/// <summary>
/// Muere el agente
/// </summary>
protected override void Die()
{
state = States.Die;
GetComponent <DinasorsAnimationCorrector>().die();
defense = 0;
if (gameObject.transform.Find("leaderLigth") != null)
{
Destroy(gameObject.transform.Find("leaderLigth").gameObject);
}
if (herd.Contains(gameObject))
{
herd.Remove(gameObject);
}
foreach (GameObject go in herd)
{
go.GetComponent <DinoObject>().herd.Remove(gameObject);
}
if (isLeader == true)
{
Prey p = gameObject.GetComponent <Prey>();
Predator pr = gameObject.GetComponent <Predator>();
if (p != null)
{
List <Prey> listHerd = new List <Prey>();
foreach (GameObject go in p.herd)
{
listHerd.Add(go.GetComponent <Prey>());
}
if (listHerd.Count > 0)
{
p.getNewLeader(listHerd);
}
else
{
p.isLeader = false;
}
}
if (pr != null)
{
List <Predator> listHerd = new List <Predator>();
foreach (GameObject go in pr.herd)
{
listHerd.Add(go.GetComponent <Predator>());
}
if (listHerd.Count > 0)
{
pr.getNewLeader(listHerd);
}
else
{
pr.isLeader = false;
}
}
}
}
public void createHerds()
{
elements = GameObject.FindObjectsOfType <Prey>();
int nElements = elements.Count();
int[] curCluster = new int[elements.Count()];
for (int i = 0; i < nElements; i++)
{
curCluster[i] = i; // We assign each element to it's cluster. Making sure this way to at least
//run 1 time the assignment step
}
// Make sure we can create the K clusters,
// If we have less than K elements then we only can create
// Number of elements clusters, each cluster with 1 element.
if (nElements < K)
{
K = nElements;
}
// Create K Herds, and K objects with the points where elements live
// Then populate the K herds using K Means.
List <List <Prey> > best = new List <List <Prey> > ();
float bestMSE = 100000000.0F;
List <List <Prey> > herds = new List <List <Prey> > ();
List <Vector3> means = new List <Vector3> ();
float minX = 1000000.0F, maxX = 0.0F, minY = 1000000.0F, maxY = 0.0F;
for (int i = 0; i < nElements; i++)
{
Vector3 pos = elements[i].transform.position;
if (pos.x < minX)
{
minX = pos.x;
}
if (pos.x > maxX)
{
maxX = pos.x;
}
if (pos.z < minY)
{
minY = pos.z;
}
if (pos.z > maxY)
{
maxY = pos.z;
}
}
for (int T = 0; T < 100; T++)
{
herds = new List <List <Prey> > ();
means = new List <Vector3> ();
curCluster = new int[elements.Count()];
for (int i = 0; i < nElements; i++)
{
curCluster[i] = i; // We assign each element to it's cluster. Making sure this way to at least
//run 1 time the assignment step
}
for (int i = 0; i < K; i++)
{
herds.Add(new List <Prey>());
means.Add(new Vector3(
(float)Random.Range((int)minX, (int)maxX),
15.0F,
(float)Random.Range((int)minY, (int)maxY)
)
);
}
float MSE = 0.0F;
float MSEPREV = 1.0F;
int epoch = 0;
while (Mathf.Abs(MSE - MSEPREV) > EPS && epoch < 100000) //Each time we find a change on the assignment
{
MSEPREV = MSE;
MSE = 0.0F;
epoch++;
//Empty the clusters
for (int i = 0; i < K; i++)
{
herds [i] = new List <Prey> ();
}
//We iterate to create the K clusters
for (int i = 0; i < nElements; i++) //For each element
{
int bestCluster = 0; // Keep the bestCluster
for (int j = 1; j < K; j++) // For each cluster
{
if (distance(elements [i].transform.position, means [j]) < distance(elements [i].transform.position, means [bestCluster])) //bestCluster
{
bestCluster = j;
}
}
if (curCluster [i] != bestCluster)
{
curCluster [i] = bestCluster; // Assign Cluster maps ;)
herds [bestCluster].Add(elements [i]); // Add the element to the cluster.
}
else
{
herds[curCluster[i]].Add(elements[i]);
}
}
// Now update the clusters
for (int i = 0; i < K; i++)
{
float newX = 0;
float newY = 0;
float newZ = 0;
for (int j = 0; j < herds[i].Count; j++)
{
newX += herds [i] [j].transform.position.x;
newY += herds [i] [j].transform.position.y;
newZ += herds [i] [j].transform.position.z;
}
int tot = herds [i].Count == 0 ? 1 : herds [i].Count; //Avoid division by 0;
means[i] = new Vector3(newX / (float)tot, means[i].y, newZ / (float)tot);
for (int j = 0; j < herds[i].Count; j++)
{
MSE += distance(herds[i][j].transform.position, means[i]);
}
}
}
if (MSE < bestMSE)
{
bestMSE = MSE;
best = herds;
for (int i = 0; i < nElements; i++)
{
elements[i].herdid = curCluster[i];
}
}
}
herds = best;
//Finally we iterate each of the final clusters and select the leader of the elements.
for (int i = 0; i < K; i++)
{
if (herds[i].Count == 0)
{
continue;
}
Prey element = herds[i][0];
List <GameObject> herdList = new List <GameObject>();
for (int j = 0; j < herds[i].Count; j++)
{
herdList.Add(herds[i][j].gameObject);
}
for (int j = 0; j < herds[i].Count; j++)
{
herds[i][j].herd = herdList;
}
element.getNewLeader(herds[i]);
}
}
