Sistemas de control con recursos restringidos

Laburpena

In this PhD thesis solutions are given for some of the problems that appear when real-time systems must attend simultaneously to multiple control tasks. Emphasis is given to those problems which affect the quality of control. A multitasking real-time system distributes one or more resources among several control loops. The shared resource can be either a processor or a network. Variations in the execution periods and in the latencies inherent to the control algorithms appear regardless of the sharing method used. When the shared resource is a processor, the problem lies in the impossibility of running two algorithms at the same time. This means a time-ordered running, and hence variations in the basic parameters of the controller. When a network is shared, no two messages can be transmitted at the same time so, once more, a time-ordering of the messages is imposed and, again, variations in the controller parameters arise. The first part of the thesis is an analysis of the effects of those variations upon the control. There are two objectives: First. To analyze the effect of using multitasking real-time systems upon the controllability and observability of the controlled systems. Second. To obtain stability analysis techniques for the real-time control systems. The results of the above are: First. Controllability and observability are maintained when the controller is implemented in a real-time platform. Second. Stability analysis for real-time control systems can be done by probabilistic techniques and by intervalar techniques. The second part of the thesis studies the efficiency of real-time implemented control loops. This is done by analyzing the best way to distribute the available resources among the control loops taking account of the state of the control. The guiding idea is "to give more resources to the control loop in more need of them". The two objectives are: Third. Management of overload conditions in control tasks. Fourth.