Heriot-Watt University
ROS: The Research Output Service. Heriot-Watt University EdinburghNot a member yet
4689 research outputs found
Sort by
Shallow neural networks for autonomous robots
Get PDFThe use of Neural Networks (NNs) in modern applications is already well established
thanks to the technological advancements in processing units and Deep Learning (DL), as
well as the availability of deployment frameworks and services. However, the embedding of
these methods in robotic systems is problematic when it comes to field operations. The use
of Graphics Processing Units (GPUs) for such networks requires high amounts of power
which would lead to shortened operational times. This is not desired since autonomous
robots already need to manage their power supply to accommodate the lengths of their
missions which can extend from hours to days. While external processing is possible,
real-time monitoring can become unfeasible where delays are present. This also applies to
autonomous robots that are deployed for underwater or space missions.
For these reasons, there is a requirement for shallow but robust NN-based solutions that
enhance the autonomy of a robot. This dissertation focuses on the design and meticulous
parametrization complemented by methods that explain hyper-parameter importance. This
is performed in the context of different settings and problems for autonomous robots in field
operations.
The contribution of this thesis comes in the form of autonomy augmentation for robots
through shallow NNs that can potentially be embedded in future systems carrying NN
processing units. This is done by implementing neural architectures that use sensor data
to extract representations for event identification and learn patterns for event anticipation.
This work harnesses Long Short-Term Memory networks (LSTMs) as the underpinning
framework for time series representation and interpretation. This has been tested in three
significant problems found in field operations: hardware malfunction classification, survey
trajectory classification and hazardous event forecast and detection
The study of developing large area (10cm × 10cm) Dye Sensitised Solar Cells and its up-scaling issues
Get PDFOver the past few years, Dye Sensitised Solar Cell (DSC) technology has attracted
much interest due to its low cost, simple production, and countless applications. The
Monolithic series interconnected DSC modules further reduce the cost and enhance the
stability of the design by providing better sealing and encapsulation for the durability of
the cell and a reduction in the module dead-area. The novelty of the method is that it
makes the interconnections possible in liquid as well as solid-state DSC configurations
using the same strategy. In this study, fabrications of Monolithically inter-connected
Dye Sensitized Solar Cell (DSC) modules were successfully carried out for mini-modules of 5cm×5cm and 10cm×10cm. Also studied are the optimal geometric factors,
i.e. cell thickness, cell width, spacer thickness, counter electrode thickness, contact
resistance and sealing methods were also investigated. A series interconnection of strips
of 10 individual cells resulting in a 10cm ×10cm module were fabricated. Having an
active area of almost 80% have been demonstrated, with an efficiency of 3.5%. In the
present study efficiency of 6% has been achieved for modules using platinum as a
catalytic and counter electrode material for an area of 5cm×5cm. The optimisation of
the geometrical parameters relative to the interconnection strategy has resulted in an
appreciably high fill factor, almost close to 70%. Novel module manufacturing methods
that are cheap and guarantee high efficiency with a drastic reduction in module overall
series resistance have been developed
Household decision-making, empowerment and risk attitudes in Indonesia
Get PDFThis thesis consists of three self-contained yet interrelated studies in household economics.
All the studies in this thesis are based on the Indonesia Family Life Survey data set. The first
study examines the impact of positive income shock, which is conditional cash transfer
(CCT) programs, on frivolous consumption among the poor households. This study combines
propensity score matching (PSM) and difference-in-differences (DID) strategy to estimate
the average treatment effect on the treated (ATET) of the CCT programs. The results detect
declining consumption on frivolous goods by the program participation. This finding
potentially comes from the enforcement to education conditionality as the beneficiary
households are found to increase expenditure in education-related goods.
