Q-learning in evolutionary rule based systems