1,721,025 research outputs found
Automatic conversion of activity diagrams into flexible smart home apps
No full textDespite the availability of a large number of sensor and actuator devices designed to co-perform in a smart home, only a few of these devices are easily integrated into a single smart home unit. However, as devices become more advanced and feature-rich, the need for smart software to orchestrate these devices to offer complex smart home services has risen. The research focus of this thesis is designing and deploying software (or apps) that works with different and changing, sensor-actuator configurations in smart-homes.
A systematic literature review was used to identify a visual design modeling framework for designing smart home apps. Behavioral models, specifically UML Activity Diagrams were identified as the most appropriate app design model due to high usability and similarity with flowcharts. The literature review also informed the key qualities of an end-to-end solution to design and deploy these smart home apps. Subsequently, we design and develop an automatic translation tool to address some key usability and deployment challenges. This tool offers a customized and fully-featured UML Activity Diagram Editor that allows non-experts to model any smart home system, such as a smart lighting system. The compiler offered by the Automatic Translation Tool accepts UML Activity Diagrams as input and generates executable Java code which can be deployed into any smart home application. An evaluation using a representative a set of case studies shows that the Automatic Translation Tool features high usability, availability, and performance
GALADE: A Round-Trip Graphical Modelling Tool for Abstraction Layered Architecture Applications
No full textIn recent years, a new software architecture, the Abstraction Layered Architecture (ALA), has emerged at Datamars Ltd to help address the issue of code bases becoming harder to maintain over time. Previous quantitative assessments, using a refined set of metrics based on the ISO/IEC 25010 and 25023 quality models, have strongly indicated that using ALA to develop an application allows for high modularity, testability, reusability, and analysability. These merits are largely supported through the separation of logical software components into artefacts called \textit{domain abstractions}. These domain abstractions are wired together at runtime through common interfaces, through which they communicate with one another.
This usage of wiring has enabled the visualisation of ALA applications as directed port graphs, and has led ALA application development to incorporate a diagram-first approach. The design of the application would be drawn first, from which the application code would be written. The problem that arose, then, was that it was a manual and time-consuming process to ensure that changes in the diagram were correctly reflected in the code, and vice versa. After seeing promising results from a prototype code generation tool, Datamars Ltd sought to develop a graphical tool that could visualise ALA diagrams, automatically generate the corresponding application code, and keep both ends synchronised.
We have systematically examined the literature, and found that no tools exist that can holistically satisfy the requirements that such a tool would impose.
Therefore, this thesis presents the Graphical Abstraction Layered Architecture Development Environment (GALADE), a novel tool to support the visualisation and maintenance of ALA applications. The creation of this tool has been the result of a productive partnership between Auckland University of Technology (AUT) and Datamars Ltd. All of the development for this tool has been performed by the author, with consultation and resources provided by Datamars Ltd.
We have used the Design Science research methodology to frame the design, development, and evaluation of GALADE. A case study of GALADE in use at Datamars Ltd suggested that it improves productivity in ALA-based development, and a qualitative and quantitative evaluation has shown that GALADE shows significant improvements to the previous diagram-first design process for ALA.
Finally, ALA is a reference software architecture that has so far shown promise in the embedded software field, and research is underway to examine its applicability in other software fields. This implies that GALADE has the potential to be a general-purpose tool for the visualisation and development of highly maintainable software, therefore GALADE may be significant for the wider community of software engineering researchers and practitioners
Improved Deep Learning Model Based on Integrated Convolutional Neural Networks and Transfer Learning for Shoeprint Image Classification
No full textMachine Learning (ML) and Deep Learning (DL) techniques have recently aided in resolving critical problems in various sectors and use cases. For example, image classification techniques based on machine learning and deep learning have proven useful in medical science and other industries. Existing work has investigated a shoeprint image classification technique to identify different classes of shoeprints for several forensic applications and wear-pattern identifications. Despite the challenges presented by the situation, there are several opportunities to explore this area. The need for more sufficient datasets is the main obstacle in this field. In addition, deep learning techniques frequently fail to achieve a high accuracy since the field still needs to be sufficiently developed. The literature has, however, used a range of traditional machine-learning algorithms.
