static void runMarkovProblemSovler()
{
MarkovProblemSolver qlearner = new QLearningAgent(10000, 10000, 1);
qlearner.accept(new GridProblem());
qlearner.solve();
}
/*
* Allow to analyze effectiveness by saving a magnets lap times to a text file.
*
* Allow learning agents to increase their age.
*/
private void saveLapTimeData()
{
if (saveLapDataFileType != null)
{
// TODO: fix
DateTime lapTime = System.DateTime.Now;
TimeSpan lapse = lapTime - lastLapTime;
lastLapTime = lapTime;
// Use for time System.DateTime.Now;
StreamWriter file = new StreamWriter(fileName, true);
file.WriteLine("Lap : " + _lap + " Time: " + lapse.TotalSeconds);
file.Close();
}
try{
//TODO: Put any other learning agents aging function calls here in the future
QLearningAgent agent = GetComponent <QLearningAgent>();
if (agent != null)
{
agent.increaseMaturity();
}
} catch (Exception e) {
// Do nothing if no learning agent is present
}
}
public void testFirstStepsOfQLAAgentWhenFirstStepTerminates()
{
QLearningAgent <CellWorldPosition, String> qla = new QLearningAgent <CellWorldPosition, String>(
fourByThree);
CellWorldPosition startingPosition = new CellWorldPosition(1, 4);
String action = qla.decideAction(new MDPPerception <CellWorldPosition>(
startingPosition, -0.04));
Assert.AreEqual(CellWorld.LEFT, action);
Randomizer betweenEightyANdNinetyPercent = new MockRandomizer(
new double[] { 0.85 }); // to force left to become an "up"
qla.execute(action, betweenEightyANdNinetyPercent);
Assert.AreEqual(new CellWorldPosition(2, 4), qla.getCurrentState());
Assert.AreEqual(-1.0, qla.getCurrentReward(), 0.001);
Assert.AreEqual(0.0, qla.getQTable().getQValue(startingPosition,
action), 0.001);
String action2 = qla.decideAction(new MDPPerception <CellWorldPosition>(
new CellWorldPosition(2, 4), -1));
Assert.IsNull(action2);
Assert.AreEqual(-1.0, qla.getQTable().getQValue(startingPosition,
action), 0.001);
}
public void testFirstStepsOfQLAAgentUnderNormalProbability()
{
QLearningAgent <CellWorldPosition, String> qla = new QLearningAgent <CellWorldPosition, String>(
fourByThree);
Randomizer alwaysLessThanEightyPercent = new MockRandomizer(
new double[] { 0.7 });
CellWorldPosition startingPosition = new CellWorldPosition(1, 4);
String action = qla.decideAction(new MDPPerception <CellWorldPosition>(
startingPosition, -0.04));
Assert.AreEqual(CellWorld.LEFT, action);
Assert.AreEqual(0.0, qla.getQTable().getQValue(startingPosition,
action), 0.001);
qla.execute(action, alwaysLessThanEightyPercent);
Assert.AreEqual(new CellWorldPosition(1, 3), qla.getCurrentState());
Assert.AreEqual(-0.04, qla.getCurrentReward(), 0.001);
Assert.AreEqual(0.0, qla.getQTable().getQValue(startingPosition,
action), 0.001);
String action2 = qla.decideAction(new MDPPerception <CellWorldPosition>(
new CellWorldPosition(1, 3), -0.04));
Assert.AreEqual(-0.04, qla.getQTable().getQValue(startingPosition,
action), 0.001);
}
static void qLearningAgentDemo()
{
CellWorld <double> cw = CellWorldFactory.CreateCellWorldForFig17_1();
CellWorldEnvironment cwe = new CellWorldEnvironment(
cw.GetCellAt(1, 1),
cw.GetCells(),
MDPFactory.createTransitionProbabilityFunctionForFigure17_1(cw),
CommonFactory.CreateRandom());
QLearningAgent <Cell <double>, CellWorldAction> qla = new QLearningAgent <Cell <double>, CellWorldAction>(
MDPFactory.createActionsFunctionForFigure17_1(cw),
CellWorldAction.None, 0.2, 1.0, 5,
2.0);
cwe.AddAgent(qla);
output_utility_learning_rates(qla, 20, 10000, 500, 20);
}
public void testQLearningAgent()
{
QLearningAgent <CellWorldPosition, String> qla = new QLearningAgent <CellWorldPosition, string>(
fourByThree);
Randomizer r = new MockRandomizer(new double[] { 0.1, 0.9, 0.2, 0.8,
0.3, 0.7, 0.4, 0.6, 0.5 });
// Randomizer r = new JavaRandomizer();
Dictionary <Pair <CellWorldPosition, String>, Double> q = null;
QTable <CellWorldPosition, String> qTable = null;
for (int i = 0; i < 100; i++)
{
qla.executeTrial(r);
q = qla.getQ();
qTable = qla.getQTable();
}
// qTable.normalize();
// System.Console.WriteLine(qTable);
// System.Console.WriteLine(qTable.getPolicy());
}
/// <summary>
/// Restarts the learning process with a new walking condition.
