Available PhD projects in Systems & Control
Dr Eduardo Alonso
Smart control for energy systems
Eduardo Alonso and Professor Keith Pullen
Smart systems, that is, systems that show autonomy in decision-making are becoming extensively used in engineering energy systems from the control of microgrids and motors for electric vehicles to smart buildings and smart meters. In particular, we have developed an innovative method, winner of the first prize of the European Institute of Innovation & Technology on Smart Energy Systems 2014, for the efficient control of grid-connected converters. In short, we train a recurrent neural network with novel adaptive dynamic programming algorithms. Results in simulation and hardware-in-the-loop equipment show that our technology consistently outperforms conventional proportional-integral controllers. We plan to apply our technology to other scenarios, such a microgrids with a flywheel.
Mathematical modelling of complex systems
Eduardo Alonso and Nikos Karcanias
Complexity is pervasive in both natural and man-made systems, which can be characterized by the emergence of properties that cannot be reduced to their constituent elements and their hierarchical structure. Complex systems have been traditionally studied as non-linear dynamic systems, paying special attention to their evolution under various constraints. Complementary to this approach we are interested in exploring the fundamental structure of complex systems using the concept of symmetry and in using abstract algebra, category theory in particular, to formalise it. The main idea is to provide a basic mathematical framework for the definition of systems as collections of elements that remain invariant under certain transformations and to built on it to represent how complexity emerges and structures in hierarchies through symmetry breaking. The analysis will be embedded in real-life systems, such as a traffic network.
Computational models of neuro-behavioural systems
Eduardo Alonso and Esther Mondragón
The last 100 years has seen the progressive refinement of our understanding of the neuro-behavioural mechanisms of classical conditioning. In such context, it is widely acknowledged the importance of studying them from a computational perspective: On the one hand, implementing a model requires precise definitions –be it in the form of a specific programming language or as a formal model. On the other hand, algorithms allow us to execute calculations rapidly and, most importantly, accurately. Automation is critical, particularly when the models are described in non-linear equations that can only be solved numerically. Crucially, the outputs of a simulation feedback the psychological models –thus becoming an essential part of the cycle of theory formation and refinement. This project aims at developing computational models and simulators of classical conditioning with special attention given to the representation of stimuli and their associations, the role of attention in learning, and how it evolves in time.
Professor Tom Chen
Cyber security of wearables and Internet of Things
Wearables are the most "personal" computing devices (such as fitness bands, smart watches, smart glasses). With increasing processing power and wireless communications, they are seeing a broad range of uses such as health monitoring, messaging, contactless payments, and communication with other smart IoT devices. Unfortunately wearables will also introduce new risks if they are stolen, lost, or compromised. Strong authentication of the owner to the wearable, and authentication between wearables and other smart devices, is essential and the focus of this project. Research may focus on biometric authentication and new authentication protocols, particularly protocols optimised for resource-constrained devices.
Adversarial risk modelling
Risk is a central concept in cyber security but has always been difficult to quantify, particularly the "likelihood" of particular threats. Adversarial risk modeling is different from the traditional approach to calculating risk by recognising that attackers and defenders are not static. A likelihood of threat is not possible to calculate because attackers and defenders are intelligent and constantly reacting to each other. Typically adversarial risk is modeled as a game between attacker and defender, each choosing the best strategy in response to the other player. This research project will advance the state-of-the-art in adversarial risk modeling by improving game theoretic models and relating games to real life environments (e.g., critical infrastructures) and real types of adversaries.
Game theory models for cyber conflicts
Whether cyber attacks are used for crime or warfare, they are a means to an end. They are part of a strategic plan to steal information, cause damage, make profits, or some other motive. This research aims to understand how strategic "cyber campaigns" are conducted by means of game theoretic models. Game theory has a long history of modeling conflicts. Game models may explain a number of important security questions, e.g. (i) when would a nation-state choose a cyber attack on an enemy state instead of traditional alternatives such as diplomacy or kinetic warfare? (ii) could deterrence work in cyber warfare? (iii) which nation-states are the most capable building up "cyber arms"? (iv) how should an organization or nation-state under cyber attack choose its "best" response?
Popular movies like Skyfall have shown the possible impact of cyber attacks carried out by terrorists. In real life, terrorists have shown an interest in developing cyber skills and are known to use the Internet for social media, recruiting, fund raising, activity planning, and gathering intelligence. But no major cyber attacks by terrorists have been seen so far. This research is interested in addressing questions such as (i) why have terrorists chosen to continue traditional attack methods instead of cyber attacks? (ii) what are terrorists planning for cyber attacks? (iii) what could be vulnerable targets for cyber terrorism? (iv) how much can monitoring of their Internet and messaging activities reveal about their organisation structure and attack plans? (v) how can terrorist use of the Internet and telecoms be an opportunity for law enforcement?
