Java调用Cplex解VRPTW问题（代码非原创）

2023-10-20 10:50
文章标签 java 代码 问题 调用 原创 vrptw cplex

本文主要是介绍Java调用Cplex解VRPTW问题（代码非原创），希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！

参考资料：微信公众号“数据魔法师”；Column Generation

模型如下：
在这里插入图片描述
在这里插入图片描述

我 在华中科技大学黄楠博士的原作 （非常感谢原作者！）的基础上，做了小修，代码如下：

整体思路：
先设置问题的背景，数据的存放和组织，具体如：客户节点、车辆数、每个车辆的路径、访问客户的时间段和时刻，这些都在类Data中。实际读入数据之后(见 函数process_solomon)，删除掉没用的边（有的边绕远路,有的端点组合的时间窗、距离明显不可搭配等等），这里所说“删除”是指data.arc[i][j]设为0,表示不可走。
然后，设置模型，具体见函数build_model;
调用cplex解模型。
最后，输出解(要考虑把每个车辆要访问的客户按顺序记录到ArrayList中等等)，见class Solution.

// VRPTW.java
package HuangVRPTW;import ilog.concert.*;
import ilog.cplex.*;import java.io.BufferedReader;
import java.io.FileNotFoundException;
import java.io.FileReader;
import java.util.ArrayList;
import java.util.Scanner;//import HuangVRPTW.Data.Solution;
public class VRPTW {static double epsilon =  0.0001;Data data;IloCplex model;public IloNumVar[][][] x;public IloNumVar[][] w;double cost; // obj valueSolution solution;public VRPTW(Data data){this.data = data;}public void solve() throws IloException { // 整理问题的答案，并输出ArrayList<ArrayList<Integer>> routes = new ArrayList<>();ArrayList<ArrayList<Double>> servetimes = new ArrayList<>();// initialize 'routes', 'servetimes'for (int k = 0; k < data.vecnum; k++) {ArrayList<Integer> r = new ArrayList<>();ArrayList<Double> t = new ArrayList<>();routes.add(r); // 一辆车有一条路径，有一条服务时间线servetimes.add(t);}if(model.solve()==false){System.out.println("The Problem cannot be solved!");return;}else{// we found a solution, so print it!for (int k = 0; k < data.vecnum; k++) {boolean terminate = true;int i=0;routes.get(k).add(0); // depot 0servetimes.get(k).add(0.0);while(terminate){for (int j = 0; j < data.vetexnum; j++) {if(data.arcs[i][j]>=0.5 && model.getValue(x[i][j][k])>=0.5){routes.get(k).add(j);servetimes.get(k).add(model.getValue(w[j][k]));i=j; // this car continues to move foward the next customerbreak;}}if(i==data.vetexnum-1){terminate=false;}}}}solution = new Solution(data,routes,servetimes);cost = model.getObjValue();System.out.println("routes= "+solution.routes);} // function solve endsprivate void build_model() throws IloException {// 流程：// 首先 建一个model;// 然后，向model里面添加决策变量；// 接着，yongmodel建立一个表达式作为目标函数，设置目标；// 最后,求解这个model// modelmodel = new IloCplex();model.setOut(null); // ?? what is set out// decision varsx = new IloNumVar[data.vetexnum][data.vetexnum][data.vecnum];w = new IloNumVar[data.vetexnum][data.vecnum];// more details about decision varsfor (int i = 0; i < data.vetexnum; i++) {for (int k = 0; k < data.vecnum; k++) {w[i][k] = model.numVar(0,1e5,IloNumVarType.Float,"w"+i+","+k);}for (int j = 0; j < data.vetexnum; j++) {if(data.arcs[i][j]==0){x[i][j]=null;}else{for (int k = 0; k < data.vecnum; k++) {x[i][j][k] = model.numVar(0,1,IloNumVarType.Int,"x"+i+","+j+","+k);}}}}// end for i// obj funcIloNumExpr obj = model.numExpr();for (int i = 0; i < data.vetexnum; i++) {for (int j = 0; j < data.vetexnum; j++) {if(data.arcs[i][j]==1){for (int k = 0; k < data.vecnum; k++) {obj = model.sum(obj,model.prod(data.dist[i][j],x[i][j][k]));}}}} // end for imodel.addMinimize(obj);// constraints// constraint 1, every customer needs to be servedfor (int i = 1; i < data.vetexnum-1; i++) {IloNumExpr expr1 = model.numExpr();for (int j = 1; j < data.vetexnum; j++) {if(data.arcs[i][j]==1){for (int k = 0; k < data.vecnum; k++) {expr1 = model.sum(expr1,x[i][j][k]);}}}model.addEq(expr1,1);}// constraint 