This thesis has applied deep learning classifiers for shoeprint identification. This study first explored traditional classification methods and deep learning Convolutional Neural Networks (CNN). Then it proposed a method for integrating CNN and Transfer Learning (CNN-TL) to improve the classification results. In CNN and CNN-TL methods, python's tensor-flow library was used. Finally, the shoeprint classifications were performed using a pre-trained and fine-tuned version of the Inception model, including comparing different pre-trained Inception V3, VGG16, and ResNet50 models. The results show that convolutional neural networks-transfer learning (CNN-TL) improved classification accuracy by approximately 3% compared to conventional convolutional neural networks (CNN).
The study employed three techniques for shoeprint classification, namely CNN, TL, and the proposed method that combined TL with CNN with various pre-trained models (Inception V3, VGG16, and ResNet50). The performance metrics of each model employed in this study produced the following individual results: CNN model (accuracy = 96.17%), CNN-TL Inception V3 model (accuracy = 92.19%), CNN-TL VGG16 model (accuracy = 96.88%), and CNN-TL ResNet50 model (accuracy = 97.14%).
The ResNet50 model achieved the highest accuracy of 97.14%, outperforming all state-of-the-art approaches in shoeprint classification. Regarding accuracy, the VGG16 model outperformed the CNN model, but the Inception V3 model performed with lower accuracy. The study highlights that the proposed methodology significantly improved the accuracy compared to previous literature. The proposed methodology is expected to open new avenues for forensic science research and deep learning approaches to image classification
DynaCool - Simulating Efficient Liquid Cooling for Current and Next Generation Large Scale Data Centres
No full textEnergy consumption in Large Scale Data Centres (LSDC’s) doubled from 2000 to 2006 reaching 61 TerraWatt-hour (TWh) per year. Most power generation sources are sadly fossil fuelled, which is increasing the effects of anthropomorphic climate change. 99-100% of the energy consumed by IT equipment is dissipated as heat from the servers, which creates a real problem of cooling in LSDC’s. Air cooling systems in LSDC’s are struggling to handle the increased cooling demands, which is why they account for 40% of the total energy consumed. The rest of the energy consumed is used to power the IT equipment and data centre infrastructure facilities like lighting. The ratio of power consumed by the data centre facility to the power consumed by the IT equipment is known as Power Usage Effectiveness (PUE), which is a metric used to measure data centre efficiency. Reducing the PUE by even fractions of percentages can prevent millions of tons of greenhouse gases from being emitted into the atmosphere. One of the methods of reducing PUE is by using alternate forms of cooling technologies like liquid cooling. This thesis explores novel optimisation methods for cooling control in liquid cooled LSDC’s. The three strategies focused on are Static Flow Rate, Variable Flow Rate (VFR) and the proposed Pulsed Variable Flow Rate (PVFR) cooling control strategies. The power consumption of coolant pumps and the effect it has on reducing energy consumption and PUE are investigated for all three cooling control strategies using computer simulations. Current simulation software were limited to air cooling and as such we needed to develop a proprietary computer simulation software.
