最短路径FloydWarshall算法

介绍

    Floyd-Warshall的算法也是运用动态规划来实现的最短路径的一种方法。

实现

typedef int (*Metrix)[VERTEX_NUMBER];

void printfMatrix(Metrix graphmetrix)
{
	for (int x = 0; x < VERTEX_NUMBER; ++ x)
	{
		for (int y = 0; y < VERTEX_NUMBER; ++ y)
		{
			cout << "\t" << graphmetrix[x][y] << " ";
		}
		cout << endl;
	}
	cout << endl;
}

int getSentinel()
{
	return 0x7fffffff;
}

    以上是辅助函数用来输出矩阵。

Metrix floydWarshall(Metrix graph)
{
	const int n = VERTEX_NUMBER;
	int *tmp = new int[n * n];
	Metrix c = (int(*)[n])tmp;
	memcpy(c, graph, n * n * sizeof(4));
	tmp = new int[n * n];
	Metrix p = (int(*)[n])tmp;
	memcpy(p, graph, n * n * sizeof(4));
	tmp = new int[n * n];
	Metrix precursor_graph = (int(*)[n])tmp;

	for (int i = 0; i < n; ++ i)
	{
		for (int j = 0; j < n; ++ j)
		{
			if (i != j && graph[i][j] != getSentinel())
			{
				precursor_graph[i][j] = i + 1;
			}else
			{
				precursor_graph[i][j] = 0;
			}
		}
	}

	printfMatrix(precursor_graph);

	for (int k = 0; k < n; ++ k)
	{
		for (int i = 0; i < n; ++ i)
		{
			for (int j = 0; j < n; ++ j)
			{
				if (getSentinel() != c[i][k] && getSentinel() != c[k][j])
				{
					p[i][j] = std::min(c[i][j], c[i][k] + c[k][j]);
					if (c[i][j] <= c[i][k] + c[k][j])
					{
						precursor_graph[i][j] = precursor_graph[i][j];
					}else
						precursor_graph[i][j] = precursor_graph[k][j];
				}
			}
		}
		cout << "shortest path: " << endl;
		printfMatrix(p);
		cout << "precursor sub graph: " << endl;
		printfMatrix(precursor_graph);
		memcpy(c, p, n * n * sizeof(4));
	}

	return c;
}

    在Floyd-Warshall算法中，采用的是自底而上的办法来计算最短路径权重，每次使用前一个的计算结果值来得出当前最短路径权重值。首先初始化一个和原图一样的矩阵结构，并初始化一个结果矩阵也和原图保持一致。接着初始化一个前趋子图矩阵，里面保存i->j的某条最短路径中结点j的前趋结点。

而我们可以用下面的程式来验证，使用如下的图来运行Floyd-Warshall算法：

Floyd-Warshall算法测试代码图

int _tmain(int argc, _TCHAR* argv[])
{
	copyright();
	cout << getSentinel() << endl;

	cout << "directed adjacency matrix: " << endl;
	int graphmetrix[VERTEX_NUMBER][VERTEX_NUMBER] = 
	{
		{0, 3, 8, 0x7fffffff, -4},
		{0x7fffffff, 0, 0x7fffffff, 1, 7},
		{0x7fffffff, 4, 0, 0x7fffffff, 0x7fffffff},
		{2, 0x7fffffff, -5, 0, 0x7fffffff},
		{0x7fffffff, 0x7fffffff, 0x7fffffff, 6, 0},
	};
	cout << "directed weight graph structure: " << endl;
	for (int x = 0; x < VERTEX_NUMBER; ++ x)
	{
		cout << "vertex " << x + 1 << "；";
		for (int y = 0; y < VERTEX_NUMBER; ++ y)
		{
			if (0 != graphmetrix[x][y])
			{
				cout << y + 1 << " ";
			}
		}
		cout << endl;
	}
	cout << "origin: " << endl;
	printfMatrix(graphmetrix);
	Metrix c = floydWarshall(graphmetrix);
	delete [] c;

	return 0;
}

    运行结果如下所示：

运行Floyd-Warshall算法之前初始化原图和前趋矩阵图

Floyd-Warshall算法运行当k在1-3时的最短路径矩阵和前趋矩阵图

Floyd-Warshall算法运行当k在4-5时的最短路径矩阵和前趋矩阵图

    接着我们来看看它的详细结果图解：

Floyd-Warshall算法的最短路径矩阵和前趋子图矩阵

    比如，我们可以根据上图，在左边矩阵中一眼就可以看出1->2的的最短路径权重就是1，因为d[0][1] = 1。而在右边矩阵图中则可以看出1->2的最短路径权重的路径。首先是找II[0][1] = 3，可以看出2的前趋是3，然后找II[0][2] = 4，即3的前趋是4，依次可以找到4的前趋是5，5的前趋是1，则回到了起点1。则1->2的最短路径就是1->5->4->3->2。