2, every cars start from depot 0for (int k = 0; k < data.vecnum; k++) {IloNumExpr expr2 = model.numExpr();for (int j = 1; j < data.vetexnum; j++) {expr2 = model.sum(expr2,x[0][j][k]);}model.addEq(expr2,1);}// constraint 3, for every car and every customer, #inFlow = #outFlowfor (int k = 0; k < data.vecnum; k++) {for (int i = 1; i < data.vetexnum-1; i++) {IloNumExpr subExpr1 = model.numExpr();IloNumExpr subExpr2 = model.numExpr();for (int j = 0; j < data.vetexnum; j++) {if(data.arcs[i][j]==1){subExpr1 = model.sum(subExpr1,x[i][j][k]);}if(data.arcs[j][i]==1){subExpr2 = model.sum(subExpr2,x[j][i][k]);}}model.addEq(subExpr1,subExpr2);}}// constraint 4, every car goes back to the depot n-1for (int k = 0; k < data.vecnum; k++) {IloNumExpr expr4 = model.numExpr();for(int i=0;i<data.vetexnum-1;i++){if(data.arcs[i][data.vetexnum-1]==1){expr4 = model.sum(expr4,x[i][data.vetexnum-1][k]);}}model.addEq(expr4,1);}// constraint 5, time window constraintdouble M=1e5;for (int k = 0; k < data.vecnum; k++) {for (int i = 0; i < data.vetexnum; i++) {for (int j=0;j<data.vetexnum;j++){if(data.arcs[i][j]==1){IloNumExpr expr5 = model.numExpr();IloNumExpr expr6 = model.numExpr();// 注意 data.s[i]+data.dist[i][j] 两项都是 数值，所以可以直接用加法expr5 = model.sum(w[i][k],data.s[i]+data.dist[i][j]); // s[i], the time period of serving customer iexpr5 = model.sum(expr5,model.prod(-1,w[j][k]));expr6 = model.prod(M,model.sum(1,model.prod(-1,x[i][j][k])));model.addLe(expr5,expr6);}}}}// constraint 6, every customer is served during his own time windowfor (int i = 1; i < data.vetexnum-1; i++) {for (int k = 0; k < data.vecnum; k++) {IloNumExpr expr5 = model.numExpr();expr5 = model.sum(expr5,w[i][k]);model.addLe(data.a[i],expr5);model.addLe(expr5,data.b[i]);}}// constraint 7, car's capacityfor (int k = 0; k < data.vecnum; k++) {IloNumExpr expr7 = model.numExpr();for (int i=1;i<data.vetexnum-1;i++){IloNumExpr expr8= model.numExpr();for (int j = 0; j < data.vetexnum; j++) {if(data.arcs[i][j]==1){expr8 = model.sum(expr8,x[i][j][k]);}}expr7 = model.sum(expr7,model.prod(data.demands[i],expr8));}model.addLe(expr7,data.cap);}}// end build_modelpublic static void process_solomon(String path, Data data, int vetexnum) throws FileNotFoundException {// 把 path指定的文件 读入到 data中String line = null;String[] substr = null;Scanner cin = new Scanner(new BufferedReader((new FileReader(path))));for (int i = 0; i < 4; i++) {line = cin.nextLine();}line = cin.nextLine();line.trim();substr = line.split("\\s+");// initialize paramsdata.vetexnum = vetexnum;data.vecnum = Integer.parseInt(substr[1]);data.cap = Integer.parseInt(substr[2]); //data.vertexs = new int[data.vetexnum][2];data.demands = new int[data.vetexnum];data.vehicles = new int[data.vecnum];data.a = new double[data.vetexnum];data.b = new double[data.vetexnum];data.s = new double[data.vetexnum]; // time period of customersdata.arcs = new int[data.vetexnum][data.vetexnum];data.dist = new double[data.vetexnum][data.vetexnum];for (int i = 0; i < 4; i++) {line = cin.nextLine();}// read the content of (data.vetexnum-1) linesfor (int i = 0; i < data.vetexnum-1; i++) {line = cin.nextLine();line.trim();substr = line.split("\\s+");data.vertexs[i][0] = Integer.parseInt(substr[2]);data.vertexs[i][1]= Integer.parseInt(substr[3]);data.demands[i] = Integer.parseInt(substr[4]);data.a[i] = Integer.parseInt(substr[5]);data.b[i]  = Integer.parseInt(substr[6]);data.s[i]  = Integer.parseInt(substr[7]);}cin.close();// depot n-1data.vertexs[vetexnum-1] = data.vertexs[0];data.demands[vetexnum-1] = 0;data.a[vetexnum-1] = data.a[0];data.b[vetexnum-1] = data.b[0];data.s[vetexnum-1] = 0; // dont need time to be serveddata.E=data.a[0];data.L = data.b[vetexnum-1];double min1 = 1e15;double min2 = 1e15;// initialize the distance matrixfor (int i = 0; i < data.vetexnum; i++) {for (int j = 0; j < data.vetexnum; j++) {if(i==j){data.dist[i][j] = 0;continue;}data.dist[i][j] = Math.sqrt((data.vertexs[i][0]-data.vertexs[j][0])*(data.vertexs[i][0]-data.vertexs[j][0])+ (data.vertexs[i][1]-data.vertexs[j][1])*(data.vertexs[i][1]-data.vertexs[j][1]));data.dist[i][j] = data.double_truncate(data.dist[i][j]);}}data.dist[0][data.vetexnum-1]=0;data.dist[data.vetexnum-1][0]=0;// triangular relationship ?? why do this // I guess it's because you don't need to take extra effort。 你没必要走弯路，如果可以抄近道的话for (int k = 0; k < data.vetexnum; k++) {for (int i = 0; i < data.vetexnum; i++) {for (int j = 0; j < data.vetexnum; j++) {if(data.dist[i][j]>data.dist[i][k]+data.dist[k][j]){ // 我后来想到这里是不是也应该用double_compare functiondata.dist[i][j]=data.dist[i][k]+data.dist[k][j];}}}}// set data.arcs// step 1,for (int i = 0; i < data.vetexnum; i++) {for (int j = 0; j < data.vetexnum; j++) {if(i!=j){data.arcs[i][j] = 1;}else {data.arcs[i][j] = 0;}}}// step 2, considering time window and capfor (int i = 0; i < data.vetexnum; i++) {for (int j = 0; j < data.vetexnum; j++) {if(data.arcs[i][j]==1){// 这里是double型时刻数据在比较大小，我觉得应该同这套代码的其他地方一样，调用double_compare函数// 这里也是  和原作不同的地方if(double_compare( // make double_compare staticdata.a[i]+data.s[i]+data.dist[i][j],data.b[j])>0||data.demands[i]+data.demands[j]> data.cap){data.arcs[i][j] = 0;}if(double_compare( // make double_compare staticdata.a[0]+data.s[i]+data.dist[0][i]+data.dist[i][data.vetexnum-1],data.b[data.vetexnum-1])>0){System.out.println("the current data is false");}}}}// step 3: considering time windowfor (int i = 1; i < data.vetexnum-1; i++) {if(double_compare( data.b[i]-data.dist[0][i],min1)<0){min1 = data.b[i]-data.dist[0][i];}if(double_compare(data.a[i]+data.s[i]+data.dist[i][data.vetexnum-1],min2)<0){min2 = data.a[i]+data.s[i]+data.dist[i][data.vetexnum-1];}}if(min1<data.E || min2>data.L){System.out.println("Duration false");}// step 4 : considering depot 0 and depot n-1 as the start and the terminaldata.arcs[0][data.vetexnum-1]=1;data.arcs[data.vetexnum-1][0]=0;for (int i = 1; i < data.vetexnum-1; i++) { // 客户点 无法回到 分配点0data.arcs[i][0] = 0;}for (int i = 1; i < data.vetexnum-1; i++) { // 分配点n-1 无法去到客户点data.arcs[data.vetexnum-1][i]=0;}}public static int double_compare(double a, double b){if(a<b-epsilon){return -1;}if(a>b+epsilon){return 1;}return 0;}public static void main(String[] args) throws FileNotFoundException, IloException {Data data = new Data();int vetexnum=102; // 100+2 // 100 customers and 2 depotsString path = "D:\\VRP问题实践\\CGVRPTW_Model-master\\CGVRPTW_Seminar\\resource\\c101-2.txt";//算例地址process_solomon(path,data,vetexnum);System.out.println("Input Successfully");System.out.println("Cplex procedure ....");VRPTW model = new VRPTW(data);model.build_model();double time1 = System.nanoTime();model.solve();model.solution.feasible();double time2 = System.nanoTime();System.out.println("Runing time is " + (time2 - time1) / 1e9 + ", best cost is " + model.cost);}
}

下面的文件同上面的文件在同一个目录下。

// Data_Solution.java
package HuangVRPTW;import ilog.concert.*;
import ilog.cplex.*;import java.util.ArrayList;
// this file defines two class, class Data and class Solution
// class Solution has a Data-class object
// class Data is used to read data from a given file
// class Solution is used to print the final solution
class Data {int vetexnum;// #customers + 2 // 2 depots double E; // start of the whole durationdouble L;// end of the whole durationint vecnum; // #vehicles double cap;// vehicle capacityint[][] vertexs;int[] demands;int[] vehicles; // carNo. 车辆编号double[] a; // ready time of Customersdouble[] b; // due time of Customersdouble[] s; // service time period of customersint[][] arcs; // arcs[i][j]=1 means this arc is feasibledouble[][] dist;public double double_truncate(double v){int iv = (int)v;if(iv+1 -v<=0.000000001){return iv+1;}double dv = (v-iv)*10;int idv = (int)dv;double rv = iv + idv/10.0; // !!/10.0 , not /10return rv;}}
class Solution{ // inner class of class Datadouble epsilon = 0.0001;Data data = new Data();ArrayList<ArrayList<Integer>> routes = new ArrayList<>();// routes of all the vehiclesArrayList<ArrayList<Double>> servetimes = new ArrayList<>();//the time point of every customer on the routepublic Solution(Data data, ArrayList<ArrayList<Integer>> routes,ArrayList<ArrayList<Double>> servetimes){super();this.data = data;this.routes = routes;this.servetimes = servetimes;}public int double_compare(double a, double b){ // Double 型数据比较大小要注意！if(a<b-epsilon){return -1;}if(a>b+epsilon){return 1;}return 0;}public void feasible() {if(routes.size()>data.vecnum){ // feasibility of the amount of carsSystem.out.println("Error: vecnum!!");System.exit(0);}// feasibility of capacity of carsfor (int k = 0; k < data.vecnum; k++) {ArrayList<Integer> route = routes.get(k);double trueCap = 0;for(int i=0;i<route.size();i++){trueCap += data.demands[route.get(i)];}if(trueCap>data.cap){System.out.println("error: cap!!");System.exit(0);}}// feasibility of time windowfor (int k = 0; k < data.vecnum; k++) {ArrayList<Integer> route = new ArrayList<>();for (int i = 0; i < route.size()-1; i++) {int origin = route.get(i);int destination  = route.get(i+1);double si = servetimes.get(k).get(i);double sj = servetimes.get(k).get(i+1);if(si<data.a[origin] || si>data.b[origin]){System.out.println("Error: serveTime");System.exit(0);}if(double_compare(si+data.dist[origin][destination],data.b[destination])>0){System.out.println(origin + ":[" + data.a[origin] + "," + data.b[origin] + "] " + si);System.out.println(destination + ":[" + data.a[destination] + "," + data.b[destination] + "] " + sj);System.out.println(data.dist[origin][destination]);System.out.println(destination + ": ");System.out.println("Error: forward servetime");System.exit(0);}if(double_compare(sj-data.dist[origin][destination],data.a[origin])<0){System.out.println(origin + ":[" + data.a[origin] + "," + data.b[origin] + "] " + si);System.out.println(destination + ":[" + data.a[destination] + "," + data.b[destination] + "] " + sj);System.out.println(data.dist[origin][destination]);System.out.println(destination + ": ");System.out.println("Error: backward serveTime");System.exit(0);}}}}
}

运行结果：

Input Successfully
Cplex procedure ....
routes= [[0, 81, 78, 76, 71, 70, 73, 77, 79, 80, 101], 
[0, 57, 55, 54, 53, 56, 58, 60, 59, 101],
[0, 98, 96, 95, 94, 92, 93, 97, 100, 99, 101], 
[0, 90, 87, 86, 83, 82, 84, 85, 88, 89, 91, 101],[0, 5, 3, 7, 8, 10, 11, 9, 6, 4, 2, 1, 75, 101], [0, 101],[0, 20, 24, 25, 27, 29, 30, 28, 26, 23, 22, 21, 101], [0, 32, 33, 31, 35, 37, 38, 39, 36, 34, 101], [0, 67, 65, 63, 62, 74, 72, 61, 64, 68, 66, 69, 101], [0, 13, 17, 18, 19, 15, 16, 14, 12, 101], [0, 101], [0, 43, 42, 41, 40, 44, 46, 45, 48, 51, 50, 52, 49, 47, 101]]
Runing time is 16.1905963, best cost is 827.3

算例文件中 限制车辆数是12，输出答案中所有车辆都有自己的路径，只是有的车不被使用，比如：
这里[0, 101] 表示对应的这辆车不被使用（因为depot 101 and depot 0 are the same one.）

