public int PickMove(string state)
{
var random = new Random();
//Sometimes make a random move
if ((random.NextDouble() < ExploreRate) || (!QTable.ContainsKey(state)))
{
return(random.Next(4));
}
var currentOptionsQVals = QTable[state];
var validActionList = new List <int>();
var maxQVal = currentOptionsQVals.Max();
//Get a list of all optimal values
for (int i = 0; i < currentOptionsQVals.Count(); i++)
{
if (currentOptionsQVals[i] == maxQVal)
{
validActionList.Add(i);
}
}
;
//pick one at random
int action = validActionList[random.Next(validActionList.Count())];
return(action);
}
public void Learn(string state, int action, double reward, string nextState)
{
if (!QTable.ContainsKey(state))
{
QTable.Add(state, new List <double>()
{
0, 0, 0, 0
});
}
if (!QTable.ContainsKey(nextState))
{
QTable.Add(nextState, new List <double>()
{
0, 0, 0, 0
});
}
var currentQ = QTable[state][action];
var newQ = reward + DiscountFactor * QTable[nextState].Max();
QTable[state][action] += LearningRate * (newQ - currentQ);
}