No description available.

No description available.

No description available.

In this thesis, contributions are made to the general problem of identifying linear, multivariate systems (with more than one input and more than one output) from operating data for purposes of control. The problem is shown to be the proposal of a suitable mathematical model structure followed by the task of estimating parameters within the structure by statistical methods. From the initial assumption that the systems of interest have a state-space description, deterministic canonical forms having fewer parameters than the original description are derived by means of linear transformations. The methods for construction of these transforms are related to the conditions for controllability and observability. The method for transforming an important new form, the A-canonical form, into a vector differential (or difference) equation in the input and output variables only is developed. The inverse problem, of transforming back into the A-canonical form is solved. Then both descriptions are extended to include structure for stochastic inputs that appear as correlated disturbances at the system output. The problem of estimating the parameters of the multivariate stochastic difference equation is investigated. It is found that estimations by conventional least squares leads to unsuitable estimates. Consequently, a new, modified method of least squares is developed and shown to be asymptotically unbiased. Algorithms for estimation by maximum likelihood methods are also developed. A number of examples are computed to illustrate the effectiveness of the algorithms and for comparison purposes. A simple example of stochastic control using a model, the parameters of which are estimated by running data, is shown. In the final chapter, some possible extensions of the work are considered.

Policies are descriptive and provide information which can be used to modify the behaviour of a system without the need of recompilation and redeployment. They are usually written in a policy definition language which allows end users to specify their requirements, preferences and constraints. Policies are used in many software application areas: network management, telecommunications, security, and access control are some typical examples. Ponder, KAoS, Rein, XACML, and WSPL are examples of policy definition languages. These languages are usually targeted at a specific domain, hence there is a plethora of languages. APPEL (the Adaptable Programmable Policy Environment Language) [69] has followed a different approach: It is a generic policy description language conceived with a clear separation between core language and its specialization for concrete domains. So far, there has not been any formal method for the extension and domain specialization of the APPEL policy language. Policy conflict can occur when a new or a modified policy is deployed in a policy server, which leads to unspecified behaviour. To make policy based systems conflict free it is necessary to detect and resolve conflicts before they occur, otherwise the intended behaviour of a policy cannot be guaranteed. We introduce a structural modelling approach to specialize the policy language for different domains, implemented in the VIATRA2 graph transformation tool. This approach is applied to APPEL. Our method for conflict analysis is based on the modelling methodology. As conflicts depend on domain knowledge, it is sensible to use this knowledge for conflict analysis. The identified conflicting actions are then encoded in the ALLOY model checker that confirm the existence of actual and potential conflicts.

This research is concerned with IS practice and how that practice can be improved through the novel notion of systems harmony. It draws on Peter Checkland's work to explore the principles and practice of aikido, the Japanese 'art of harmony'. The thesis investigates systems harmony in three stages. Firstly, it looks at aikido, to uncover its underlying philosophical foundations and techniques of harmonisation. The empirical focus of this phase concentrates primarily on the lived experiences of aikido practitioners. The key theoretical output from this stage is a conceptual model, the AikiSys model, which describes the complex interconnected web of factors that contribute to the dynamic emergence of harmony within a martial encounter. Next, the AikiSys model is used to fonn the basis of a practical sense-making mechanism, the information Systems Harmonisation (ISH) framework. This framework enables the transfer of the essentially personal constructs of aikido into the IS domain. This conceptual linking is realised by associating elements of the core AikiSys model to analogous concepts and techniques operating within IS practice. The final stage of the research uses the ISH framework in two modes - one diagnostic and retrospective, the other guiding and proactive - to inquire into IS practice. Four case studies are presented to reveal the new perspectives offered by the ISH framework. The study introduces a framework that provides a way of engaging with IS situations in an action-oriented fashion. This framework includes iteration between hard and soft approaches and makes specific reference to non-technical manifestations of harmony such as respect, trust, multiple-perspectives, elegance, emergence, self-organisation, training and awareness. The research suggests that these elements, far from being peripheral, are in fact crucial to successful real-world IS practice.

