Jitter Compensation in Real-Time Control systems

In control theory, sampling and actuation are generally assumed to be synchronous and periodic, and control computations are assumed to be instantaneous. However, when a control algorithm is implemented as a task (or a set of subtasks) in a real-time system, these assumptions cannot be met during the execution of the tasks due to the jitters that the schedule introduces on each task instance execution. This causes degradation in control performance and maybe even instability. We show that by accepting those jitters in the controller design, we compensate for the control system performance degradation, providing also new flexible timing constraints for the schedulability problem.In this paper, we discuss the types of jitters incurred by real-time systems and present an integrated approach to improve control performance. We address implementation aspects and. propose timing constraints to increase schedulability utilizing the integrated method. Similar to a schedulability test, our approach provides guarantees offline, that the system will be stable at runtime, if temporal requirements are met at runtime, even when actual execution patterns are not known beforehand. It performs offline schedulability and control analysis as input to the calculation of adjusted controller parameters for jitter values. With our approach, we can address the problems due to (a) sampling jitters and (b) varying delays between sampling and actuation - not addressable using traditional EDF and FPS based scheduling, or by previous real-time and control integration approaches.We achieve this by making use of control properties and through online compensation. Our experimental results indicate that our approach to compensations not only guarantees stability by compensating for the degraded control performance for EDF and FPS, but also leads to better schedulability due to the new flexible timing constraints offered by the application of our compensation approach.