Real-time Control Systems Co-design Based On Fixed Preemption Points Scheduling

Input-Output Delay(IO Delay)and Sampling Delay are very important timing attributes in the real-time control systems,which can affect the performance and stability of the control systems.Thus,that timing attributes should be considered in the design process.Fixed Preemption Points Scheduling has been proposed to reduce the run-time overhead,improve the predictability of worst-case execution time(WCET)and the schedulability of the system.However,the advantages of fixed preemption points scheduling are not limited to that.Applying the fixed preemption points scheduling,especially when the last non-preemptive region is setting the largest one,the IO Delay can be minimum.However,it will block higher priority tasks and bring more blocking time to them.In addition,the length of last non-preemptive region will affect that task's Sampling Delay.In this thesis,we deal with the analysis of 10 Delay and Sampling Delay for fixed priority tasks with last non-preemptive region(FP-LNPR),which is a variant of the fixed preemption points scheduling.Based on the FP-LNPR scheduling algorithm,one scheduling algorithm named FP-LNPR*is proposed.For the fully preemptive schedulable tasks under the fixed priority scheduling algorithm,the FP-LNPR*and FP-LNPR scheduling algorithm can be applied to real-time control systems to improve the control performance.The experiment results show that when comparing with the fully preemptive fixed priority scheduling algorithm and FP-LNPR scheduling algorithm,FP-LNPR*can obtain the lowest IO Delay and Sampling Delay.For a benchmark control system,FP-LNPR*can obtain the best control performance.When the preemption overhead cannot be ignored,one algorithm that combine the branch and bound,exhaustive search method to select the optimal preemption points has been proposed based on the IO Delay analysis.The main contributions of this thesis are list following:1.The Input-Out Delay Analysis.We first define the timing instant "IO-critical instant"that cause the largest IO Delay and find that instant under the FP-LNPR scheduling algorithm.One method to calculate the largest 10 Delay has been proposed and depend on that,the following conclusion can get:Decreasing the length of last non-preemptive region,the worst-case 10 Delay of that task cannot decrease.2.Sampling Delay Analysis.We first define the Sampling Delay Self-pushing Phenomenon and compare it with the Self-pushing Phenomenon.We testify that there is no Sampling Delay Self-pushing phenomenon for the fully preemptive schedulable tasks under the fixed priority scheduling.The property of the ?-critical instant has been studied under the FP-LNPR scheduling.Based on that,one method to calculate the Sampling Delay has been proposed.We testify that the Sampling Delay can be decreased by decreasing the length of non-preemptive region for non-control tasks.One important conclusion has been derived that,for tasks that do not exist the Sampling Delay Self-pushing Phenomenon,the Sampling Delay of that task cannot be affected by the length of last non-preemptive region.3.Based on the analysis of Input-Output Delay,We propose the FP-LNPR*scheduling algorithm.We set the largest non-preemptive region for control tasks and the non-control tasks are setted fully preemptive.4.When the preemption overhead cannot be ignored,we propose the analysis of Input-Output Delay with the preemption overhead.We propose one method to select the optimal preemption points based on the branch and bound method,exhaustive search.