virtual protected Reward Playout(IWorldModel initialPlayoutState)
{
IWorldModel prevState = initialPlayoutState.GenerateChildWorldModel();
CurrentDepth = 0;
//Perform n playouts for each state [to deal with stochastic nature]
while (!prevState.IsTerminal() && CurrentDepth < MaxPlayoutDepthAllowed)
{
GOB.Action[] actions = prevState.GetExecutableActions();
int randomAction = RandomGenerator.Next(actions.Length);
prevState = StochasticPlayout(actions[randomAction], prevState, MaxPlayoutSimulations);
prevState.CalculateNextPlayer();
CurrentDepth++;
}
Reward reward = new Reward(prevState, prevState.GetNextPlayer());
return(reward);
}
protected override Reward Playout(IWorldModel initialPlayoutState)
{
IWorldModel state = initialPlayoutState.GenerateChildWorldModel();
CurrentDepth = 0;
while (!state.IsTerminal() && CurrentDepth < MaxPlayoutDepthAllowed)
{
List <KeyValuePair <int, GOB.Action> > actions = new List <KeyValuePair <int, GOB.Action> >();
foreach (GOB.Action action in state.GetExecutableActions())
{
actions.Add(new KeyValuePair <int, GOB.Action>((int)action.GetHValue(state), action));
}
actions.Sort(
delegate(KeyValuePair <int, GOB.Action> p1, KeyValuePair <int, GOB.Action> p2)
{
return(p1.Key.CompareTo(p2.Key));
}
);
if (actions.Count == 0)
{
break;
}
int randomValue = this.RandomGenerator.Next((int)actions[actions.Count - 1].Key);
foreach (KeyValuePair <int, GOB.Action> pair in actions)
{
if (pair.Key > randomValue)
{
state = StochasticPlayout(pair.Value, state, MaxPlayoutSimulations);
pair.Value.ApplyActionEffects(state);
break;
}
}
state.CalculateNextPlayer();
CurrentDepth++;
}
Reward reward = new Reward(state, state.GetNextPlayer());
return(reward);
}