algorithm and harmony search algorithm are described,and the research status of multi-modal and multi-objective optimization is elaborated in detail.Secondly,an adaptive multi-strategy flower pollination algorithm is proposed to solve the single-objective optimization problem.An excellent characteristic of good point set can be used to initialize population to enhance the diversity of population;an adaptive switch probability is designed to balance the exploration and exploitation abilities of algorithm;a novel step size adjustment strategy is introduced for fine global search capability,and learns from global optimal solution to improve local depth search capability of algorithm.Then the improved flower pollination algorithm is applied to multimodal multi-objective optimization problems.Thirdly,a performance enhanced niching multi-objective bat algorithm is proposed.The dynamic ring topology and the stagnation detection strategy are applied.The dynamic ring topology makes use of the neighborhood information to form stable niches.This algorithm is tested on the multimodal multi-objective test functions,including the comparison of the test results between the proposed algorithm and other algorithmsThen,a niching multi-objective harmony search algorithm is proposed.It adopts the neighborhood information to build dynamic harmony memory for maintaining the population diversity.A new memory consideration rule is also applied to prevent the algorithm be trapped into local optimal solution.The parameters self-adaptive mechanism implies a better balance between exploration and exploitation.To verify the effectiveness of the algorithm,the proposed algorithm has been evaluated on the twelve test functions and compared with a number of multi-objective algorithms.Finally,a summary of this thesis and future work are given,then this thesis points out the further improving direction.