A*算法概述
算法背景:
Dijkstra算法类似BFS算法,每次找到跟起点最近的顶点然后往外扩展。这种往外扩渣的思路有点盲目,可能会导致路线搜索方向跑偏。没有考虑顶点到终点的距离。比如如下从S点到T点,Dijkstra算法会先遍历1,2,3这些顶点。
算法核心:估算顶点到终点的距离,增加启发函数/估价函数。
算法对Dijkstra改造之处:
1、顶点Vertex类定义多了x,y坐标,以及f(i)估价函数
2、优先级队列构造方式不同,A算法根据f值f=g+h来构建优先级队列,其中g是起始点与顶点的距离,h是顶点到结束点的距离。D算法是根据g值来构建。
3、A算法在更新顶点dist值的时候,也会同步更新f值
4、循环结束条件不一样,D算法是在终点出队列的时候才结束,A*算法是一旦遍历到终点就结束,所以未必是最短路径
A*算法代码实现
#include<iostream>
#include<vector>
#include<limits.h>
#include<cmath>
using namespace std;
class Edge {
public:
int sid; // 边的起始顶点编号
int tid; //边的终止顶点编号
int w; //权重
Edge(int sid, int tid, int w) {
this->sid = sid;
this->tid = tid;
this->w = w;
}
};
class Vertex {
public:
int id; //顶点编号
int dist; //从起始顶点到这个顶点的距离
int x; // 比astar算法多x,y坐标
int y; // 比astar算法多x,y坐标
int f; // f = g + h; g = dist; h = |x1 - x2| + |y1 - y2|
Vertex (int id, int dist, int x, int y) {
this->id = id;
this->dist = dist;
}
Vertex () {
this->id = 0;
this->dist = INT_MAX;
this->f = INT_MAX;
this->x = 0;
this->y = 0;
}
};
class Graph {
private:
vector<Edge *> * adj; //邻接表
int v; //顶点个数
public:
Graph(int v) {
this->v = v;
this->adj = new vector<Edge *>[v];
}
~Graph() {
if (this->adj != NULL) {
for (int i = 0; i < v; ++i) {
for (int j = 0; j < adj[i].size(); ++j) {
delete adj[i][j];
}
}
delete [] adj;
}
}
void AddEdge(int s, int t, int w) {
Edge * e = new Edge(s, t, w);
adj[s].push_back(e);
}
void astar(int s, int t);
void Print(int s, int t, int * predecessor);
int hManhattan(Vertex s, Vertex t);
};
class PriorityQueue {
private:
Vertex * nodes; //顶点信息
int size; //堆可以存储的最大数据个数
int count; //堆中已经存储的数据个数
public:
PriorityQueue(int v) {
nodes = new Vertex[v+1];
size = v + 1;
count = 0;
}
~PriorityQueue() {
if (nodes != NULL) {
delete [] nodes;
}
count = 0;
}
void Add(Vertex* ver);
void Update(Vertex* ver);
bool IsEmpty();
void Print();
Vertex Poll();
void Adjust(int index);
void Clear();
};
void PriorityQueue::Clear(){
if (nodes != NULL) {
delete [] nodes;
}
count = 0;
}
Vertex PriorityQueue::Poll() {
Vertex top(nodes[1].id, nodes[1].dist, nodes[1].x, nodes[1].y);
nodes[1] = nodes[count];
count--;
int index = 1;
while(true) {
int maxpos = index;
if (2 * index < count && nodes[2 * index].f < nodes[maxpos].f) {
maxpos = 2 * index;
}
if (2 * index + 1 < count && nodes[2 * index + 1].f < nodes[maxpos].f) {
maxpos = 2 * index + 1;
}
if (maxpos == index) {
break;
}
int tmp_f = nodes[index].f;
int tmp_id = nodes[index].id;
int tmp_dist = nodes[index].dist;
nodes[index].f = nodes[maxpos].f;
nodes[index].dist = nodes[maxpos].dist;
nodes[index].id = nodes[maxpos].id;
nodes[maxpos].f = tmp_f;
nodes[maxpos].dist = tmp_dist;
nodes[maxpos].id = tmp_id;
}
return top;
}
bool PriorityQueue::IsEmpty() {
return count == 0;
}
void PriorityQueue::Add(Vertex* ver) {
