【作者】 胡丹；

【导师】 杨晓波； 刘晓雁；

【作者基本信息】 湖南大学 ， 计算机技术， 2010， 硕士

【摘要】 虽然目前多目标优化在工程、工业和科学领域获得广泛应用。但是由于问题本身的复杂性，多目标优化的相关技术目前仍不够完善和成熟，还存在许多值得研究的问题，如：收敛速度、局部最优、参数控制、多目标之间如何进行折衷等。如何快速、有效地实现多目标优化成为工程应用中的研究热点问题。常用的多目标优化方法自身的不足及其在实际应用中存在的诸多困难，一直阻碍着多目标优化方法的向前发展。研究结果表明蚁群算法在大部分多目标优化问题上比传统进化算法具有更好性能。本文首先介绍蚁群算法的概念、背景、模型以及未来发展趋势，然后介绍了目前常用多目标优化方法，并对现有的蚁群多目标优化的模型和方法进行详细分析。研究表明蚁群算法解决一些复杂多维问题的能力不强，容易陷入局部最优，造成算法早熟。为有效克服以上缺点，更好解决实际中的优化问题，本文将量子计算理论引入蚁群算法，提出一种基于量子衍生方法的多目标蚁群算法。该算法先采用量子遗传算法生成信息素分布，然后利用蚁群算法正反馈求精确解,力求优势互补。算法将量子比特的两个概率幅看作是蚂蚁当前的位置信息，在蚂蚁数目相同时，使搜索空间加倍。能较好的解决蚁群算法在求解问题时收敛速度慢和易于陷入局部最优的问题。多目标0-1背包问题是个复杂的NP难问题，它能够很好的检验多目标进化算法的优劣。最后将本文算法用于多维0-1背包问题的求解，并与MOA及经典算法NSGA2的试验结果进行对比分析，结果表明：本文算法不仅能更快更精确地逼近Pareto最优前端，同时能够维持Pareto最优解分布的均匀性。本文算法是将量子计算与蚁群算法相结合的一种崭新的优化方法。由于量子算法中融入了量子力学的许多基本特性，极大地提高了计算效率与搜索效率且能弥补蚁群算法的不足，具有广泛的研究前景。更多还原

【Abstract】 Multi-objective optimization is widely applied in engineering, industrial andscientific fields, such as: electronic engineering, hydraulic engineering, design andmanufacturing, computer science, robotics and control. Multi-objective optimizationneeds to optimize multiple objectives simultaneously. There are some constraintsamong multiple targets and sometimes a number of objective constraints. As thecomplexity of the problem itself, relevant technology of multi-objective optimizationis still not perfect and mature, there are still many issues to be studied, such as:convergence rate, local optimization, parameter control and how to balance multipleobjectives. How to achieve a multi-objective optimization quickly and effectively isone of hot spots in engineering application.Commonly used multi-objective optimization methods and its own shortcomingsexist in the practical application of many difficulties, has been hampered bymulti-objective optimization method for forward. Research results show that the antcolony optimization problem in most multi-objective evolutionary algorithm than thetraditional better the performance, but solutions of complex multi-dimensionalproblems of the strong, avoiding local optimum, resulting Sauna’s premature. In orderfor ant colony algorithm can effectively overcome these shortcomings, the better tosolve the practical optimization problems. Firstly, the concept of ant colonyoptimization algorithm, background, and future trends of algorithm model based onthe introduction of the multi-objective optimization methods currently used on theexisting multi-objective ant colony optimization models and methods for detailedanalysis, for multi-target ants swarm optimization for solving multi-objectiveoptimization problem shortcomings; then this will be the introduction of ant colonyoptimization theory of quantum computing, quantum derivative method is proposedbased on multi-objective ant colony algorithm. Using quantum genetic algorithm togenerate the initial pheromone distribution, post positive feedback seeking the antcolony algorithm for exact solutions, and strive to complement each other. Algorithmincreases the probability of two qubits as ant current location information, the samenumber in the ant, so that the search space doubled. Can better solve the problem ofant colony algorithm for solving the slow convergence and easy to fall into localoptimum. Finally, the algorithm is used for solving multi-dimensional0-1knapsackproblem. The algorithm and the classical algorithm NSGA2MOA and compare testresults of each method were compared, simulation results show that: the proposedalgorithm can not only faster and more accurate approximation of Pareto optimal front,while able to maintain uniform distribution of Pareto optimal solutions sex. Quantumcomputing algorithm is combined with the ant colony algorithm for a newoptimization method. As the quantum algorithm into a number of basic features ofquantum mechanics, which greatly improves the computational efficiency and searchefficiency and can make up for lack of the ant colony algorithm has broad prospectsfor the study.更多还原