To see the other types of publications on this topic, follow the link: Edge devices.
Dissertations / Theses on the topic 'Edge devices'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 dissertations / theses for your research on the topic 'Edge devices.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
1
Choi, Myung Cheon. "Traffic flow management for RSVP/ATM edge devices." Diss., Georgia Institute of Technology, 1999. http://hdl.handle.net/1853/15520.
Full text
2
Barnes, Brian E. "Real-time resource management for RSVP/ATM edge devices." Diss., Georgia Institute of Technology, 1999. http://hdl.handle.net/1853/15395.
Full text
3
Pinkney, Thomas David. "Wound-edge protection devices to reduce surgical site infection." Thesis, University of Birmingham, 2017. http://etheses.bham.ac.uk//id/eprint/7588/.
Full textAbstract:
This thesis provides an overview of the design, conduct and analysis of a multicentre phase III trial exploring the clinical effectiveness of a novel in-theatre intervention to try to reduce rates of post-operative surgical site infection (SSI). The pitfalls inherent in the conduct of research into SSI are discussed, along with measures to try and overcome these. The wound-edge protection device (WEPD) intervention is defined and the published evidence assessing its clinical effectiveness systematically appraised. Clinical surgical research and its difficulties are described, and the paradigm-shift bought about by the new trainee-led research collaborative model introduced. The design considerations involved creating a pragmatic and simple trial within the complex intervention that is surgery are explored in the context of the creation of the ROSSINI trial. This trial successfully recruited ahead of time and target and robustly proved that WEPDs are not clinically effective in reducing SSI. It also demonstrated the power and ability of this new collaborative model, as witnessed by both the citations of the results paper and the exponential growth in similar collaborative ventures. Finally, lessons learned about SSI research and clinical surgical research are summarised, and plans for future research presented.
4
Kaarlela, H. (Heikki). "Edge adaptive filtering of depth maps for mobile devices." Master's thesis, University of Oulu, 2019. http://jultika.oulu.fi/Record/nbnfioulu-201910313045.
Full textAbstract:
Abstract. Mobile phone cameras have an almost unlimited depth of field, and therefore the images captured with them have wide areas in focus. When the depth of field is digitally manipulated through image processing, accurate perception of depth in a captured scene is important.
Capturing depth data requires advanced imaging methods. In case a stereo lens system is used, depth information is calculated from the disparities between stereo frames. The resulting depth map is often noisy or doesn’t have information for every pixel. Therefore it has to be filtered before it is used for emphasizing depth. Edges must be taken into account in this process to create natural-looking shallow depth of field images.
In this study five filtering methods are compared with each other. The main focus is the Fast Bilateral Solver, because of its novelty and high reported quality. Mobile imaging requires fast filtering in uncontrolled environments, so optimizing the processing time of the filters is essential.
In the evaluations the depth maps are filtered, and the quality and the speed is determined for every method. The results show that the Fast Bilateral Solver filters the depth maps well, and can handle noisy depth maps better than the other evaluated methods. However, in mobile imaging it is slow and needs further optimization.Reunatietoinen syvyyskarttojen suodatus mobiililaitteilla. Tiivistelmä. Matkapuhelimien kameroissa on lähes rajoittamaton syväterävyysalue, ja siksi niillä otetuissa kuvissa laajat alueet näkyvät tarkennettuina. Digitaalisessa syvyysterävyysalueen muokkauksessa tarvitaan luotettava syvyystieto.
Syvyysdatan hankinta vaatii edistyneitä kuvausmenetelmiä. Käytettäessä stereokameroita syvyystieto lasketaan kuvien välisistä dispariteeteista. Tuloksena syntyvä syvyyskartta on usein kohinainen, tai se ei sisällä syvyystietoa joka pikselille. Tästä syystä se on suodatettava ennen käyttöä syvyyden korostamiseen. Tässä prosessissa reunat ovat otettava huomioon, jotta saadaan luotua luonnollisen näköisiä kapean syväterävyysalueen kuvia.
Tässä tutkimuksessa verrataan viittä suodatusmenetelmää keskenään. Eniten keskitytään nopeaan bilateraaliseen ratkaisijaan, johtuen sen uutuudesta ja korkeasta tuloksen laadusta. Mobiililaitteella kuvantamisen vaatimuksena on nopea suodatus hallitsemattomissa olosuhteissa, joten suodattimien prosessointiajan optimointi on erittäin tärkeää.
Vertailuissa syvyyskuvat suodatetaan ja suodatuksen laatu ja nopeus mitataan jokaiselle menetelmälle. Tulokset osoittavat, että nopea bilateraalinen ratkaisija suodattaa syvyyskarttoja hyvin ja osaa käsitellä kohinaisia syvyyskarttoja paremmin kuin muut tarkastellut menetelmät. Mobiilikuvantamiseen se on kuitenkin hidas ja tarvitsee pidemmälle menevää optimointia.
5
Brada, Matej. "Graphene electronic devices in magnetic field." Thesis, Loughborough University, 2016. https://dspace.lboro.ac.uk/2134/21800.
Full textAbstract:
This thesis discusses the two dimensional allotrope of carbon known as graphene in presence of magnetic field, with special focus on edge states. The structure of graphene is described in detail and from the structure, two models are formed. The Dirac equation is a good description of graphene for large samples, far away from edges, where the boundaries can be ignored. However, it causes problems with most types of edge and hard wall approximation has to be implemented. The Dirac equation is described in detail and used to obtain an energy spectrum, wavefunction and density of states for graphene edge in a strong magnetic field. For comparison, a Bohr-Sommerfield approximation was used to find the dispersion relation and compare it to the results obtained numerically from the Dirac equation. The second model, better fitting for nano-scale systems, is the tight binding model. This model was utilized to find Energy spectrum for graphene flakes in magnetic field, which resembles Hofstadter's butterfly spectrum. The spectrum was analyzed and periodic oscillations of magnetisation dependent on magnetic field (known as the de Haas-van Alphen effect) were described. The oscillation of magnetisation depends on the shape of the dot, even though the main properties remain the same: at low magnetic field, periodic oscillations due to Aharonov-Bohm effect, turning into more chaotic oscillations depending on the boundary conditions of the given quantum dot.
6
Sankin, Igor. "Edge termination and RESURF technology in power silicon carbide devices." Diss., Mississippi State : Mississippi State University, 2006. http://library.msstate.edu/etd/show.asp?etd=etd-12162005-141206.
Full text
7
Hansson, Gustav. "Computation offloading of 5G devices at the Edge using WebAssembly." Thesis, Luleå tekniska universitet, Institutionen för system- och rymdteknik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:ltu:diva-85898.
Full textAbstract:
With an ever-increasing percentage of the human population connected to the internet, the amount of data produced and processed is at an all-time high. Edge Computing has emerged as a paradigm to handle this growth and, combined with 5G, enables complex time-sensitive applications running on resource-restricted devices. This master thesis investigates the use of WebAssembly in the context of computa¬tional offloading at the Edge. The focus is on utilizing WebAssembly to move computa¬tional heavy parts of a system from an end device to an Edge Server. An objective is to improve program performance by reducing the execution time and energy consumption on the end device. A proof-of-concept offloading system is developed to research this. The system is evaluated on three different use cases; calculating Fibonacci numbers, matrix multipli¬cation, and image recognition. Each use case is tested on a Raspberry Pi 3 and Pi 4 comparing execution of the WebAssembly module both locally and offloaded. Each test will also run natively on both the server and the end device to provide some baseline for comparison.
8
Rajakaruna, A. (Archana). "Lightweight edge-based networking architecture for low-power IoT devices." Master's thesis, University of Oulu, 2019. http://jultika.oulu.fi/Record/nbnfioulu-201906072483.
Full textAbstract:
Abstract. The involvement of low power Internet of Things (IoT) devices in the Wireless Sensor Networks (WSN) allow enhanced autonomous monitoring capability in many application areas. Recently, the principles of edge computing paradigm have been used to cater onsite processing and managing actions in WSNs. However, WSNs deployed in remote sites require human involvement in data collection process since internet accessibility is still limited to population dense areas. Nowadays, researchers propose UAVs for monitoring applications where human involvement is required frequently. In this thesis work, we introduce an edge-based architecture which create end-to-end secure communication between IoT sensors in a remote WSN and central cloud via UAV, which assist the data collection, processing and managing procedures of the remote WSN. Since power is a limited resource, we propose Bluetooth Low Energy (BLE) as the communication media between UAV and sensors in the WSN, where BLE is considered as an ultra-low power radio access technology. To examine the performance of the system model, we have presented a simulation analysis considering three sensor nodes array types that can realize in the practical environment. The impact of BLE data rate, impact of speed of the UAV, impact of distance between adjacent sensors and impact of data generation rate of the sensor node have been analysed to examine the performance of system. Moreover, to observe the practical functionality of the proposed architecture, prototype implementation is presented using commercially available off-the-shelf devices. The prototype of the system is implemented assuming ideal environment.
