void Start()
{
world = new World(mapSize, winDistance, winReward, stepReward, loseReward);
nextState = GetRandomState();
actions = new List <Action>();
for (int x = -1; x <= 1; x++)
{
for (int y = -1; y <= 1; y++)
{
if (x != 0 || y != 0)
{
Debug.Log("Action: (" + x + "," + y + ")");
actions.Add(new Action(new Vector2(x, y) / inverseDeltaTime, Vector2.zero));
}
}
}
learner = new QLearner(actions, hiddenSize, numInputs, seed);
GameObject goal = Instantiate(goalPrefab);
goal.transform.position = nextState.GetGoalPos();
goalPrefab = goal;
GameObject player = Instantiate(playerPrefab);
player.transform.position = nextState.GetPlayerPos();
playerPrefab = player;
background.transform.position = (Vector3)(Vector2.one * mapSize / 2) + Vector3.forward;
FindObjectOfType <Camera>().transform.position = (Vector3)(Vector2.one * mapSize / 2) + Vector3.forward * FindObjectOfType <Camera>().transform.position.z;
FindObjectOfType <Camera>().orthographicSize = mapSize / 2.0f;
}
public AIAgent(int player) : base(player)
{
playerNum = player;
opponentNum = playerNum == 1 ? 2 : 1;
// Create a new, default QLearner. Usually want to replace this value for more control
QLearner = new QLearner();
IsLearning = false;
// Set strategy
strategy = Strategy.StrategyWithType(playerNum, StrategyType.RunAway);
level1Searcher = new DiscreteAdversarialSearch(playerNum,
strategy.Level1Heuristic,
getFillerAction,
getNewPathIndex,
Level1StepSize,
4);
decisionTimer = 0;
level2Searcher = new AStar(Level2MaxNodesInPrioQueue, Level2MaxExpansions, strategy.Level2CostFunction,
strategy.Level2GoalFunction, strategy.Level2HeuristicFunction);
fillerAction = WorldAction.NoAction;
isFirstTime = true;
strategyTimer = MaxStrategyTime;
calculatePathNextFrame = false;
}
void Start()
{
world = new World(mapSize, winDistance, winReward, stepReward, loseReward);
nextState = GetRandomState();
actions = new List <Action>();
for (int x = -1; x <= 1; x++)
{
for (int y = -1; y <= 1; y++)
{
if (x != 0 || y != 0)
{
Debug.Log("Action: (" + x + "," + y + ")");
actions.Add(new Action(new Vector2(x, y) / inverseDeltaTime, Vector2.zero));
}
}
}
learner = new QLearner(eValue, alpha, gamma, eta, actions, hiddenSizes, numInputs, seed);
GameObject goal = Instantiate(goalPrefab);
goal.transform.position = nextState.GetGoalPos();
goalPrefab = goal;
GameObject player = Instantiate(playerPrefab);
player.transform.position = nextState.GetPlayerPos();
playerPrefab = player;
}
public void Start()
{
// Set up the world with the initial state
currentWorld = new World();
// Create QLearner object
qLearner = new QLearner(alpha: 0.3f, epsilon: 0.15f, discount: 0.66f);
qLearner.OpenSavedData();
// Create 2 ai agents
agentList = new List <IAgent>();
AIAgent ai1 = new AIAgent(1);
ai1.IsLearning = true;
ai1.QLearner = qLearner;
AIAgent ai2 = new AIAgent(2);
ai2.IsLearning = true;
ai2.QLearner = qLearner;
agentList.Add(ai1);
agentList.Add(ai2);
// Specify number of games
numberOfGames = 10;
// Specify which iteration of games we are on
gameIteration = 1;
gameFrames = 0;
Debug.Log("Beginning learning. Simulating " + numberOfGames.ToString() + " games before writing.");
}
// First-time setup
void Start()
{
// Set up the world with the initial state
currentWorld = new RenderedWorld(this);
currentWorld.Display();
Gui.SetMode(0);
restartTimer = -1;
winner = 0;
agentList = new List <IAgent>();
#if AIEnabled
// Create q-learning object for the AI, and pick up learning where we left off
