Cooperative control of dynamical systems

The progressive automation of technical systems requires new perspectives on the design of controllers. In the classical point of view, a controller is designed to achieve a desired behaviour of a plant, e. g. set-point following or attenuation of disturbances [2]. Modern concepts, on the other hand, focus on more global aims, e. g. synchronisation of multiple oscillators. For this purpose, it is not possible to focus on a single plant, which leads to the concept of multi-agent systems [1]. Figure 1 shows a multi-agent system, which consists of physically uncoupled plants Pi, i ∈ {1, 2, . . . , N}, which are controlled by local controllers Ci. The combination of plant and controller is denoted by Σ̄i. These controlled subsystems are able to communicate with each other via a given communication network to exchange control variables and set-points or possibly model information in a more general framework.