Decentralized control is a promising method with many advantages while an effective design method for it is desired. Design in conventional methods is complicated because the correspondence between a system role and its controllers is unclear. This paper aims at establishing a design framework for decentralized control systems that deals with task shift, faults, performance limits, and environmental variation in a unified manner. We have proposed functionbased controller design for a bilateral control system, the simplest form of a decentralized control system. In this paper, the method is expanded to a decentralized control system with multiple subsystems. "Function" is defined as a minimum component of a system role. Conversely, the system role is described as a combination of functions. Controllers are designed based on functions instead of control objects. This idea solves the problem of design complexity in decentralized control systems.In this study, robot information is transformed to new coordinate space based on functions to associate function-based controllers with control objects. The controllers are designed in the new coordinate space based on functions. This new coordinate space is called function coordinate space. On the other hand, the original coordinate space is called robot coordinate space. The overview of the control system is shown in Fig. 1. The robot coordinate is transformed to function coordinate by transformation matrix T.The dynamics in each coordinate does not interact with each other. Therefore, the control system shown in Fig. 1 could be treated as a decoupled system. Essential behavior of the function in the function coordinate space is simple although one function-based controller affects multiple robots. The behavior realized by each function-based controller is named "function mode". The behavior of the entire system in robot coordinate space shows up as superposition of function modes.The design flow of function-based control system is shown in Fig. 2. This method makes the controller design explicit since system role and individual controllers are related directly. The controller design also becomes simple since each function is a minimum component of the system role. These features provide some advantages for the system. For example, flexibility is improved because of simple design for task shift. This system can also handle The main idea of function-based controller design is to design each controller as a detachable component. Many kinds of functionbased controllers are designed in advance like peripheral equipment. Among them, requisite functions are exerted depending on the varying system role. Great patterns of tasks are realized with such a framework. In sum, this framework is useful for control of robots adaptive to complicated environments since it solves the issues of task variation and exception handling of complicated systems. Results on experiments and simulations show validity of the proposed method.-13 -
Paper
A Controller Design Method of Decentralized Control...