public void SelectParents(IEnumerable <Genotype> pop, out Genotype parent1, out Genotype parent2)
{
parent1 = SelectParent(pop);
var p1Tmp = parent1;
parent2 = SelectParent(pop.Where(x => x != p1Tmp));
}
public void SetGenotype(Genotype g)
{
m_Genotype = g;
body.transform.localScale = new Vector3(g.bodySizeX, g.bodySizeY, 1);
fin.transform.localScale = new Vector3(g.finSizeX, g.finSizeY, 1);
var motor = joint.motor;
motor.motorSpeed = g.motorForce;
joint.motor = motor;
}
public void SetGenotype(Genotype g)
{
this.genotype = g;
// ... use genotype to set parameters
body.transform.localScale = new Vector3(g.bodySizeX, g.bodySizeY, 1);
fin.transform.localScale = new Vector3(g.finSizeX, g.finSizeY, 1);
var motor = joint.motor;
motor.motorSpeed = g.motorForce;
joint.motor = motor;
}
private void Start()
{
// Create initial population
List <Genotype> initialGenotypes = new List <Genotype>();
for (int i = 0; i < populationSize; i++)
{
Genotype g = new Genotype();
g.bodySizeX = Random.Range(1, 10);
g.bodySizeY = Random.Range(1, 10);
g.finSizeX = Random.Range(1, 10);
g.finSizeY = Random.Range(1, 10);
g.motorForce = Random.Range(0, 5000);
initialGenotypes.Add(g);
}
StartCoroutine(GeneticCO(initialGenotypes));
}
IEnumerator GeneticCO(List <Genotype> initialGenotypes)
{
List <Genotype> currentGenotypes = initialGenotypes;
// Do N generations
for (int i = 0; i < nGenerations; i++)
{
// Spawn fishs
fishSpawner.SpawnPopulation(currentGenotypes);
// ... wait for simulation
yield return(new WaitForSeconds(3.0f));
fishSpawner.EndSimulation();
// Selection
currentGenotypes.Sort((x, y) => (int)((y.fitness - x.fitness) * 100));
List <Genotype> selectedGenotypes = currentGenotypes.GetRange(0, (int)(0.3f * populationSize));
// Crossover
List <Genotype> crossoverGenotypes = new List <Genotype>();
for (int ci = 0; ci < selectedGenotypes.Count / 2; ci++)
{
Genotype p1 = selectedGenotypes[ci * 2];
Genotype p2 = selectedGenotypes[ci * 2 + 1];
Genotype c1 = new Genotype();
c1.bodySizeX = Random.value > 0.5f ? p1.bodySizeX : p2.bodySizeX;
c1.bodySizeY = Random.value > 0.5f ? p1.bodySizeY : p2.bodySizeY;
c1.finSizeX = Random.value > 0.5f ? p1.finSizeX : p2.finSizeX;
c1.finSizeY = Random.value > 0.5f ? p1.finSizeY : p2.finSizeY;
c1.motorForce = Random.value > 0.5f ? p1.motorForce : p2.motorForce;
Genotype c2 = new Genotype();
c2.bodySizeX = Random.value > 0.5f ? p1.bodySizeX : p2.bodySizeX;
c2.bodySizeY = Random.value > 0.5f ? p1.bodySizeY : p2.bodySizeY;
c2.finSizeX = Random.value > 0.5f ? p1.finSizeX : p2.finSizeX;
c2.finSizeY = Random.value > 0.5f ? p1.finSizeY : p2.finSizeY;
c2.motorForce = Random.value > 0.5f ? p1.motorForce : p2.motorForce;
crossoverGenotypes.Add(c1);
crossoverGenotypes.Add(c2);
}
List <Genotype> mutatedGenotypes = new List <Genotype>();
int nMutated = populationSize - selectedGenotypes.Count - crossoverGenotypes.Count;
for (int mi = 0; mi < nMutated; mi++)
{
Genotype p = crossoverGenotypes[Random.Range(0, crossoverGenotypes.Count)];
Genotype m = new Genotype();
m.bodySizeX = p.bodySizeX;
m.bodySizeY = p.bodySizeY;
m.finSizeX = p.finSizeX;
m.finSizeY = p.finSizeY;
m.motorForce = p.motorForce;
int choice = Random.Range(0, 5);
