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 static string InitFileStructure(QTable Table)
{
string res = "[\n ";
for (int i = 0; i < Table.tableStructure.states; i++)
{
for (int j = 0; j < Table.tableStructure.actions; j++)
{
res += $"{Table.tableStructure.actionValues[i][j]}, ";
}
res += "\n ";
}
return(res + "\b\b]");
}
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);
}
static void Main(string[] args)
{
var actionList = new List <double[]>()
{
};
int states = 16;
int actions = 4;
for (int i = 0; i < states; i++)
{
actionList.Add(QTable.Zeros(actions));
}
table = new QTable(states, actions, new string[] { "0", "1", "2", "3", "4", "5", "6", "7", "8", "9", "10", "11", "12", "13", "14", "15" }, actionList); //Actions: UP, DOWN, RIGHT, LEFT
table.ReadExistingStructure();
Console.WriteLine(QTable.InitStructure(table));
//for (int repeater = 0; repeater < 10; repeater++)
//{
StartProcess();
//}
Console.WriteLine(QTable.InitStructure(table));
table.SaveQTableToFile();
Console.ReadLine();
}