qLearner = new QLearner(alpha: 0.3f, epsilon: 0.15f, discount: 0.66f);
qLearner.OpenSavedData();
#endif
#if AIP1
// Create AI for player 1
AIAgent player1Ai = new AIAgent(1);
player1Ai.ResourceScript = this; // For debug rendering
player1Ai.QLearner = qLearner;
player1Ai.IsLearning = true;
agentList.Add(player1Ai);
#else
// Create the human agent
agentList.Add(new WASDFAgent(1));
#endif
#if AIP2
// Create AI for player 1
AIAgent player2Ai = new AIAgent(2);
player2Ai.ResourceScript = this; // For debug rendering
player2Ai.QLearner = qLearner;
player2Ai.IsLearning = true;
agentList.Add(player2Ai);
#else
// Create the human agent
agentList.Add(new ArrowsAgent(2));
#endif
}
void Start()
{
world = new World(mapSize);
List <Action> actions = new List <Action>();
actions.Add(new Action(Vector2.up));
actions.Add(new Action(Vector2.down));
actions.Add(new Action(Vector2.left));
actions.Add(new Action(Vector2.right));
List <int> tableDimensions = new List <int>();
tableDimensions.Add(mapSize);
tableDimensions.Add(mapSize);
tableDimensions.Add(mapSize);
//tableDimensions.Add(mapSize);
tableDimensions.Add(actions.Count);
learner = new QLearner(actions, tableDimensions, startingQ);
nextState = GetRandomState();
playerPrefab = Instantiate(playerPrefab);
counter = Time.time;
background.transform.position = Vector2.one * mapSize / 2;
FindObjectOfType <Camera>().transform.position = (Vector3)(Vector2.one * mapSize / 2) + Vector3.forward * FindObjectOfType <Camera>().transform.position.z;
FindObjectOfType <Camera>().orthographicSize = mapSize / 2.0f;
}
/// <summary>
/// Basic initialisation.
/// </summary>
void Start()
{
QLearner.SetParamReceiver(this);
label = this.GetComponent <Text>();
}
private void Start()
{
_qLearner = GetComponent <QLearner>();
_qLearner.OnStep += Step;
_qLearner.OnEpisodeComplete += EpisodeComplete;
}
public override void Initialise(GameState state)
{
Globals.state = state;
Globals.QLearnerFolder = FindQLearnerFolder();
Globals.random = new Random(state.PlayerSeed);
Globals.pathFinder = new Pathfinder(state.Width, state.Height);
Globals.enemyInfluence = new InfluenceMap(state.map);
Globals.friendlyInfluence = new InfluenceMap(state.map);
Globals.hillInfluence = new InfluenceMap(state.map);
Globals.foodInfluence = new InfluenceMap(state.map);
string backupFile = Globals.QLearnerFolder + "\\" + "Qbackup.Q";
string originalFile = Globals.QLearnerFolder + "\\" + "QData.Q";
if(File.Exists(originalFile))//previously run, there is learning data
{
BinaryFormatter formatter = new BinaryFormatter();
Stream learnerStream = new FileStream(originalFile, FileMode.Open);
Learner = (QLearner)formatter.Deserialize(learnerStream);
learnerStream.Close();
if (File.Exists(backupFile))
{
FileInfo originalInfo = new FileInfo(originalFile);
FileInfo backupInfo = new FileInfo(backupFile);
if (originalInfo.Length > backupInfo.Length)//the more the bots learn the bigger this should get, if it is suddenly smaller there has been a serialisation error, do not backup
System.IO.File.Copy(originalFile, backupFile, true);//in case of deserialisation error, your progress is not lost
}
else
{
System.IO.File.Copy(originalFile, backupFile, true);//in case of deserialisation error, your progress is not lost
}
Learner.GamesPlayed++;
}
else
{
Learner = new QLearner(0.9f, 0.8f, 0.9f, state.PlayerSeed);//very first game, no learning done previously
}
}