The software we developed was DynaCool and the simulation data we gathered was used to analyse the effectiveness of the different cooling control strategies. DynaCool was built using requirements engineering and model driven design to ensure the validity of the software, as these methodologies are commonly used in large complex industrial systems. The data analysed indicated a PUE reduction of at least 15.4% for the novel PVFR liquid cooling control strategy over the static and VFR cooling control strategies. This reduction equates to savings of 2.84 million tons of greenhouse gas emissions and 18.788 TWh of power consumption per year for an adoption rate of 100%. Realistically speaking, an adoption rate of 10% would yield power saving of 1.88 TWh or 284 thousand tons of greenhouse gas emission per year. This adoption rate is easily achievable by the industry as recent trends indicate data centre operators are moving towards alternate cooling technologies. This is evidenced by Google aiming to achieve carbon neutrality by 2017 using liquid cooling technologies
Computer-Aided Diagnosis for Early Detection of Melanoma Based on Deep-Learning Techniques
No full textMelanoma is the deadliest form of skin cancer, with a high mortality rate every year, and New Zealand is known to be one of the countries with the highest incidence of this disease. Overexposure to ultraviolet sun rays causes the upper layers of the skin to produce a pigment known as melanin, the primary cause of melanoma. An early diagnosis and prognosis of melanoma can improve survival rates before it becomes dangerous. There is a high demand for computerized automated skin lesion analysis techniques at low cost since trained specialists are in limited supply, and manual diagnosis involves high costs. The increasing use of non-invasive methods in diagnosing malignant melanoma reduces the need for biopsies.
A primary objective of this thesis is the development of computerized detection systems for analyzing lesions and distinguishing melanoma from other types of skin cancer. The thesis focuses mainly on designing novel methods for three major phases of automatic melanoma diagnosis process: a) Pre-processing deals with removal of noise artefacts such as hairlines and improving image’s contrast, b) Lesion segmentation to accurately extract lesion region using deep learning, and c) Melanoma classification using deep learning for a fast and accurate detection.
For the first phase, an algorithm Intensity Adjustment-based Hair Removal (IA-HR) employing morphological operators is developed to remove the hairlines, which are a significant problem in skin image samples. Additionally, a Multi-scale Context Aggregation Convolutional Neural Network (MCACNN) is used to enhance the contrast and resolution of images. A cleaned dataset is then generated using these pre-processing methods. A class imbalance problem is also addressed using data augmentation methods.
To determine lesion borders and extract lesion information, deep learning networks are used. Two network designs were constructed; one was based on encoder-decoder layered patterns (EDNet) and the other on atrous convolutions (DilatedSkinNet). As DilatedSkinNet shows higher average accuracy, thus it is preferred to EDNet for segmentation tasks.
Finally, in the classification stage, we present the design of a multi-layer deep convolutional neural network (DCNN) named as LCNet to distinguish melanoma from benign tumors. The designed classification network is a lightweight network having a smaller number of learnable parameters. Due to the less complex architecture of classification network, it takes less inference time. Moreover, the network can also diagnose diseases efficiently and accurately without pre-processing and segmentation techniques, unlike traditional machine learning classifiers that rely heavily upon these initial steps.
The study examined, however, the effects of applying designed pre-processing and segmentation methods on segmentation and classification performance. The aim of this effort is to further enhance the performance of the classifier by preparing more rich and clean data for training. The classification network is fed with hairlines-free, high-contrast, and segmented images, and its performance is compared to raw images. Our experiments showed that the classification network efficiently processes raw and complex data by offering an accuracy (ACC) of 90.92±1.0%. The accuracy performance of classification model is improved with pre-processed data as 92.47% and with pre-processed+segmented data as 93.40% indicating the classification model’s performed higher with pre-processing and segmentation operations to distinguish melanoma vs benign. Additionally, it is observed that noise-free and cleaned data using IA-HR and MCACNN methods improved the performance of the segmentation approach.
In our study, we found that the proposed pre-processing and segmentation methods could improve the performance of deep learning-based classifiers. Furthermore, there are many areas of melanoma diagnosis process where the proposed approaches can be successfully applied. The denoising method can be used to clean skin samples without causing any discomfort to the patients such as hairlines can be automatically eliminated from images and contrast can be enhanced. Another use of segmentation model to extract lesion region and to perform detailed analysis of it. Additionally, segmentation method can be employed to generate accurate and smooth ground truth labels for new samples that are currently annotated manually by experts. The classification model can be used to classify melanoma in less time such as in 1.3 seconds as predicted by our model. This method may also be used to generate a handcrafted feature extraction process if a machine learning-based classifier is employed to diagnose skin cancer. The DCNN classification model also showed more improvement in diagnosing melanoma when trained on a large, balanced, and pre-processed dataset. In its future scope, embedded systems such as FPGA based system-n-chip and other resource-constrained implementations can benefit from the designed classification network
Automatic Conversion of Activity Diagrams Into Flexible Smart Home Apps
No full textDespite the availability of a large number of sensor and actuator devices designed to co-perform in a smart home, only a few of these devices are easily integrated into a single smart home unit. However, as devices become more advanced and feature-rich, the need for smart software to orchestrate these devices to offer complex smart home services has risen. The research focus of this thesis is designing and deploying software (or apps) that works with different and changing, sensor-actuator configurations in smart-homes.
