Scheduling of a limited communication channel for optimal control

Abstract

In this paper a method for optimal off-line scheduling of a limited resource used for control purposes is presented. For various reasons, real-time communication channels are prone to have limited bandwidth. To overcome this obstacle, the rate of actions must be chosen accordingly at design time, both with respect to the limitation of the resource and to control performance. A resulting off-line schedule implements the rate of actions as a repetitive sequence of communication instants. Periodic control theory is used to define a cost functional for LQ-control, that measures the performance of a sampled-data system in relation to a desired continuous time performance. In contrast to uniform sampling, the communication sequence is here allowed to be time-varying. This approach results in a complex combinatorial optimization problem, whose solution gives the optimal off-line schedule, i.e., the sequence in which the actions should take place. The optimization problem is solved by a neighborhood search method where a heuristic method is used to generate initial guesses close to the optimum. The optimal schedule is typically such that the sampling is non-uniform, but the resulting LQ-control law is time-varying and takes this non-uniform sampling into account.