Skip to Main Content
This research tries to construct a new architecture of intelligent information processing imitating that of human thought and having ability of autonomous learning. Its processing is divided into logical and intuitive processing, which are selected flexibly and properly for the changing situation. To investigate its effectiveness, it is applied to a robot playing a Tetris game, which is regarded as a symbol of human intellectual behavior. The robot consists of three elements: recognition part, thinking part, and action part. The thinking part consists of logical and intuitive processing. Logical processing has the ability of accurate and slow processing, and intuitive processing has the ability of rough and quick processing. The learning method of logical processing consists of fuzzy Q-learning with neural network and intuitive processing consists of neural network, which is trained by empirical knowledge of logical processing. The combination of logical and intuitive processing leads to the improvement of game playing results for the reason that intuitive and logical processing built along with playing time becomes more applicable. As a result, it is shown that the proposed method is a useful approach.