Dr Lambros Ekonomou
Intelligent solutions for electricity transmission and distribution networks
Electricity transmission and distribution networks continue to function in a manner that optimize cost and environmental performance without giving up traditionally high security and quality of supply, while host very large and further increasing penetration of renewable and distributed (dispersed) generation. Stimulation of local production of renewable energy requires the emergence of more intelligent transmission and distribution networks in a view of accommodating variable generation from multiple sources and a growing demand for renewable energy. The project aims to provide intelligent and innovative solutions that can be applied on electricity transmission and distribution networks to support these changes.
Professor George Halikias
Cooperative network control
The research aims to develop a methodology for tackling network modelling and control problems involving cooperating agents. In contrast to general systems, networks are typically characterised by a loose structure with reconfigurable interconnections and the interaction of numerous autonomous agents, each with its individual strategy and objectives. Cooperation is enforced using techniques which provide rapid agreement and teamwork between agents allowing effective task performance in self-organising networks (”consensus algorithm”). The ultimate objective is to control the emerging global network properties by implementing appropriate rules to structure the interaction between agents. The area has many important applications, such as flow control in communications networks, data fusion from multiple sensors, flocking of autonomous agents (mobile robots, UAV’s, highway platoons), supply chains, synchronisation of oscillators, distributed algorithms, etc..
Analysis and synthesis of piecewise-linear/switched dynamic systems
The proposed work aims to develop modelling and control design methods for piecewise linear dynamic systems. These are systems which switch abruptly at specific times or when the state vector crosses a region (e.g. hyperplane) in state-space and have a wide range of applicability in engineering, e.g. diodes, transistors, amplifiers, actuators with saturating nonlinearities, etc.. In addition, many smooth nonlinear systems are often approximated by piecewise-linear systems, e.g. mechanical or civil engineering structures are often modelled separately in their ”elastic” and ”inelastic” regions. In the proposed work it will be attempted to develop a fundamental theoretical understanding of the dynamic behaviour of piecewise linear dynamic systems and obtain results related to system properties such as stability, controllability, and observability. The work will examine problems of stabilisation and is likely to involve techniques arising in robust control, e.g. quadratic stability, dissipativity and optimisation of Lyapunov functions.
Optimisation methods for robust control
The project will investigate convex relaxation methods for calculating the structured singular value arising in robust control. Recent work has concentrated on techniques for breaching the duality gap between the structured singular value and its convex upper bound. It was shown that this is possible provided a reduced-rank problem s.s.v can be solved (at least approximately) along with a low-dimensional eigenvalue calculation. The proposed work will attempt to extend these results for various uncertainty structures and will concentrate on techniques applicable to low rank s.s.v problems.
Strong stability and stabilisation
The concept of strong stability has been recently introduced for linear and nonlinear autonomous systems. The concept applies to natural state-space descriptions and is related to non-overshooting responses. The project will investigate the concept originating from different choice of norms and its applicability to control design with limits in transient energy for both linear and nonlinear systems.
Modelling and control of supply chains
The project concentrates on modelling, simulation and control design work for supply chain models. The main aims are: (i) To develop fundamental understanding of the supply-chain “bullwhip-effect” and other dynamic instability phenomena inb SC’s related to the increased volatility in order sizes and inventory levels experienced by suppliers at remote locations in the chain. (b) To apply advanced control techniques for efficient supply-chain management: The main problem here is the decentralised nature in which control decisions are taken and the absence of global information.
Congestion-control of communication networks
The main objective of this work is to model and analyse typical flow control problems arising in communication networks, such as the Internet. Existing communication protocols, such as TCP, are capable of regulating the transmission rate of the networks sources, using feedback information obtained from a congestion window of transmitted packets. With the anticipated future growth in network complexity and the supported traffic, the capabilities of simple protocols will be stressed to their limit. The project investigates some recent approaches in this area, based on certain ideas of economic theory related to the decentralising-regulatory role of prices in competitive economies. The approach essentially involves the Lagrangian formulation of a constrained optimisation problem
(aggregate utility maximisation subject to capacity constraints), solved in its dual form via a distributed algorithm (e.g gradient projection or Newton-like algorithms). Specifically, each network link is associated with a price reflecting the instantaneous level of local traffic, and each source receives aggregate price information corresponding to those links to which it is connected. The overall aim of the resulting decentralised non-linear feedback scheme is to allow flows to dynamically adapt to external disturbances, related to varying traffic demand. Additional issues that need to be addressed
include the analysis of the dynamic behaviour of the network, especially its performance and stability (in the presence of uncertain time-delays), and issues related to scalability, fairness, quality of service and the distributed implementation of the algorithm.
