Scouting smartly?

This topic is 2158 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

Recommended Posts

Update: thanks for all the no-reply, after some reading it seems that this is traveling salesman problem...

Edited by Moon Dancer

Share on other sites

/Update:

wrote a simple recursive search, it's kinda fcking slow =d

        class rs_data
{
public List<ExplorerPoint> bestpath = null;
public float bestpath_hdist = float.MaxValue;
}

private float recursive_search(float traveleddist, Vector3D lastpoint, List<ExplorerPoint> xplist, rs_data rs_data, int deep=0, List<ExplorerPoint> chain=null)
{
if (rs_data.bestpath_hdist < traveleddist)
return traveleddist;
if (chain == null)
chain = new List<ExplorerPoint>();
if (xplist.Count == 0)
{
if (traveleddist < rs_data.bestpath_hdist)
{
rs_data.bestpath_hdist = traveleddist;
rs_data.bestpath = chain.ToList();
//backtrack last ?
}
return traveleddist;
}
var newlist = xplist.OrderBy(
w => { return heuristicDistance((float)lastpoint.X, (float)lastpoint.Y, (float)w.Pos.X, (float)w.Pos.Y); });
ExplorerPoint bestnode = null;
float bestnode_dist = float.MaxValue;
var i = 1;
if (deep < 10)
i = 2;

foreach (var xp1 in newlist)
{
i--;
if (i == -1)
break;
var hdist = (float)Math.Sqrt(heuristicDistance((float)xp1.Pos.X, (float)xp1.Pos.Y, (float)lastpoint.X, (float)lastpoint.Y));
if ((traveleddist + hdist) > rs_data.bestpath_hdist)
continue;

var n = newlist.ToList();
n.Remove(xp1);
var res = recursive_search(traveleddist + hdist, xp1.Pos, n, rs_data, deep + 1, chain);
chain.RemoveAt(chain.Count - 1);
if (res < bestnode_dist)
{
bestnode = xp1;
bestnode_dist = res;
}
}
return bestnode_dist;
}



Share on other sites

Let me try to understand your problem, because it's not immediately clear from the pictures. You have a finite set of points (the red dots), you are currently at one of them and you want to design a path through all the other nodes, minimizing total distance travelled. Is that it? If so, ask Google about the "travelling salesman problem": The original problem includes returning to the starting point at the end of the path, but you can probably use pretty much the same algorithms in your case.

If that is not your problem, please try to state it as clearly as possible.

Share on other sites

Never delete your original post and say "never mind, I've solved it". It just confuses people who come in later.

Share on other sites

Also, other people might have the same or a very similar problem. So solutions can be valuable to more than one person.

Edited by AngleWyrm

• 21
• 11
• 9
• 17
• 13