测绘学报 ›› 2014, Vol. 43 ›› Issue (11): 1197-1203.doi: 10.13485/j.cnki.11-2089.2014.0184

• 学术论文 • 上一篇    下一篇

交通网络旅行商路径优化的遗传禁忌搜索算法

余丽,陆锋   

  1. 中国科学院地理科学与资源研究所
  • 收稿日期:2013-12-17 修回日期:2014-06-16 出版日期:2014-11-20 发布日期:2014-12-02
  • 通讯作者: 余丽 E-mail:yul@lreis.ac.cn
  • 基金资助:

    国家863计划项目;国家863计划项目;国家自然科学基金项目

A Hybrid Algorithm for Traveling Salesman Problem in Road Network

YU Li1, 2   

  1. 1. The Institute of Geographic Sciences and Natural Resources Research
    2.
  • Received:2013-12-17 Revised:2014-06-16 Online:2014-11-20 Published:2014-12-02
  • Contact: YU Li E-mail:yul@lreis.ac.cn

摘要:

旅行商路径优化问题是经典的网络分析问题之一。由于旅行商问题具有NP Hard特性,主要通过智能优化方法或启发式算法来获得近似最优解。然而,单一智能优化方法存在运算量过大、参数选择苛刻,对初值依赖性强等缺陷,很难快速实现全局优化。结合多种优化机制和邻域搜索结构设计混合启发式算法可在一定程度上解决这一问题。本文结合遗传算法的全局寻优能力和禁忌搜索的记忆功能,设计实现了一种基于分散集中策略的禁忌遗传算法,即采用遗传变异算子作为分散策略构造邻域,开辟新的搜索空间,有效提升获得全局最优解的概率;将禁忌搜索作为集中策略进行局部寻优,避免迂回探测,充分体现禁忌搜索较强的“爬山”能力,并通过实际交通网络和不同规模的节点集合,从求解精度、稳定性和效率三个方面对算法进行了评价。结果表明,本文提出的交通网络旅行商路径优化的禁忌遗传算法平均求解精度比禁忌搜索算法提高了9%,略优于ArcGIS;当与ArcGIS求解的TSP路径长度差异在1%以内时,禁忌搜索算法已经难以获得对应精度的TSP路径,而禁忌遗传算法效率比遗传算法提高了50%。且禁忌遗传算法具有很好的并行化潜力。

关键词: 旅行商问题, 交通网络, 禁忌搜索, 遗传算法, 启发式算法

Abstract:

Traveling salesman problem is one of the classic network analysis problems. Due to it is NP Hard, people mainly take optimization method or heuristic algorithm to obtain the approximate optimal solution. However, single heuristic algorithms have some drawbacks, such as large computational complexity, rigorous parameter selection, strong dependence for the initial value and so on, which are difficult to quickly achieve whole optimization. Hybrid heuristic algorithm can solve this problem to some extent, which combines with a variety of optimization mechanisms and structures of neighborhood search. Considering the global optimization ability of genetic algorithm and the memory function of tabu search, this paper designed and implemented a tabu genetic algorithm based on decentralized and centralized strategy. Genetic mutation operator as a decentralized strategy was used to construct neighborhood, which exploited the new search space and enhanced the probability to obtain the global optimal solution. Tabu search as a centralized strategy was used to local search, which avoided circuitous detection and embodied the powerful "mountain climbing" ability. Moreover, this paper evaluated the algorithm from three aspects of accuracy, stability and efficiency using different scale of traffic network data. The results show that the tabu-genetic algorithm has higher accuracy which is increased by 9% than the tabu search algorithm, when the accuracy error is within 1%. And it can reduce the time consumption over 50% compared with genetic algorithm. Moreover, the tabu-genetic algorithm has the potential of parallelism.

Key words: traveling salesman problem, transportation network, tabu search, genetic algorithm, heuristic algorithm

中图分类号: