在城市智能交通中,经常会用到最短路径的问题,比如找最佳的行车路线等,Dijkstra算法做为最经典的求解方法,为我们指明了方向.不过真正想让我了解该算法的原因是在学习ICTCLAS的N-最短路径算法,虽然和我们常用的案例有一点区别,但基本相同,为了更好的理解N-最短路径算法,我又重新把大学时代的数据结构知识搬了出来。
在网上找到一篇文章,非常详细生动(有FLASH动画演示)的描述了该算法的实现,不过第一页右下角的图终点那一列2和3弄反了,看的时候要注意 ,具体的算法描述不再赘述,请参考:
http://student.zjzk.cn/course_ware/data_structure/web/tu/tu7.5.1.htm
下面给出我的算法实现具体代码,为了更好的验证程序的正确性,在原来的基础上我又多加了几条边
package sinboy.datastructure; import java.util.ArrayList; public class Dijkstra ...{ static ArrayList map = null; static ArrayList redAgg = null; static ArrayList blueAgg = null; static Side[] parents = null; public static void main(String[] args) ...{ // 初始化顶点集 int[] nodes = ...{ 0, 1, 3, 2, 4, 5,6 }; // 初始化有向权重图 map = new ArrayList(); map.add(new Side(0, 1, 10)); map.add(new Side(0, 3, 30)); map.add(new Side(0, 4, 100)); map.add(new Side(1, 2, 50)); map.add(new Side(2, 4, 10)); map.add(new Side(3, 2, 20)); map.add(new Side(3, 4, 60)); map.add(new Side(4, 5, 50)); map.add(new Side(3, 5, 60)); map.add(new Side(5, 6, 10)); map.add(new Side(3, 6, 80)); // 初始化已知最短路径的顶点集,即红点集,只加入顶点0 redAgg = new ArrayList(); redAgg.add(nodes[0]); // 初始化未知最短路径的顶点集,即蓝点集 blueAgg = new ArrayList(); for (int i = 1; i < nodes.length; i++) blueAgg.add(nodes[i]); // 初始化每个顶点在最短路径中的父结点,及它们之间的权重,权重-1表示无连通 parents = new Side[nodes.length]; parents[0] = new Side(-1, nodes[0], 0); for (int i = 0; i < blueAgg.size(); i++) ...{ int n = blueAgg.get(i); parents[i + 1] = new Side(nodes[0], n, getWeight(nodes[0], n)); } // 找从蓝点集中找出权重最小的那个顶点,并把它加入到红点集中 while (blueAgg.size() > 0) ...{ MinShortPath msp = getMinSideNode(); if(msp.getWeight()==-1) msp.outputPath(nodes[0]); else msp.outputPath(); int node = msp.getLastNode(); redAgg.add(node); // 如果因为加入了新的顶点,而导致蓝点集中的顶点的最短路径减小,则要重要设置 setWeight(node); } } /** *//** * 得到一个节点的父节点 * * @param parents * @param node * @return */ public static int getParent(Side[] parents, int node) ...{ if (parents != null) ...{ for (Side nd : parents) ...{ if (nd.getNode() == node) ...{ return nd.getPreNode(); } } } return -1; } /** *//** * 重新设置蓝点集中剩余节点的最短路径长度 * * @param preNode * @param map * @param blueAgg */ public static void setWeight(int preNode) ...{ if (map != null && parents != null && blueAgg != null) ...{ for (int node : blueAgg) ...{ MinShortPath msp=getMinPath(node); int w1 = msp.getWeight(); if (w1 == -1) continue; for (Side n : parents) ...{ if (n.getNode() == node) ...{ if (n.getWeight() == -1 || n.getWeight() > w1) ...{ n.setWeight(w1); n.setPreNode(preNode);//重新设置顶点的父顶点 break; } } } } } } /** *//** * 得到两点节点之间的权重 * * @param map * @param preNode * @param node * @return */ public static int getWeight(int preNode, int node) ...{ if (map != null) ...{ for (Side s : map) ...{ if (s.getPreNode() == preNode && s.getNode() == node) return s.getWeight(); } } return -1; } /** *//** * 从蓝点集合中找出路径最小的那个节点 * * @param map * @param blueAgg * @return */ public static MinShortPath getMinSideNode() ...