Overlay network resource allocation using a decentralized market-based approach

We present a decentralized market-based approach to resource allocation in a heterogeneous overlay network. This resource allocation strategy dynamically assigns resources in an overlay network to requests for service based on current system utilization, thus enabling the system to accommodate fluctuating demand for its resources. Our approach is based on a mathematical model of this resource allocation environment that treats the allocation of system resources as a constrained optimization problem. From the solution to the dual of this optimization problem, we derive a simple decentralized algorithm that is extremely efficient. Our results show the near optimality of the proposed approach through extensive simulation of this overlay network environment. The simulation study utilizes components taken from a real-world middleware application environment and clearly demonstrates the practicality of the approach in a realistic setting.     

Centralized and decentralized asynchronous optimization ofstochastic discrete-event systems

We propose and analyze centralized and decentralized asynchronous control structures for the parametric optimization of stochastic discrete-event systems (DES) consisting of K distributed components. We use a stochastic approximation type of optimization scheme driven by gradient estimates of a global performance measure with respect to local control parameters. The estimates are obtained in distributed and asynchronous fashion at the K components based on local state information only. We identify two verifiable conditions for the estimators and show that if they, and some additional technical conditions, are satisfied, our centralized optimization schemes, as well as the fully decentralized asynchronous one we propose, all converge to a global optimum in a weak sense. All schemes have the additional property of using the entire state history, not just the part included in the interval since the last control update; thus, no system data are wasted. We include an application of our approach to a well-known stochastic scheduling problem and show explicit numerical results using some recently developed gradient estimators     

Sequential stability and optimization of large scale decentralized systems

The notion of sequential stability in the synthesis of decentralized control for large scale systems is introduced in this paper. This notion is concerned with the property of a synthesis technique which allows the decentralized controllers of a large scale system to be connected to the systems one at a time (in a sequential way) such that the controlled system remains stable at all times. The motivation for introducing this constraint is that in practical terms, it is generally impossible to connect all decentralized controllers to a system simultaneously (due to the difficulties of communication etc.). A practical design procedure for the synthesis of a decentralized robust regulator for the servomechanism problem, based on a sequential approach to system design, is then given. The design procedure proceeds in two stages: (1) decentralized controllers are initially connected to the system in a sequential way to guarantee stability; (2) the parameters of the decentralized controllers are then sequentially adjusted, in a way to guarantee stability, so as to optimize a given performance index for the system. Applications of the above procedure are then made to the synthesis of centralized multivariable controllers and to the decentralized robust control of unknown systems.A simple example is given to illustrate the design synthesis.     

Learning users' interests by quality classification in market-based recommender systems

Recommender systems are widely used to cope with the problem of information overload and, to date, many recommendation methods have been developed. However, no one technique is best for all users in all situations. To combat this, we have previously developed a market-based recommender system that allows multiple agents (each representing a different recommendation method or system) to compete with one another to present their best recommendations to the user. In our system, the marketplace encourages good recommendations by rewarding the corresponding agents who supplied them according to the users' ratings of their suggestions. Moreover, we have theoretically shown how our system incites the agents to bid in a manner that ensures only the best recommendations are presented. To do this effectively in practice, however, each agent needs to be able to classify its recommendations into different internal quality levels, learn the users' interests for these different levels, and then adapt its bidding behavior for the various levels accordingly. To this end, in this paper, we develop a reinforcement learning and Boltzmann exploration strategy that the recommending agents can exploit for these tasks. We then demonstrate that this strategy does indeed help the agents to effectively obtain information about the users' interests which, in turn, speeds up the market convergence and enables the system to rapidly highlight the best recommendations.     

Practical stabilization of robotic systems by decentralized control

The practical stability of large-scale robotic systems with variable parameters is considered. The control should ensure the system state to belong to a finite region around the nominal trajectory for various values of parameters. The robotic system is considered as a set of decoupled subsystems each of which corresponds to one degree of freedom. For each decoupled subsystem a local controller is synthesized ensuring the practical stability of free subsystem. Then the practical stability of the coupled global system is analysed for various values of mechanical parameters. This permits the synthesis of decentralized control which provides practical stabilization of robotic systems in given finite regions and for the given set of allowable parameter values. Global control is also introduced. Decentralized control for a manipulation robot with variable payload is synthesized.     

