Networked control systems differ from classical control systems in that their control loops are closed around real-time communication networks. The flexible architectures of these systems, together with their reduced installation and maintenance time and costs, make them an appealing option in a broad range of industrial applications [4]. For example, in large-scale plants and geographically distributed systems, where the number and locations of controlled subsystems make the use of single wires to interconnect the control systems expensive, the flexibility and ease of maintenance using a network to transfer information is an appealing goal. Systems designed in this manner allow for easy modification of the control strategy by rerouting signals, having redundant systems that can be activated automatically when component failure occurs [2], and in general they allow having a high level supervisor control over the entire plant.

The use of a shared network in contrast to using several dedicated independent connections, however, introduces new challenges and makes the analysis and design of the control system more complex. One of these challenges is the limited number of channels that communication networks have, which implies that at any one time only some of the sensors and actuators can exchange information with the plant supervisor, while others must wait. Such medium access constraints occur independently of any other communication constraints, such as possible packet losses or transmission delays. Under these constraints, the performance of the control loops is dependent not only on the controller design for the individual units but also on the selection of the medium access strategy that assigns a transmission schedule to each transmission entity (sensor, controller, actuator) on the network based on a scheduling algorithm (a set of rules that, at any time, determines the order in which messages are transmitted). Traditionally, the design of each of these two components has been studied separately by selecting one while assuming the other is given. More recent efforts on this problem, however, have been guided by the realization that the design of both the communication protocols and the interacting controlled system should not be treated as separate (e.g., [1], [5], [6]). In the co-design approach, network issues such as band-width, quantization, and message delays need to be considered simultaneously with controlled system issues such as stability, performance and fault-tolerance. Existing results on this problem, however, have focused mainly on linear plants under centralized control or plants with non-interacting subsystems.

In this work, we focus on the problem of scheduling a set of interconnected nonlinear subsystems sharing a single communication network and competing for network access. To illustrate the main ideas in the proposed approach, we consider first the case of ideal network transmission and study, as an example, a plant configuration where the local controllers of all the units are assumed to be collocated with the supervisor (i.e., the supervisor has un-interrupted access to the control actuators of each unit) and only the sensor suites vie for access to the supervisor. The plant-wide communication network is assumed to provide connectivity between the sensors and the supervisor in a discrete manner (on/off). Communication is limited in the sense that only a certain number of units may transmit their measurements to the supervisor at any one time. The objective is to find a sequence for establishing and terminating communication between the sensor suites of the different units and the supervisor in a way that stabilizes all the interconnected subsystems in the collection. To address the problem, we initially determine an upper bound on the period that each unit must transmit its sensors' data at for the overall plant to be stable. Then, the rate monotonic scheduling algorithm [3] is used to develop a schedulability test for the set of sensor suites in terms of the transmission periods and times. In this algorithm, priorities are assigned to tasks before execution and do not change over time. It is shown that a unit with faster dynamics (i.e., requiring higher transmission rate) is given higher priority over a slower one. Next, the assumptions of ideal transmission are relaxed to study the above network scheduling problem with network-induced delays and packet dropouts taken into account. Finally, the results are applied to the problem of controlling a network of fuel cell plants that are connected to a power distribution system and communicate with a central computer through a real-time communication network.

References:

[1] Branicky, M. S., S. M. Phillips and W. Zhang, "Scheduling and feedback co-design for networked control systems", Proceedings of the 41st IEEE Conference on Decision and Control, pages 1211-1217, Las Vegas, NV, 2002.

[2] El-Farra, N. H., ``Failure Analysis in Networked Process Control Systems with Control and Communication Constraints," AIChE Annual Meeting, paper 454g, Cincinnati, OH, 2005.

[3] Liu, C. L. and J.W. Layland, "Scheduling algorithm for multiprogramming in a hard-real-time environment," J. ACM, 20:46-61, 1973.

[4] Murray, R. M., K. J. Astrom, S. P. Boyd, R. W. Brockett and G. Stein, "Control in an information rich world," IEEE Contr. Syst. Mag., 23: 20-33, 2003.

[5] Zhanga, L. and D. Hristu-Varsakelis, "Communication and control co-design for networked control systems," Automatica, 42: 953-958, 2006.

[6] Walsh, G. and H. Ye, "Scheduling of networked control systems," IEEE Contr. Syst. Mag., 21: 57–65, 2001.

Extended Abstract Status: Not Uploaded