The second study investigates the role of mothers’ bargaining power on children health
outcome. Using instrumental variables (IVs) approach, not only using individual and
community attributes but also this study accounts for cultural aspect, namely patrilocality
degree for ethnic group, to determine a direct measure of women’ bargaining power in the
household. This study finds that mothers with higher bargaining power improve long term
measure of child health status as indicated by higher height-for-age z-scores (HAZ) and lower
prevalence of stunting. Further analysis reveals that only boys seem to enjoy the benefit of
greater bargaining power of the mother while girls do not.
The third study discusses whether and to what extent risk preferences of the parents,
especially the household head, influence education outcome of the children as measured by
child cognition. This study adds to the existing literature by providing supportive evidence
of an inverse link between parental risk aversion and child cognitive ability in the developing
country context. This finding indicates that education is deemed a risky investment. Further
investigation exposes that girls are more sensitive toward risk-averse parents rather than boys
in terms of education outcome, especially in matrilocal communities.Heriot-Watt Universit
Investigating human factors of automated driving
Get PDFAutonomous vehicles have become a reality. This novel form of transport is expected to
enhance road safety, provide comfort for drivers and decrease traffic problems. From a
human factors’ perspective, however, this advanced technology can give rise to negative
effects. Considerable human factors’ research on automated driving has already been
undertaken on issues such as loss of situation awareness, vigilance decrement, suboptimal
mental workload, over trust, distrust and acceptance of automated driving systems.
Nevertheless several gaps still remain, particularly, how people in different age spectrum
consider this novel technology?: what is the main factor to predict acceptance of the
technology?: and how can the transition between automated and manual driving be
managed safely in ways that drivers find acceptable?. To answer these empirical
questions, four detailed studies were performed: i) public opinion; ii) factors influencing
users’ acceptance; iii) the impact of control transition between automated and manual
driving on drivers’ mental workload and driving performance; and iv) appropriate
strategies to deal with system failure. To specifically answer those initial research
questions and achieve the accurate results, the whole research process has been
undertaken in four individual studies, using various research designs and methods of data
collection (focus groups, a survey and two experiments using a driving simulator) in order
to address these gaps in existing research. The specific research questions and objectives
are pointed out in individual studies. Results of the focus group study revealed the
different opinions toward autonomous driving and provided suggestions to develop the
technology in the future. Key among these were the participants cited almost equal
numbers of positive and negative opinions on the technology. Older people favoured
automated driving technology more than others and younger people considered more
negative opinions than other groups. Moreover, various options of autonomy, specific
software for people with a disability or older people and ease of use were pointed out for
the system development. Findings from the survey presented factors that predict users’
intention to use autonomous driving. The predictive models between the U.K and
Thailand are different. In addition, the participants from Thailand were willing to use
highly automated driving than the U.K.’s participants. The results of the focus group and
survey also revealed that people were concerned about the technology especially when
the system failure. Thus, the first driving simulator study was conducted to investigate
the effects of control transition between automated and manual driving. The findings
found the negative effects of transition control between automated and manual driving on drivers’ mental overload and driving performance degradation. Finally, the second
driving simulator study reported the situation criticality and difficulty when the driver
took over control from automated driving. It found that the autonomous system bringing
the vehicle to a controlled stop is an appropriate strategy when the autonomous system
fails. These main findings have important implications for consideration by technology
makers and designers when designing automated vehicles. For example, at present, the
dominant strategy in the case of automation failures is to hand back control to the driver.
The results of this research suggest that the vehicle should bring itself to a controlled stop
Mathematical modelling and analysis of sheared energetic materials
Get PDFThe large stores of chemical energy within energetic materials mean that their improper handling poses a serious safety concern, and so appropriate safety protocols
need to be put in place. To do so requires understanding of the materials, but
their mechanical properties cannot be easily determined through rheometric testing
precisely because of the safety risks involved.
To proceed, mathematical models can be created that simulate the mechanical
behaviour. The core objective of the present work has been to develop reduced and
thus mathematically tractable models that, by focusing on general physical principles, capture the essential aspects of the behaviour of sheared energetic materials.
