Altuğ İftar

Rate-based flow controllers for communication networks in the presence of uncertain time-varying multiple time-delays

An H^~ based robust controller is designed for a rate-feedback flow-control problem in single-bottleneck communication networks. The controller guarantees stability robustness to uncertain time-varying multiple time-delays in different channels. It also brings the queue length at the bottleneck node to the desired steady-state value asymptotically and satisfies a weighted fairness condition. Lower bounds for stability margins for uncertainty in the time-delays and for the rate of change of the time-delays are derived. A number of simulations are included to demonstrate the time-domain performance of the controller. Trade offs between robustness and time-domain performance are also discussed.

Decentralized estimation and control with overlapping input, state, and output decomposition

Abstract The extension principle, which was introduced recently, is generalized to the case of input, state, and output expansion. Estimator and controller design problems are considered within the framework of the extension principle. Decentralized estimator and controller design with overlapping decompositions is also discussed within the same framework. It is shown that if the extension principle is used then any estimator or controller designed in the expanded spaces is contractible to the original spaces for implementation. Furthermore, it is shown that if an estimator designed for the expanded system achieves good estimation and/or asymptotic estimation then the contracted estimator also achieves good estimation and/or asymptotic estimation for the original system. Similarly, it is also shown that if a controller designed for the expanded system achieves stability and/or good performance then the contracted controller achieves stability and/or good performance for the original system.

Overlapping decentralized dynamic optimal control

Abstract An optimal dynamic output feedback controller design approach is proposed for systems which are required to have an overlapping feedback structure. Such a structure may be either imposed due to some design considerations or selected to match the overlapping structure of the given system. The proposed approach first involves transforming the given system into a larger dimensional system. Decentralized optimal controllers are then designed for this expanded system. These controllers are contracted for implementation on the original system in the final phase. It is shown that if the controller designed for the expanded system achieves stability and satisfies the necessary conditions of optimality for the expanded system, then the contracted controller achieves stability and satisfies the necessary conditions of optimality for the original system. Furthermore, the costs for the original and the expanded systems are equal. The details of overlapping controller design for a particular pattern of overla...

Overlapping decompositions, expansions, contractions, and stability of hybrid systems

Inclusion and extension principles are defined and studied for hybrid systems. The controller design problem for hybrid systems is then considered within the framework of the extension principle. It is shown that if the extension principle is used, then any controller designed in the expanded spaces is contractible to the original spaces for implementation. The stability of hybrid systems is also considered. A number of definitions of stability for hybrid systems are given and it is shown that if a system, which includes another system, is stable, then the included system is also stable. Furthermore, it is shown that, if a controller designed for the expanded system achieves stability, then the contracted controller also achieves stability for the original system. Decentralized controller design with overlapping decompositions is also discussed within the framework of extension. An example is provided to illustrate the proposed approach.

Overlapping decompositions and expansions of Petri nets

Overlapping decompositions and expansions are considered to design decentralized controllers for discrete-event systems (DESs) modeled by Petri nets. The inclusion principle for Petri nets is first defined. It is shown that properties like boundedness, reversibility, and liveness (with a mild additional condition) carry over from the including net to the included net. Moreover, a new property called obstruction, is introduced for the including net, and it is shown that if obstruction does not occur in the including net, then deadlock does not occur in the included net. An expansion procedure, which guarantees inclusion for an overlappingly decomposed Petri net, is then introduced.

Decentralized Robust Control for Dynamic Routing of Large Scale Networks

A variable structure decentralized controller for dynamic routing in multi-destination large-scale data communication networks is presented. A dynamic model, which can incorporate different processing delays at different nodes, is developed to describe the network dynamics. It is shown that the proposed controller guarantees stability and clears the queues of the system in the absence of external input flows. The controller also keeps the queue lengths bounded in the presence of external flow rates, which do not exceed a certain maximum flow rate (obtained by solving a LP optimization problem - there is no solution to the routing problem if these rates are exceeded). Furthermore, the controller is also robust with respect to variations in the processing delay constants. Some simulation results are presented to illustrate the controllers performance in a number of practical cases.

A decentralized discrete-time controller for dynamic routing

A decentralized controller for dynamic routing in multi-destination large-scale data communication networks is presented. A dynamic model, which can incorporate possibly time-varying different processing delays at different nodes, is developed to describe the network dynamics. A knowledge of these delays is, however, not required for the actual implementation of the controller. It is assumed that the rate of messages being sent out from one node to another (the control signals) are updated at discrete time instants, which is the usual case in practice. The proposed controller is decentralized in the sense that all on-line computations can be done locally at the individual nodes without any information transfer from the other nodes. The design of the controller involves a linear programming optimization problem, which can be solved off-line. The resulting controller has the property that it allows the maximum possible magnitude on each external message arrival rate of the system to occur, without violating...

Centralized and decentralized supervisory controller design to enforce boundedness, liveness, and reversibility in Petri nets

Supervisory controller design to enforce boundedness, liveness, and reversibility in Petri nets is considered. The Petri nets considered may have non-unity weight arcs and both controllable and uncontrollable transitions. Algorithms for a centralized controller design approach are first developed. The developed algorithms always find a controller whenever it exists. This controller enforces boundedness, liveness, and reversibility; it also avoids deadlock. Furthermore, it is shown that the controller obtained is the least restrictive controller among all controllers which enforce desired properties. A decentralized controller design approach, based on overlapping decompositions, is then introduced. Algorithms to design decentralized controllers based on this approach are also developed. These controllers, when they exist, also guarantees boundedness, liveness, reversibility and deadlock freeness. The decentralized controllers have two main advantages over the centralized ones. First, they have reduced on-line computation and communication requirements. Second, the computational time required to design decentralized controllers is considerably less than that required for centralized controllers.

Robust rate-based flow controllers for high-speed networks: the case of uncertain time-varying multiple time-delays

An H/sup /spl infin// based robust controller is designed for a rate-feedback flow-control problem in single-bottleneck data-communication networks. The controller is robust to uncertain time-varying multiple time-delays in different channels. The controller forces the queue length at the bottleneck node to the desired steady-state value asymptotically and also satisfies a weighted fairness condition. Some simulations are also presented to demonstrate the time domain performance of the controller.

Decentralized routing controller design using overlapping decompositions

The use of overlapping decompositions in designing routing controllers for large-scale networks is considered and a decentralized dynamic routing controller design strategy is proposed. The strategy is based on the overlapping decompositions method and can be used for large-scale networks which have more than two overlapping subnetworks and/or have more than one node in the overlapping part. The controller obtained by the proposed strategy is decentralized in the sense that all on-line computations can be done locally at the individual nodes without any information transfer from the other nodes. The controller satisfies all the routing control constraints and it avoids any looping. The controller also clears the queues in finite time in the absence of external arrivals and it keeps the queue lengths bounded in the presence of external arrival rates which do not exceed a certain maximum rate. In order to illustrate the proposed controllers performance, the controller is applied to an example network and s...