private QualityLearningState Train()
{
_learningState = QualityLearningState.Training;
for (int epoch = 0; epoch < maxEpochs; ++epoch)
{
_environmentMaze.SetAllStateToUncharted();
int currentState = Random.Next(0, _stateRewards.Length);
_environmentMaze.GoToState(currentState, _learningState);
do
{
int nextState = GetRandNextState(currentState, _environmentMatrix);
List <int> possibleActions = GetPossibleActions(nextState, _environmentMatrix);
double maxQ = MaxQ(_qMatrix, possibleActions, nextState);
_qMatrix[currentState][nextState] =
((1 - learningRate) * _qMatrix[currentState][nextState]) +
(learningRate * (_stateRewards[nextState] + (discountRate * maxQ)));
currentState = nextState;
_environmentMaze.GoToState(currentState, _learningState);
if (_stateRewards[currentState] < 50.0f)
{
break;
}
} while (currentState != goalState);
} // for
_learningState = QualityLearningState.Trained;
return(_learningState);
}
public void GoToState(int state, QualityLearningState learningState)
{
SpriteRenderer spriteRenderer = environmentMazeStates[state].GetComponent <SpriteRenderer>();
if (!spriteRenderer)
{
Debug.LogError($"SpriteRenderer for state: {state} is null!");
return;
}
spriteRenderer.color = learningState == QualityLearningState.Training ? trainingColor : trainedColor;
}
