Internal observation and mutual adaptation in human-robot cooperation

Abstract

"Internal observation" means that an agent observes environment in inseparable mutual interference between the agent and its environment, and similar situations are widely observed in cooperative behavior in multi-agent system, group robots, human interface and so on. However, conventional framework of system theory which is based on separation between controller and controlled object can not be applicable to these mutually interferent situations. Under this kind of dynamical complexity, the agent can't completely identify its environment, and it should generate internal model to solve such ill-posed problem. Thus, to overcome this essential difficulty, we propose "dual-center model" based autonomous agent which is composed of two different dynamics. One is to self-organize coherent relationship in mutual interference, and the other is to separate the coherence into two one-sided interactions by using internal model. Through mutual constraint between these two different dynamics, the agent generates hypothetical constraints as an internal model which represents interaction between the agent and its environment. By using this model, internal observation under mutual interference becomes possible. As an example of it, mutual adaptation in cooperative behavior between human and robotic agent was realized.