9
Wood, Michael G. "Active Silicon Photonic Devices Based on Degenerate Band Edge Resonances." The Ohio State University, 2016. http://rave.ohiolink.edu/etdc/view?acc_num=osu1480432902683812.
Full text
10
Tania, Zannatun Nayem. "Machine Learning with Reconfigurable Privacy on Resource-Limited Edge Computing Devices." Thesis, KTH, Skolan för elektroteknik och datavetenskap (EECS), 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-292105.
Full textAbstract:
Distributed computing allows effective data storage, processing and retrieval but it poses security and privacy issues. Sensors are the cornerstone of the IoT-based pipelines, since they constantly capture data until it can be analyzed at the central cloud resources. However, these sensor nodes are often constrained by limited resources. Ideally, it is desired to make all the collected data features private but due to resource limitations, it may not always be possible. Making all the features private may cause overutilization of resources, which would in turn affect the performance of the whole system. In this thesis, we design and implement a system that is capable of finding the optimal set of data features to make private, given the device’s maximum resource constraints and the desired performance or accuracy of the system. Using the generalization techniques for data anonymization, we create user-defined injective privacy encoder functions to make each feature of the dataset private. Regardless of the resource availability, some data features are defined by the user as essential features to make private. All other data features that may pose privacy threat are termed as the non-essential features. We propose Dynamic Iterative Greedy Search (DIGS), a greedy search algorithm that takes the resource consumption for each non-essential feature as input and returns the most optimal set of non-essential features that can be private given the available resources. The most optimal set contains the features which consume the least resources. We evaluate our system on a Fitbit dataset containing 17 data features, 4 of which are essential private features for a given classification application. Our results show that we can provide 9 additional private features apart from the 4 essential features of the Fitbit dataset containing 1663 records. Furthermore, we can save 26:21% memory as compared to making all the features private. We also test our method on a larger dataset generated with Generative Adversarial Network (GAN). However, the chosen edge device, Raspberry Pi, is unable to cater to the scale of the large dataset due to insufficient resources. Our evaluations using 1=8th of the GAN dataset result in 3 extra private features with up to 62:74% memory savings as compared to all private data features. Maintaining privacy not only requires additional resources, but also has consequences on the performance of the designed applications. However, we discover that privacy encoding has a positive impact on the accuracy of the classification model for our chosen classification application.Distribuerad databehandling möjliggör effektiv datalagring, bearbetning och hämtning men det medför säkerhets- och sekretessproblem. Sensorer är hörnstenen i de IoT-baserade rörledningarna, eftersom de ständigt samlar in data tills de kan analyseras på de centrala molnresurserna. Dessa sensornoder begränsas dock ofta av begränsade resurser. Helst är det önskvärt att göra alla insamlade datafunktioner privata, men på grund av resursbegränsningar kanske det inte alltid är möjligt. Att göra alla funktioner privata kan orsaka överutnyttjande av resurser, vilket i sin tur skulle påverka prestanda för hela systemet. I denna avhandling designar och implementerar vi ett system som kan hitta den optimala uppsättningen datafunktioner för att göra privata, med tanke på begränsningar av enhetsresurserna och systemets önskade prestanda eller noggrannhet. Med hjälp av generaliseringsteknikerna för data-anonymisering skapar vi användardefinierade injicerbara sekretess-kodningsfunktioner för att göra varje funktion i datasetet privat. Oavsett resurstillgänglighet definieras vissa datafunktioner av användaren som viktiga funktioner för att göra privat. Alla andra datafunktioner som kan utgöra ett integritetshot kallas de icke-väsentliga funktionerna. Vi föreslår Dynamic Iterative Greedy Search (DIGS), en girig sökalgoritm som tar resursförbrukningen för varje icke-väsentlig funktion som inmatning och ger den mest optimala uppsättningen icke-väsentliga funktioner som kan vara privata med tanke på tillgängliga resurser. Den mest optimala uppsättningen innehåller de funktioner som förbrukar minst resurser. Vi utvärderar vårt system på en Fitbit-dataset som innehåller 17 datafunktioner, varav 4 är viktiga privata funktioner för en viss klassificeringsapplikation. Våra resultat visar att vi kan erbjuda ytterligare 9 privata funktioner förutom de 4 viktiga funktionerna i Fitbit-datasetet som innehåller 1663 poster. Dessutom kan vi spara 26; 21% minne jämfört med att göra alla funktioner privata. Vi testar också vår metod på en större dataset som genereras med Generative Adversarial Network (GAN). Den valda kantenheten, Raspberry Pi, kan dock inte tillgodose storleken på den stora datasetet på grund av otillräckliga resurser. Våra utvärderingar med 1=8th av GAN-datasetet resulterar i 3 extra privata funktioner med upp till 62; 74% minnesbesparingar jämfört med alla privata datafunktioner. Att upprätthålla integritet kräver inte bara ytterligare resurser utan har också konsekvenser för de designade applikationernas prestanda. Vi upptäcker dock att integritetskodning har en positiv inverkan på noggrannheten i klassificeringsmodellen för vår valda klassificeringsapplikation.
11
Eriksson, Fredrik, and Sebastian Grunditz. "Containerizing WebAssembly : Considering WebAssembly Containers on IoT Devices as Edge Solution." Thesis, Linköpings universitet, Institutionen för datavetenskap, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-177581.
Full textAbstract:
This paper will explore the speed of execution, memory foot-print and the maturity of WebAssembly Runtimes (WasmRT).For this study, the WasmRT will be Wasmer1and Wasmtime.2Initially, benchmarks were run on a Raspberry Pi 3 model Bto simulate a more hardware capable IoT-device. Tests per-formed on a Raspberry Pi shows that there are many instanceswhere a WasmRT outperforms a similar Docker+C solution.WasmRT has a very clear use case for IoT devices, specifi-cally short jobs, the results from our research will show thatWasmRT can be up to almost 70 times as fast as a similarDocker solution. WasmRT has a very strong use case thatother container solutions can not contend with. This paperwill show how effective a lightweight, portable, and fast Was-merRT can be, but also to highlight its pain points and whenother container solutions may make more sense
12
Lajux, Vincent. "Methodology for the design of leading edge devices applied to variable camber." Thesis, Cranfield University, 2007. http://hdl.handle.net/1826/2548.
Full textAbstract:
This thesis will describe a new and innovative way of approaching the design of leading
edge devices; this done using new tools and state of the art software. This innovative
design methodology for leading edge devices has a special focus on the application of
variable camber technology.
A precise description is given of the way to check that methodology acts as a means of
validation.
A case study shows how to apply this methodology and examines a variable camber
application. This case study provides the basis for the understanding of how to apply the
design methodology and give results generated from its different parts. An analysis of the
structure and aerodynamic performances of a new type of leading edge device was
performed in addition to moving the overall design towards an optimum solution in term
of mass, reliability and cost.
The case study is showing that using a variable camber leading edge device could be
beneficial as it proves a better option than more classical concepts. Some of the results
show that the aerodynamics implications of using this type of device at the leading edge
can be beneficial in cruise and also during take off and landing. Also on the structural
side of things, it is possible to see that the deployment trajectory as been optimised to fit
to the required trajectory and the structure is able to resist critical loading.
Finally there is a discussion on the obtained results and on the overall methodology to
make conclusions on the overall meaning of this research and the possible impact of the
new design methodology as well as implication on the design of variable camber leading
edge devices.
13
Šalucha, Darius. "Passivation of the p-n junction edge in high-power semiconductor silicon devices." Doctoral thesis, Lithuanian Academic Libraries Network (LABT), 2009. http://vddb.library.lt/obj/LT-eLABa-0001:E.02~2009~D_20090707_154834-90672.
