【计算机科学】【2008.05】【含源码】多目标随机路径规划研究

在这里插入图片描述

本文为美国德克萨斯A&M大学(作者:SUMANTRA DASGUPTA)的硕士论文,共76页。

本研究将路径规划作为一个多目标随机边缘参数的优化问题。第一部分介绍了PP问题的不同变体,并讨论了该问题的现有解决方案。下一部分将介绍并解决本研究范围内的各种版本的PP模型。前三个版本的模型描述了从单个源节点到单个目标节点的单个实体。在第一个版本中,实体有一个单一的目标,并受多个约束的限制。第二个版本针对具有多个目标和多个约束的实体。第三个版本是对第二个版本的修改,在第二个版本中,沿边缘的实际行程时间概率分布是已知的。第四个和最后一个版本针对从多个源(供应节点)沿容量边缘路由到多个目的地(需求节点)的多个异构实体。这些公式都是通过使用精确算法或本研究开发的启发式算法来解决的,最后一部分将讨论每种算法/启发式算法的性能。本研究的主要贡献如下:

  1. 提供了一个框架来分析存在多目标和随机边缘参数的PP问题。
    
  2. 确定可应用基于聚类的多级编程以消除不可行边缘的候选约束。
    
  3. 为建立冗余最短路径提供精确的O(V.E)算法。
    
  4. 提供了一种O(V.E+C2)启发式方法,在多目标存在时生成Pareto最优最短路径,其中C是路径长度的上界。通过使用Pareto边界的图形读取,可以进一步将复杂性降低到O(V.E)。
    
  5. 提供了一种能够捕获边缘变量多个关键概率分布参数的代价结构,这与通常只捕获单个参数(如分布的平均值或方差)的技术不同。
    
  6. 给出了一个具有多个决策变量、随机需求和不确定边缘/路径容量的多商品运输问题的MIP公式。
    
  7. 为经典的二元设备选择问题提供了一个替代公式。
    

The present research formulates the pathplanning as an optimization problem with multiple objectives and stochasticedge parameters. The first section introduces different variants of the PPproblem and discusses existing solutions to the problem. The next sectionintroduces and solves various versions of the PP model within the scope of thisresearch. The first three versions describe a single entity traveling from asingle source to a single destination node. In the first version, the entityhas a single objective and abides by multiple constraints. The second versiondeals with an entity traveling with multiple objectives and multipleconstraints. The third version is a modification of the second version wherethe actual probability distributions of travel times along edges are known. Thefourth and final version deals with multiple heterogeneous entities routed frommultiple sources (supply nodes) to multiple destinations (demand nodes) alongcapacitated edges. Each of these formulations is solved by using either exactalgorithms or heuristics developed in this research. The performance of eachalgorithm/heuristic is discussed in the final section. The main contributionsof this research are:

  1. Provide a framework for analyzing PP inpresence of multiple objectives and stochastic edge parameters. 2. Identifycandidate constraints where clustering based multi-level programming can beapplied to eliminate infeasible edges. 3. Provide an exact O (V.E) algorithmfor building redundant shortest paths. 4. Provide an O (V.E+C 2 ) heuristic forgenerating Pareto optimal shortest paths in presence of multiple objectiveswhere C is the upper bound for path length. The complexity can be furtherreduced to O (V.E) by using graphical read-out of the Pareto frontier. 5.Provide a cost structure which can capture multiple key probability distributionparameters of edge variables. This is in contrast with usual techniques whichjust capture single parameters like the mean or the variance of distributions.6. Provide a MIP formulation to a multi-commodity transportation problem withmultiple decision variables, stochastic demands and uncertain edge/routecapacities. 7. Provide an alternate formulation to the classic binary facilityselection problem.

1 引言:问题描述与以前工作的回顾
2 路径规划模型与解决方案
2.1 单目标多概率约束
2.2 多目标多约束
2.3 基于概率的行程时间分布
2.4 多来源/目的地/实体的路径规划
3 结果与结论
3.1 单一实体
3.2 多实体
3.3 结论
附录A MATLAB源码
附录B 第3.2节案例A的仿真结果

下载英文原文地址:

http://page2.dfpan.com/fs/7l4cbj72624182e9169/

更多精彩文章请关注微信号:在这里插入图片描述

猜你喜欢

转载自blog.csdn.net/weixin_42825609/article/details/88362458