{ MinShortPath minMsp = null; if (blueAgg.size() > 0) ...{ int index = 0; for (int j = 0; j < blueAgg.size(); j++) ...{ MinShortPath msp = getMinPath(blueAgg.get(j)); if (minMsp == null || msp.getWeight()!=-1 && msp.getWeight() < minMsp.getWeight()) ...{ minMsp = msp; index = j; } } blueAgg.remove(index); } return minMsp; } /** *//** * 得到某一节点的最短路径(实际上可能有多条,现在只考虑一条) * @param node * @return */ public static MinShortPath getMinPath(int node) ...{ MinShortPath msp = new MinShortPath(node); if (parents != null && redAgg != null) ...{ for (int i = 0; i < redAgg.size(); i++) ...{ MinShortPath tempMsp = new MinShortPath(node); int parent = redAgg.get(i); int curNode = node; while (parent > -1) ...{ int weight = getWeight(parent, curNode); if (weight > -1) ...{ tempMsp.addNode(parent); tempMsp.addWeight(weight); curNode = parent; parent = getParent(parents, parent); } else break; } if (msp.getWeight() == -1 || tempMsp.getWeight()!=-1 && msp.getWeight() > tempMsp.getWeight()) msp = tempMsp; } } return msp; } } /** *//** * 图中的有向边,包括节点名及他的一个前向节点名,和它们之间的权重 * */ class Side ...{ private int preNode; // 前向节点 private int node;// 后向节点 private int weight;// 权重 public Side(int preNode, int node, int weight) ...{ this.preNode = preNode; this.node = node; this.weight = weight; } public int getPreNode() ...{ return preNode; } public void setPreNode(int preNode) ...{ this.preNode = preNode; } public int getNode() ...{ return node; } public void setNode(int node) ...{ this.node = node; } public int getWeight() ...{ return weight; } public void setWeight(int weight) ...{ this.weight = weight; } } class MinShortPath ...{ private ArrayList nodeList;// 最短路径集 private int weight;// 最短路径 public MinShortPath(int node) ...{ nodeList = new ArrayList(); nodeList.add(node); weight = -1; } public ArrayList getNodeList() ...{ return nodeList; } public void setNodeList(ArrayList nodeList) ...{ this.nodeList = nodeList; } public void addNode(int node) ...{ if (nodeList == null) nodeList = new ArrayList(); nodeList.add(0, node); } public int getLastNode() ...{ int size = nodeList.size(); return nodeList.get(size - 1); } public int getWeight() ...{ return weight; } public void setWeight(int weight) ...{ this.weight = weight; } public void outputPath() ...{ outputPath(-1); } public void outputPath(int srcNode) ...{ String result = "["; if (srcNode != -1) nodeList.add(srcNode); for (int i = 0; i < nodeList.size(); i++) ...{ result += "" + nodeList.get(i); if (i < nodeList.size() - 1) result += ","; } result += "]:" + weight; System.out.println(result); } public void addWeight(int w) ...{ if (weight == -1) weight = w; else weight += w; } }
运行结果如下:
[0,1]:10
[0,3]:30
[0,3,2]:50
[0,3,2,4]:60
[0,3,5]:90
[0,3,5,6]:100
总结:最短路径算法关键先把已知最短路径顶点集(只有一个源点)和未知的顶点分开,然后依次把未知集合的顶点按照最短路径(这里特别强调一下是源点到该顶点的路径权重和,不仅仅是指它和父结点之间的权重,一开始就是在没有这个问题弄清楚)加入到已知结点集中。在加入时可以记录每个顶点的最短路径,也可以在加入完毕后回溯找到每个顶点的最短路径和权重。