Full textAbstract:
Thin dielectric passivation layer is one of the basic construction elements in semiconductor device technology. There are few materials, from which the layers may be manufactured. They are oxides mainly, with Si02 as the most popular of them, although, the phosphor- and boron-silicon glasses are used as passivation layers, too. In choosing a passivant of power thyristors and diodes, there are two important considerations in addition to the usual requirement for providing uniform high breakdown voltage via substrate. One consideration is the thermal stability of the passivant to subsequent high-temperature processes. The other consideration is the bias-temperature stability of the passivation layers affecting the operation life expectancy of a device. In the technology of thyristors and diodes on silicon substrates the bias-breakdown voltage is not uniform over substrate due to non-homogeneity of passivated surface of the p-n junction.
In this work, passivation of moat surface by means of electrochemical etching, formation of hydrogen-rich porous silicon layers and glass in-melting steps has been investigated. Passivation quality was controlled by the measurements of surface recombination characteristics after each technological step using a non-invasive technique, which employed microwave probed photoconductivity transients (MW-PCT). It has been shown that electrochemical etching - glass melting steps involved in passivation technological procedures resulted in a decrease of... [to full text]Puslaidininkinių prietaisų pramušimo įtampos valdymas formuojant griovelį periferiniame perimetre yra viena iš labiausiai paplitusių technologinių operacijų, gaminant galios diodus bei tiristorius Si pagrindu. Aukštavolčių didelės galios puslaidininkinių prietaisų, kurie dirba kelių tūkstančių amperų diapazone, o uždarymo įtampa iki kelių tūkstančių voltų, didelė problema elektrinio lauko pasiskirstymas ties kristalo briauna, kur p-n sandūra išeina į paviršių ir kur vyksta griūtinis krūvininkų skaičiaus didėjimas. Darbo stabilumui užtikrinti būtina pasyvuoti paviršių kristalo periferijoje, ant profiliuoto krašto. Šiame darbe išanalizuota galingų puslaidininkinių struktūrų konstrukcija, pagrindinės charakteristikos, parametrų tarpusavio ryšis, taip pat technologinis procesas ir jo ypatumai. Išanalizuotos technologinio gamybos maršruto silpniausios pozicijos. Nustatyta izoliacinių griovelių ėsdinimo charakteristikų priklausomybė nuo ėsdiklio sudėties, nuo ėsdinimo įrenginio struktūros ir nuo ėsdiklio temperatūros kitimo. Sukurta stiklo pasyvacijos difuzinės krosnies monitoringo sistema, kuri skirta aukštų temperaturų ir dujų srautų matavimui proceso metu. Rekombinacijų charakteristikų kitimo pagalba, matuojant be kontakte MW-PCT technika, įvertinama izoliacinių griovelių pasyvacijos kokybė. Technologiniame gamybos maršrute, po izoliacinio griovelio ėsdinimo operacijos, prieš stiklo pasyvaciją sudarinėjamas porėtojo silicio sluoksnis, taip pat siūloma įvesti homogeniškumo... [toliau žr. visą tekstą]
14
Jang, Minsung. "Virtual platforms: System support to enrich the functionality of end client devices." Diss., Georgia Institute of Technology, 2015. http://hdl.handle.net/1853/53914.
Full textAbstract:
Client devices operating at the edges on the Internet, in homes, cars, offices, and elsewhere, are highly heterogeneous in terms of their hardware configurations, form factors, and capabilities, ranging from small sensors to wearable and mobile devices, to the stationary ones like smart TVs and desktop machines. With recent and future advances in wireless networking allowing all such devices to interact with each other and with the cloud, it becomes possible to combine and augment capabilities of individual devices via services running at the edge - in edge clouds - and/or via services running in remote datacenters.
The virtual platform approach to combining and enhancing such devices developed in this research makes possible the creation of innovative end user services, using low-latency communications with nearby devices to create for each end user exactly the platform needed for current tasks, guided by permissions and policies controlled by remote, cloud-resident social network services (SNS). To end users, virtual platforms operate beyond the limitations of individual devices, as natural extensions of those devices that offer improved functionality and performance, with ease-of-use provided by cloud-level global context and knowledge.
15
Yu, Jie. "Novel swing arm mechanism design for trailing edge flaps on commercial airliner." Thesis, Cranfield University, 2008. http://dspace.lib.cranfield.ac.uk/handle/1826/9586.
Full textAbstract:
This thesis will describe the works had been done by the author in the Flying Crane
aircraft group design project and the new design of a novel swing arm mechanism
which can be applied in the trailing edge high lift devices for this aircraft concept.
Flying Crane aircraft is a new generation commercial airliner concept as the result of
group design project conducted by China Aviation Industry Corporation I (AVIC I) and
Cranfield University. At the end of the group design project, parameters such as take-off
and landing distance, trailing edge flap type and deflection in take-off and landing
configuration of the Flying Crane concept have been determined. These parameters are
design input of the novel trailing edge high lift device mechanism for this aircraft
concept.
The idea of this innovative mechanism comes from the research achievement of a
previous MSc student, Thomas Baxter, which applied swing arm mechanism into a
passenger aircraft's leading edge slat. This thesis applied this idea to trailing edge flap
and modeled the mechanism on CATIA software to yield a kinematic simulation for the
purpose of check motion trail and force transfer in this mechanism. Relevant works
such as actuation, mass and stress analysis are also involved.
As the result of this research project, it was found that swing arm mechanism trends to
require relatively small fairings for supports and attachments due to its high stowed
space utilizing efficiency. Initial mass estimation carried out in this thesis also indicates
that the new design takes advantage in terms of weight comparing with traditional
trailing edge flap mechanisms. Thus. swing arm mechanism is supposed to show great
competitive potential for commercial airliner's trailing edge flaps after further analysis
has been done in the detail design phase.
16
Smith, Stephen James. "Development of transition edge sensor distributed read-out imaging devices for applications in X-ray astronomy." Thesis, University of Leicester, 2006. http://hdl.handle.net/2381/30586.
Full textAbstract:
This thesis is concerned with the development of, position-sensitive, Transition Edge Sensors (TES) operating at cryogenic temperatures (~ 0.1 K). The Distributed Read-Out Imaging Device (DROID) concept uses TES read-out at both ends of a linear X-ray absorber, to derive, through heat diffusion, both spectral and spatial information. Potentially, DROIDs offer a simpler technological alternative to the development of large area pixel arrays for future X-ray space observatories (Chapter 1). Chapter 2 presents a comprehensive review of the theory of microcalorimeter detectors. A finite-element model is established in Chapter 3, to numerically simulate the response of the DROID to an X-ray photon, as well as the noise spectral density at the detector outputs, including frequency dependent correlations between the two TESs. In Chapter 4 this model is used to implement pre-existing algorithms, based on the use of the optimal filter, to calculate the position and energy resolution along the length of experimental DROID designs. We show that these algorithms do not lead to optimum performance under all conditions and then derive the true optimal filters, based on least-squares minimisation of the total noise power in the DROID. By numerical simulation we show that significant improvements in both the energy and in particular, the position resolution, are theoretically possible. The design and experimental characterisation of two prototype DROIDs are described in Chapter 6 and 7, using the equipment described in Chapter 5. The first X-ray results from a prototype TES DROID, using single TES read-out, are reported. These results demonstrate proof of concept, confirming spatial sensitivity along the DROID, through the actual spectral and spatial resolutions are limited by the availability of only a single read-out channel.
17
Arnesson, Pontus, and Johan Forslund. "Edge Machine Learning for Wildlife Conservation : Detection of Poachers Using Camera Traps." Thesis, Linköpings universitet, Reglerteknik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-177483.
Full textAbstract:
This thesis presents how deep learning can be utilized for detecting humans ina wildlife setting using image classification. Two different solutions have beenimplemented where both of them use a camera-equipped microprocessor to cap-ture the images. In one of the solutions, the deep learning model is run on themicroprocessor itself, which requires the size of the model to be as small as pos-sible. The other solution sends images from the microprocessor to a more pow-erful computer where a larger object detection model is run. Both solutions areevaluated using standard image classification metrics and compared against eachother. To adapt the models to the wildlife environment,transfer learningis usedwith training data from a similar setting that has been manually collected andannotated. The thesis describes a complete system’s implementation and results,including data transfer, parallel computing, and hardware setup. One of the contributions of this thesis is an algorithm that improves the classifi-cation performance on images where a human is far away from the camera. Thealgorithm detects motion in the images and extracts only the area where thereis movement. This is specifically important on the microprocessor, where theclassification model is too simple to handle those cases. By only applying theclassification model to this area, the task is more simple, resulting in better per-formance. In conclusion, when integrating this algorithm, a model running onthe microprocessor gives sufficient results to run as a camera trap for humans.However, test results show that this implementation is still quite underperform-ing compared to a model that is run on a more powerful computer.
