<p>The majority of computers around us are embedded in productsand dedicated to perform certain tasks. A specific task is thecontrol of a dynamic system. The computers are ofteninterconnected by communication networks forming a distributedsystem. Vehicles and manufacturing equipment are two types ofmechatronic machines which often host dedicated computercontrol systems. A research problem is how the real-timebehaviour of the computer system affects the application,especially the control of the dynamic system.</p><p>If the internal or external conditions varies over time, itbecomes difficult to assign a fixed resource reservation thatwill work well in all situations. In general, the more time anapplication gets of a resource, the better its gauged orperceived quality will be. A strategy is to alter the resourcereservation when the condition changes. This can be constructedas a negotiation between competing applications, a method forwhich the term<i>quality of control</i>, QoC, has been coined. Scalability isthe ability to change the structure and configuration of asystem. It promotes evolving systems and a can help manage acomplex product family. An architecture for a QoC middleware ontop of a scalable computer system, has been proposed.</p><p>As a<i>quality measure</i>of a control application, the well-knownweighted quadratic loss function used in optimal control, hasbeen revised to encompass a subset of the so called timingproperties. The timing properties are the periods and thedelays in the control loop, including time-varying period anddelay. They are the interface between control and computerengineering, from a control engineering viewpoint. The qualitymeasure can be used both offline and on-line given a model ofthe sampled-data system and an appropriate description of thetiming properties.</p><p>In order to use a computer system efficiently and toguarantee its responsiveness, real-time scheduling is a must.In fixed priority scheduling each task arrives periodically andhas a fixed priority. A task with a high priority can preempt alow priority task and gain access to the resource. Thebest-case response time characterizes the delays in the system,which is useful from a control viewpoint. A new algorithm tocalculate the<i>best-caseresponse</i>time has been derived. It is based on ascheduling scenario which yields a recurrence equation. Themodel is dual to the well-known worst-case response timeanalysis.</p><p>Besides the dynamic fixed priority scheduling algorithm,optimal control using<i>static scheduling</i>has been studied, assuming a limitedcommunication. In the static schedule, which is constructedpre-runtime, each task is assigned a time window within aschedule repeated in eternity. The optimal scheduling sequenceis sought by optimizing the overall control performance. Aninteresting aspect is that the non-specified control periodfalls out as a result of the<i>optimal schedule</i>. The time-varying delay is accountedfor in the control design.</p><p><b>Keywords:</b>Real-time scheduling, sampled-data control,performance measure, quality of control, limited communication,time-varying delay, jitter.</p>
Identifer | oai:union.ndltd.org:UPSALLA/oai:DiVA.org:kth-3785 |
Date | January 2004 |
Creators | Sanfridson, Martin |
Publisher | KTH, Machine Design, Stockholm : Maskinkonstruktion |
Source Sets | DiVA Archive at Upsalla University |
Language | English |
Detected Language | English |
Type | Doctoral thesis, comprehensive summary, text |
Relation | Trita-MMK, 1400-1179 ; 2004:7 |
Page generated in 0.0028 seconds