/// </summary>
public void BeginNewGame()
{
sendAction = 0;
SetUp();
agent = new QLearningAgent();
// if we continue a previous training
if (this.loadExistingQTable)
{
Debug.Log(this.loadExistingQTable);
#if UNITY_EDITOR
((QLearningAgent)agent).LoadQTable("Assets/Resources/qtable.txt", envParameters.state_size, envParameters.action_size);
#endif
((QLearningAgent)agent).action_size = envParameters.action_size;
}
else
{
agent.SendParameters(envParameters);
}
Reset();
}
private void resetModel()
{
_agent = new QLearningAgent();
trackBarAlpha.Value = 2;
trackBarGamma.Value = 8;
trackBarEpsilon.Value = 1;
_agent.alpha = 0.2;
_agent.gamma = 0.8;
_agent.epsilon = 0.05;
// regiser state-qvalue
foreach (var userHp in Enumerable.Range(0, _maxUserHp + 1))
{
foreach (var userPos in Enumerable.Range(0, _maxUserPos + 1))
{
foreach (var towerHp in Enumerable.Range(0, _maxTowerHp + 1))
{
var state = new GameState(userHp, userPos, towerHp);
var compactState = new CompactGameState(userHp, userPos, towerHp);
foreach (var action in state.GetActionSet())
{
if (checkBoxCompact.Checked)
{
_agent.registerStateQValue(compactState, action, 0);
}
else
{
_agent.registerStateQValue(state, action, 0);
}
}
}
}
}
}
public abstract void OnAgentExited(QLearningAgent agent);
public abstract void OnAgentEntered(QLearningAgent agent);
static void AbsoluteMazeProblem()
{
//Declare random
MyRandom.RandomPool.Declare("action",(int)DateTime.Now.Ticks);
//環境
MazeEnvironment mazeEnv = new MazeEnvironment();
mazeEnv.Map = new MazeMap(new int[,]{
{ 1, 0, 0, 0, 0 },
{ 0, 0, 0, 0, 0 },
{ 0, 0, 0, 0, 0 },
{ 0, 0, 0, 0, 3 },
{ 0, 0, 3, 0, 2 },
}) ;
//エージェントを宣言
var mazeAgent
= new QLearningAgent<MazeEnvironment,PositionState,MoveAction>();
//= new ActorCritic<MazeEnvironment,PositionState,MoveAction>();
//エージェントに環境をセット
mazeAgent.Environment = mazeEnv;
//環境にエージェントをセット
mazeEnv.AddAgent(mazeAgent);
mazeAgent.Id = 1;
//エージェントを宣言
var mazeAgent2
//= new QLearningAgent<MazeEnvironment,PositionState,FourDirectionAction>();
= new ActorCritic<MazeEnvironment,PositionState,MoveAction>();
//エージェントに環境をセット
mazeAgent2.Environment = mazeEnv;
//環境にエージェントをセット
mazeEnv.AddAgent(mazeAgent2);
mazeAgent2.Id = 2;
//初期状態をセット
mazeAgent2.CurrentState = mazeEnv.StartState;
//初期状態をセット
mazeAgent.CurrentState = mazeEnv.StartState;
//View のセット
MazeView mazeView = new MazeView();
mazeView.Maze = mazeEnv;
while(true)
{
mazeAgent.Act();
mazeAgent2.Act();
mazeView.show();
Console.WriteLine();
Console.WriteLine();
Console.WriteLine();
mazeAgent.Show();
Console.WriteLine();
mazeAgent2.Show();
string command = Console.ReadLine(); //を、キーをおすたびにやる。
if (command.Equals("init"))
{
mazeAgent.Init();
mazeAgent2.Init();
}
else if (command.Equals("q"))
{
mazeAgent.Show ();
continue;
}
try //数字としてパースしてみて
{ //パースできたらその回数ステップ進める
int num = int.Parse(command);
foreach (int j in Enumerable.Range(0, num))
{