Dr Ioannis Kaparias
Multi-objective optimisation of urban mobility management policy- and decision-making using key performance indicators
Cities today share common transport problems and objectives with respect to mobility management, and put great focus on Intelligent Transport Systems (ITS). The market offers decision-makers a variety of solutions, from which they are required to choose the most suitable and effective ones. Making this choice is a non-trivial task, however, especially given that transport problems are multi-dimensional by nature, and various objectives need to be met, such as traffic efficiency, safety, pollution reduction and social inclusion. As it is likely that trade-offs will have to be made between meeting different objectives to varying extents, the decision-making process can become a highly complex problem. The objective of this project is, hence, to develop and test a multi-objective optimisation algorithm for multi-objective urban mobility management decision-making. Objectives are to be defined on the basis of the recently formulated CONDUITS performance evaluation framework, which consists of a set of Key Performance Indicators (KPIs) for various themes of urban mobility management, and it is expected that the algorithm will be implemented in the purpose-developed CONDUITS-DST decision-support tool, which will also enable its validation through a real-world case study.
User-oriented travel time reliability modelling for public transport networks
The fact that travel time in road networks is not constant, but entails an element of variability, resulting in uncertainty when attempting to predict it, has been recognised for a long time. In fact, much research has highlighted the importance of this variability, and has concluded that it is often a greater concern for travellers than travel time itself. Hence, several methods have been developed for modelling what has been termed “travel time reliability”, defined as the probability of encountering delays, concentrating mostly on road traffic. As opposed to road networks, however, where traffic congestion can be easily identified as the sole source of uncertainty, passengers in public transport networks of large cities may be exposed to delays arising from a number of sources, such as service reliability and overcrowding. For example, important sources of delay in public transport, currently not considered by journey planners, are the passenger queuing and vehicle dwell times at stops and stations, which, from an operational perspective, may lead to passengers changing their route and mode choice, and sometimes even their final destination. Hence, the objective of this project is to extend the concept of travel time reliability to public transport networks with a view of developing more advanced and accurate algorithms for public transport journey planning and network operation.
Modelling and prevention of cyclist accidents through the Cyclist 360° Alert system
Cycling is an increasingly popular mode of travel in cities due to the great advantages that it offers in terms of space consumption, health and environmental sustainability, and is therefore favoured and promoted by many city authorities worldwide. The large number of recently introduced schemes in many cities worldwide (such as the Cycle Hire and the Cycle Super-Highway schemes in London) demonstrates this trend. However, the relatively low perceived safety of cycling from the users’ side currently presents itself as a major hurdle to the desired uptake of cycling as a real alternative to the private car. Accident numbers, unfortunately, confirm this perception as reality: as reported in the Times, in 2012 alone there were 122 cyclist fatalities in the whole of Britain. Recent trends in the development of ubiquitous computing and sensor technologies, however, offer ways to develop low-cost Intelligent Transport Systems (ITS) solutions for the prevention of such accidents. One such system is Cyclist 360° Alert, developed at City, the first stage of which has involved the instrumentation of a Santander Cycle Hire bicycle with MEMS sensors (iBike), enabling its tracking to a high positional accuracy. The objective of this project is, hence, to further contribute to the development of Cyclist 360° Alert by formulating and testing a modelling framework and user interface for the prediction and prevention of cyclist accidents. The project will use the iBike configuration to collect large samples of data from cyclist volunteers for the calibration of the model and will also involve real-world validation.
Professor Nicos Karcanias
Structure evolving systems in the context of passive electrical networks: The problem of growth and death
The project aims at investigating the modeling of systems whose structure is not fixed, but evolves during the system lifecycle. The modeling of passive electrical networks with variability in the interconnection topology and/or nature and value of their elements is to be considered and an appropriate representation of system variability is sought. A new implicit description is to be developed and issues of defining sets of inputs and outputs to shape the corresponding model and properties are to be investigated. The project is of theoretical nature and will involve network theory, graph analysis and control theory.
