A communication-based distributed model predictive control scheme with time-varying consistency constraints

A communication-based distributed model predictive control scheme for a set of dynamically decoupled autonomous systems (agents) is proposed. The individual dynamics are described by discrete-time linear systems. Locally, each agent obeys a model predictive controller which minimizes a given local cost-function. Coupling between the agents occurs only through state-constraints. It is assumed that an initial feasible solution for the centralized optimization problem can only be found in a suboptimal manner due to computational limits. Then, the agents can decrease their local costs by replanning in each time step. This has to be done in a consistent way which is obtained by communicating local predicted trajectories under consideration of communication delays.