On regularizing singular systems by decentralized output feedback

This paper considers linear time-invariant decentralized singular systems which are either nonregular or, if they are regular, they have impulsive modes. It derives algebraic necessary and sufficient conditions for making a singular system both regular and impulse-free by decentralized output feedback control laws and decentralized proportional-plus-derivative output feedback control laws     

Optimal observations in decentralized systems

We study here the problem of improving the performance of decentralized regulators by optimizing the constrained observation matrices. A numerical example illustrates the solution.     

Stabilization of decentralized control systems by means of periodic feedback

This paper deals with structurally constrained periodic control design for interconnected systems. It is assumed that the system is linear time-invariant (LTI), observable and controllable, and that its modes are distinct and nonzero. It is shown that the notions of a quotient fixed mode (QFM) and a structured decentralized fixed mode (SDFM) are equivalent for this class of systems. Then, it is proved that if the system is decentrally stabilizable, then one candidate for the decentralized stabilizing controller is a time-varying one consisting of a decentralized LTI discrete-time compensator and a zero-order hold. More specifically, the non-quotient fixed modes of the system will be eliminated via sampling for almost all sampling periods, while any QFM will still remain a fixed mode. The results obtained are ultimately extended to the case when the system has some repeated modes, none of which is a DFM.     

A decentralized control of interconnected systems using neural networks.

We develop a decentralized neural-network (NN) controller for a class of large-scale nonlinear systems with the high-order interconnections. The controller is a mixed NN comprised of a conventional NN and a special NN. The conventional NN is used to approximate the unknown nonlinearities in the subsystem, while a special NN is used to counter the high-order interconnections. We prove that this NN structure can achieve a stable controller for the large-scale systems.     

Reliable decentralized stabilization of linear systems

Reliable stabilization of linear time-invariant multi-input/multi-output plants is considered using a two-channel decentralized controller configuration. Necessary and sufficient conditions are obtained for existence of reliable controllers that maintain stability under the possible failure of either one of the two controllers. All decentralized controllers that achieve reliable stabilization are characterized     

Fixed zeros of decentralized control systems

Considers the notion of decentralized fixed zeros for linear, time-invariant, finite-dimensional systems. For an N-channel plant that is free of unstable decentralized fixed modes, an unstable decentralized fixed zero of channel i (1⩽i⩽N) is defined as an element of the closed right half-plane, which remains as a blocking zero of that channel under the application of every set of N-1 controllers around the other channels, which make the resulting single-channel system stabilizable and detectable. The paper gives a complete characterization of unstable decentralized fixed zeros in terms of system-invariant zeros     

On decentralized stabilization of interconnected systems

A decentralized control scheme is proposed for stabilization of interconnected systems consisting of arbitrarily connected, linear, time-invariant multivariable subsystems. Sufficient conditions are given for an interconnected system to be stabilized using only local state feedback. The obtained results are illustrated by an example.     

Control and stabilization of decentralized systems

The control and stabilization of decentralized systems attracted a great deal of attention in the late 1970s and 1980s, in part because the microprocessor revolution made it feasible to envision the control of large, complex systems. Most of the stability research undertaken during this era was analytic, focusing for example on various interpretations of weak or strong coupling between subsystems, the existence of multiple time-scales, etc. Motivated by the so-called ‘twin-lift’ problem, a problem of stabilizing the motion of two helicopters manoeuvring a single load, we develop a methodology for stabilizing classes of decentralized systems based on a more algebraic approach involving the external symmetries of decentralized systems. The study of the symmetry algebra of classes of such systems was initiated by Hazewinkel and Martin and in this language, we derive stabilizing, local feedback laws for any class of decentralized systems having a semi-simple algebra of symmetries. The twin-lift problem as well as certain problems involving stabilization of discretizations of distributed parameter problems have semi-simple algebras of symmetries.     