This thesis is concerned with the theory and simulation of the static and dynamic properties of discrete-potential systems. Discrete potentials are a class of models used to describe the interactions between particles in many-body systems and have applications in many fields. To investigate the dynamics of systems with discrete potentials, DYNAMO, a general event-driven molecular dynamics simulation package was developed. This software is one of the first event driven implementations to exhibit linear scaling of the computational cost. with the system size, which is the theoretically optimal scaling. DYNAMO also implements stochastic interactions, including a novel extension of DSMC proposed to attempt to include pre-collision velocity correlations, which are significant in dense or complex fluids. Using a phenomenological correction for the dynamical spatial effects, the predictions for the self diffusion coefficient of the hard sphere fluid are improved over Enskog theory at low to moderate densities. DYNAMO was used to examine a number of systems to determine their behaviour and to test theoretical predictions. Accurate simulations of large systems of hard spheres were used to test recent equations of state including a novel exponential closure of the virial series. This closure performs well in comparison to other theoretical equations of state for the fluid phase, and a revised semi-empirical equation of state is proposed for the solid phase. The collision statistics of sheared granular materials were studied to highlight the fundamental differences between elastic and inelastic systems and to critically compare Enskog theory and its approximate solutions. Overall, Enskog theory performs well; however, care must be taken with the predictions of the energy anisotropy and collision statistics. The effect of nano-colloids on the transport properties of a fluid were investigated using the binary hard sphere model. A 'fines' effect in the viscosity and a strong thermal conductivity enhancement was observed in the simulations, which are absent from the Enskog theory predictions. Finally, a simple square-well homopolymer model is examined using constant temperature molecular dynamics, coupled with replica exchange techniques. This polymer forms helical structures, and the control parameters of the model (e.g., chain stiffness, range of the attractive interactions) were completely characterised to investigate the nature and stability ofthis surprising folding behaviour. The results of all of these systems highlight the speed and versatility of both the DYNAMO simulation package and the underlying richness of discrete potential models.

The plan of the thesis is as follows: The introductory chapters review the literature pertaining to human prediction and predictive control models (Chapter 1), and to engineering aspects of predictive displays (Chapter 2). Chapter 3 describes a fundamental study of predictive display parameters in a laboratory scheduling task, Chapter 4 attempts to verify these findings using test data from an actual job shop scheduling problem. Chapter 5 branches into the area of continuous control with a pilot study of predictive displays in a laboratory simulated continuous stirred-tank chemical reactor. Chapter 6 uses the experience gained in the pilot study as the basis for a comprehensive study of predictive display parameters in a further laboratory study of a simplified dual-meter monitoring and control task, and Chapter 7 attempts to test the optimal design in a part-simulated semi-batch chemical reactor using real plant and experienced operators in an industrial setting. The results of the experimental programme are summarized for convenience in Chapter 8. Chapter 9 draws together the threads from the various experiments and discusses the findings in terms of a general hierarchical model of an operator's control and monitoring behaviour. Finally, Chapter 10 presents conclusions and recommendations from the programme of research, together with suggestions for further work.

This thesis explores the application of Chaos Theory to forecast urban traffic conditions. The research takes advantage of a highly resolved temporal and spatial data available from the Split Cycle Optimisation Technique (SCOOT) system, in order to overcome the limitations of previous studies to investigate applying Chaos Theory in traffic management. This thesis reports on the development of a chaos-based algorithm and presents results from its application to a SCOOT controlled region in the city of Leicester, UK. A Phase Space Reconstruction method is used to analyse non-linear data from the SCOOT system, and establishes that a 20 second resolved data is suitable for understanding the dynamics of the traffic system. The research develops the Lyapunov exponent as a chaos-based parameter to forecast link occupancy using a multiple regression model based on the temporal and spatial relationships across the links in the network. The model generates a unique forecast function for each link for every hour of the day. The study demonstrates that Lyapunov exponents can be used to predict the occupancy profile of links in the network to a reasonably high level of accuracy (R-values generally greater than 0.6). Evidence also suggests that the predictions from the Lyapunov exponents (rather than occupancy) make it possible to report on the impending conditions over a wider part of the network so that imminent congested conditions can be foreseen in advance and mitigation measures implemented. Thus, the thesis concludes that incorporating chaos-based algorithms in this way can enable urban traffic control systems to be one-step ahead of traffic congestion, rather than one-step behind. This would improve the management of traffic on a more strategic level rather than purely within smaller network regions thus playing an important role in improving journey times and air quality and making a vital contribution to mitigating climate change.

No description available.