Discussion on: "Development and Experimental Verification of a Mobile Client-Centric Networked Controlled System"
European Journal of Control
Tzes et al. present the experimental verification of a mobile client-centric networked control system (NCS) implemented over a General Purpose Radio Service (GPRS) communication channel. The system is set up by defining a client, the controller, which sends information over the network to the server, the plant and actuator. The characteristics of GPRS make the system different from other NCSs studied, which are often implemented over Ethernet or DeviceNet. For security reasons, mobile-phone
... ice providers preclude a server from independently sending information back to the client. In this network, all actions must be initiated by the controller creating a client-centric NCS. The mobile NCS time delays are composed of both encoding/decoding delays and as well the transmission times of data through the mobile network. The transmission delays are highly uncertain and depend on a number of factors that include the number of users, loss of packets and the existence of higher priority transmissions. As explained in the paper, GSPR transmissions have a lower priority than GSM-based voice calls. A useful contribution of the paper is the characterization of transmission delays in both UDP and FTP connections through GSPR. The measurements show that the average bit rates are significantly lower than the theoretical limits of GSPR. The stability of the closed-loop is guaranteed by implementing results from time-delayed systems and by solving a set of linear matrix inequalities. The authors state that the maximum time delay, max , calculated is conservative. Walsh, et al.  present an analytical proof for the global stability of the closed loop of the NCS based on perturbation theory; however, their results also yield a conservative max . In most cases, an NCS controller is designed so that it broadcasts updates to the plant every T seconds, where T < max to insure stability. However, the most effective way to improve the performance of an NCS is to minimize time delays by reducing network traffic. Finding a tighter bound on max is an important research question since it would relax the minimum controller transmission frequency and reduce network traffic. An investigation of how to systematically reduce the node transmission frequency in NCSs using variable deadbands is presented in Ref. . Researchers have often modelled NCSs as a variant of time-delayed systems and designed controllers that withstand the worst possible time delay. Although this approach simplifies analysis, it does not consider the effects of the controller on the network. In order to improve NCS performance, there is a need to formulate controllers that utilize the packet structure of network communication as well as monitor traffic to adjust their operating mode. Depending on the protocol, each packet has a minimum size requirement that in control applications is often not efficiently used. Georgiev and Tilbury  exploit the relative large size of Ethernet packets by sending several control commands in a single packet and thus reducing network traffic. Tang and de Silva  propose using a modified model predictive control strategy to encode a sequence of commands in a packet to reduce communication. As the characteristics of the network traffic change, the length of the prediction time horizon is adjusted to minimize the effects of time