CLASSIFY SITUATIONS INTO A REINFORCEMENT BASED CONTROL SYSTEM: UNSUPERVISED LEARNING VERSUS HEURISTIC