我最近采用了A *搜索算法。我之前尝试过但无济于事,但这次我取得了一定程度的成功。它总是找到一条路径,除非它(显然)不能并且它通常接近最短路径。其他时候它的行为非常棘手,因为添加了太多次,以锯齿形图案,随机向错误的方向移动。这很奇怪。屏幕截图here.
以下代码:
int manhattan( Coord a, Coord b )
{
int x = abs(b.x-a.x);
int y = abs(b.y-a.y);
return x+y;
}
std::vector<Coord> AStar( std::vector< std::vector< int > > grid, Point start, Point end )
{
//The current 'focal' point.
Point *cur;
//The open and closed lists.
std::vector< Point* > closed;
std::vector< Point* > open;
//Start by adding the starting position to the list.
open.push_back( &start );
//Just so it knows whether or not to try and reconstruct a path.
bool error = true;
while( open.size() > 0 )
{
//The current point is the first entry in the open list.
cur = open.at(0);
if( cur->getPos() == end.getPos() )
{
error = false;
break;
}
//Add in all the neighbors of the current point.
for( int y = -1; y <= 1; y++ )
{
for( int x = -1; x <= 1; x++ )
{
int curX = cur->getPos().x+x;
int curY = cur->getPos().y+y;
int movCost = 10;
//If it is a diagonal, make it cost 14 instead of 10.
if( (y == -1 && x == -1)||
(y == 1 && x == -1)||
(y == -1 && x == 1)||
(y == 1 && x == 1))
{
movCost = 14;
//continue;
}
Coord temp( curX, curY );
bool make = true;
//If it is outside the range of the map, continue.
if( curY >= grid.size() ||
curX >= grid.size() )
{
continue;
}
/*
These two loops are to check whether or not the point's neighbors already exist.
This feels really sloppy to me. Please tell me if there is a better way.
*/
for( int i = 0; i < open.size(); i++ )
{
if( temp == open.at(i)->getPos() )
{
make = false;
break;
}
}
for( int i = 0; i < closed.size(); i++ )
{
if( temp == closed.at(i)->getPos() )
{
make = false;
break;
}
}
//If the point in the map is a zero, then it is a wall. Continue.
if( (grid.at(temp.x).at(temp.y) == 0 ) ||
( temp.x<0 || temp.y < 0 ) )
{
continue;
}
//If it is allowed to make a new point, it adds it to the open list.
if( make )
{
int gScore = manhattan( start.getPos(), Coord( curX, curY ) );
int hScore = manhattan( end.getPos(), Coord( curX, curY ) );
int tileCost = grid[curX][curY];
int fScore = gScore+hScore+tileCost;
open.push_back( new Point( curX, curY, fScore, cur ) );
}
}
}
//It then pushes back the current into the closed set as well as erasing it from the open set.
closed.push_back( cur );
open.erase( open.begin() );
//Heapsort works, guranteed. Not sure if it's a stable sort, though. From what I can tell that shouldn't matter, though.
open = heapsort( open );
}
std::vector<Coord> path;
if( error )
{
return path;
}
//Reconstruct a path by tracing through the parents.
while( cur->getParent() != nullptr )
{
path.push_back( cur->getPos() );
cur = cur->getParent();
}
path.push_back( cur->getPos() );
return path;
}
反正!谢谢你提前帮忙!如果你想给我一些有用的提示或任何其他帮助,那将是非常棒的!非常感谢! :^)
答案 0 :(得分:2)
我可以看到你在这里试图使对角线更加昂贵:
int movCost = 10;
//If it is a diagonal, make it cost 14 instead of 10.
if( (y == -1 && x == -1)||
(y == 1 && x == -1)||
(y == -1 && x == 1)||
(y == 1 && x == 1))
{
movCost = 14;
//continue;
}
但您实际上并未在代码中的其他位置使用movCost
。
相反,您的成本函数仅使用曼哈顿距离:
int gScore = manhattan( start.getPos(), Coord( curX, curY ) );
int hScore = manhattan( end.getPos(), Coord( curX, curY ) );
int tileCost = grid[curX][curY];
int fScore = gScore+hScore+tileCost;
这解释了对角锯齿形的路径:
顺便提一下,你的代码中还有一个逻辑错误:在A *中,g-cost应该计算为从开始到当前节点的实际成本,而不是像你使用{{1功能。您应该在开放和关闭集合中保存成本以及您的积分。
将来,您应该打开所有编译器警告,不要忽略它们。这将捕获容易错过的错误,如未使用的变量。