Study On The Co-design Methods Of Real-time Control Systems

Real-time control systems (RTCS) generally performs multiple parallel control tasks on a real-time operating systems (RTOS), where multiple tasks compete for limited processing resource. The co-design of the real-time control system integrates the control theory and scheduling strategies. In order to achieve the given performance metrics, it is necessary to consider the inherent jitters caused by the scheduling and their influence on the performance when the controller is designed. On the other hand, the controller attributes would be also taken into account in the study of scheduling policy, so that both scheduling flexibility and control performance would be improved. This dissertation studies collaborative method combining efficient scheduling and controller design, aiming at improving the performance of RTCS under the limited processing resource. The main content and innovative work of this dissertation are briefly described as follows:1. The mutual influence between the control and the scheduling in RTCS was analyzed, so as to provide the baseline support for the collaborative design. The impact from control to scheduling mainly indicates that the different processor load or the different control performance requirement brings different tasks'schedulability in control systems. Otherwise, the sampling jitters and input-output delay jitters caused by scheduling affect the control system performance in various degrees.2. Focusing on the variation of workload caused by the uncertainty of control tasks in RTCS, the dissertation presents an adaptive feedback elastic period model to deal with the system overload. In this proposed model, CPU utilization is estimated by measuring the execution time of each task, and the system load is dynamically adjusted through self-adaptively modifying the task sample periods online. Thus the schedulability is improved as well as the control performance is guaranteed in RTCS. The simulation results show that this co-design model is effective and useful. It can handle overload situations in a flexible way.3. To guarantee the schedulability of the critical tasks and enhance the control performance, a FSD (fuzzy scheduling design) algorithm was proposed based on time-triggering sampling computer control systems. The FSD algorithm configures the control index and the idle time of the control tasks with fuzzy design, and deduces a multi-rules optimal scheduling method. In a normal workload, it efficiently decrease the deadline loss ratio and improve the system schedulability. In a overload situation, the performance of the critical task is guaranteed.4. Focused on the discrete state system whose jitters are not exactly known but within bounded intervals, a stability condition is proposed based on a compensated control algorithm. It represents the discrete system of the embedded real-time multi-tasks scheduling systems as a closed-loop interval matrix by the interval algebra method, and deduces a sufficient robust stability condition using Lyapunov theory and the infinite norm of the matrix. The simulation results show that the method is more computationally effective and less conservative compared to the conventional one.5. A co-design approach combining predictive control compensation and network scheduling is presented in network-based RTCS to overcome the negative influences of the network long latency, e.g. the latency is longer than a sampling period. When the control signal is lost due to a long time delay or packet losses, the control performance is improved by predicting the latest control value and applying it to the control objective. When the stability criteria under the predictive control compensation algorithm can not be guaranteed, the task period is adjusted through the network scheduling to decrease the network workload. The simulations show that the presented method can improve the performance of the network control systems with the bounded network time-delay.6. Taking the ocean mining vehicle real-time control system as a practical research example, a co-design method of the ocean mining vehicle real-time control system was presented. Based on the adaptive feedback elastic period model, the schedulability is improved and the plant control performance is optimized with automatically adapting their periods to the situation online. Simulations illustrate how the proposed algorithms apply to real-time industrial problems, which provides guidances for the co-design of the practical systems.