Sensitivity Analysis of Dropped Samples for Performance-Oriented Controller Design

When designing networked control system (NCS), a main design goal is to reduce the resource usage as far as possible while guaranteeing a predefined control performance. A sensor node that skips samples reduces the resource usage of the control application, but also causes a performance degradation of the NCS. To obtain an efficient co-design, this paper makes use of an expressive model that bounds the occurrence of dropped samples in the interval domain. With such a model, the borderline of the specified control performance is reached while introducing minimal pessimism. Furthermore, we present in this paper a sensitivity analysis that derives the maximum density of dropped samples from the impulse response behavior of a control system with respect to a predefined performance threshold. This bound on the density of dropped samples allows a resource reduction, which can be used by sporadic tasks in mixed criticality systems without endangering the control performance threshold.