Measuring relative strength of systems properties and design strategies for global instrumentation
The problem of studying the relative degree of presence of system properties such as controllability, observability etc and then linking to the selection of input, output model structure is considered here. The objective is to develop criteria which may allow the linking of these properties to the state space model parameters and then develop design strategies. Amongst the issues to be considered is the role of feedback on the relative measures for system properties and the characterization, if possible of invariant measures. The project will examine various system properties and try to link the easy, or difficult nature of control design to the strength of presence of such properties.
Matrix pencils and robustness in geometric system theory
The subject of “approximate algebraic computations” provides a natural environment for approximating exact algebraic problems, by handling them in a robust numerical way. This involves the approximation of the fundamental algebraic notions such as that of the greatest common divisor and (GCD) and Least Common Multiple (LCM) which underpin the development of many crucial representations of algebraic system theory. The central goal of this research is to combine the results of the Matrix Pencil Approach to Geometric Theory with those of the Approximate Algebraic Computations in order to produce a new “robust” version of Geometric Theory, that can be supported by a suitable numerical computations framework. This is essential for transforming exact synthesis methods of the Geometric Systems Theory to design tools. From the conceptual viewpoint, this work aspires to extend the powerful geometric concepts of different types of invariant spaces to models characterised by parameter uncertainty and provide design versions to exact control synthesis procedures.
Simplification of complexity, and model structure evolution
The need for developing simple models for different stages of design, decision making and diagnostics is examined. Simplification can take place at the level of conceptual and qualitative modelling, physical modelling, data, signals and available complex quantitative models. Different lines of research may be pursued within this general framework. Amongst them some possible directions are: (i) Simplification of graphs and study of evolution in a set of models parametrized by some measure of complexity and study of graph based properties; (ii) Simplification of input-output models based on truncation of Laurent expansions and evolution of structural properties; (iii) Simplification of physical models based on assumptions on the properties of components and evolution of system properties; (iv) Impact of model simplification in optimisation problems.
System of systems: Towards a formal characterisation
The concept of “System of Systems” (SoS) has emerged in many fields of applications. Within this new challenging paradigm the notion of emergence is also frequently used in a rather loose way. The research aims to formalise the loose concept of “System of Systems” (SoS) within the context of Systems Theory whilst exploring and developing a conceptual framework for emergence that is suitable for further development and valid independently from the particular field of applications. The development of the of the overall area requires a formal definition of SoS beyond the phenomenological description, the development of a proper taxonomy for the notion of emergence as well as development of metrics that may provide suitable quantitative evaluation of the notion. We aim to place the concept of “System of Systems” within the standard framework of Systems Theory that is suitable for some further formal development (mathematical formulation) subsequently, as well as provide a characterisation of the notion of emergence that allows the definition of appropriate metrics.
Dr Stathis Milonidis
Sampling theory, structural properties and control design of discretised system models
The use of digital computers as means of controlling continuous time plants is well established. In their use as controllers (digital control) two distinct approaches are followed: the first is to design a continuous time controller and then discretise it and implement it as a computer algorithm, and the second is to discretise first the plant and then design directly a digital controller. The implementation of digital control schemes involves issues as fixed-point arithmetic, quantisation, round off effects, and the selection of sampling. The selection of sampling is crucial for both digital control schemes and especially for the second one, since sampling defines the properties of the discretised plant on which the design is based. The selection of sampling so far is dominated by the rules of signal processing and heuristics. The development of a theory for sampling selection based on the quality of the discretised model is a long-term goal of this research area. The present proposal is focused mainly on the effect of sampling on the structural properties and performance indicators of the discretised system. As a final aim it has the development of an integrated methodology for sampling selection based on signal-recovery and model-based criteria.
Simultaneous finite settling time stabilisation: The multivariable case
With many systems (physical, financial, biological), we would like to have a settled response to a constant change in demand, in finite or if possible in minimum time. We call this performance requirement Finite Settling Time (FST) stabilisation and in the case of optimum time dead-beat control. FST stabilisation can be achieved by linear controllers in the case of discrete – time systems, i.e. systems controlled by digital computers and there is a nice algebraic framework, developed among others by the applicant, for the solution of this problem. Also, systems in their majority are nonlinear and exhibit quite diverse behaviours for different “operating conditions”. To control such systems we usually linearise them around each operating condition and design a specific controller for each one of them. So we end up with as many controllers as operating conditions and it is a common practice to do gain scheduling when switching between different operating conditions.