mazeAgent.Act();
mazeAgent2.Act();
}
}
catch (Exception)
{
}
}
}
/// <summary>
/// エージェントを設定する
/// </summary>
private void SetAgent()
{
EnvironmentMaze env = new EnvironmentMaze(EnvMaze, IndexData.GoalReword);
Agent = new QLearningAgent(env, IndexData.Alpha, IndexData.Gamma, IndexData.Epsilon);
}
public void testFirstStepsOfQLAAgentWhenFirstStepTerminates()
{
QLearningAgent<CellWorldPosition, String> qla = new QLearningAgent<CellWorldPosition, String>(
fourByThree);
CellWorldPosition startingPosition = new CellWorldPosition(1, 4);
String action = qla.decideAction(new MDPPerception<CellWorldPosition>(
startingPosition, -0.04));
Assert.AreEqual(CellWorld.LEFT, action);
Randomizer betweenEightyANdNinetyPercent = new MockRandomizer(
new double[] { 0.85 }); // to force left to become an "up"
qla.execute(action, betweenEightyANdNinetyPercent);
Assert.AreEqual(new CellWorldPosition(2, 4), qla.getCurrentState());
Assert.AreEqual(-1.0, qla.getCurrentReward(), 0.001);
Assert.AreEqual(0.0, qla.getQTable().getQValue(startingPosition,
action), 0.001);
String action2 = qla.decideAction(new MDPPerception<CellWorldPosition>(
new CellWorldPosition(2, 4), -1));
Assert.IsNull(action2);
Assert.AreEqual(-1.0, qla.getQTable().getQValue(startingPosition,
action), 0.001);
}
public void testFirstStepsOfQLAAgentUnderNormalProbability()
{
QLearningAgent<CellWorldPosition, String> qla = new QLearningAgent<CellWorldPosition, String>(
fourByThree);
Randomizer alwaysLessThanEightyPercent = new MockRandomizer(
new double[] { 0.7 });
CellWorldPosition startingPosition = new CellWorldPosition(1, 4);
String action = qla.decideAction(new MDPPerception<CellWorldPosition>(
startingPosition, -0.04));
Assert.AreEqual(CellWorld.LEFT, action);
Assert.AreEqual(0.0, qla.getQTable().getQValue(startingPosition,
action), 0.001);
qla.execute(action, alwaysLessThanEightyPercent);
Assert.AreEqual(new CellWorldPosition(1, 3), qla.getCurrentState());
Assert.AreEqual(-0.04, qla.getCurrentReward(), 0.001);
Assert.AreEqual(0.0, qla.getQTable().getQValue(startingPosition,
action), 0.001);
String action2 = qla.decideAction(new MDPPerception<CellWorldPosition>(
new CellWorldPosition(1, 3), -0.04));
Assert.AreEqual(-0.04, qla.getQTable().getQValue(startingPosition,
action), 0.001);
}
public void testQLearningAgent()
{
QLearningAgent<CellWorldPosition, String> qla = new QLearningAgent<CellWorldPosition,string>(
fourByThree);
Randomizer r = new MockRandomizer(new double[] { 0.1, 0.9, 0.2, 0.8,
0.3, 0.7, 0.4, 0.6, 0.5 });
// Randomizer r = new JavaRandomizer();
Dictionary<Pair<CellWorldPosition, String>, Double> q = null;
QTable<CellWorldPosition, String> qTable = null;
for (int i = 0; i < 100; i++)
{
qla.executeTrial(r);
q = qla.getQ();
qTable = qla.getQTable();
}
// qTable.normalize();
// System.Console.WriteLine(qTable);
// System.Console.WriteLine(qTable.getPolicy());
}