if(count < size) {
count++;
nodes[count].dist = ver->dist;
nodes[count].f = ver->f;
nodes[count].id = ver->id;
//自下而上进行堆化
Adjust(count);
} else {
cout << "error! size is full!" << endl;
}
}
void PriorityQueue::Adjust(int index) {
int i = index;
while (i / 2 > 0 && nodes[i / 2].f > nodes[i].f) {
int tmp_f = nodes[i / 2].f;
int tmp_id = nodes[i / 2].id;
nodes[i / 2].f = nodes[i].f;
nodes[i / 2].id = nodes[i].id;
nodes[i].f = tmp_f;
nodes[i].id = tmp_id;
}
}
void PriorityQueue::Update(Vertex* ver) {
bool isupdate = false;
for (int i = 1; i <= count; ++i) {
if (nodes[i].id == ver->id) {
nodes[i].f = ver->f;
isupdate = true;
Adjust(i);
break;
}
}
if (isupdate == false) {
cout << "error! not find id:" << ver->id << endl;
}
}
void PriorityQueue::Print() {
for (int i = 1; i <= count; ++i) {
cout << nodes[i].f << " ";
}
cout << endl;
}
int Graph::hManhattan(Vertex s, Vertex t) {
return abs(s.x- t.x) + abs(s.y - t.y);
}
void Graph::astar(int s, int t) {
// predecessor 存放前驱节点,用来还原最短路径
int * predecessor = new int[v];
Vertex* vers = new Vertex[v];
// inqueue 标记是否进入过队列
bool * inqueue = new bool[v];
for (int i = 0; i < v; ++i) {
vers[i].id = i;
inqueue[i] = false;
vers[i].x = i;
vers[i].y = i;
}
//小顶堆
PriorityQueue* q = new PriorityQueue(v);
q->Print();
vers[s].dist = 0;
vers[s].f = 0;
inqueue[s] = true;
q->Add(&vers[s]);
q->Print();
while (!q->IsEmpty()) {
Vertex minV = q->Poll(); //取出堆顶元算并删除
if (minV.id == t) break; // 这就是最短路径
for (int i = 0; i < adj[minV.id].size(); ++i) {
Edge * e = adj[minV.id][i];
Vertex nextV = vers[e->tid];
cout << "minV.id: " << minV.id << "nextV.id: " << nextV.id << " begin" << endl;
if (minV.dist + e->w < nextV.dist) {
nextV.dist = minV.dist + e->w;
nextV.f = nextV.dist + hManhattan(nextV, vers[t]);
predecessor[nextV.id] = minV.id;
if (inqueue[nextV.id] == true) {
q->Update(&nextV);
q->Print();
} else {
q->Add(&nextV);
q->Print();
inqueue[nextV.id] = true;
cout << "update inqueu nextV.id : " << nextV.id << endl;
}
}
cout << "minV.id: " << minV.id << "nextV.id: " << nextV.id << " end" << endl;
if (nextV.id == t) { // 只要到达t就结束while
q->Clear(); // 清空queue,才能退出while循环
break;
}
}
}
cout << s;
Print(s, t, predecessor);
cout << endl;
}
void Graph::Print(int s, int t, int * predecessor) {
if (s == t) return;
Print(s, predecessor[t], predecessor);
cout << "->" << t;
}
int main() {
Graph g(6);
g.AddEdge(0, 1, 10);
g.AddEdge(0, 4, 15);
g.AddEdge(1, 2, 15);
g.AddEdge(1, 3, 2);
g.AddEdge(2, 5, 5);
g.AddEdge(3, 2, 1);
g.AddEdge(3, 5, 12);
g.AddEdge(4, 5, 10);
g.astar(0, 5);
}
A*算法应用
可以应用于游戏寻路,游戏中的角色从一个地方到另外一个地方,鼠标点击后角色就走过去,其背后的路径选择可以用A星算法。这里把地图抽象成图,把整个地图分割成一个一个的小方块,在某个方块上的人物只能往上下左右四个方向的方块上移动。每个方块看作为一个顶点。两个方块相邻,则连两条有向边,边权重为1.这样就转化为从一个顶点到另一个顶点的路径问题,进而可以使用A*算法求解。