A spiking neural network for behavior learning of a mobile robot in a dynamic environment

It is very difficult to design the learning structure of a robot beforehand in an unknown and dynamic environment, because the dynamics of the environment is unknown. Therefore, this paper proposes a spiking neural network (SNN) for behavior learning of a mobile robot in a dynamic environment including multiple moving obstacles. Furthermore, a behavior should be acquired and improved according to previously acquired behaviors. This indicates the robot should accumulate behavior knowledge through the lifetime learning. Therefore, we have proposed the concept of structured intelligence including three components; intuitive inference, logical inference, and self-consciousness. Intuitive inference is realized by neural computing based on numerical processing, while the logical inference is realized by fuzzy computing based on linguistic and symbolic processing. In this paper, we propose a learning method of SNN by using outputs from fuzzy controllers as teaching signals. The simulation shows the robot can extract and modifies the behavior knowledge of fuzzy controllers according to the spatiotemporal context of the facing environment.