Reduction of Complexity in Q-Learning a Robot Control for an Assembly Cell by using Multiple Agents