抄録
In the large scale complex systems, it is hard to control the whole system by only one controller and it is required to attain variety, flexibility and fault tolerability, etc. For this purpose, an autonomous decentralized system has been studied recently. On one hand, organisms actually have these properties, and this is due to a structure constructed by a lot of homogeneous subsystems. In this paper, a gate pattern generator system of a quadruped animal is considered as one example of autonomous decentralized systems. It is well known that this system is made by neural oscillator networks.
First, gait patterns of a quadruped animal are investigated for a construction of pattern generator. Second, coordinations of two coupled oscillators are considered. The order parameter of this system is the relative phase between two oscillators, and it can be controlled by interaction functions. A potential function of relative phase is very usefull for determining good interactions.
Some computer simulations show that this system generates some gait patterns (e.g., walk, trot, rotatory gallop) suitable for its moving speed.