switch (choice)
{
case 0:
m.bodySizeX = Random.Range(1, 10);
break;
case 1:
m.bodySizeY = Random.Range(1, 10);
break;
case 2:
m.finSizeX = Random.Range(1, 10);
break;
case 3:
m.finSizeY = Random.Range(1, 10);
break;
case 4:
m.motorForce = Random.Range(0, 5000);
break;
}
mutatedGenotypes.Add(m);
}
currentGenotypes = new List <Genotype>();
currentGenotypes.AddRange(selectedGenotypes);
currentGenotypes.AddRange(crossoverGenotypes);
currentGenotypes.AddRange(mutatedGenotypes);
Debug.Log("Generation " + i + " - Best Fitness: " + selectedGenotypes[0].fitness + " - Second best Fitness: " + selectedGenotypes[1].fitness);
SetBestIndividualMarker(selectedGenotypes[0].fitness);
}
}
IEnumerator GeneticCO(List <Genotype> initialGenotypes)
{
List <Genotype> currentGenotypes = initialGenotypes;
for (int i = 0; i < nGenerations; i++)
{
fishSpawner.SpawnPopulation(currentGenotypes);
yield return(new WaitForSeconds(generationLT));
fishSpawner.EndSimulation();
// Selection
currentGenotypes.Sort((x, y) => (int)(100 * y.fitness - 100 * x.fitness));
List <Genotype> selectedGenotypes = currentGenotypes.GetRange(0, (int)(0.3f * populationSize));
// crossover
List <Genotype> crossoverGenotypes = new List <Genotype>();
for (int ci = 0; ci < selectedGenotypes.Count / 2; ci++)
{
Genotype p1 = selectedGenotypes[ci * 2];
Genotype p2 = selectedGenotypes[(ci * 2) + 1];
Genotype c1 = new Genotype();
c1.bodySizeX = Random.value > 0.5 ? p1.bodySizeX : p2.bodySizeX;
c1.bodySizeY = Random.value > 0.5 ? p1.bodySizeY : p2.bodySizeY;
c1.finSizeX = Random.value > 0.5 ? p1.finSizeX : p2.finSizeX;
c1.finSizeY = Random.value > 0.5 ? p1.finSizeY : p2.finSizeY;
c1.motorForce = Random.value > 0.5 ? p1.motorForce : p2.motorForce;
Genotype c2 = new Genotype();
c2.bodySizeX = Random.value > 0.5 ? p1.bodySizeX : p2.bodySizeX;
c2.bodySizeY = Random.value > 0.5 ? p1.bodySizeY : p2.bodySizeY;
c2.finSizeX = Random.value > 0.5 ? p1.finSizeX : p2.finSizeX;
c2.finSizeY = Random.value > 0.5 ? p1.finSizeY : p2.finSizeY;
c2.motorForce = Random.value > 0.5 ? p1.motorForce : p2.motorForce;
crossoverGenotypes.Add(c1);
crossoverGenotypes.Add(c2);
}
// mutation
List <Genotype> mutatedGenotypes = new List <Genotype>();
int nMutated = populationSize - selectedGenotypes.Count - crossoverGenotypes.Count;
for (int mi = 0; mi < nMutated; mi++)
{
Genotype p = crossoverGenotypes[Random.Range(0, crossoverGenotypes.Count)];
Genotype m = new Genotype();
m.bodySizeX = p.bodySizeX;
m.bodySizeY = p.bodySizeY;
m.finSizeX = p.finSizeX;
m.finSizeY = p.finSizeY;
m.motorForce = p.motorForce;
int choice = Random.Range(0, 5);
switch (choice)
{
case 0:
m.bodySizeX = Random.Range(1, 20);
break;
case 1:
m.bodySizeY = Random.Range(1, 20);
break;
case 2:
m.finSizeX = Random.Range(1, 20);
break;
case 3:
m.finSizeY = Random.Range(1, 20);
break;
case 4:
m.motorForce = Random.Range(0, 5000);
break;
default:
break;
}
mutatedGenotypes.Add(m);
}
currentGenotypes = new List <Genotype>();
currentGenotypes.AddRange(selectedGenotypes);
currentGenotypes.AddRange(crossoverGenotypes);
currentGenotypes.AddRange(mutatedGenotypes);
Debug.Log("Generation " + i + " best fitness = " + selectedGenotypes[0].fitness);
}
}
