In research on networked control systems (NCS), the performance of the control system is assumed to be affected by network-induced parameters such as delays, jitters, packet losses and link failures. In this paper, two delay compensation methods are studied, adaptive Smith Predictor and robust control, in the framework of Networked Control Systems while considering the protocol specifications. First, a clock synchronization protocol is implemented in each device of the network, thereby providing the possibility of end-to-end delay measurement, and a compensation strategy based on this measurement and the adaptive Smith Predictor is applied.Secondly, the end-to-end upper bound delay estimate is calculated on the basis of the network architecture and the traffic estimation, and then subsequently used in a robust control based compensation strategy. The switched Ethernet network is used as an example of the communication network of the NCS.

In an industrial or technical context, some of the communications are strongly time-constrained. Our objective is to evaluate the service offered by a switched Ethernet architecture. We have chosen the network calculus theory because it gives a good representation of the exchanges and enables to determine the maximum end to end delays. Previous models implementing a strict priority policy presented in [7] are improved in order to take into account the Weighted Fair Queueing mechanism which enables to balance the allocation of the network to the different traffics regarding the time contraints they have to respect. This paper also presents the new analytical formulas which upper bound the maximum end-to-end delays of the time-critical messages over the whole network. This enables to determine the optimal bandwith dedicated to both critical and best effort traffics. On an industrial scenario, Strict Priority and Weighted Fair Queueing policies are finally compared.

SUMMARYThe Ethernet network is more and more used to interconnect industrial devices. The objective of this paper is to study the performances of such a network to support real-time communications. For this, first we propose a general representation to model a switched Ethernet network by using a sequence of elementary components such as buffers, multiplexers, etc. Second, we aggregate the individual temporal properties of each component given in the Cruz' survey to obtain a global formula enabling one to calculate the maximum end-to-end delay for any industrial communication scenarios. Finally, we deduce the limits of the switched Ethernet network regarding the number of input/output cards connected to the network and to the sizes of periodic and aperiodic messages.