private void button1_Click_1(object sender, EventArgs e)
{
char test = 'a';
StateTree<string> tree = new StateTree<string>(EqualityComparer<string>.Default);
for (int i=0; i<26; i++)
{
char test2 = (char)(test + i);
tree.AddState(test2.ToString());
}
MessageBox.Show(String.Join(",",tree.GetLevel0Children("a", 3)));
MessageBox.Show(String.Join(",",tree.GetChildren("a", 3)));
}
public override double update(StateTransition <int[], actionType> transition)
{
//if (currentGoal.goalState == null)
//{
// Console.WriteLine("Goal: null");
//}
//else
//{
// Console.WriteLine("Goal: Level " + currentGoal.level + ", at " + String.Join(",", currentGoal.goalState) + ", value: " + currentGoal.value);
//}
//Console.WriteLine(" current state: " + String.Join(",", transition.newState) + " / " + String.Join(",", stateTree.GetParentState(transition.newState, currentGoal.level)));
// update the stateTree
stateTree.AddState(transition.newState);
// perform model updates
double returnValue = models[0].update(transition);
if (transition.absorbingStateReached) // if this was the end of the simulation // ****** do we need this?
{
for (int i = 1; i < models.Length; i++)
{
if (transitions[i].oldState != null)
{
//transitions[i].reward += transition.reward;
transitions[i].reward = Math.Max(transitions[i].reward, transition.reward);
models[i].update(transitions[i]);
transitions[i] = new StateTransition <int[], actionType>(null, default(actionType), double.NegativeInfinity, null);
}
}
currentGoal.goalState = null;
for (int i = 0; i < subgoals.Length; i++)
{
subgoals[i].Clear();
}
return(returnValue);
}
else
{
for (int i = 1; i < models.Length; i++)
{
// get the representation of state at this level
int[] thisOldState = stateTree.GetParentState(transition.oldState, i);
int[] thisNewState = stateTree.GetParentState(transition.newState, i);
// check if the local transition represents a transition at this level
if ((!stateComparer.Equals(thisOldState, thisNewState)) || i == 0)// transition - update the model at this level, and start a new accumulation of reward
{
if (transitions[i].oldState != null)
{
models[i].update(transitions[i]);
}
transitions[i] = new StateTransition <int[], actionType>(thisOldState, transition.action, transition.reward, thisNewState); // this line might only apply to grid-worlds
}
else
{
// accumulate the reward
//transitions[i].reward += transition.reward;
transitions[i].reward = Math.Max(transitions[i].reward, transition.reward);
}
}
}
if (currentGoal.goalState == null)
{
return(returnValue);
}
if (stateComparer.Equals(stateTree.GetParentState(transition.newState, currentGoal.level), currentGoal.goalState)) // if the current goal has been reached
{
for (int i = 0; i < subgoals.Length; i++)
{
subgoals[i].Clear();
}
if (currentGoal.level == 0) // if the current goal is at level 0, select a new goal from scratch
{
currentGoal.goalState = null;
}
else // otherwise do a contrained selection of a new goal
{
currentGoal = selectGoal(transition.newState, currentGoal.level, availableActions);
subgoals[currentGoal.level].Add(currentGoal);
}
}
else if (subgoals[0].Count == 0) // error check ****************
{
currentGoal.goalState = null;
for (int i = 0; i < subgoals.Length; i++)
{
subgoals[i].Clear();
}
}
else if (stateComparer.Equals(subgoals[0][0].goalState, transition.newState)) // if the next step at level 0 has been reached successfully
{
for (int i = 0; i < currentGoal.level; i++)
{
subgoals[i].RemoveAt(0);
if (subgoals[i].Count > 0)
{
break;
}
}
}
else if (minLevel > 0 && !stateComparer.Equals(stateTree.GetParentState(transition.newState, currentGoal.level), currentGoal.startState)) // if we're no longer in the start state at the goal level (for dH lesions)
{
currentGoal.goalState = null;
for (int i = 0; i < subgoals.Length; i++)
{
subgoals[i].Clear();
}
}
else if (!stateComparer.Equals(subgoals[0][0].goalState, transition.newState)) // if we're not where we expected to be at level 0
{
if (currentGoal.level == 0)
{
currentGoal.goalState = null;
for (int i = 0; i < subgoals.Length; i++)
{
subgoals[i].Clear();
}
}
else
{
subgoals[0].Clear();
//for (int i = 0; i <= currentGoal.level; i++)
//{
// subgoals[i].Clear();
//}
//subgoals[currentGoal.level].Add(currentGoal);
}
}
return(returnValue);
}
