In this thesis, efficient methods are presented to calibrate large or small aperture array systems containing different types of uncertainties. specifically the challenge of reducing the number of external sources required to calibrate an array is addressed and array calibration methods suitable for use when sources may be operating in the "near-far" field of the array are developed. Together, this can ease the overheads involved in calibrating and recalibrating an array system. In addition to presenting novel array calibration algorithms, this thesis also presents a novel transformation allowing a planar array to be expressed as a virtual uniform linear array of a much larger number of elements. This allows the array manifold of a planar array, which in general consists of non-hyperhelical curves, to be expressed using a number of hyperhelices which each correspond to the array manifold of a linear array. This hyperhelical structure has the potential to ease calibration overheads as well as having many other potential applications in array processing. This thesis presents novel pilot and auto array calibration schemes for estimating different types of array uncertainties. A novel pilot calibration algorithm is proposed whereby a single source transmitting from a known location (i.e. a pilot) at two carrier frequencies is used to estimate geometrical uncertainties in a planar array. This is achieved by exploiting the frequency dependence on the boundary between the "near-far" and "far" field of the array. In addition, an auto-calibration method is presented which doesn't require any external sources to estimate array uncertainties. Here, geometrical, complex gain and local oscillator (i.e. frequency and phase) uncertainties associated with the array elements are considered. In this approach, array elements transmit in turn to the others which operate as an array receiver. Large and small array apertures are investigated. Throughout the thesis, extensive computer simulations are presented to analyse the performance of the algorithms developed.

This thesis describes the design and analysis of a novel wideband circularly-polarized textile antenna to power up a wearable wirelessly-powered sensor system operating in the 2.45 GHz ISM band (2.4-2.5 GHz) and the building of the whole system. The system is constructed using off-the-shelf components and it is shown that the wirelessly-powered sensor system is able to operate when just a few mW are transmitted from a base station at a distance over a metre. Initially, standard linearly-polarized patch antennas are used for power transmission. However, the antennas have to be aligned perfectly for the best efficiency. Subsequently, a circularly-polarized antenna is proposed for enhanced wireless-power transfer due to the freedom of orientation. A wide-slot antenna without a ground plane has been chosen for its simplicity and wide impedance band. The geometry is firstly optimized for wide impedance and 3-dB axial ratio bandwidth on FR-4. The experimental and simulation results have been studied to analyse the characteristics of such an antenna. The wideband circularly-polarized antenna is then constructed using a conductive textile and re-optimized for on-body applications. With a simple antenna geometry and only a single layer of conductive textile layer, the axial ratio and impedance bandwidths are wide enough to cover the whole 2.45 GHz ISM band with plenty of margin and are significantly wider than any other on-body circularly-polarized textile patch antennas which have been reported. The characteristics of this wideband circularly-polarized antenna under different conditions on the human body have been measured and then connected to the wirelessly-powered sensor system to demonstrate the effectiveness of power transfer to the human body.

This thesis considers the use of medium-voltage power electronic compensation at distribution network voltage levels (11kV) for the improvement of power quality, reliability, and to accommodate growth in customer demand or distributed generation capacity. Specifically, power electronic compensators connecting two or more nodes of previously isolated radial feeders are considered. This type of device can be considered as an alternative to normally-open points, which connect two nodes with mechanical switchgear. Rather than connecting these nodes directly, power-electronics are placed between them. This type of device will be deemed a soft-open point (SOP) in this thesis. Several compensator topologies which can achieve the functionality of a SOP are considered. The feature criteria used to choose which compensators are suitable for use as a SOP are: the ability to transfer active power between feeders; the ability to resupply (post-fault) adjacent feeders connected via the compensator; an inherent or controlled disturbance rejection or fault current limiting between adjacent feeders. Modified versions of some existing flexible AC transmission system (FACTS) or custom power devices o er the potential to meet these criteria. The compensator topologies considered include: static synchronous series compensators, unified power flow controllers, back-to-back connected voltage-sourced converters (VSCs) or multi-terminal connected VSCs. In order to quantify and compare the benefits of these compensator topologies, their relative performance on UK distribution networks is assessed based on load flow and optimal power flow case studies performed on datasets representing several hundred UK distribution networks. Benefits quantified include an increase in customer reliability ratings, prevention or deferral of asset replacement, reduction in conductor losses, accommodation of increased distributed generation, and accommodation of increased customer demand. The benefit analyses show that multi-terminal VSC based SOPs provide the greatest flexibility, but one must recognize that associated cost and right-of-way issues associated with distribution networks can be prohibitive. Series and series-shunt compensators are shown to offer an an adequate amount of control, achieving reasonable levels of load and generation growth with lower overall estimates for cost. Several control strategies and converter topologies are considered for use in SOP implementation under a number of scenarios. The use of multi-terminal VSCs is also verified through implementation in a prototype network and through time-domain simulations. These demonstrations serve as a proof of concept for SOP operation in scenarios relevant to their intended use in distribution networks. Also considered is the use of SOPs to directly compensate overload substation transformers, for which it is found that SOPs can very effectively mitigate overload events at the expense of increased cumulative losses. Different high-level control schemes are shown to reduce the impact of the additional converter losses.