18
Reba, Ramona Andris. "A theoretical study of leading-edge devices for active and passive control of wake-airfoil interaction noise." Diss., The University of Arizona, 2000. http://hdl.handle.net/10150/284280.
Full textAbstract:
A theoretical analysis is developed for control of wake-airfoil interaction noise by the use of actuators on the airfoil surfaces near the leading edge. The objective is to eliminate the sound radiation by canceling the wake-airfoil noise near its source (the leading edge). Actuators are mounted on the upper and lower surfaces of the airfoil and driven out of phase to match the antisymmetric nature of the wake-airfoil noise field. The analysis is based on linearization about a two-dimensional compressible subsonic mean flow past a semi-infinite flat-plate airfoil. Asymptotic solutions are developed for the unsteady boundary-layer flow over surface-mounted actuators which are long compared to the triple-deck scale. The analysis shows that boundary layer effects play a significant role in determining the acoustic field, and that traditional approaches used to account for acoustically thin boundary layers are not always adequate. In addition to the acoustic monopole field associated with the actuator volume flux, acoustic monopole and dipole fields are produced at leading order by deflection of the boundary layer into the oncoming free stream. For a specified actuator size and location (relative to the leading edge), the analysis determines the actuator amplitude and phase which minimizes the total sound radiation. The cancellation is most effective when the directivity pattern for the control field closely matches that for the wake-airfoil noise field. At low Mach numbers, this is achieved by mounting the actuators sufficiently close to the leading edge. At high subsonic Mach numbers, the sound cancellation that can be achieved with a single actuator on each surface is limited by mismatch in the directivity patterns. However, a match in directivity patterns can be recovered in part by using two actuators on each surface. The use of passive treatment in the leading-edge region to locally control wake-airfoil interaction noise is also discussed.
19
Neges, Matthias, and Jan Luca Siewert. "Interaktive Initialisierung eines Echtzeit 3D-Trackings für Augmented Reality auf Smart Devices mit Tiefensensoren." Saechsische Landesbibliothek- Staats- und Universitaetsbibliothek Dresden, 2016. http://nbn-resolving.de/urn:nbn:de:bsz:14-qucosa-215154.
Full textAbstract:
Zusammenfassung
Heutige Ansätze des 3D-Trackings für die Registrierung in der realen Welt zum Einsatz von Augmented Reality lassen sich in modellbasierte und umgebungsbasierte Verfahren unterteilen. Umgebungsbasierte Verfahren nutzen den SLAM-Algorithmus zur Erzeugung dreidimensionaler Punktwolken der Umgebung in Echtzeit. Modellbasierte Verfahren finden Ihren Ursprung im Canny edge detector und nutzen aus den CAD-Modellen abgeleitete Kantenmodelle. Wird das modellbasierte Verfahren über Kantendetektion und das umgebungsbasierte Verfahren über 3DPunktewolken kombiniert, ergibt sich ein robustes, hybrides 3D-Tracking. Die entsprechenden Algorithmen der verschiedenen Verfahren sind in heute verfügbaren AR-Frameworks bereits implementiert. Der vorliegende Betrag zeigt zwar, welche Effizienz das hybride 3D-Tracking aufweist, jedoch auch die Problematik der erforderlichen geometrischen Ähnlichkeit von idealem CAD-Modell, bzw. Kantenmodell, und realem Objekt. Bei unterschiedlichen Montagestufen an verschiedenen Montagestationen und mit wechselnden Anwendern ist beispielsweise eine erneute Initialisierung erforderlich. Somit bedingt das hybride 3D-Tracking zahlreiche Kantenmodell, die zuvor aus der jeweiligen Montagestufe abgeleitet werden müssen. Hinzu kommen geometrische Abweichungen durch die Fertigung, die je nach Größe der branchenspezifischen Toleranzen keine hinreichend hohe Übereinstimmung mit den abgeleiteten Kantenmodellen aus den idealen CAD-Modellen aufweisen. Die Autoren schlagen daher den Einsatz parametrisch aufgebauter Mastermodelle vor, welche durch eine interaktive Initialisierung geometrisch Instanziiert werden. Zum Einsatz kommt hier ein mobiler Tiefensensor für Smart Devices, welcher mit Hilfe des Anwenders eine Relation der realen geometrischen Merkmale mit den Idealen des CAD-Modells ermöglicht. Des Weiteren wird in dem dargestellten Konzept die Nutzung von speziellen Suchalgorithmen basierend auf geometrischen Ähnlichkeiten vorgeschlagen, sodass eine Registrierung und Instanziierung auch ohne hinterlegtes Mastermodell ermöglicht wird. Der Beitrag fokussiert sich bei der Validierung auf die interaktive Initialisierung anhand eines konkreten anwendungsnahen Beispiels, da die Initialisierung die Grundlage für die weitere Entwicklung des Gesamtkonzeptes darstellt.
20
Erlandsson, Niklas. "Utilizing machine learning in wildlife camera traps for automatic classification of animal species : An application of machine learning on edge devices." Thesis, Linnéuniversitetet, Institutionen för datavetenskap och medieteknik (DM), 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:lnu:diva-104952.
Full textAbstract:
A rapid global decline in biodiversity has been observed in the past few decades, especially in large vertebrates and the habitats supporting these animal populations. This widely accepted fact has made it very important to understand how animals respond to modern ecological threats and to understand the ecosystems functions. The motion activated camera (also known as a camera trap) is a common tool for research in this field, being well-suited for non-invasive observation of wildlife. The images captured by camera traps in biological studies need to be classified to extract information, a traditionally manual process that is time intensive. Recent studies have shown that the use of machine learning (ML) can automate this process while maintaining high accuracy. Until recently the use of machine learning has required significant computing power, relying on data being processed after collection or transmitted to the cloud. This need for connectivity introduces potentially unsustainable overheads that can be addressed by placing computational resources on the camera trap and processing data locally, known as edge computing. Including more computational power in edge and IoT devices makes it possible to keep the computation and data storage on the edge, commonly referred to as edge computing. Applying edge computing to the camera traps enables the use of ML in environments with slow or non-existent network accesss since their functionality does not rely on the need for connectivity. This project shows the feasibility of running machine learning algorithms for the purpose of species identification on low-cost hardware with similar power to what is commonly found in edge and IoT devices, achieving real-time performance and maintaining high energy efficiency sufficient for more than 12 hours of runtime on battery power. Accuracy results were mixed, indicating the need for more tailor-made network models for performing this task and the importance of high quality images for classification.
21
Kheffache, Mansour. "Energy-Efficient Detection of Atrial Fibrillation in the Context of Resource-Restrained Devices." Thesis, Luleå tekniska universitet, Datavetenskap, 2019. http://urn.kb.se/resolve?urn=urn:nbn:se:ltu:diva-76394.
Full textAbstract:
eHealth is a recently emerging practice at the intersection between the ICT and healthcare fields where computing and communication technology is used to improve the traditional healthcare processes or create new opportunities to provide better health services, and eHealth can be considered under the umbrella of the Internet of Things. A common practice in eHealth is the use of machine learning for a computer-aided diagnosis, where an algorithm would be fed some biomedical signal to provide a diagnosis, in the same way a trained radiologist would do. This work considers the task of Atrial Fibrillation detection and proposes a novel range of algorithms to achieve energy-efficiency. Based on our working hypothesis, that computationally simple operations and low-precision data types are key for energy-efficiency, we evaluate various algorithms in the context of resource-restrained health-monitoring wearable devices. Finally, we assess the sustainability dimension of the proposed solution.
22
Richards, Mark Andrew. "An intuitive motion-based input model for mobile devices." Queensland University of Technology, 2006. http://eprints.qut.edu.au/16556/.
Full textAbstract:
Traditional methods of input on mobile devices are cumbersome and difficult to use. Devices have become smaller, while their operating systems have become more complex, to the extent that they are approaching the level of functionality found on desktop computer operating systems. The buttons and toggle-sticks currently employed by mobile devices are a relatively poor replacement for the keyboard and mouse style user interfaces used on their desktop computer counterparts. For example, when looking at a screen image on a device, we should be able to move the device to the left to indicate we wish the image to be panned in the same direction.