A typical physical system that it is both controlled by digital computers and it is highly nonlinear, exhibiting hundreds of operating conditions, is a modern fighter aircraft. In addition, it is highly desirable to have finite settling time responses. Instead of designing a dedicated FST controller for each one operating condition, we would like to have a single, common controller for a set of operating conditions and if possible for the whole set of them. We cal this problem Simultaneous FST Stabilisation (S-FSTS) problem and this proposal aims into contributing to its solution. The S-FSTS problem is an extremely challenging theoretical problem and it remains unsolved, for the general case of stabilisation, since the 80s.
Professor Martin Newby
Reliability, maintainability and quality for complex systems
I am interested in analysis, design and optimisation for complex systems. The research exploits the properties of stochastic processes, statistical analysis, optimisation, and decision theory to determine optimum policies for system management and optimum structures for the organisation of the system. Past studies have been in aerospace, defense, manufacturing, automotive, and logistics.
Dr Veselin Rakocevic
Distributed protocols for spatiotemporal understanding of nodes in vehicular ad hoc networks
This project includes research on distributing location-specific information between nodes in a distributed network of moving objects (e.g. vehicles). The research aims to develop an optimal solution for the trade-off between the communication complexity and the accuracy of the exchanged information. A fully connected network of vehicles requires complex and continuous exchange of data packets, while the accuracy of the exchanged location-specific data can be sacrificed in certain situations.
Low latency network control for machine-to-machine communication networks
Machine-to-machine (M2M) communication systems have been gaining increased interest in the academic and business communities in recent years. There are several competing definitions of M2M communication systems, with most of them identifying that M2M systems consist of devices and software applications that enable systems or machines to communicate with each other, without the necessary human input or control. M2M communication systems can find application in a range of industries, including oil, gas, water, and electric companies, and transportation. This project will investigate the solutions for delay-controlled M2M networks, and will aim to design new network control algorithms and protocols which deliver bounded data transfer delay in large-scale wireless M2M networks.
Professor David Stupples
Ubiquitous radar system - Range performance
Systems performance modelling to improve the detection range of this new radar concept to exceed 200 nautical miles. The research project will be to represent the radar using both mathematical modelling and Matlab. Much of the work will be in signals processing using advanced graphics processes.
Ubiquitous radar - Advanced target recognition
Systems performance modelling to enhance the radars 'stare' capability to detect very slow moving targets such as almost stationary small drone air vehicles (low radar cross-section area) using advanced and enhanced doppler techniques. Research will requires both Matlab modelling and design of signal processing prototypes; excursions into Space Time Adaptive Processing (STAP) will be required.
Ubiquitous radar - Detection of stealth aircraft at long range
Systems performance modelling of the radar's 'stare' capability to detect stealth aircraft with very low 'radar cross section' area at long range. The research will involve optimising the radar's RF performance with advances signal processing - this will be a state of the art research undertaking using modified Swerling Case models combined with STAP techniques.
Dr Maria Tomas-Rodriguez
Offshore floating wind energy turbines nonlinear dynamical modelling
Offshore wind turbines (OWTs) are becoming an accepted method for generating electricity. The environmental conditions of offshore locations often impose high wind and wave forces on OWTs making them susceptible to intense loading and undesirable vibrations. It has therefore become necessary to utilise mechanical techniques for making OWTs more adapted to external conditions. One method to reduce system vibrations is through the use of structural control devices typically utilised in civil structures. This project will explore the effect of using a passive mechanical damping methods. The main idea behind this work would be to study the possible benefits of including these type of damping mechanisms when fitted in OWTs. For this purpose, a detailed nonlinear modeling of the offshore structures is also needed.
Motorcycle alternative suspension design for stability purposes
One of the most important factors on motorcycles stability is the front end. It links the front wheel with the main frame and has two main functions: the suspension of the front wheel and the steering of the motorcycle. Up to this date, several suspension systems have been developed to reach the best behaviour of the front end, being the telescopic fork the most extended one. It consists in a couple of fork tubes which contain the suspension components (coil springs and damper) internally. Conventionally, the fork stanchions are at the top, clamped to a pair of triple trees and the sliders are at the bottom, attached to the front wheel spindle. This system keeps the suspension's displacement of the front wheel in a straight line parallel to the steering axis, in such a way that the geometry of the assembly is modified with the suspension travel. The idea of this project would be to investigate alternative suspension designs that would allow different trajectories of the front wheel with the suspension travel. They can be designed to achieve a desirable performance in terms of manoeuvrability parameters when suspension action takes place. The purpose of this work would be to study the impact of these type of suspensions on the dynamical properties and stability of sport high-performance motorcycles.