Embedding Kalman techniques in the one-shot task model when non-uniform samples are corrupted by noise

The performance of several closed-loop systems whose controllers concurrently execute in a multitasking real-time system may be deteriorated due to timing uncertainties in taskséxecutions, problem known as scheduling jitters. Recently, the one-shot task model, that combines irregular sampling, a predictor observer, and strictly periodic actuation, was presented in order to remove the negative effects of jitters. However, its successful application required noise-free samples. In this paper we extend the one-shot task model to the case of noisy measurements. In particular, we embed a Kalman filter into the model taking into account that the available measurements are not periodic. This poses the problem of adapting the standard discrete-time Kalman filter to the case under study, and decide when to apply the prediction and the correction phase. Two different strategies are presented, and their control performance and computation demand are analyzed through real experiments.