This research investigates a new input model based on the natural hand motions and reactions of users. The model developed by this work uses the generic embedded video cameras available on almost all current-generation mobile devices to determine how the device is being moved and maps this movement to an appropriate action.
Surveys using mobile devices were undertaken to determine both the appropriateness and efficacy of such a model as well as to collect the foundational data with which to build the model. Direct mappings between motions and inputs were achieved by analysing users' motions and reactions in response to different tasks.
Upon the framework being completed, a proof of concept was created upon the Windows Mobile Platform. This proof of concept leverages both DirectShow and Direct3D to track objects in the video stream, maps these objects to a three-dimensional plane, and determines device movements from this data.
This input model holds the promise of being a simpler and more intuitive method for users to interact with their mobile devices, and has the added advantage that no hardware additions or modifications are required the existing mobile devices.
23
Barutcu, Burcu. "The Design And Production Of Interference Edge Filters With Plasma Ion Assisted Deposition Technique For A Space Camera." Master's thesis, METU, 2012. http://etd.lib.metu.edu.tr/upload/12614574/index.pdf.
Full textAbstract:
Interference filters are multilayer thin film devices. They use interference effects between
the incident and reflected radiation waves at each layer interface to select wavelengths. The
production of interference filters depend on the precise deposition of thin material layers on
substrates which have suitable optical properties. In this thesis, the main target is to design
and produce two optical filters (short-pass filter and long-pass filter) for the CCDs that will
be used in the electronics of a space camera. By means of these filters, it is possible to take
image in different bands (RGB and NIR) by identical two CCDs. The filters will be fabricated
by plasma ion-assisted deposition technique.
24
Martins, Pires Rui Miguel. "Design methodology for wing trailing edge device mechanisms." Thesis, Cranfield University, 2007. http://hdl.handle.net/1826/3393.
Full textAbstract:
Over the last few decades the design of high lift devices has become a very
important part of the total aircraft design process. Reviews of the design process
are performed on a regular basis, with the intent to improve and optimize the
design process.
This thesis describes a new and innovative methodology for the design and
evaluation of mechanisms for Trailing Edge High-Lift devices. The initial research
reviewed existing High-Lift device design methodologies and current flap systems
used on existing commercial transport aircraft. This revealed the need for a design
methodology that could improve the design process of High-Lift devices, moving
away from the conventional "trial and error" design approach, and cover a wider
range of design attributes. This new methodology includes the use of the
innovative design tool called SYNAMEC. This is a state-of-the-art engineering
design tool for the synthesis and optimizations of aeronautical mechanisms. The
new multidisciplinary design methodology also looks into issues not usually
associated with the initial stages of the design process, such as Maintainability,
Reliability, Weight and Cost.
The availability of the SYNAMEC design tool and its ability to perform Synthesis
and Optimization of mechanisms led to it being used as an important module in
the development of the new design methodology. The SYNAMEC tool allows
designers to assess more mechanisms in a given time than the traditional design
methodologies.
A validation of the new methodology was performed and showed that creditable
results were achieved.
A case study was performed on the ATRA
-
Advance Transport Regional Aircraft,
a Cranfield University design project, to apply the design methodology and select
from within a group of viable solutions the most suitable type of mechanism for the
Variable Camber Wing concept initially defined for the aircraft. The results show
that the most appropriate mechanism type for the ATRA Variable Camber Wing is
the Link /Track Mechanism. It also demonstrated how a wide range of design
attributes can now be considered at a much earlier stage of the design.
25
Henry, Michael Maier. "Two-Dimensional Shock Sensitivity Analysis for Transonic Airfoils with Leading-edge and Trailing-edge Device Deflections." Thesis, Virginia Tech, 2001. http://hdl.handle.net/10919/30924.
Full textAbstract:
This investigation, in consideration of the sudden separation increase involved in wing drop, was to determine if the incorporated 2-D airfoil exhibits abnormal shock sensitivity. A comparative airfoil study was used to determine if this particular transonic airfoil is prone to abrupt shock movement, resulting in increased regions of separation.Master of Science
26
Woolridge, Daniel 'Shane'. "An Open Architecture Approach to Networked Telemetry System." International Foundation for Telemetering, 2010. http://hdl.handle.net/10150/605942.
Full textAbstract:
ITC/USA 2010 Conference Proceedings / The Forty-Sixth Annual International Telemetering Conference and Technical Exhibition / October 25-28, 2010 / Town and Country Resort & Convention Center, San Diego, CaliforniaWhen designing data transport systems, Telemetry and Communications engineers always face the risk that their chosen hardware will not be available or supported soon after the hardware has been installed. The best way to reduce this risk and ensure the longevity of the system is to select an open architecture standard that is supported by multiple manufacturers. This open architecture should also have the ability to be easily upgraded and provide for all of the features and flexibility that are required to be a reliable carrier-grade edge-device. The PCI Industrial Computer Manufacturers Group (PICMG) developed the MicroTCA open standard to address the specific needs of these Communications and Network System Engineers. This paper describes the MicroTCA architecture and how it can be applied as the ideal edge-device solution for Networked Telemetry Systems applications.
27
Broumas, Ioannis. "Design of Cellular and GNSS Antenna for IoT Edge Device." Thesis, Högskolan i Halmstad, Akademin för informationsteknologi, 2019. http://urn.kb.se/resolve?urn=urn:nbn:se:hh:diva-39239.
Full textAbstract:
Antennas are one of the most sensitive elements in any wireless communication equipment. Designing small-profile, multiband and wideband internal antennas with a simple structure has become a necessary challenge. In this thesis, two planar antennas are designed, simulated and implemented on an effort to cover the LTE-M1 and NB-IoT radio frequencies. The cellular antenna is designed to receive and transmit data over the eight-band LTE700/GSM/UMTS, and the GNSS antenna is designed to receive signal from the global positioning system and global navigation systems, GPS (USA) and GLONASS. The antennas are suitable for direct print on the system circuit board of a device. Related theory and research work are discussed and referenced, providing a strong configuration for future use. Recommendations and suggestions on future work are also discussed. The proposed antenna system is more than promising and with further adjustments and refinement can lead to a fully working solution.
28
Sarazin, Yanick. "Etude de la turbulence de bord dans les plasmas de tokamaks." Université Joseph Fourier (Grenoble), 1997. http://www.theses.fr/1997GRE10265.
Full textAbstract:
Cette these remet en cause la description classique du transport turbulent en terme de diffusion, et propose un nouveau cadre d'analyse theorique. Dans les plasmas, l'approche standard de la turbulence suppose que la taille et la frequence caracteristiques des fluctuations sont tres inferieures a celles de l'equilibre. Cette hypothese permet une etude avec un forcage par un gradient d'equilibre constant. Notre etude privilegie le forcage par un flux, levant ainsi la contrainte de separabilite d'echelle. Nous developpons un modele cellulaire 1d auto-organise critique ou soc. Le transport local a seuil en gradient est couple a l'etat des sites voisins. Le systeme est force par une source coherente en volume, l'ionisation des neutres. En accord avec l'experience, le transport moyen peut etre decrit en termes de diffusion et de convection, bien que le transport local soit d'une toute autre nature. L'instabilite d'interchange est etudiee sur les lignes de champ ouvertes connectees a la paroi. Le modele 2d de cette instabilite a seuil est utilise dans les deux cas de forcage, par un gradient et par un flux. Dans le premier cas, le transport est homogene et diffusif. La taille des cellules turbulentes est fixee par l'allure du taux de croissance et les conditions aux limites imposees par l'interface plasma-paroi ; les processus de cascade d'energie jouent un role negligeable. Lorsque le systeme est force par un flux constant, le transport est inhomogene et caracterise par un flux sortant convectif intermittent, des avalanches. Une statistique sur des particules test ne rend que partiellement compte du transport fluide observe. Le modele reproduit les observations experimentales :les fluctuations relatives de densite sont maximales au bord, et le profil exhibe une decroissance exponentielle, a nouveau compatible avec un transport diffusif, bien que le transport reel ne le soit pas.
29
Rakhshani, Bassam. "Multidisciplinary investigations of aerodynamic noise from high - lift leading - edge device." Thesis, University of Manchester, 2009. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.500776.