A systematic literature review was used to identify a visual design modeling framework for designing smart home apps. Behavioral models, specifically UML Activity Diagrams were identified as the most appropriate app design model due to high usability and similarity with flowcharts. The literature review also informed the key qualities of an end-to-end solution to design and deploy these smart home apps. Subsequently, we design and develop an automatic translation tool to address some key usability and deployment challenges. This tool offers a customized and fully-featured UML Activity Diagram Editor that allows non-experts to model any smart home system, such as a smart lighting system. The compiler offered by the Automatic Translation Tool accepts UML Activity Diagrams as input and generates executable Java code which can be deployed into any smart home application. An evaluation using a representative a set of case studies shows that the Automatic Translation Tool features high usability, availability, and performance
An Optimized Hardware System on Chip for a Support Vector Machine Classifier: A Case Study on Melanoma Detection
No full textSupport Vector Machine (SVM) is a robust machine learning model used for efficient classification with high accuracy. SVM is widely utilized for online classification in various embedded applications. However, implementing the SVM classification algorithm for an embedded system or application is challenging, due to intensive and complicated computations required. This increases the importance of implementing SVM on hardware platforms for achieving high performance computing at low cost and power consumption.
Field-Programmable Gate Array (FPGA) is a powerful parallel processing reconfigurable device that is widely used for achieving essential performance of embedded systems, while effectively utilizing hardware resources, offering low cost and low power consumption. Accordingly, FPGA is a promising hardware platform for implementing an efficient embedded SVM classification system, while achieving vital embedded system constraints.
SVM has shown high accuracy for classifying melanoma (skin cancer) clinical images within a computer-aided diagnosis system used by dermatologists to detect melanoma early and save lives. This research aims to develop an optimized FPGA-based SVM classifier to be embedded within a low-cost handheld medical scanning device that runs an embedded SVM-based diagnosis system dedicated to early detection of melanoma in primary care. We aim to consider meeting significant constraints of embedded systems, while achieving efficient classification with high accuracy rate.
A hardware/software co-design for implementing an SVM classifier onto FPGA is proposed to realize melanoma detection on a chip. This SVM implementation achieves efficient melanoma classification on a recent FPGA-based hybrid platform “Zynq SoC” designed using the latest UltraFast High-Level Synthesis design methodology. The hardware implementation results demonstrate classification accuracy of 97.9% and a significant hardware acceleration rate of up to 37x with only 2.7% resource utilization and 1.69 watts for power consumption.
Furthermore, a scalable multi-core architecture is proposed to achieve multi-purpose classification on a single chip/device, which has been validated with a 2-stage cascade classifier implementation with accuracies of 98 % and 73%, to enhance melanoma detection. A simple hardware-friendly design is proposed for the building SVM core of the multi-core architecture, aiming to reduce hardware complexity and optimize implementation results for achieving an efficient classification performance.
A novel dynamic hardware system is also proposed for implementing a cascade SVM classifier on FPGA for early melanoma detection. The hardware implementation results are optimized by using the powerful dynamic partial reconfiguration technology, where very low resource utilization of 1% slices and power consumption of 1.5 watts are achieved.
The implemented SVM classification systems on Zynq SoC using the proposed hardware designs have shown the least power consumption results among other related implementations, in addition to significantly low hardware resource utilization and processing time with significant speedups and high classification accuracy rates at low cost. Consequently, the implemented Zynq systems meet crucial embedded system constraints of high performance and low cost, resource utilization and power consumption, while achieving efficient classification with high classification accuracy, which promises realization of a cost- and energy-efficient handheld medical scanning device for early detection of melanoma
An Optimized Hardware System on Chip for a Support Vector Machine Classifier: a Case Study on Melanoma Detection
No full textSupport Vector Machine (SVM) is a robust machine learning model used for efficient classification with high accuracy. SVM is widely utilized for online classification in various embedded applications. However, implementing the SVM classification algorithm for an embedded system or application is challenging, due to intensive and complicated computations required. This increases the importance of implementing SVM on hardware platforms for achieving high performance computing at low cost and power consumption.
Field-Programmable Gate Array (FPGA) is a powerful parallel processing reconfigurable device that is widely used for achieving essential performance of embedded systems, while effectively utilizing hardware resources, offering low cost and low power consumption. Accordingly, FPGA is a promising hardware platform for implementing an efficient embedded SVM classification system, while achieving vital embedded system constraints.
SVM has shown high accuracy for classifying melanoma (skin cancer) clinical images within a computer-aided diagnosis system used by dermatologists to detect melanoma early and save lives. This research aims to develop an optimized FPGA-based SVM classifier to be embedded within a low-cost handheld medical scanning device that runs an embedded SVM-based diagnosis system dedicated to early detection of melanoma in primary care. We aim to consider meeting significant constraints of embedded systems, while achieving efficient classification with high accuracy rate.
A hardware/software co-design for implementing an SVM classifier onto FPGA is proposed to realize melanoma detection on a chip. This SVM implementation achieves efficient melanoma classification on a recent FPGA-based hybrid platform “Zynq SoC” designed using the latest UltraFast High-Level Synthesis design methodology. The hardware implementation results demonstrate classification accuracy of 97.9% and a significant hardware acceleration rate of up to 37x with only 2.7% resource utilization and 1.69 watts for power consumption.
Furthermore, a scalable multi-core architecture is proposed to achieve multi-purpose classification on a single chip/device, which has been validated with a 2-stage cascade classifier implementation with accuracies of 98 % and 73%, to enhance melanoma detection. A simple hardware-friendly design is proposed for the building SVM core of the multi-core architecture, aiming to reduce hardware complexity and optimize implementation results for achieving an efficient classification performance.
A novel dynamic hardware system is also proposed for implementing a cascade SVM classifier on FPGA for early melanoma detection. The hardware implementation results are optimized by using the powerful dynamic partial reconfiguration technology, where very low resource utilization of 1% slices and power consumption of 1.5 watts are achieved.
The implemented SVM classification systems on Zynq SoC using the proposed hardware designs have shown the least power consumption results among other related implementations, in addition to significantly low hardware resource utilization and processing time with significant speedups and high classification accuracy rates at low cost. Consequently, the implemented Zynq systems meet crucial embedded system constraints of high performance and low cost, resource utilization and power consumption, while achieving efficient classification with high classification accuracy, which promises realization of a cost- and energy-efficient handheld medical scanning device for early detection of melanoma
TORUS: Tracing Complex Requirements for Large Cyber-physical Systems
No full textCyber-Physical Systems are embedded computers that control complex, physical processes
via autonomous peripherals while cooperating as agents in distributed networks.
Due to the scale and complexity of the interactions that occur within cyber-physical
systems, tracing system requirements accurately and appropriately is extremely hard.
The literature confirms that they are even harder to maintain and keep up-to-date during
the life of the project.
However, the information that requirements traceability provides is a crucial part of
determining the completeness of an application. Existing requirements management
systems do not scale well and traceability is difficult in such highly heterogeneous
environments.
This research presents TORUS (Traceability Of Requirements Using Splices), a
novel traceability framework that operates outside of, yet connects to, diverse requirements
and development environments. Our approach introduces Splices, autonomous
traceability data structures that persist trace information through the inevitable changes
that occur during system design and development.
A Design Science research methodology was adopted to show how the TORUS
framework can be applied to cyber-physical systems that employ the IEC 61499 Function
Block Architecture. A mechanical item sorting machine is modeled, the requirements
of which are described initially using CESAR (Cost-efficient Methods and Processes for Safety-relevant Embedded Systems) requirement templates. These templates
help to formalize the pre-Requirement Specification’s free-form text into less
ambiguous requirements statements. A domain ontology is defined before modeling the
requirements further within the Sparx Enterprise Architect Requirements Management
system. Enterprise Architect uses SysML diagrams to capture each requirement in
context with its acceptance tests, non-functional and safety requirements while the
model can be persisted for later use.
Formal mathematical models of requirements, function blocks and splices are
presented to show how this trace information can be mined, delivering important project
metrics to stakeholders. By capturing not only the current state of the system but also
by preserving historic traces, TORUS allows project teams to see a much richer view of
their system’s artifacts.
In parallel with the creation of these models, prototypes of TORUS were created
in Java to explore the proposed splice metadata model. These demonstrated that it is
possible to extract trace information directly from both Enterprise Architect models and
the nxtStudio IEC 61499 object repository. Using the relationships expressed by these formalisms, the resulting metadata information model for splices is extended to demonstrate how these entities can capture the status of each requirement. We define a set of splices as being the Skein of the system;
the set of traces that connect the model and application artifacts together like warp and
weft of the threads in a tapestry. Information aggregated in this way is important since
it provides quantifiable metrics that can be used to provide an empirically-determined
overall state of the system under examination.
The results indicate that the TORUS framework scales well and that the skein and
splices can provide metrics that should allow us to perform code-level validation and
completeness checking in the future
Light-Weight Active Security Solutions for Resource-Constrained ICPS
No full textIndustrial Cyber-Physical Systems (ICPS) are driving the 4th Industrial Revolution, significantly impacting the productivity and efficiency of all sectors, including industrial automation. Central to this revolution is the networking and digitisation of multi-domain and large-scale physical systems within the industrial context. However, the seamless convergence of the digital and physical world has made ICPS vulnerable to new and sophisticated security threats. Ensuring the security requirements of ICPS is paramount, especially against cyber attacks that can significantly impact the availability of critical ICPS applications. The ICPS applications traditionally execute on resource-constrained devices like PLCs. These devices have limited resources, and standard security measures are inadequate to safeguard them due to the resource limitations. Balancing security requirements with distinctive characteristics of resource-constrained ICPS is pivotal for maintaining the performance, availability, and robustness of ICPS applications.
In this research, the significant contributions of our works are framed as research objectives and achieved using design science research methodology. We have presented several novel light-weight active security solutions developed to address the current gaps against cyber attacks, specifically Distributed Denial of Service (DDoS) attacks on the resource-constrained ICPS. DDoS attacks are the most reported attacks that disrupt or degrade the availability of systems, either by overloading them with a flood of packets or exploiting vulnerabilities. Considering the disruptive and degrading impacts of DDoS attacks on the normal operations of the resource-constrained ICPS, this thesis focuses on detecting such attacks using light-weight active security solutions. The light-weight active security solutions are considered generic and programmable security measures that can proactively protect the devices with minimal overhead on their performance.
Systematic Mapping Study (SMS) and Systematic Literature Review (SLR) results have shown that light-weight active security solutions are crucial for resource-constrained ICPS to deal with DDoS attacks. We have also proposed the generic active security technique for detecting DDoS attacks on resourceconstrained ICPS. Moreover, the notable inclusion of a novel multi-vector and cross-domain DDoS attack taxonomy helps us to devise the solutions for multi-scale flooding attacks and attack volumes and binary and multi-class slow-rate attack detection frameworks. The proposed works’ effectiveness was determined using PLCs and publicly available datasets. The evaluations show noteworthy accuracy, low prediction time, and distinguished performance over existing state-of-the-art mechanisms
- …
