Prediction Schemes for Compensating Variable Delay for Improving Performance of Real-Time Control Tasks

Period selection based techniques for resource distribution among concurrent control tasks have gained significant research interest in the last decade. The application of these techniques in the networked control environment may give rise to an asynchronous data communication between the sensor and the controller and may result in a variable delay in the loop. This work investigates the degrading effects of such variable delay on the control performance and presents novel delay compensation methods to effectively reduce the performance degradation. A linear estimation based approach is presented first where the time-stamp information of network messages is used to estimate the sensor value at the control task execution instants. Further, a periodic pattern in the successive delay values has been identified and a delay compensation scheme by online learning of the delay pattern is proposed. The application of the proposed schemes on an automotive wheel slip control system is shown to achieves significant performance improvement.