On an adaptation in distributed system based on a gradient dynamics

Distributed system is usually constructed from many homogeneous subsystems with their local couplings, and does not have a control center. From the structural similarity to biological systems, the distributed system is expected to possess the flexibility or adaptability. To construct a distributed system, two problems arise: One is how to control the whole system by only using local interactions based on local couplings, and another is how to design the adaptation mechanisms. To the former problem, a method based on the gradient system has been discussed in our previous paper. In this paper, we develop this method further and propose an new adaptation rule executable in the subsystems with local couplings. The essence of the adaptation is to adjust the desired system states so that interactions among subsystems are made small. We show a simulation result of a coupled oscillator system where the oscillation pattern changes adaptively with environmental conditions.