int controls_evaluate(DateTime currentTime, DateTime elapsedTime, double tStep)
//
// Input: currentTime = current simulation date/time
// elapsedTime = decimal days since start of simulation
// tStep = simulation time step (days) //(5.0.013 - LR)
// Output: returns number of new actions taken
// Purpose: evaluates all control rules at current time of the simulation.
//
{
int r; // control rule index
int result; // TRUE if rule premises satisfied
struct TPremise* p; // pointer to rule premise clause
struct TAction* a; // pointer to rule action clause
DateTime theDate = floor(currentTime);
DateTime theTime = currentTime - floor(currentTime);
// --- evaluate each rule
if ( RuleCount == 0 ) return 0;
clearActionList();
for (r=0; r<RuleCount; r++)
{
// --- evaluate rule's premises
result = TRUE;
p = Rules[r].firstPremise;
while (p)
{
if ( p->type == r_OR )
{
if ( result == FALSE )
result = evaluatePremise(p, theDate, theTime,
elapsedTime, tStep);
}
else
{
if ( result == FALSE ) break;
result = evaluatePremise(p, theDate, theTime,
elapsedTime, tStep);
}
p = p->next;
}
// --- if premises true, add THEN clauses to action list
// else add ELSE clauses to action list
if ( result == TRUE ) a = Rules[r].thenActions;
else a = Rules[r].elseActions;
while (a)
{
updateActionValue(a, currentTime, tStep); //(5.0.012 - LR)
updateActionList(a);
a = a->next;
}
}
// --- execute actions on action list
if ( ActionList ) return executeActionList(currentTime);
else return 0;
}
Action RolloutAgent::getAction(Tetris *board)
{
vector<Action > bestActions;
float bestValue = -99999999;
//For all valid rotations
for (int r = 0; r < NUM_ROTATIONS; r++) {
Rotation rot = (Rotation) r;
//Make sure that we only check valid columns
int maxColumn = board->highestValidColWithRot(rot) + 1;
//For all valid columns for each rotation
for (int col = 0; col < maxColumn; col++) {
//Create the action for this move (will be cleaned up by playing it)
float actVal = 0;
//TODO: Look at all possible next pieces instead of random sampling
for (int w = 0; w < W; w++) {
Action a = Action(rot, col);
//Copy the board -- this makes it have random next pieces
Tetris *trajectorySim = board->gameCopy();
// trajectorySim->playAction(a, false);
actVal += heurAgent->valueOfActionOnBoard(a, trajectorySim);
for (int k = 0; k < K; k++) {
// trajectorySim->setPiece(1);
// trajectorySim->printBoard();
Action heuristicAct = heurAgent->getAction(trajectorySim);
// trajectorySim->playAction(heuristicAct, false);
actVal += pow(GAMMA, k + 1) * heurAgent->valueOfActionOnBoard(heuristicAct, trajectorySim);
}
// actVal += heurAgent->valueBetweenBoards(board, trajectorySim);
// trajectorySim->printBoard();
delete trajectorySim;
}
//Average over the W K-length trajectories
actVal /= (float)W;
// actionSim->printBoard();
// cout << "PLAY WITH ROW: " << r << " AND COL: " << col << endl;
// cout << "SCORE: " << actVal << endl << endl;
// exit(1);
if (actVal > bestValue) {
bestValue = actVal;
foundNewBestAction(bestActions, rot, col);
} else if (actVal == bestValue) {
foundTiedAction(bestActions, rot, col);
}
}
}
//Choose the best action
Action a = pickRandomAction(bestActions);
//Clean up the actions
clearActionList(bestActions);
//Play the action
return a;
}
void UserControlsEngine::endGame()
{
hasEnd = true;
clearActionList();
killTimer(idTimer);
}
