//At the algorithm manager, generate step is ambiguous with actually stepping through the algorithm,
//Or starting the algorithm, and making the first history entry at step 0.
//Here, a step only happens when we have been asked by the manager to *actually* take a step.
public void TakeStep()
{
//Get step from qmatrix. Being randomly generated for now.
move_this_step = live_qmatrix.GenerateStep(bender_perception_starting); //Tentative; we'll attempt this later. just a random move for now.
result_this_step = board_data.ApplyMove(move_this_step); //The move should be performed now, if possible.
obtained_reward = ReinforcementFactors.list[result_this_step]; //Get the reward for this action
episode_rewards += obtained_reward; //Update the rewards total
if (result_this_step == MoveResult.can_collected())
{
++cans_collected;
}
location_result = new int[2] {
board_data.bender.x_coordinate, board_data.bender.y_coordinate
};
bender_perception_ending = board_data.bender.get_perception_state();
live_qmatrix.UpdateState(bender_perception_starting, bender_perception_ending, move_this_step, obtained_reward);
//give the value to the q matrix to digest
if (GetStepNumber() == Qmatrix.step_limit && GetEpisodeNumber() > Qmatrix.episode_limit)
{
algorithm_ended = true;
}
}
