高维多目标进化算法研究及其在复杂车辆路径问题中的应用

Multiobjective optimization is an important topic in the science and engineering fields. Many problems have many objectives, thus many-objective optimization becomes the hotspot of research. This project firstly studies the source of difficulties of many-objective optimization problem, and characteristics of the existing many-objective optimization algorithms. Then, this project proposes the method to measure the diversity and convergence and to control the balance of the diversity and convergence in many-objective optimization algorithm. Thirdly, how to utilize neighborhood and direction information to guide the generation of offspring solution is studied. Furthermore, a cooperative evolutionary framework combing single-objective and many-objective optimization algorithms, and a cooperative evolutionary framework combing different kinds of many-objective optimization algorithms are proposed. At last, the proposed methods are applied to a real-world many-objective optimization problem---vehicle routing problem in logistic field. This project studies many-objective optimization algorithm and real-world many-objective optimization problem, thus it has great scientific significance and engineering application value.

多目标优化问题是科学与工程领域，和社会科学领域中一个非常重要的研究课题。很多问题的目标数是高维的，开发高维多目标优化算法不仅是当今智能优化领域新的研究热点和难点，更是现实应用的迫切需求。本项目研究高维多目标优化问题的难度和特性，和现有高维多目标优化算法中的有效组件和机制，然后提出高维多目标优化算法中的关键技术---收敛性和多样性的度量和控制方法。并从收敛性和多样性平衡的角度出发，提出充分利用邻域和方向信息产生新解的方法。进一步的提出单多目标算法和不同特性多目标算法的协同演化框架。最后把开发的新算法应用于解决实际的高维多目标优化问题---现代物流中的车辆路径问题。本项目对高维多目标算法和实际高维多目标问题进行深入研究，有较高的科学意义和工程应用价值。

研究高维多目标进化算法，设计开发了一套高维多目标进化算法，并用于解决实际的多目标复杂车辆路径优化与调度问题。具体而言，研究主要围绕算法框架、典型问题、算例构造、解集分析、并行实现和新一代算法等方面展开研究，取得了系列成果，在IEEE汇刊等发表系列文章。项目成果可以应用于智慧物流中的调度优化问题。