Control loop scheduling paradigm in distributed control systems

J. Lopez, P. Marti, J.M. Fuertes
IECON'03. 29th Annual Conference of the IEEE Industrial Electronics Society (IEEE Cat. No.03CH37468)  
Abstracf-The performance Distributed Control System (DCS) depends not only on the operation of the individual components, but also on their interaction and cooperation. Therefore, the rules that allow the exchange of messages, i.e. message scheduling policy, is a key issue in terms of system performance. For control applications where control loops are closed over Communication networks, this is especially true. Traditional scheduling policies are based on models and techniques that do not take
more » ... es that do not take application demands into account. This precludes the dynamic adaptation of the use of the communication bandwidth according to the application needs. To overcome this problem and focusing on control applications, we present an early specification of a novel scheduling technique: Large Error First (LEF). This scheduling algorithm uses feedback information from the application in order to assign communication bandwidth to each individual component. We studied the performance of a distributed application when the messages are sent according to this novel scheduling policy and encouraging simulation results have been obtained. I. INTRODUCTION Distributed Control Systems (DCS) are gaining increased popularity kecause they offer several advantages such as modular and flexible system design (e.g. distributed processing and interoperability), simple and fast implementation (e.g., small volume of wiring and powerful configuration tools), and powerful system diagnosis and maintenance utilities (e.g., alarm handling and supervisory packets). Distributed Control Systems contain a large number of interconnected devices that exchange data through communication networks; examples include industrial automation, building automation, office and home automation, intelligent vehicle systems, and advanced aircraft and spacecraft. The specific application imposes different degrees of timing requirements to the DCS implementation. Control loops are the applications that pose the mst stringent timing constraints because control theory assumes a highly deterministic timing on any implementation [2] . Consequently, the insertion of the communication network in to control loops makes the analysis and design of such applications complex [12] . Therefore, the performance of a DCS is dependent ofmany variables, such as the network traffic, the network devices, distributed control architecture, communication protocols, and the controller and its implementation. Thus, the performance of a DCS depends not only on the performance of its individual components hut also on their interaction and cooperation. In addition, the successful implementation of control loops closed over communication networks require the adequate integration of several disciplines, such as communication systems, real-time computation systems and control systems [13]. Traditionally, each one of these disciplines has been developed separately. This has allowed each discipline to focus on its own problems domain without considering any interaction from the others (for example, control algorithms have been designed witbout consideration of their implementation details). However, as it bas been claimed in recent works (see for example [7], [IO], [Ill, and [19]), integrated approaches combining different issues from the three disciplines end up in better designs for DCS. Refer to section I1 for a more detailed discussion of the state-of-the-art, In this effort, we focus our attention on the interaction of the scheduling of messages and the performance of control loops closed over communications networks. It is well known that randombased scheduling policies (e.g. Ethemet [SI) conflict with the determinism that control theory imposes on an implementation. This conflicting situation degrades system performance. To overcome this problem we can think of well-known communication networks with deterministic scheduling policies such as table driven (e.g. WorldFlP [I]), masterislave (e.g., Profibus [ 171) or norrpreemptive fixed priority-based (e.g. CAN [I611 scheduling policies. Such
doi:10.1109/iecon.2003.1280270 fatcat:vqdw5tbtrbe6fd2exlijstzbmq