A robust and reliable noise estimation algorithm is required in many speech enhancement systems. The aim of this thesis is to propose and evaluate a robust noise estimation algorithm for highly non-stationary noisy environments. In this work, we model the non-stationary noise using a set of discrete states with each state representing a distinct noise power spectrum. In this approach, the state sequence over time is conveniently represented by a Hidden Markov Model (HMM). In this thesis, we first present an online HMM re-estimation framework that models time-varying noise using a Hidden Markov Model and tracks changes in noise characteristics by a sequential model update procedure that tracks the noise characteristics during the absence of speech. In addition the algorithm will when necessary create new model states to represent novel noise spectra and will merge existing states that have similar characteristics. We then extend our work in robust noise estimation during speech activity by incorporating a speech model into our existing noise model. The noise characteristics within each state are updated based on a speech presence probability which is derived from a modified Minima controlled recursive averaging method. We have demonstrated the effectiveness of our noise HMM in tracking both stationary and highly non-stationary noise, and shown that it gives improved performance over other conventional noise estimation methods when it is incorporated into a standard speech enhancement algorithm.

Robot learning from demonstrations (RLD), also known as imitation learning, plays an important role in human-robot interactions (HRI) because it creates a userfriendly environment for non-expert end-users to teach a robot new skills. Recent advances in sophisticated humanoid robots provide an opportunity for these robots to manipulate everyday tools. This thesis focuses on some key aspects in enabling humanoid robots to learn manipulation of tools using an RLD approach as a threefold investigation. Firstly, this work investigates the outstanding issues in existing RLD frameworks for applications in a real-world HRI environment and presents an RLD model using template-based approach integrated with an online learning algorithm to address these issues. Further investigations probe into biological systems for inspirations on improving the reusability of learned skills which results in an integrated framework using the proposed RLD model as the base model. Attention is then turned to an important aspect tightly coupled with RLD learning for tool manipulation - recognition of tools. A biologically-inspired framework based on RLD-learned tool affordance is proposed to address the issues of applying to tools the traditional static feature matching approach for object recognition. In a series of experiments, we show that our RLD model is capable of learning skills without demanding the users for repeated demonstrations. The learned skills can be generalised to similar tasks with different environmental constraints and updated as more demonstrations of the skill are presented. We also demonstrate the reusability of the complete framework in comparison to the base model by a realworld HRI application in playing the tic-tac-toe game. Through the experiments on tool recognition using a set of human demonstrations, the benchmarked performance indicates that our proposed tool representation framework is a suitable supplement to the current object recognition models on tools.

One major hindrance to the advent of novel bio-implantable sensor technologies is the need for a reliable power source and data communications platform capable of continuously, remotely, and wirelessly monitoring deeply implantable biomedical devices. This research proposes the feasibility and potential of combining well established, ‘human-friendly' inductive and ultrasonic technologies to produce a proof-of-concept, generic, multi-tier power transfer and data communication platform suitable for low-power, periodically-activated implantable analogue bio-sensors. In the inductive sub-system presented, 5 W of power is transferred across a 10 mm gap between a single pair of 39 mm (primary) and 33 mm (secondary) circular printed spiral coils (PSCs). These are printed using an 8000 dpi resolution photoplotter and fabricated on PCB by wet-etching, to the maximum permissible density. Our ultrasonic sub-system, consisting of a single pair of Pz21 (transmitter) and Pz26 (receiver) piezoelectric PZT ceramic discs driven by low-frequency, radial/planar excitation (-31 mode), without acoustic matching layers, is also reported here for the first time. The discs are characterised by propagation tank test and directly driven by the inductively coupled power to deliver 29 μW to a receiver (implant) employing a low voltage start-up IC positioned 70 mm deep within a homogeneous liquid phantom. No batteries are used. The deep implant is thus intermittently powered every 800 ms to charge a capacitor which enables its microcontroller, operating with a 500 kHz clock, to transmit a single nibble (4 bits) of digitized sensed data over a period of ~18 ms from deep within the phantom, to the outside world. A power transfer efficiency of 83% using our prototype CMOS logic-gate IC driver is reported for the inductively coupled part of the system. Overall prototype system power consumption is 2.3 W with a total power transfer efficiency of 1% achieved across the tiers.

No description available.

In this thesis, examination of polycrystalline organic based devices such as the Schottky diode and MOS capacitor is carried out. The data is interpreted in terms of a polycrystalline model based mainly on a conventional polysilicon model with slight modifications to fit the organic properties. A brief introduction to the existing charge transport models for organic materials is presented. The most dominant being the variable range hopping model for disordered materials. The disorder analysis is appropriate in the grain boundaries of a polycrystalline material. The distribution of the traps of density of states (DOS) is commonly described by the Gaussian distribution and the associated exponential approximate at low energies. This is a valid assumption for organic semiconductors with low carrier mobility values [N. Sedghi et al., J. Non Crys. Solids 352, 1641, 2006]. Detailed investigation on the temperature effects of the polycrystalline Schottky diode leads to determination of important electrical parameters. Such studies are essential in understanding the conduction processes of the organic device, particularly in terms of trapping effects, which is essential in the development of device models for organic circuitry. Several parameters such as dopant (ND) and carrier concentrations (p), effective mobility (μeff), depletion width (Wdep), effective Debye length (LDe), Meyer Neldel energy (MNE) and the characteristic temperature of the carriers (T0) are extracted from the current-voltage characteristics of the diode. For a soluble derivative of pentacene, 6, 13-triisopropylsilyethynyl pentacene (TIPS) blended with Polytriarylamine (PTAA), the respective values extracted at room temperature are found to be approximately 1017 cm-3, 1.8x10-2 cm2V-1s-1, 185 nm, 11 nm, 31.5 meV and 780 K, respectively. As the temperature falls, the values of most parameters remain constant until a critical temperature. The activation energy also remains constant at approximately 0.3 eV for various applied voltages in saturation. Below this critical temperature, Wdep, LDe and T0 increase whilst μeff, ND/p and characteristic temperature of the states (TC) decrease. Similar analysis is carried out on doped layers of TIPS with a different insulating binder Poly-alpha methylstyrene (PAMS). The value of Wdep, LDe, T0, μeff, ND/p, MNE and TC obtained from such doped Schottky diode are approximately 100 nm, 5 nm, 1200 K, 1x10-2 cm2V-1s-1, 2x1017 cm-3, 35 meV and 400 K, respectively. The Capacitance-Voltage (C-V) analysis on polycrystalline Schottky diodes provides ND of approximately 7.6 x 1016 cm-3, 5.2 x 1014 cm-3 and 2.98 x 1014cm-3 at 500 Hz, 1 kHz and 2 kHz respectively. These ND values are lower than those extracted from current-voltage characteristics and decrease with increasing frequency. This is thought to be due to the low mobility of holes, unable to respond to the signal at higher frequencies. The conduction in polycrystalline organic Schottky diode is proposed using a 2-dimensional (2D) model, which focuses on both the lateral and vertical conduction paths. The organic semiconductor layer is assumed to be relatively thin so that only a single layer of the grain exists between adjacent grain boundaries for the vertical conduction. A two dimensional situation is treated as two separate one dimensional problems that are positioned at right angles to each other. The grain and grain boundaries in the polycrystalline material are explained in terms of two boundary conditions. The variation of potential in grain boundary is the basis in defining the variation of potential in the grains. Conduction under forward bias in the grain and grain boundary is thus established assuming two distributions for the DOS, namely the Gaussian and Laplace. In comparison, Laplace DOS is believed to be a better representation of the distribution of states, where a large number of energy levels are being scanned with applied voltage. The ac properties of a polycrystalline based MOS capacitor are investigated. The frequency and temperature effects on the C-V characteristics of MOS capacitor based on another soluble derivative of pentacene, refered here as S1150, are studied. Equivalent circuits which include the effects of bulk and series resistance due to contact effects are analysed. The bulk resistance (Rb), bulk capacitance (Cb) and series resistance (RS) are found to be approximately 13 kΩ, 760 pF and less than 309 Ω respectively, for an organic film thickness (tOSC) of 27nm. For ND ≈ 3.6 x 1017cm-3 at 1 kHz, the hole mobility is found to be approximately 4.6x10-7cm2V-1s-1. As expected the mobility decreases with increase in frequency. Furthermore, the temperature study of inverse square space charge capacitance (1/CS2) against absolute temperature (T) provides an intercept close to T (~ 330K) instead of TC. A low intercept value indicates a decrease in disorder which suggests that the large grains may be dominating the capacitance.

This thesis concerns the problem of prototype reduction in instance-based learning. In order to deal with problems such as storage requirements, sensitivity to noise and computational complexity, various algorithms have been presented that condense the number of stored prototypes, while maintaining competent classification accuracy. Instance selection, which recovers a smaller subset of the original training set, is the most widely used technique for instance reduction. But, prototype abstraction that generates new prototypes to replace the initial ones has also gained a lot of interest recently. The major contribution of this work is the proposal of four novel frameworks for performing prototype reduction, the Class Boundary Preserving algorithm (CBP), a hybrid method that uses both selection and generation of prototypes, Instance Seriation for Prototype Abstraction (ISPA), which is an abstraction algorithm, and two selective techniques, Spectral Instance Reduction (SIR) and Direct Weight Optimization (DWO). CBP is a multi-stage method based on a simple heuristic that is very effective in identifying samples close to class borders. Using a noise filter harmful instances are removed, while the powerful heuristic determines the geometrical distribution of patterns around every instance. Together with the concepts of nearest enemy pairs and mean shift clustering this algorithm decides on the final set of retained prototypes. DWO is a selection model whose output set of prototypes is decided by a set of binary weights. These weights are computed according to an objective function composed of the ratio between the nearest friend and nearest enemy of every sample. In order to obtain good quality results DWO is optimized using a genetic algorithm. ISPA is an abstraction technique that employs the concept of data seriation to organize instances in an arrangement that favours merging between them. As a result, a new set of prototypes is created. Results show that CBP, SIR and DWO, the three major algorithms presented in this thesis, are competent and efficient in terms of at least one of the two basic objectives, classification accuracy and condensation ratio. The comparison against other successful condensation algorithms illustrates the competitiveness of the proposed models. The SIR algorithm presents a set of border discriminating features (BDFs) that depicts the local distribution of friends and enemies of all samples. These are then used along with spectral graph theory to partition the training set in to border and internal instances.

Intelligent Transportation Systems (ITS) apply Information and Communication Technologies (ICT) to improve safety and efficiency as well as the passenger experience in modern transport systems. It is envisaged that dynamic vehicular networks, particularly, Vehicular Ad-hoc Networks (VANETs) based on dedicated short-range communications (DSRC) and cellular networks, will be important parts of the future ITS. Unlike traditional communication networks, VANETs are highly dynamic systems resulting in significant reliability issues for the communication protocols. In addition, cellular networks incur notable usage cost. Motivated by this, we investigate efficient and reliable geo-routing and transport protocols aimed at VANETs and VANET/cellular hybrid architectures. Specifically, first we develop an innovative, unicast, cross-layer, weighted, position-based routing protocol (CLWPR) that takes into account mobility and cross layer information about neighbour nodes. A heuristic algorithm based on analytic hierarchy process (AHP) is employed to combine multiple decision criteria into a unique weight parameter used to select the node to which the packet is forwarded. Comprehensive simulations are performed in realistic representative urban scenarios with synthetic and real traffic. Insights on the effect of different communication and environment parameters are obtained. The results demonstrate that the proposed protocol outperforms existing routing protocols for VANETs, including ETSI's proposed greedy routing protocol, GyTAR, and AGF in terms of combined packet delivery ratio, end-to-end delay, and overhead. To efficiently distribute location information, required for the proper functionality of geo-routing, we develop a centralised Location Service. Exploiting the availability of two interfaces (DSRC and LTE) in a hybrid system, we propose separation of signalling and data traffic. The former is transferred over a cellular network and the later over a short range ad-hoc network. For the evaluation of the proposed scheme, we develop an analytical model of the upper bound delay based on stochastic network calculus (SNC) theory. We compare the upper-bounds of three networks, namely a pure short-range ad-hoc network, a pure cellular based on 3GPP LTE and the proposed hybrid with signalling on cellular and data on ad-hoc network. The results of our investigation suggest that hybrid networks can significantly improve performance of vehicular networks in terms of end-to-end delay both for data and signalling traffic. In the light of these findings, we investigate transport protocols for hybrid networks benefiting from multi-homing support. As Stream Control Transmission Protocol (SCTP) is one IETF standard that supports multi-homing, we develop an analytical model for throughput calculation of a round-trip time (RTT)-aware SCTP variant. Finally, we propose a novel SCTP scheme that takes into account not only path quality but also the cost of using each network. We show that the combination of QoS and cost information increases economic benefits for provider and end-users, while providing increased packet throughput.