Full text
30
Eliasson, Anton. "Anomaly Detection in Industrial Networks using a Resource-Constrained Edge Device." Thesis, Luleå tekniska universitet, Institutionen för system- och rymdteknik, 2019. http://urn.kb.se/resolve?urn=urn:nbn:se:ltu:diva-74530.
Full textAbstract:
The detection of false data-injection attacks in industrial networks is a growing challenge in the industry because it requires knowledge of application and protocol specific behaviors. Profinet is a common communication standard currently used in the industry, which has the potential to encounter this type of attack. This motivates an examination on whether a solution based on machine learning with a focus on anomaly detection can be implemented and used to detect abnormal data in Profinet packets. Previous work has investigated this topic; however, a solution is not available in the market yet. Any solution that aims to be adopted by the industry requires the detection of abnormal data at the application level and to run the analytics on a resource-constrained device. This thesis presents an implementation, which aims to detect abnormal data in Profinet packets represented as online data streams generated in real-time. The implemented unsupervised learning approach is validated on data from a simulated industrial use-case scenario. The results indicate that the method manages to detect all abnormal behaviors in an industrial network.
31
Le, Minh. "Universal Mobile Service Execution Framework for Device-To-Device Collaborations." DigitalCommons@USU, 2018. https://digitalcommons.usu.edu/etd/7032.
Full textAbstract:
There are high demands of effective and high-performance of collaborations between mobile devices in the places where traditional Internet connections are unavailable, unreliable, or significantly overburdened, such as on a battlefield, disaster zones, isolated rural areas, or crowded public venues. To enable collaboration among the devices in opportunistic networks, code offloading and Remote Method Invocation are the two major mechanisms to ensure code portions of applications are successfully transmitted to and executed on the remote platforms. Although these domains are highly enjoyed in research for a decade, the limitations of multi-device connectivity, system error handling or cross platform compatibility prohibit these technologies from being broadly applied in the mobile industry.
To address the above problems, we designed and developed UMSEF - an Universal Mobile Service Execution Framework, which is an innovative and radical approach for mobile computing in opportunistic networks. Our solution is built as a component-based mobile middleware architecture that is flexible and adaptive with multiple network topologies, tolerant for network errors and compatible for multiple platforms. We provided an effective algorithm to estimate the resource availability of a device for higher performance and energy consumption and a novel platform for mobile remote method invocation based on declarative annotations over multi-group device networks. The experiments in reality exposes our approach not only achieve the better performance and energy consumption, but can be extended to large-scaled ubiquitous or IoT systems.
32
Millar, Alasdair J. "Step edge Josephson junctions and high temperature superconducting quantum interference device (SQUID) gradiometers." Thesis, University of Strathclyde, 2002. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.249130.
Full textAbstract:
This thesis is concerned with the development of Superconducting Quantum Interference Device (SQUID) gradiometers based on the high temperature superconductor YBa2Cu3O7-o (YBCO). A step-edge Josephson junction fabrication process was developed to produce sufficiently steep (> 60°) step-edges such that junctions exhibited RSJ-like current-voltage characteristics. The mean I(RN product of a sample of twenty step-edge junctions was 130jtV. Step-edge dc SQUIDs with inductances between 67pH and 114pH were fabricated. Generally the SQUIDs had an intrinsic white flux noise in the 10-30μ(Do/ Hz range, with the best device, a 70pH SQUID, exhibiting a white flux noise of 5μ4bo/ Hz. Different first-order SQUID gradiometer designs were fabricated from single layers of YBCO. Two single-layer gradiometer (SLG) designs were fabricated on 10 x 10mm2 substrates. The best balance and lowest gradient sensitivity measured for these devices were 1/300 and 308fT/cm Hz (at 1kHz) respectively. The larger baseline and larger flux capture area of the pick-up loops in a large area SLG design, fabricated on 30 x 10mm2 substrates, resulted in significant improvements in the balance and gradient sensitivity with 1/1000 and 50fT/cm Hz (at 1kHz) measured respectively. To reduce the uniform field effective area of SLGs and therefore reduce the direct pick-up of environmental field noise when operated unshielded, a novel gradiometric SQUID (G-SQUID) device was developed. Fabricated from a single layer of YBCO, the G-SQUIDs, with inductances of 67pH, had small uniform field effective areas of approximately 2μm2 - more than two orders of magnitude smaller than the uniform field effective areas of conventional narrow-linewidth SQUIDs of similar inductance. Two designs of G-SQUID SLGs were fabricated on 10 x 10mm2 substrates. Due to their small effective areas, when cooled unshielded these devices showed no increase in their white flux noise. The best balance achieved for a G-SQUID SLG was approximately 1/5000 - an order of magnitude better than the balance of similar SLGs incorporating conventional narrow-linewidth SQUIDs.
33
Applebaum, Stephen I. "The anti-wire device : cutting edge transport in the integer quantum hall effect /." Diss., Connect to a 24 p. preview or request complete full text in PDF format. Access restricted to UC IP addresses, 1999. http://wwwlib.umi.com/cr/ucsd/fullcit?p9935467.
Full text
34
Mastricola, Nicholas Palma. "Nonlinear Stiffness and Edge Friction Characterization of Coned Disk Springs." The Ohio State University, 2016. http://rave.ohiolink.edu/etdc/view?acc_num=osu1480346443676492.
Full text
35
Vadivelu, Somasundaram. "Sensor data computation in a heavy vehicle environment : An Edge computation approach." Thesis, KTH, Skolan för elektroteknik och datavetenskap (EECS), 2018. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-235486.
Full textAbstract:
In a heavy vehicle, internet connection is not reliable, primarily because the truck often travels to a remote location where network might not be available. The data generated from the sensors in a vehicle might not be sent to the internet when the connection is poor and hence it would be appropriate to store and do some basic computation on those data in the heavy vehicle itself and send it to the cloud when there is a good network connection. The process of doing computation near the place where data is generated is called Edge computing. Scania has its own Edge computation solution, which it uses for doing computations like preprocessing of sensor data, storing data etc. Scania’s solution is compared with a commercial edge computing platform called as AWS (Amazon Web Service’s) Greengrass. The comparison was in terms of Data efficiency, CPU load, and memory footprint. In the conclusion it is shown that Greengrass solution works better than the current Scania solution in terms of CPU load and memory footprint, while in data efficiency even though Scania solution is more efficient compared to Greengrass solution, it was shown that as the truck advances in terms of increasing data size the Greengrass solution might prove competitive to the Scania solution.One more topic that is explored in this thesis is Digital twin. Digital twin is the virtual form of any physical entity, it can be formed by obtaining real-time sensor values that are attached to the physical device. With the help of sensor values, a system with an approximate state of the device can be framed and which can then act as the digital twin. Digital twin can be considered as an important use case of edge computing. The digital twin is realized with the help of AWS Device shadow.I ett tungt fordonsscenario är internetanslutningen inte tillförlitlig, främst eftersom lastbilen ofta reser på avlägsna platser nätverket kanske inte är tillgängligt. Data som genereras av sensorer kan inte skickas till internet när anslutningen är dålig och det är därför bra att ackumulera och göra en viss grundläggande beräkning av data i det tunga fordonet och skicka det till molnet när det finns en bra nätverksanslutning. Processen att göra beräkning nära den plats där data genereras kallas Edge computing. Scania har sin egen Edge Computing-lösning, som den använder för att göra beräkningar som förbehandling av sensordata, lagring av data etc. Jämförelsen skulle vara vad gäller data efficiency, CPU load och memory consumption. I slutsatsen visar det sig att Greengrass-lösningen fungerar bättre än den nuvarande Scania-lösningen när det gäller CPU-belastning och minnesfotavtryck, medan det i data-effektivitet trots att Scania-lösningen är effektivare jämfört med Greengrass-lösningen visades att när lastbilen går vidare i Villkor för att öka datastorleken kan Greengrass-lösningen vara konkurrenskraftig för Scania-lösningen. För att realisera Edge computing används en mjukvara som heter Amazon Web Service (AWS) Greengrass.Ett annat ämne som utforskas i denna avhandling är digital twin. Digital twin är den virtuella formen av någon fysisk enhet, den kan bildas genom att erhålla realtidssensorvärden som är anslutna till den fysiska enheten. Med hjälp av sensorns värden kan ett system med ungefärligt tillstånd av enheten inramas och som sedan kan fungera som digital twin. Digital twin kan betraktas som ett viktigt användningsfall vid kantkalkylering. Den digital twin realiseras med hjälp av AWS Device Shadow.