In particular, a dynamic model has been developed to examine the phenomenon of
shear banding, motivated by its ability to cause extreme shear rates within shrinking
regions and the resultant likelihood of hotspot generation.
In the model, dynamic shear banding is treated by considering simple shear of a
slab of shear-softening, bi-viscous fluid, producing a parabolic free-boundary problem for the diffusion of shear stress through the shear band and surrounding unyielded material. The behaviour of the system is analysed using precise asymptotic
methods, and expansions involving material parameters are obtained for the growth
of the shear band over time, and the spacetime variation of shear stress within the
band and the surrounding region. In turn, the latter allows for determination of
the spacetime variation of the rate of local heating due to mechanical dissipation
within the band and the hard region, giving order-of-magnitude predictions for temperature increases likely to occur within the material. Since the expansions are in
terms of material parameters, an opportunity for comparison against experiment
arises. We expect good qualitative agreement of the model with empirical results
after assignment of suitable parameter values for a given material sample.UK Engineering and Physical Sciences Research Council (grant EP/L016508/01
Development of a broadband visible to infrared astrocomb
Get PDFMotivated by the applications of high-resolution spectroscopy in astrophotonics, this
thesis presents the experimental work and development of a broadband wide-mode-spacing astrocomb system as part of the UK consortium for the High Resolution
Spectrograph (HIRES) in the forthcoming Extremely Large Telescope (ELT).
A 1-GHz Ti:sapphire laser served as the master source comb of the system,
pumping a visible broadband supercontinuum and a phase-coherent PPKTP-based
degenerate optical parametric oscillator (OPO), spectrally broadened in highly non-linear fibre. The system was fully stabilised, and provided an atomically-traceable
1-GHz astrocomb across nearly two octaves from 500–2200 nm.
Residual noise from the dither-locked OPO hindered subsequent modal filtering
in a Fabry P´erot cavity motivating development of a dither-free technique exploiting the always-present parasitic sum-frequency light. This locking protocol provided
a sixfold lower relative intensity noise and nearly five times less power fluctuation
than the dither-locked version, leading to successful modal filtering from 1150–1800
nm. The spectra of the filtered output were recovered using Fourier transform spectroscopy, with the 10-GHz comb mode spacings directly resolved optically. Through
linear regression, fCEO was found to be 565.7 MHz ± 64.3 MHz and frep was calculated as 992.1 MHz ± 352 Hz. Individual comb mode values were identified with
high accuracy using the comb equation. This technique was verified with a heterdyne experiment between the Ti:sapphire comb source and a Rb-referenced CW
laser to which the FTS was referenced.
An alternative approach for visible wavelength generation was investigated using
a PPLN waveguide. Using a sample optimised for 1560 nm, 1.5 % of our coupled
degenerate OPO input was up-converted to second harmonic generation as well as
sum-frequency mixing and showed a good agreement with modelled results, providing confidence in the ability to use such simulations to design fully grating-engineered
PPLN waveguide for broadband conversion
Queering the academy : UK academics’ negotiation of heteronormativity at work
Get PDFWhilst ‘queering’ theory has been accepted in higher education institutions, queering the
organization is not, and more research is needed on lesbian, gay, bisexual, trans*, queer, intersex
and asexual (LGBTQIA) and heterosexual academics’ experiences working in Higher Education
(Rumens, 2016a). In this manner, research omits to examine how LGBTQIA academics are
marginalized and the ways in which sexuality and gender are inextricably “linked within a
dominant heterosexual masculinity” in which academics are judged (Fisher, 2007:512). Research
drawing on queer theory is thus needed in order to expose the oppressive systems conditioned and
(re)produced by ‘compulsory heterosexuality’, also termed heteronormativity, within academic
institutions (Ozturk and Rumens, 2014). Although queer theory recognizes multiple and
intersecting identities, it often tends to “reinforce simple dichotomies between heterosexuality and
everything ‘queer’” (Gamson and Moon, 2004:52) hence neglects and subsumes differences in
terms of marginalized sexualities (Angelides, 2006) and how non-normative sexualities can render
gender as a ‘stable’ category unstable (Butler, 1999).
Given the lack of research ‘queering’ the academy, the aim of the study was to understand how
academics navigate working life in workplaces where discourses around ‘heterosexuality’ are the
implicit norm. Using snowball sampling to facilitate recruitment of participants, this study is based
on 30 qualitative semi-structured interviews with academics identifying across the sexuality and
gender spectrum working in UK higher education. All of the transcribed interviews were analysed
for emerging discourses relating to LGBTQIA academics’ workplace experience and
heteronormativity. Additionally, to explore how sexualities negotiate the heterosexual/homosexual
binary and the possibility to deconstruct essentialism, Willig’s (2013) Foucauldian discourse
analysis was utilised.
Rather than challenging normative ideals of gender and sexuality, the study found academia, as an
heteronormative institution, (re)produces knowledge, practices and norms that marshals academics
‘with a sexuality’ into a ‘queer friendly closet’ which restricts opportunities to ‘bring sexuality to
work’. This was found to present particular challenges for self-identified bisexual academics that
further have to negotiate the heterosexual/homosexual dichotomy and its underpinning norms
rendering non-binary sexualities ‘invisible’. From a queer theoretical perspective, this poses
problems for the opportunity to deconstruct essentialism and ‘queer’ the academy away from
binary and normative thinking
Numerical simulation and optimisation of polymer flooding in a heterogenous reservoir : constrained versus unconstrained optimisation
Get PDFPolymer flooding offers the potential to recover more oil from reservoirs but requires significant
investments which necessitate a robust analysis of economic upsides and downsides. Key
uncertainties in designing a polymer flood are often reservoir geology and polymer degradation.
The objective of this study is to understand the impact of geological uncertainties and history
matching techniques on designing the optimal strategy for, and quantifying the economic risks of,
polymer flooding in a heterogeneous clastic reservoir.
We applied two different history matching techniques (adjoint-based and a stochastic algorithm)
to match data from a prolonged waterflood in the Watt Field, a semi-synthetic reservoir that
contains a wide range of geological and interpretational uncertainties. Next, sensitivity studies
were carried out to identify first-order parameters that impact the Net Present Value (NPV). These
parameters were then deployed in an experimental design study using Latin Hypercube Sampling
to generate training runs from which a proxy model was created using polynomial regression. A
particle swarm optimization algorithm was employed to optimize the NPV for the polymer flood.
The same approach was used to optimize a standard water flood for comparison. Optimizations of
the polymer flood and water flood were performed for the history matched model ensemble and
the original ensemble.
The Adjoint technique yielded a better quality match compared to stochastic history matching,
whereas, the stochastic history matching resulted in a more diverse set of history matched
ensemble. The optimal strategy to deploy the polymer flood and maximize NPV varies based on
the history matching technique. The average NPV and the variance is predicted to be higher by
4% (149 million) respectively in the stochastic history matching
compared to the adjoint technique. This difference is due to the ability of the stochastic algorithm
to explore the parameter space more broadly, which created situations where the oil in place was
shifted upwards, resulting in higher NPV. Optimizing a history matched ensemble leads to a
narrower range in absolute NPV compared to optimizing the original ensemble. This difference is
because the uncertainties associated with polymer flooding are not captured during history
matching. The result of cross comparison, where an optimal polymer design strategy for one
ensemble member is deployed to the other ensemble members, predicted a decline in NPV but
surprisingly still shows that the overall NPV is higher than for an optimized water food, even for
sub-optimal polymer injection strategies. This observation indicates that a polymer flood could be
beneficial compared to a water flood, even if geological uncertainties are not captured properly.
This thesis reported the bias of stochastic algorithm by creating reservoir models where oil in place
were shifted upwards. This can be further investigated and addressed
Bayesian modelling of Critical Illness Insurance claim rates
Get PDFIn this thesis we present methods and results for the estimation of claim rates for
Critical Illness Insurance (CII) in the UK and an actuarial application (pricing) of
these methods and results. This is the first study to provide a Bayesian stochastic
model for the claim rates for CII products in the UK, taking potential overdispersion into account. The data have been supplied by the Continuous Mortality
Investigation (CMI) and relate to claims settled in the years 2007–2010. First, we
develop a classical Poisson model and a classical negative binomial model for the
claim rates to determine the age-specific structure of the related models using a
stepwise regression where BIC is the selection criterion. Three Bayesian models
are then built to model the crude rates where the observed numbers of claims are
assumed to have, Poisson, Poisson-lognormal, and negative binomial distributions.
Variable selection is applied to the covariates using Bayesian methodology to obtain the best model with different prior distribution setups for the parameters. The
claim rates are then estimated using their posterior distributions and Markov Chain
Monte Carlo methodology under the three models with the best predictor. The
three models are assessed using Bayesian residual analysis, Bayes factors and the
Deviance Information Criterion. The net premium rates for certain types of policies
are calculated using Markov Chain Monte Carlo methodology and the estimated
claim rates. Our estimated rates are compared with rates from an earlier period
(1999–2005) and show a moderate increase over time, while comparisons with rates
published by the Continuous Mortality Investigation (CMI) reveal consistent trends
with some distinct differences
Optimising the use of zebrafish embryos and in vitro models as alternatives to rodent testing for assessing neutrophil responses to nanomaterials
Get PDFNanomaterials (NMs) exhibit unique properties making them highly attractive for use in
a variety of consumer products. There are uncertainties regarding the potential adverse
effects of engineered NMs on human health, thus their safety must be thoroughly assessed
to ensure the responsible exploitation of nanotechnology. Assessment of NM safety
currently relies on the use of rodents, where neutrophil accumulation is commonly used
as a marker of NM toxicity. However, there is an urgent need to replace, reduce, and
refine the use of animals in research. Thus, zebrafish (Danio rerio) embryos and in vitro
models should be utilised more widely when screening NM toxicity.
In this thesis, non-protected stages of transgenic zebrafish larvae with fluorescentlylabelled neutrophils were exploited to assess inflammatory responses to NMs using two
exposure methods: aqueous exposure of injured fish (using the tail fin injury model), and
microinjection into the otic vesicle. Inflammatory responses were monitored at 4, 6, 8, 24
and 48 hours. In the tail fin injury model it was shown that post-injury aqueous treatment
with silver (Ag) and zinc oxide (ZnO) NMs stimulated an enhanced and sustained
neutrophilic inflammatory response. Following microinjection, Ag NMs stimulated a
neutrophilic inflammatory response at the otic vesicle. These results show that the
zebrafish larvae can be used to quickly determine the pro-inflammatory effects of NMs.
An in vitro investigation of neutrophil responses was performed to investigate the
migration of primary human neutrophils and a neutrophil-like cell line (HL-60) towards
conditioned medium (CM) generated by lung epithelial cells (A549 cells) exposed to Ag
NMs. It was found that the migration of neutrophils was not observed in response to CM.
These results highlight that further optimisation is required to improve upon existing
neutrophil in vitro models, in order to fully assess the pro-inflammatory effects of NMs
in vitro.
Overall, the results demonstrate that zebrafish larvae are a suitable alternative model for
assessing NM toxicity, and that they should be prioritised for use by the wider scientific
community for assessing chemical safety (e.g. NMs, pathogens, pharmaceuticals). In
addition, the responses of neutrophils in vitro can be studied to quickly assess NM
toxicity. The more widespread use of such alternative, non-rodent models will make
nanotoxicology testing quicker, cheaper, and more ethical
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!