Stability analysis of a multiple-model controller for constrained uncertain nonlinear systems

This paper outlines the construction and the analysis of a multiple-model-based controller in order to deal with the stability of uncertain systems subject to constraints on its state and its control. It is assumed that the process can be defined by a finite number of models. The multiple-model approach is first formally introduced, and then the stability analysis is performed by using the vector norms and overvaluing model frameworks. We demonstrate through original theorems how the multiple-model approach improves the sizes of the stability and attractive subsets first when only one model is sufficient to represent the process and then when a collection of its models is used.     

The power dissipation method and kinematic reducibility of multiple-model robotic systems

This paper develops a formal connection between the power dissipation method (PDM) and Lagrangian mechanics, with specific application to robotic systems. Such a connection is necessary for understanding how some of the successes in motion planning and stabilization for smooth kinematic robotic systems can be extended to systems with frictional interactions and overconstrained systems. We establish this connection using the idea of a multiple-model system, and then show that multiple-model systems arise naturally in a number of instances, including those arising in cases traditionally addressed using the PDM. We then give necessary and sufficient conditions for a dynamic multiple-model system to be reducible to a kinematic multiple-model system. We use this result to show that solutions to the PDM are actually kinematic reductions of solutions to the Euler-Lagrange equations. We are particularly motivated by mechanical systems undergoing multiple intermittent frictional contacts, such as distributed manipulators, overconstrained wheeled vehicles, and objects that are manipulated by grasping or pushing. Examples illustrate how these results can provide insight into the analysis and control of physical systems.     

Application of a multiple-model adaptive filter method to inertial navigation systems

The multiple-model adaptive filter (MMAF) method is applied to the estimation of error states of inertial navigation systems (INS). Monte Carlo simulations are performed to evaluate the sensitivity of several MMAFs to uncertainties in flight condition, where a Doppler radar receiver or Omega receiver is considered as the reference information source. It is shown that the MMAF method is useful not only for a case where the actual system model is included within the candidate models, but also for a case where the actual system model is not included within the candidate models.     

Optimal decentralized control of dynamic systems

In this paper, several aspects of decentralized control theory applied to dynamic systems are studied. First of all, some classical definitions about matricial functions and new results on gradient calculations are presented. In the following we generalize to matricial problems the method of gradient projection of Rosen. Finally, some aspects of stability, initialization and initial condition independence are studied in detail, and two numerical examples are considered in order to emphasize the advantages of the given procedure: the decentralized Kalman filter and the optimal power-frequency control.     

Feedback gain optimization in decentralized eigenvalue assignment

This paper develops a new design procedure for minimizing the norm of a decentralized output feedback matrix which assigns a user specified set of eigenvalues. Two distinct approaches which can be meshed together as a unified design tool are derived. Assuming one has computed a decentralized feedback matrix which assigns a desired spectrum, the first approach describes an iterative algorithm which reduces an algebraic cost function (e.g. the Frobenius norm) of the feedback gains while maintaining the desired spectrum. This algorithm allows for small movements in the eigenvalues. The iteration step is based on the first order variational behaviour of the eigenvalue-eigenvector equations. The second algorithm modifies a continuation method for decentralized eigenvalue assignment to include an optimizing factor. Numerical considerations for the design procedures are discussed. An example showing the improvement possible by the application of the procedure is also given.     

Elimination of fixed modes in decentralized systems by means of sampling

In this contribution it is shown that fixed modes in continuous-time decentralized control systems can be eliminated by the use of sampling. Criteria are derived which distinguish the fixed modes which are still present in the sampled-data system from those which are eliminated.     

Robust stabilization of nonlinearly perturbed large-scale systems by decentralized observer-controller compensators

This paper is concerned with the problem of robust stabilization for nonlinearly perturbed large-scale systems via decentralized observer-controller compensators. The large-scale system is composed of several interconnected perturbed subsystems, each containing a nonlinearly perturbed plant and an observer-controller compensator. Here a robust stability criterion for the perturbed large-scale system is introduced, and a simple but useful inequality is derived for synthesizing the decentralized observer-controller compensators. The main features of this paper are as follows: (i) the nominal plant of each subsystem is not constrained to be stable and/or minimum phase; (ii) the perturbations of the plants and/or interconnections are considered; and (iii) a two degree observer-controller compensating scheme is employed to treat the perturbed large-scale systems.