这篇关于Java调用Cplex解VRPTW问题（代码非原创）的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！


http://www.chinasem.cn/article/246616。

相关文章

JVM 的类初始化机制

JVM 的类初始化机制

前言 当你在 Java 程序中new对象时，有没有考虑过 JVM 是如何把静态的字节码（byte code）转化为运行时对象的呢，这个问题看似简单，但清楚的同学相信也不会太多，这篇文章首先介绍 JVM 类初始化的机制，然后给出几个易出错的实例来分析，帮助大家更好理解这个知识点。 JVM 将字节码转化为运行时对象分为三个阶段，分别是：loading 、Linking、initialization
阅读更多...
Spring Security 基于表达式的权限控制

Spring Security 基于表达式的权限控制

前言 spring security 3.0已经可以使用spring el表达式来控制授权，允许在表达式中使用复杂的布尔逻辑来控制访问的权限。 常见的表达式 Spring Security可用表达式对象的基类是SecurityExpressionRoot。 表达式描述hasRole([role])用户拥有制定的角色时返回true （Spring security默认会带有ROLE_前缀）,去
阅读更多...
浅析Spring Security认证过程

浅析Spring Security认证过程

类图 为了方便理解Spring Security认证流程，特意画了如下的类图，包含相关的核心认证类 概述 核心验证器 AuthenticationManager 该对象提供了认证方法的入口，接收一个Authentiaton对象作为参数; public interface AuthenticationManager {Authentication authenticate(Authenti
阅读更多...
Spring Security--Architecture Overview

Spring Security--Architecture Overview

1 核心组件 这一节主要介绍一些在Spring Security中常见且核心的Java类，它们之间的依赖，构建起了整个框架。想要理解整个架构，最起码得对这些类眼熟。 1.1 SecurityContextHolder SecurityContextHolder用于存储安全上下文（security context）的信息。当前操作的用户是谁，该用户是否已经被认证，他拥有哪些角色权限…这些都被保
阅读更多...
Spring Security基于数据库验证流程详解

Spring Security基于数据库验证流程详解

Spring Security 校验流程图 相关解释说明（认真看哦） AbstractAuthenticationProcessingFilter 抽象类 /*** 调用 #requiresAuthentication(HttpServletRequest, HttpServletResponse) 决定是否需要进行验证操作。* 如果需要验证，则会调用 #attemptAuthentica
阅读更多...
Spring Security 从入门到进阶系列教程

Spring Security 从入门到进阶系列教程

Spring Security 入门系列 《保护 Web 应用的安全》 《Spring-Security-入门（一）：登录与退出》 《Spring-Security-入门（二）：基于数据库验证》 《Spring-Security-入门（三）：密码加密》 《Spring-Security-入门（四）：自定义-Filter》 《Spring-Security-入门（五）：在 Sprin
阅读更多...
Java架构师知识体认识

Java架构师知识体认识

源码分析 常用设计模式 Proxy代理模式Factory工厂模式Singleton单例模式Delegate委派模式Strategy策略模式Prototype原型模式Template模板模式 Spring5 beans 接口实例化代理Bean操作 Context Ioc容器设计原理及高级特性Aop设计原理Factorybean与Beanfactory Transaction 声明式事物
阅读更多...
Java进阶13讲__第12讲_1/2

Java进阶13讲__第12讲_1/2

多线程、线程池 1.  线程概念 1.1  什么是线程 1.2  线程的好处 2.   创建线程的三种方式 注意事项 2.1  继承Thread类 2.1.1 认识  2.1.2  编码实现  package cn.hdc.oop10.Thread;import org.slf4j.Logger;import org.slf4j.LoggerFactory
阅读更多...
好题——hdu2522（小数问题：求1/n的第一个循环节）

好题——hdu2522（小数问题：求1/n的第一个循环节）

好喜欢这题，第一次做小数问题，一开始真心没思路，然后参考了网上的一些资料。 知识点***********************************无限不循环小数即无理数，不能写作两整数之比*****************************（一开始没想到，小学没学好） 此题1/n肯定是一个有限循环小数，了解这些后就能做此题了。 按照除法的机制，用一个函数表示出来就可以了，代码如下
阅读更多...
hdu1043（八数码问题，广搜 + hash（实现状态压缩） ）

hdu1043（八数码问题，广搜 + hash（实现状态压缩） ）

利用康拓展开将一个排列映射成一个自然数，然后就变成了普通的广搜题。 #include<iostream>#include<algorithm>#include<string>#include<stack>#include<queue>#include<map>#include<stdio.h>#include<stdlib.h>#include<ctype.h>#inclu
阅读更多...

Copyright China编程 23002807@qq.com

沪ICP备2023033072号-2