A structured organisation of tasks, possibly hierarchical, is necessary in a BISDN network due to the complexity of the system, its large dimension and its physical distribution in space. Feedback (possibly supplemented by feedforward) control has an essential role in the effective and efficient control of BISDN. Additionally, due to the nonstationarity of the network and its complexity, a number of different (dynamic) modelling techniques are required at each level of the hierarchy. Also, to increase the efficiency of the network and allow flexibility in the control actions (by extending the control horizon) the (dynamic) tradeoff between service-rate, buffer-space, cell-delay and cell-loss must be exploited. In this thesis we take account of the above and solve three essential control problems, required for the effective control of BISDN. These solutions are suitable for both stationary and nonstationary conditions. Also, they are suitable for implementation in a decentralised coordinated form, that can form a part of a hierarchical organisation of control tasks. Thus, the control schemes aim for global solutions, yet they are not limited by the propagation delay, which can be high in comparison to the dynamics of the controlled events. Specifically, novel control approaches to the problems of Connection Admission Control (CAC), flow control and service-rate control are developed. We make use of adaptive feedback and adaptive feedforward control methodologies to solve the combined CAC and flow control problem. Using a novel control concept, based on only two groups of traffic (the controllable and uncontrollable group) we formulate a problem aimed at high (unity) utilisation of resources while maintaining quality of service at prescribed levels. Using certain assumptions we have proven that in the long term the regulator is stable and that it converges to zero regulation error. Bounds on operating conditions are also derived, and using simulation we show that high utilisation can be achieved as suggested by the theory, together with robustness for unforeseen traffic connections and disconnections. Even with such a high efficiency and strong properties on the quality of service provided, the only traffic descriptor required from the user is that of the peak rate of the uncontrollable traffic. A novel scheme for the dynamic control of service-rate is formulated, using feedback from the network queues. We use a unified dynamic fluid flow equation to describe the virtual path (VP) and hence formulate two illustrative examples for the control of service-rate (at the VP level). One is a nonlinear optimal multilevel implementation, that features a coordinated decentralised solution. The other is a single level implementation that turns out to be computationally complex. Therefore, for the single level implementation the costate equilibrium solution is also derived. For the optimal policies derived, we discuss their implementation complexity and provide implementable solutions. Their performance is evaluated using simulation. Additionally, using an ad hoc approach we have extended previous published works on the decentralised coordinated control of large scale nonlinear systems to also deal with time-delayed systems.
Copyright © 1993 Andreas Pitsillides.
Submitted in fulfillment of the requirements for the degree of Doctor of Philosophy, Swinburne University of Technology, 1993.