{[ promptMessage ]}

Bookmark it

{[ promptMessage ]} - In IFAC Control Engineering Practice 2000...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon
In IFAC Control Engineering Practice 2000 A FEEDBACK SCHEDULER FOR REAL-TIME CONTROLLER TASKS Johan Eker Per Hagander Karl-Erik Årzén Department of Automatic Control Lund Institute of Technology Box 118, S-221 00 Lund Sweden Fax: +46-[0]46 13 81 18 johan.eker|karlerik.arzen|[email protected] Keywords Real-time; Feedback scheduling; Linear quadratic control; Opti- mization. Abstract The problem studied in this paper is how to dis- tribute computing resources over a set of real- time control loops in order to optimize the to- tal control performance. Two subproblems are investigated: how the control performance de- pends on the sampling interval, and how a re- cursive resource allocation optimization routine can be designed. Linear quadratic cost functions are used as performance indicators. Expressions for calculating their dependence on the sam- pling interval are given. An optimization rou- tine, called a feedback scheduler, that uses these expressions is designed. 1. INTRODUCTION Control design and task scheduling are in most cases today treated as two separate is- sues. The control community generally assumes that the real-time platform used to implement the control system can provide deterministic, fixed sampling periods as needed. The real-time scheduling community, similarly, assumes that all control algorithms can be modeled as pe- riodic tasks with constant periods, hard dead- lines, and known worst case execution times. This simple model has made it possible for the control community to focus on its own problem domain without worrying how scheduling is be- ing done, and it has released the scheduling community from the need to understand how scheduling delays impact the stability and per- formance of the plant under control. From a his- torical perspective, the separated development of control and scheduling theories for computer based control systems has produced many use- ful results and served its useful purpose. Upon closer inspection it is, however, quite clear that neither of the above assumptions need necessarily be true. Many of the comput- ing platforms that are commonly used to im- plement control systems, are not able to give any deterministic guarantees. This is especially the case when commercial off-the-shelf operat- ing systems, e.g. Windows NT or Linux, are used. These systems are, typically, designed to achieve good average performance rather than high worst-case performance. Many control al- gorithms are not periodic, e.g. internal com- bustion engine control, or they may switch be- tween a number of different fixed sampling pe- riods. Control algorithm deadlines are not al- ways hard. On the contrary, many controllers are quite robust towards variations in sampling period and response time. It is in many cases also possible to compensate for the variations on-line by, e.g., recomputing the controller pa- rameters, see Nilsson ( 1998 ) . It is also possi- ble to consider control systems that are able
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full Document Right Arrow Icon
Feedback Scheduler Sampling Rates Desired Load Performance Exec. Times Fig. 1
Background image of page 2
Image of page 3
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}