public void mixWeights(AnimalBrain parent, float chance) //jaka jest szansa na zmiane genow
{
for (int i = 0; i < hiddenLayerSize; i++)
{
if (roll(chance))
{
hiddenLayerBias[i] = parent.hiddenLayerBias[i];
}
if (roll(chance))
{
secondHiddenLayerBias[i] = parent.secondHiddenLayerBias[i];
}
for (int j = 0; j < hiddenLayerSize; j++)
{
if (roll(chance))
{
hiddenLayerSynapsesWeights[i, j] = parent.hiddenLayerSynapsesWeights[i, j];
}
}
for (int j = 0; j < input.Length; j++)
{
if (roll(chance))
{
inputSynapsesWeights[i, j] = parent.inputSynapsesWeights[i, j];
}
}
for (int j = 0; j < output.Length; j++)
{
if (roll(chance))
{
outputSynapsesWeights[i, j] = parent.outputSynapsesWeights[i, j];
}
}
}
}
public void DeepCopyFrom(AnimalBrain source)
{
// inputSynapsesWeights = source.inputSynapsesWeights.Clone() as float[,];
// outputSynapsesWeights = source.outputSynapsesWeights.Clone() as float[,];
// hiddenLayerSynapsesWeights = source.hiddenLayerSynapsesWeights.Clone() as float[,];
// secondHiddenLayerBias = source.secondHiddenLayerBias.Clone() as float[];
// hiddenLayerBias = source.hiddenLayerBias.Clone() as float[];
inputSynapsesWeights.CopyFrom(source.inputSynapsesWeights);
outputSynapsesWeights.CopyFrom(source.outputSynapsesWeights);
secondHiddenLayerBias.CopyFrom(source.secondHiddenLayerBias);
hiddenLayerSynapsesWeights.CopyFrom(source.hiddenLayerSynapsesWeights);
hiddenLayerBias.CopyFrom(source.hiddenLayerBias);
}
public void InitializeMe()
{
sensory = gameObject.GetComponent<AnimalSensory>();
body = gameObject.GetComponent<AnimalBody>();
brain = gameObject.GetComponent<AnimalBrain>();
stateMachine = gameObject.GetComponent<AnimalStateMachine>();
possesor = GameObject.FindGameObjectWithTag("Player").GetComponent<PhantomController>();
moveSpeed = body.moveSpeed;
isInitialized = true;
if(stateMachine.myType == AnimalType.prairieDog)
jumpController = gameObject.GetComponent<JumpController>();
}
public AnimalData Mate()
{
//krzyzowanie polega na "konkursie" wynialajacym dwoch rodzicow na jednego osobnika potomnego
//osobniki o wiekszym fitness maja wieksza szanse na zostanie wybranym do krzyzowania
//losowana jest wartosc pomiedzy minimalna wartoscia z fitness list a maksymalna wartoscia z fitness list
//losowany index osobnika dopoki wylosowana wczesniej wartosc nie jest mniejsza od wartosci fitness danego indexu
//losowani dwaj rodzice, rozmnazanie jak wczesniej
float crossoverPower = 0.5f;
AnimalData childData = new AnimalData();
int parent1Index = ChooseParent();
float parent1MGene = _currentGeneration[parent1Index].animalData.mGene;
childData.mGene = parent1MGene;
int parent2Index = ChooseParent();
float chance;
int currentIteration = 0;
while (true)
{
parent2Index = ChooseParent();
float parent2MGene = _currentGeneration[parent2Index].animalData.mGene; //jeden z rodzicow przekazuje mGene
currentIteration++;
if (parent1MGene != parent2MGene) //jesli to ten sam index to taki sam mGene, dlatego nie trzeba tego sprawdzac
{
break;
}
else if (currentIteration >= _matingMaxIterations)
{
if (parent1Index != parent2Index)
{
break;
}
}
}
AnimalBrain childBrain = new AnimalBrain();
childBrain.DeepCopyFrom(_currentGeneration[parent1Index].animalData.animalBrain);
childData.partsMass = MixDictionaries <float>(_currentGeneration[parent1Index].animalData.partsMass, _currentGeneration[parent2Index].animalData.partsMass, crossoverPower);
childData.partsScaleMultiplier = MixDictionaries <System.Numerics.Vector3>(_currentGeneration[parent1Index].animalData.partsScaleMultiplier, _currentGeneration[parent2Index].animalData.partsScaleMultiplier, crossoverPower);
childData.targetJointsVelocity = MixDictionaries <System.Numerics.Vector3>(_currentGeneration[parent1Index].animalData.targetJointsVelocity, _currentGeneration[parent2Index].animalData.targetJointsVelocity, crossoverPower);
childData.limbsPositionMultiplier = MixDictionaries <System.Numerics.Vector3>(_currentGeneration[parent1Index].animalData.limbsPositionMultiplier, _currentGeneration[parent2Index].animalData.limbsPositionMultiplier, crossoverPower);
childBrain.mixWeights(_currentGeneration[parent2Index].animalData.animalBrain, crossoverPower);
float mutationRate;
float physicalMutationRate;
if (_ifAdaptive)
{
physicalMutationRate = childData.physicalMutationRate;
mutationRate = childData.weightsMutationRate;
}
else
{
physicalMutationRate = _globalPhyscialMutationRate;
mutationRate = _globalMutationRate;
}
for (int i = 0; i < _numberOfWeightsMutations; i++)
{
chance = Random.Range(0.0f, 1.0f);
if (chance < mutationRate) //obsluga mutacji - mutacja obejmuje zmiane indexu genu z losowym innym genem
{
childBrain.MutateWeights();
}
}
chance = Random.Range(0.0f, 1.0f); //szanse na mutacje wag i mutacje fizycznych wlasciwosci sa inne
if (chance < physicalMutationRate)
{
childData.MutateData();
}
childData.animalBrain = childBrain;
return(childData);
}
void Start()
{
brain = gameObject.GetComponent<AnimalBrain>();
sensory = gameObject.GetComponent<AnimalSensory>();
}
