We have been developing a reinforcement learning technique called BRL as an approach to autonomous specialization, a new concept for cooperative multi-robot systems. BRL has the mechanism for segmenting continuous state and action spaces autonomously. Also, we have two types of extended BRL. The first one is that the state space is covered with redundant rules that avoid rule deletion to overcome the over-fitting problem. The second one is that, instead of generating a random action when a robot encounters an unknown situation, an action is calculated as the linear interpolation among the rules that have high similarity to the current sensory input. BRL with both extensions is examined by arm-type autonomous robots, the task of which is lifting an object without tilting it. The experiments conducted illustrate the high incremental learning performance of the proposed method by developing various types of cooperative behavior as a result of autonomous specialization.