Biography:

In the past K. Warwick has collaborated on articles with C. Kambhampati and E. Al-Gallaf. One of their most recent publications is Brief paperThe relative order and inverses of recurrent networks☆. Which was published in journal Automatica.

More information about K. Warwick research including statistics on their citations can be found on their Copernicus Academic profile page.

K. Warwick's Articles: (9)

Brief paperThe relative order and inverses of recurrent networks☆

AbstractDifferential geometry is used to investigate the structure of neural-network-based control systems. The key aspect is relative order—an invariant property of dynamic systems. Finite relative order allows the specification of a minimal architecture for a recurrent network. Any system with finite relative order has a left inverse. It is shown that a recurrent network with finite relative order has a local inverse that is also a recurrent network with the same weights. The results have implications for the use of recurrent networks in the inverse-model-based control of nonlinear systems.

Brief paperThe relative order and inverses of recurrent networks☆

AbstractDifferential geometry is used to investigate the structure of neural-network-based control systems. The key aspect is relative order—an invariant property of dynamic systems. Finite relative order allows the specification of a minimal architecture for a recurrent network. Any system with finite relative order has a left inverse. It is shown that a recurrent network with finite relative order has a local inverse that is also a recurrent network with the same weights. The results have implications for the use of recurrent networks in the inverse-model-based control of nonlinear systems.

Brief paperThe relative order and inverses of recurrent networks☆

AbstractDifferential geometry is used to investigate the structure of neural-network-based control systems. The key aspect is relative order—an invariant property of dynamic systems. Finite relative order allows the specification of a minimal architecture for a recurrent network. Any system with finite relative order has a left inverse. It is shown that a recurrent network with finite relative order has a local inverse that is also a recurrent network with the same weights. The results have implications for the use of recurrent networks in the inverse-model-based control of nonlinear systems.

Force distribution in manipulation by a robot hand with equality and inequality constraints

AbstractThe problem of the appropriate distribution of forces among the fingers of a four-fingered robot hand is addressed. The finger-object interactions are modelled as point frictional contacts, hence the system is indeterminate and an optimal solution is required for controlling forces acting on an object. A fast and efficient method for computing the grasping and manipulation forces is presented, where computation has been based on using the true model of the nonlinear frictional cone of contact. Results are compared with previously employed methods of linearizing the cone constraints and minimizing the internal forces.

The Error Function in Low Order Modelling of Discrete-Time Systems

AbstractThe modelling and adaptive control of high order discrete time systems is investigated in terms of the time domain error between the system and its associated model. The model is assumed to be of order less than that of the original system and the error is regarded as an input function filtered by characteristics of both system and model. It is shown how reduced order models can be obtained by means of specifying certain of the parameters contained within the error function and it is considered how standard order reduction techniques, such as Padé approximation and Markov parameter matching, can be carried out by employment of this method. The modelling procedure can also be incorporated in on-line, real-time adaptive controllers, a section is therefore devoted to the description of a self-tuning control scheme which is used to control a high order plant with a low order controller.

USE OF STOCHASTIC NEURAL NETWORKS FOR PROCESS CONTROL

AbstractProcess control engineers now have a powerful tool, in the form of Artificial Neural Networks. These networks can be made to model arbitrary non-linear functions. This capability of modelling non-linear mappings is of immensense value to the process engineer in as much that he now has a tool which is capable of reflecting most, if not all, of the complexities of chemical processes. Most neural networks utilise sigmoid type activation functions. The use of these activation functions, leads to complicated learning rules like the back propagation algorithm. However, if Gaussian (stochastic) activation functions are utilised the learning algorithms are greatly simplified. In this paper networks with Guassian activation functions are used to identify tnot only he input-output relationship of a CSTR, but also the inverse relationship. Identification of the inverse relationship helps in formulating an IMC type controller. The performance of the neural based IMC controller is compared t o a neural based predictive controller.

A stable one-step-ahead predictive control of non-linear systems☆

AbstractIn this paper stability of one-step ahead predictive controllers based on non-linear models is established. It is shown that, under conditions which can be fulfilled by most industrial plants, the closed-loop system is robustly stable in the presence of plant uncertainties and input–output constraints. There is no requirement that the plant should be open-loop stable and the analysis is valid for general forms of non-linear system representation including the case out when the problem is constraint-free. The effectiveness of controllers designed according to the algorithm analyzed in this paper is demonstrated on a recognized benchmark problem and on a simulation of a continuous-stirred tank reactor (CSTR). In both examples a radial basis function neural network is employed as the non-linear system model.

Automatic Integrated System Load Forecasting Using Mutual Information and Neural Networks

AbstractThis paper describes a novel application of ANNs for electrical load forecasting. The production of a forecast is divided into two phases: Self Organising Feature Maps are used to mine the available data, identifying measurements relevant to the forecast. These measurements are then used to train a Multi Layer Perceptron to provide load forecasts. The resultant non-linear forecaster is compared with a similar, but linear system employing multivariate linear regression and results are presented.

Advertisement
Join Copernicus Academic and get access to over 12 million papers authored by 7+ million academics.
Join for free!

Contact us