36
Wylie, Diana. "The urban edge : a spatial planning tool or device for land development management : a Western Cape perspective." Master's thesis, University of Cape Town, 2016. http://hdl.handle.net/11427/20775.
Full textAbstract:
The regulation of South African land use planning law is challenging; it is a field that is complex and not fully understood.1 One reason for the complexity in the past was the number of laws that were in force.2 Other factors include the number of authorities that are involved, the irregular implementation of the land use planning tools and the gap that is prevalent generally between planning theory and practise.34 The assortment of laws is implemented by authorities using land use planning tools.5 An array of land use planning tools, such as zoning and urban edge boundaries, are used in the planning process to distinguish the various aspects of development from one another. The combined English and Roman Dutch sources of our planning law passed down traditional land use planning devices such as: zoning schemes, subdivision and title deed restrictions.6 A range of unique South African tools, such as; guide plans, regional plans and urban structure plans were used for regional planning during apartheid times.7 Several new planning tools have been created since 1994 to give effect to changing policy, such as; land development objectives, environmental impact assessments (EIAs), integrated development plans (IDPs), spatial development frameworks (SDFs), the designation of different types of protected areas; urban edge lines and marine set back delineations.8 This thesis will analyse the status of the urban edge as a land use planning tool.
37
Golfinopoulos, Theodore. "The Shoelace Antenna : a device to induce short-wavelength fluctuations in the edge plasma of the Alcator C-Mod Tokamak." Thesis, Massachusetts Institute of Technology, 2014. http://hdl.handle.net/1721.1/91031.
Full textAbstract:
Thesis: Sc. D., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2014.This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections.
140
Cataloged from student-submitted PDF version of thesis.
Includes bibliographical references (pages 305-318).
The "Shoelace" antenna is a unique device built to induce short-wavelength fluctuations in the edge plasma of the Alcator C-Mod tokamak, at a wave number and in the frequency range associated with the Quasi-Coherent Mode (QCM). The QCM is a continuous, drift-mode-like fluctuation, restricted to the low-field side of the tokamak in a 3 mm region around the last closed flux surface, and spanning both open and closed field lines. The study presented here is motivated by the fact that the QCM plays a crucial role in regulating particle transport across the plasma boundary in the Enhanced D[alpha](EDA) H-mode. It is this transport channel which sustains the EDA H-mode, flushing impurities from the plasma without the appearance of bursting Edge Localized Modes (ELMs). Because of the damage they cause to first-wall components, large-amplitude ELMs do not extrapolate to a full-size, steady state fusion reactor, and so it is of critical importance for the worldwide fusion research endeavor to identify, understand, and exploit ELM-free mechanisms of impurity flushing. It is in this context that the antenna's mission is defined. The Shoelace antenna is wound with field-aligned rungs spaced to produce a perpendicular wave number, k = 1.5 ± 0.1 cm-1, that precisely matches the QCM spatial structure, while the power system, with custom matching network, provides up to 2 kW of radio-frequency source power at any frequency in the band, 45 < f < 300 kHz. Initial experiments show that when the antenna is energized into L-mode plasmas, it produces a steady response in poloidal magnetic field, only. However, after transition to H-mode, the antenna drives both field and electron density fluctuations that are aligned with, and guided by, the background equilibrium field, propagate in the electron diamagnetic drift direction in the laboratory frame, have amplitude comparable to that of the intrinsic QCM, and display a weakly-damped resonance ([gamma]/[omega] ~ 5-10%). In EDA H-mode, the resonance is centered on the QCM frequency, but in ELM-free H-mode, it persists in the same frequency range, even in the absence of a QCM. This result is significant, offering the possibility that externally-driven modes might be used to enhance particle transport. However, additional measurements are required before a definitive statement can be made regarding transport resulting from the antenna-driven mode, as well as the driven mode's relationship with the QCM. This work has been scheduled for the 2014 Alcator C-Mod experimental campaign as part of a broader exploration of the plasma response to the Shoelace antenna.
by Theodore Golfinopoulos.
Sc. D.
38
Olsson, Andreas. "Optimering av kupmätare : Automatisk igenkänning av egenskaper i brädors ändträ med bildbehandling." Thesis, Högskolan i Gävle, Avdelningen för elektronik, matematik och naturvetenskap, 2016. http://urn.kb.se/resolve?urn=urn:nbn:se:hig:diva-21804.
Full textAbstract:
Within the timber industry the processing of sawn wood boards must be done in the right way to ensure that the product fulfills the requirements. Correct processing is crucial for wooden structures consisting of these will live up to their expectations. How the mounting of the board in wooden structures is performed to obtain satisfactory results is depending on how it is physically cupped. Due to this a curvature measuring device is used in the timber industry to detect how a sawn board is physically cupped. After detection, a proper processing can be performed to give a satisfactory product. The Swedish company Nolyx AB currently uses a curvature measuring device consisting of a smart camera with the task of taking a digital picture of the board end grain to determine its cupping. The smart camera currently has deficiencies that this work will investigate. The deficiencies are that the smart camera’s processing of images with certain properties do not give satisfactory results. The algorithms lack the robustness needed to cope with the variations of the item that might arise in the process. The desire of this study is that the smart cameras correctness in terms of variations in the object will increase, leading to financial gains for the company Nolyx AB and increased utilization of raw material for their customers. The result of this work is an algorithm that incrementally extracts and identifies the growth rings in the end grain of the board. The correctness of the image processing in this study is 82%, which is 22% higher compared to the smart camera.Inom träindustrin måste bearbetning av sågade träbrädor ske på rätt sätt för att produkten skall uppfylla kraven. En korrekt bearbetning är avgörande för att träkonstruktioner bestående av dessa ska leva upp till sina förväntningar. Hur monteringen av brädan vid byggnation av träkonstruktioner utförs för att erhålla tillfredställande resultat beror på hur den fysiskt är kupad. På grund av detta används kupmätare inom träindustrin för att detektera brädans fysiska kupning. Efter detektering kan en korrekt bearbetning utföras vilket ger en tillfredsställande produkt. Företaget Nolyx AB använder idag en kupmätare som består av en smartkamera vars uppgift är att ta en digital bild av brädans ändträ för att avgöra dess kupning. Kupmätaren har idag brister som detta arbete skall angripa. Nämligen att smartkamerans behandling av bilder med vissa egenskaper inte ger tillfredsställande resultat. Algoritmerna saknar den robusthet som krävs för att klara de variationer på objektet som kan uppkomma i processen. Önskan med denna studie är att smartkamerans felfrihet vad gäller variationer i objektet ska öka, vilket leder till ekonomiska vinster för företaget Nolyx AB och ökat utnyttjande av råvaran för sina kunder. Resultatet av det här arbetet är en algoritm som stegvis extraherar och identifierar årsringarnas struktur. Felfriheten för bildbehandlingen i denna studie är 82 % vilket är 22 % högre jämfört med smartkameran.
39
Maeda, Takuya. "Study on Avalanche Breakdown in GaN." Doctoral thesis, Kyoto University, 2020. http://hdl.handle.net/2433/253283.
Full textAbstract:
京都大学0048
新制・課程博士
博士(工学)
甲第22447号
工博第4708号
新制||工||1735(附属図書館)
京都大学大学院工学研究科電子工学専攻
(主査)教授 木本 恒暢, 教授 山田 啓文, 准教授 船戸 充
学位規則第4条第1項該当
Doctor of Philosophy (Engineering)
Kyoto University
DFAM
40
Musil, Martin. "Přenosy rastrových dat v FPGA." Master's thesis, Vysoké učení technické v Brně. Fakulta informačních technologií, 2012. http://www.nusl.cz/ntk/nusl-236507.
Full textAbstract:
This work deals with the design and implementation of high-speed communication interfaces into FPGA chip and their utilizing for image transmission and processing. In the implementation part has been created PCI Express endpoint device, which provides data transfers between the FPGA chip and computer RAM memory. As a source of image data for further processing was connected the Unicam M621 camera throught the Ethernet interface to FPGA chip. The project was implemented on the Xilinx SP605 development board. Using both of the the interfaces were demonstrated on the application of edge detection using Sobel operator. The PCI Express endpoint device driver for the Linux operating system and a simple application interface in C language was also created within this project.
41
Oksenhendler, Vincent. "Développement d'un système opto-électronique de vision 3-D pour l'établissement en temps réel de cartes de distances." Rouen, 1997. http://www.theses.fr/1997ROUES047.
Full textAbstract:
Dans ce mémoire, est présenté un capteur de vision 3-D dont l'objectif est de permettre l'assistance à la conduite de véhicule. Cela entraîne un certain nombre de contraintes telles que la fiabilité, la rapidité et l'auto-adaptabilité. Ainsi, le capteur que nous avons développé est fondé sur la stéréovision passive et la mise en correspondance de contours verticaux. L'utilisation d'une configuration spécifique des caméras nous permet de limiter l'espace de recherche des primitives stéréocorrespondantes aux lignes images de même numéro. La réalisation d'un tel capteur nous a amené à développer tout particulièrement 3 parties de ce dernier. La première partie concerne la réalisation d'un détecteur de contour monodimensionnel rapide, auto-adaptatif, de résolution sub-pixel et qui est fonde sur le seuillage des déclivités étendues. Les résultats obtenus sur des scènes réelles et synthétiques montrent que celui-ci peut être utilise afin de réaliser un capteur de précision. La seconde partie est constituée par un processus de correction et d'étiquetage des contours 3-D obtenus à la suite de la mise en correspondance. Ce traitement exploite la continuité de la disparité le long des contours permettant ainsi de rectifier la majorité des points 3-D issus de mauvais appariements. La dernière partie porte sur l'implémentation parallèle du processus de stéréovision sur 3 DSP tms320c31 hébergés dans un système de traitement d'imaging technology 150/40.
42
Lima, Diogo Emanuel da Costa. "Framework generativa para edge devices." Master's thesis, 2015. http://hdl.handle.net/1822/54739.
Full textAbstract:
Dissertação de mestrado integrado em Engenharia Eletrónica Industrial e ComputadoresNos dias de hoje, as empresas de software tentando captar o maior número de clientes, lançam para o mercado uma vasta gama de produtos. Metas temporais têm que ser cumpridas pela empresa aquando do desenvolvimento de um novo produto, de forma a evitar que empresas concorrentes cheguem primeiro ao mercado com um produto de especificações semelhantes. Todavia, o desenvolvimento de um produto em tempo reduzido pode levar a imperfeições no mesmo. E para atenuar este problema, as empresas são obrigadas a adotar novos processos que reduzam o time-to-market sem afetar a qualidade do produto. Evitando assim, que o produto entre no mercado com problemas que poderiam resultar em prejuízos, como: muitos recursos humanos alocados à manutenção pós venda do produto, perda de clientes ou até indemnizações aos clientes lesados. Normalmente, os produtos desenvolvidos por uma empresa pertencem à mesma família de produtos, onde a base dos diversos produtos é comum, só variando algumas partes dos mesmos, para adaptar os produtos aos diferentes níveis de exigência dos clientes. Desta forma, partes de um produto que já se encontra desenvolvido, testado e a ser comercializado podem ser reutilizadas para o desenvolvimento de novos produtos, acelerando assim o processo de desenvolvimento. Um bom exemplo de família de produtos onde se verifica o padrão apresentado anteriormente são as redes de sensores sem fios (WSNs). Estes tipos de redes são hoje em dia amplamente utilizadas em diversas aplicações, tais como: domésticas, industriais, engenharia civil, militares, monitorização ambiental e saúde. Sendo a base da rede semelhante às várias aplicações, e só variam algumas configurações específicas, tipos de sensores e tipos de atuadores atribuídos aos nós da rede (edge devices). Para acelerar o processo do desenvolvimento de aplicações que utilizam WSNs, é proposto nesta dissertação a elaboração de uma framework generativa, que permita obter um bom compromisso entre a gestão de variabilidade e o desempenho dos edge devices. Na fase de design da framework serão utilizadas técnicas de Software Product Lines (SPL) para identificar o grau de variabilidade das funcionalidades dos edge devices. E na fase de implementação será utilizado template metaprogramming em C++ para gerir essa variabilidade.
Nowadays, software companies trying to reach the largest possible number of customers, launch a wide range of products. Deadlines must be complied when developing a new product, in order to prevent that competitors arrive earlier to market with a product of similar specifications. However, the development of a product in a short timeframe may lead to imperfections. In order to mitigate this problem, software houses are forced to adopt new processes that reduce time-to-market without affecting product quality. This is done in order, to prevent the product entering the market with bugs that could result in losses, such as: many human resources allocated to the after-sales maintenance of the product, loss of customers or even legal damages. Usually, the products developed by a company belong to the same family of products, with a common base to all the products, varying some parts to adapt the products to different demand levels. Thus, parts of a product which were already developed, tested and marketed can be reused for the development of new products, thus speeding up the development process. A good example of a family of products where there is the pattern previously presented are wireless sensor networks (WSNs). These types of networks are nowadays widely used in various applications, such as: domestic, industrial, civil engineering, military, environmental monitoring and health. Being the base of the network common to the various applications and vary only some specific configurations, sensor types and actuators assigned to the nodes in the network (edge devices). To accelerate the development of applications using WSNs, it is proposed in this dissertation the development of a generative framework, enable to obtain a good compromise between the variability management and edge device’s performance. In the design phase of the framework, Software Product Lines techniques will be used to identify the degree of variability of the functionalities of edge devices. In the framework’s implementation phase template metaprogramming in C++ will be used to manage this variability.
43
Etkin, Grisha. "Cove-Edge Graphene Nanoribbon Semiconductors: from Molecules to Devices." Thesis, 2018. https://doi.org/10.7916/D8515FVD.
Full textAbstract:
This dissertation presents research conducted on the structure-property relationships of cove-edge graphene nanoribbon (GNR) semiconductors from the scale of molecular conformation to device performance. The ribbons described here are made derived from perylene-3,4,9,10-tetracarboxylic acid diimide (PDI) and adopt a helical conformation so we call them helical PDI (hPDI). They are n-type semiconductors with exceptional performance in field-effect transistors (FETs), organic photovoltaics (OPVs), narrowband photodetectors, and electron transporting materials in perovskite solar cells. In this work, reaction chemistry is used to design and synthesize new derivatives of hPDI to shine light on their molecular, bulk, and device properties.
The first chapter concerns the incorporation of hPDI into alternating donor- acceptor (D-A) macromolecules to create materials with internal charge transfer (CT). Computational and spectroscopic techniques, including femtosecond transient absorption spectroscopy (fsTA), are used to probe the CT character of these materials. A large dihedral angle between donor and acceptor portions limits orbital overlap, leading to lowest energy excited state with HOMO localized on the donor and LUMO localized on the acceptor. Notably, internal CT improves the OPV performance of these oligomers over their parent hPDI, while analogous macromolecules without internal CT exhibit reduced OPV performance.
Chapter 2 details a method for side chain engineering of hPDI by installing the side chain in the final step of the synthesis, rather than the first. The aromatic core of hPDI is built up with esters, rather than imides, appending the edges of the ribbons. The ester-appended ribbons are readily transformed into a late-stage intermediate for divergent installation of any desired side chains, including those that pose synthetic challenges when they are introduced into the parent PDI from the beginning. These side chains have a profound effect on the optical, thermal, and charge transport properties of hPDI in the solid state. This strategy of introducing imide side-chains into PDI-based materials in the final step can be generalized to other systems.
Chapter 3 demonstrates a method for controlling the conformation of cove-edge GNRs by changing the chemical substitution pattern at their edges. All-sp2 substituents that lock adjacent edge positions into a ring rigidify the aromatic core of these ribbons. When substituents at adjacent edge positions are no longer locked into a ring, the aromatic core becomes flexible. Modulating this flexibility dictates how these ribbons contort to accommodate their cove-edges, with rigid cores contorting into chiral helixes, and flexible cores contorting into a butterfly conformation. This may point the way forward for the use of GNRs in applications that rely on precise control of molecular conformation
44
(10716561), Sanket Ramesh Joshi. "HBONEXT: AN EFFICIENT DNN FOR LIGHT EDGE EMBEDDED DEVICES." Thesis, 2021.
Find full textAbstract:
Every year the most effective Deep learning models, CNN architectures are showcased based on their compatibility and performance on the embedded edge hardware, especially for applications like image classification. These deep learning models necessitate a significant amount of computation and memory, so they can only be used on high-performance computing systems like CPUs or GPUs. However, they often struggle to fulfill portable specifications due to resource, energy, and real-time constraints. