Heriot-Watt University
ROS: The Research Output Service. Heriot-Watt University EdinburghNot a member yet
4689 research outputs found
Sort by
New algorithms and practical implementations that revolutionise frequency-domain ambient backscatter communication
No full textThere is a growing demand for energy-efficient communication technologies to enhance the sustainability of the increasing number of wireless-connected Internet-of-Things (IoT) devices and tags in terms of both energy and hardware cost. Ambient Backscatter Communication (ABC), particularly notable for its low power
consumption and potential for battery-free operation, offers a promising solution.
ABC exploits a range of opportunistic ambient wireless signals, including Television (TV) broadcasts, Frequency Modulation (FM) radio, Long Range (LoRa),
WiFi, Bluetooth, and cellular signals. These signals serve as Radio Frequency (RF)
carriers, which facilitate information transfer during the electromagnetic scattering
process. Recent works have increasingly focused on the out-of-band ABC scheme,
which delivers extended communication distances, robust throughput, and compatibility with commercial off-the-shelf devices. However, the deployment of this
technology in practical IoT networks encounters challenges such as low spectrum
efficiency and high power consumption.
This thesis explores the adoption of spectral-efficient in-band ABC systems,
where backscattered signals and ambient carrier signals share the same spectrum
resources. We specifically use binary frequency shift keying (2FSK) modulation
across various ambient carrier signals. This thesis also evaluates compatibility with
commercial off-the-shelf receivers (Rx) and battery-free operation to meet practical IoT requirements. The first contribution is the development of a 2FSK-based
in-band FM ABC system optimised for outdoor applications, achieving communication ranges exceeding 100 metres. This in-band approach demonstrates superior spectrally efficiency, reduced power consumption, and enhanced compatibility
with all FM radios compared to traditional out-of-band systems. This achievement mainly contributes to a newly proposed Quadrature Demodulation (QD)
In-phase/Quadrature (I/Q) processing technique to mitigate self-interference and
improve Signal to Interference Plus Noise Ratio (SINR), along with a moving-window demodulation method that is tolerant to Carrier Frequency Offsets (CFO).
Secondly, this thesis extends in-band ABC systems to indoor scenarios using
more complex and prevalent WiFi carrier signals. To tackle the self-interference issue, a novel Conjugate Multiplication (CM) I/Q processing technique is proposed and validated. This method not only improves the performance of the
in-band WiFi ABC system but also shows versatility across other common ambient carriers, such as Bluetooth, Zigbee, and cellular networks, greatly expanding
its potential applications. The efficacy of the in-band WiFi ABC system has been
validated through comprehensive simulations and indoor experiments, both Line-of-Sight (LoS) and Non-Line-of-Sight (NLoS). These experiment results demonstrate a communication range of up to 15.5 metres at a bit rate of 200 bits per
second (bps) and up to 12 metres at 1 kbps.
Additionally, this work facilitates the integration of off-the-shelf commercial
Bluetooth and LoRa Rx with the proposed CM-based in-band WiFi ABC system.
To accomplish this, a hardware solution is introduced by developing a RF domain CM processing prototype, which can be seamlessly integrated into commercial chipsets. The experiments demonstrate that Bluetooth and LoRa compatible
symbols can be successfully generated by using the RF CM prototype. To support
battery-free operation, a novel RF-powered tag structure is proposed. This design
integrates an RF Energy Harvesting (EH) circuit with the backscatter modulator,
enabling real-time RF power allocation between the BackCom modulator and the
RF EH circuit through the use of a square wave with variable duty cycles. This
innovative tag structure optimises energy efficiency and system adaptability under
varying ambient RF conditions.
Overall, this thesis aims to revolutionise the frequency modulation-based in-band ABC system by introducing two innovative I/Q sample processing techniques
for self-interference cancellation and one novel RF-powered tag design. The envisioned ABC system is designed to be spectrally efficient, self-powered, compatible
with commercial Rx, and operable with ubiquitous ambient signals. These advancements are intended to facilitate the deployment of ABC technology in IoT
environments, contributing to the realisation of environmentally friendly communication solutions
SE for urban poverty mitigation : a proposed framework
Full text linkThis research addresses the understudied intersection of Social Entrepreneurship (SE) and
poverty mitigation, specifically within the Malaysian context. Literature gaps reveal limited
knowledge on SEs' poverty-mitigating methods/practices, multidimensional poverty lenses,
and urban poverty mitigation. The study aimed to identify poverty initiatives, types of social
impact, and influencing factors for Malaysian SEs. Findings show SEs employ diverse
interventions, categorized into employment, empowerment, circumstantial grievance
alleviation, and socio-emotional practices. Beneficiaries experienced improved quality of life
and psychological well-being. However, prevailing notions of SEs as potent poverty reducers
are challenged. The study identifies gaps in recognizing emotional-relational practices and
understanding human capital factors' impact on SE beneficiaries. Additionally, it questions
normative views on SEs' social branding, revealing potential stakeholder disinterest due to
contextual stereotypes. Implications for practitioners include refining strategies and
uncovering novel approaches, with awareness of SE-related factors. The nascent Malaysian
SE ecosystem and COVID-19 limitations caution against generalization. Future research
should validate the framework, explore psychological-based interventions, and address
human capital factors in SE operations. The study contributes to understanding SEs' role in
poverty alleviation and highlights avenues for further research
Exploring applications of geometric deep learning for medical imaging
Full text linkCardiovascular disease and cancer are the two main causes of death worldwide. Although the development of medical imaging has improved the diagnosis and treatment
of these conditions, the required analysis is cumbersome and demands high levels of expertise. The development of deep learning technology, which automatically replicates
expert judgment, can have a positive impact on the management of these conditions by
providing more efficient ways of completing repetitive and time-consuming processes.
Nevertheless, deep learning faces many challenges. The best performing models often are
the most computationally expensive, and the model predictions may be too general and
miss important details. This thesis approaches such challenges by leveraging a branch
of deep learning that accounts for the symmetries in data: Geometric Deep Learning.
Specifically, we used Graph Neural Networks to build efficient methods for tumour segmentation and cancer survival prediction. In a tumour segmentation and survival prediction challenge, the proposed segmentation approach required on average three times less
memory than the other competing algorithms, and the survival prediction method ranked
7th out of 12 challenge participants. Furthermore, to accurately separate arteries from
veins in retinal Fundus images, we embedded rotational symmetry in a neural network
by using an orientation sensitive filter, achieving better topological accuracy than other
state-of-the-art methods. In conclusion, this thesis showcases three scenarios in which
Geometric Deep Learning proved advantageous for medical image analysis: in building
efficient segmentation models, in developing accurate survival prediction methods, and
in obtaining topologically accurate segmentations of vascular structures
Influence of reservoir heterogeneity on the implementation of polymer Enhanced Oil Recovery (EOR) in the Niger Delta reservoirs
Full text linkThe Niger Delta, a significant oil and gas province, has long contributed to global
conventional hydrocarbon production. Despite its importance, low recovery rates persist due
to complex geological factors. This research delves into the analysis of how polymer
Enhanced Oil Recovery techniques can improve oil recovery in the Niger Delta fields. A
high resolution (‘truth’) modelling approach was designed to address this research,
identifying which sub-types of reservoirs in the Niger Delta are prime candidates for
benefiting from polymer EOR and to what extent.
Diverse data sources, including core-plug data, analogue outcrop data, and outcrop photos
were integrated, spanning various depositional environments from shoreface to mouth bar to
channel deposits. This comprehensive dataset facilitates the identification of five critical
modelling elements. These are grouped into: mudstones and siltstones, very fine sandstones,
fine sandstones, fine-medium sandstones and medium-coarse sandstones. Fine scale 'truth'
waterflooding models yielded an oil recovery ranging from 60% to 65% of the STOIIP.
Models were upscaled to a full-field scale, shedding light on the profound impact of
upscaling ratios on oil recovery. Recovery from the fine scale waterflooding models was
enhanced using polymer flooding strategies, uncovering the influence of archetype-dependent
factors, stemming from variations in reservoir heterogeneity. Considering various polymer
parameters, an impressive increase in oil recovery, ranging from 10% to 22% across the
models was observed. These recoveries translated into an average incremental Net Present
Value (NPV) of 50 per barrel for the economic viability of polymer projects in Niger Delta
reservoirs
Investigating black spot shell disease and claw deformities in commercially important crab species in the Orkney islands, Scotland
Full text linkCommercial velvet crab Necora puber (L.) and brown crab Cancer pagurus (L.) fisheries are
highly valued in the Orkney Islands and the British Isles. However, understanding of disease
and deformities which potentially affect their health and marketability is poorly understood.
Concerns related to increased incidences of black spot shell disease (BSSD) and claw
deformities prompted a study to determine drivers of these conditions and to provide a
platform for local fishers to provide anecdotal evidence. Prevalence and severity of BSSD
and claw deformities were assessed in 3,385 N. puber across Orkney and links to location and
elevated metal characterisation were investigated. A questionnaire was administered to gain
insight from fishers which suggested that BSSD was considered more common, and more of
a significant problem in C. pagurus, whereas claw deformities are observed more frequently
in N. puber, and thus considered by fishers to be more of a significant problem for this
fishery. More experienced fishers who had been at sea for longer, tended to hold more
negative views on the significance of each condition to each respective fishery.
Despite concerns from local fishers at the Bay of Isbister, data did not support increased
incidence of BSSD or claw deformities at this location. In N. puber tissues, lead (Pb) was
elevated in a small number of crabs from one industrial location, but no other crab tissues had
significantly elevated concentrations. Significant variations in elemental metals and locations
were observed, with female crabs more likely to have elevated tissue levels. Magnesium
(Mg) was the only metal element found to influence the probability of crabs having BSSD.
This research has provided the first baseline evidence on these conditions and metal
accumulation in commercially important crabs in the Orkney Islands and will inform
stakeholders and government on crab health
Application of physics-based artificial intelligence for improved surrogate modelling in gas and gas-condensate reservoirs
Full text linkNumerical simulations of subsurface flow can be computationally expensive due to
the complexity of the flow domain characteristics and the corresponding governing
equations. Furthermore, highly compressible fluids, like gases and gas condensates,
exhibit highly nonlinear flow behaviour in space and time, resulting in additional
computations during linearisation. Artificial intelligence (AI)-based surrogate reservoir
models (SRMs) can provide computationally feasible and accurate approximations to
these numerical simulations.
In this research, AI-based SRMs were developed using deep learning architectures
with skip connections to capture the nonlinear dynamics in gas and gas-condensate flow
through porous media. Two approaches were explored: non-physics-based supervised
learning, which trained the AI-based surrogates using snapshots from numerical
simulations with an added L2-norm regularization; and physics-based supervised learning.
The latter exploits a discretized partial differential equation of the flow domain,
initial and boundary conditions. These equations were formulated as physics-based
regularizations, with the L2-norm regularization added, and trained without external
simulation data. A uniqueness of this approach is the application of additional artificial
neural network configurations and regularizations to improve learning. Key innovations
include: a three-module artificial neural architecture comprising of pressure, fluid
property and time step modules, with skip connections integrated in their architectures; a
trainable layer for hard enforcing initial conditions; and physics-based regularizations for
tank material balance and time-discretization errors.
The developed AI-based SRMs demonstrate strong agreement with numerical
simulator results. The non-physics-based training with skip connections and the L2-norm
regularization improves predictions up to 1.44 times the maximum training time point.
Considering the physics-based training, including the tank material balance regularization
accelerates the learning and improves the reliability of predictions; the trainable hard
enforcement improves predictions compared to other hard enforcement techniques. Also,
including the L2-norm regularization during the physics-based supervision increases the
extent of predictions in the unseen space-time domain. The variable time step provides
the best rounding-truncation error trade-off compared to those of fixed time steps.
The predictions are timely, with the AI-based SRM 10 times faster than the
numerical simulator during predictions, based on the considered dataset and available
computing architecture. The only significant time expense is during the training, which
is higher for the physics-based than the non-physics-based supervised training.
Nevertheless, the training time can be reduced by advancing the computing and memory
architecture.
The effectiveness of the proposed techniques, including the custom architecture,
trainable hard enforcement, variable time steps, and well-adapted regularizations, proves
to be a reliable AI-based approach for reservoir performance sensitivity analysis. These
can also be used in developing state-dependent functions for processes like data
assimilation
Performing the festival : an experiential autoethnography of the festival of Sant’Efisio in Sardinia
Full text linkThis research project considers festivals as sites of transformation, adaptation and
negotiation for the communities interacting with their social environment. Employing a
case-study strategy, the thesis carries out an in-depth exploration of one of the most
celebrated events in the island of Sardinia (Italy): the Festival of Sant’Efisio. This complex
celebration has been performed for 368 years to fulfil a vow in honour of the martyr and
saint Efisio, who is believed to have saved Sardinia from the plague in the 17th century. The
festival includes a multitude of secular and religious events and ceremonies which take place
around a four-day pilgrimage. This study is placed within an interpretative
phenomenological framework, underpinned by a feminist approach throughout, that
considers “performance” as the key theoretical lens to inform the analysis of the following
socio-cultural issues in festivals: 1) the display of cultural heritage; 2) community
construction and conflict; and 3) gendered practices. Based on ethnographic fieldwork, both
in person and online, this project investigates the effects of social and cultural
transformations in relation to these issues within the Festival of Sant’Efisio, by addressing
how the festival is interpreted, experienced, felt and performed by the people involved. The
researcher’s perspective and experience are central to this enquiry and are discussed
throughout by means of autoethnography. I suggest that the way people feel in festivals is
crucial to understand their socio-cultural significance, as well as their survival through the
change of time.Heriot-Watt University scholarshi
Design and development of a nanosecond diode-pumped solid state laser system operating at 10 J, 100 Hz
No full textAbstract and full text unavailable. Restricted access until 07.09.2025. Please refer to PD
Mathematical modeling of collective cell migration : cell trait structures and intracellular variables
Full text linkCollective cell migration is a complex biological phenomenon observed, for example,
in cancer and embryonic development. A simplifying modeling assumption is to
consider a homogeneous population, where the individual members of a group behave
identically. The aim of this thesis is to shed some light onto the collective cell
migration of heterogeneous populations.
Collective cell migration is promoted by different cell-cell interactions, such as
co-attraction and contact inhibition of locomotion. These mechanisms act on cell
polarity, crucial for directed cell migration, through modulating the intracellular
dynamics of small GTPases such as Rac1. We propose a biased random walk model,
where the bias depends on the internal state of Rac1, and the Rac1 state is influenced by cell-cell and cell-environment interactions. We demonstrate the scope and
applicability of the model in various scenarios in an extensive simulation study. Furthermore, we derive a corresponding system of partial differential equations. Using
this model, we successfully replicated key observations from biological experiments.
Consistent with these observations, contact inhibition of locomotion seemed crucial
for successful collective migration. Additionally, we established a link between the
natural deactivation rate of the intracellular state and the persistence of directional
movement.
We introduce a trait-structured Keller-Segel model to account for heterogeneity
in migrating cell populations. The cell trait is given by the proportion of membrane
receptors occupied by ligands, and cells change their trait by attaching or detaching
ligands to or from their receptors. We assume that the trait is linked to the phenotype of a cell and, with that, to its ability to perform chemotaxis or proliferate. We
formally derive properties of traveling wave solutions using the Hopf-Cole transformation and compare our analytical findings to results from numerical simulations.
The derivation of this novel model is a key accomplishment of this thesis. A significant finding was the explicit expression for the dominant trait within invading waves
of heterogeneous cell populations under specific parameter regimes. Additionally,
we identified a theoretical minimal wave speed for traveling waves. Under trade-of assumptions between chemotactic ability and proliferation, we discovered a distinct
structure within the traveling waves, with proliferative cells located at the back and
migratory cells at the front.
For a modified trait-structured Keller-Segel model, we use a linear stability analysis to investigate (in-)stability conditions for a system of Keller-Segel models that
stems from discretising the trait variable in the original model. For the simplest,
two-state model, we derive instability conditions. We deduce corresponding criteria
for cases with more than two states, and support these by numerical simulations.
The main result is a novel criterion for Turing instabilities in specific parameter
regimes, stemming from our model’s explicit consideration of ligand-receptor bindings
Multi-user applications in photonic quantum networks
Full text linkThe field of quantum information science aims to exploit the principles of quantum
mechanics to achieve tasks such as quantum communication—a pair of individuals securely communicating over unsecure channels. Scaling this to multiple users,
forming a quantum network, is a notable experimental challenge due to noise and
lower efficiencies for increased number of users. Tackling this is two-fold: adaptations of the task itself to be more accepting of errors and experimental developments
which directly characterises the sources of error in the network. Further, the type
of resource distributed over the quantum network directly relates to the efficiency
of the protocol. This thesis demonstrates, using photonic quantum networks, that
anonymous protocols built around multi-partite resources, instead of the traditional
bi-partite resources, see dramatic efficiency advantages and further enable tasks that
are not possible with bi-partite resources alone. This tackles the question of how
feasible current quantum technology is at addressing advanced tasks in quantum
information science.
Starting with addressing quantum measurements, an experiment is realised that
achieves the opposite of intuitively attempting to directly identify a quantum state
through an optimised measurement procedure, by re-constructing the premise to
identify what the quantum state is not. This leads to outcomes that are not possible by employing discriminatory methods, in that with a single measurement one
can exclude a subset of arbitrary quantum states with unit certainty, under certain
circumstances. This has roots in quantum foundations, and is further used to investigate interpretations of the quantum state—does a measurement of a quantum
state correspond to an element of reality or is it just updating our limited knowledge
of the system at hand.
Following this, an advanced quantum communication protocol is realised, where
general quantum key distribution schemes are adapted to not only distribute keys to
multiple users in a network, but also provide said users with anonymity. This work is
targeting what resources future quantum networks should ideally host. The results
show that if multi-partite entangled states are used as the resource over the traditional bi-partite entangled states currently employed in quantum networks, then
there are dramatic advantages in the key rates. This advantage has been demonstrated in multi-user key agreement protocols, yet by including classical cryptography protocols for features like user anonymity, the advantage becomes far larger.
This is demonstrated on a six-user network, realised through a six-photon maximally
entangled state.
Finally, using a similar experimental setup, a novel quantum sensing protocol
that insures a level of security to each sensor is performed. The security in this work
takes the form of the composable security framework featured in quantum key dis tribution, yet here the quantum metrology toolbox is adopted as the mathematical
construction. This links the precision and accuracy of the quantum sensor readings
with the security achievable for each sensor. We find the current theoretical flaws in
this particular protocol make it intractable to realistically guarantee security, and
further analyse where to direct the research efforts to resolve this.
In conclusion, this thesis presents a narrative that joins many facets of experimental quantum information together from investigations into measurement strategies to advanced quantum communication schemes
Access Repository Dashboard
Do you manage ROS: The Research Output Service. Heriot-Watt University Edinburgh ? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!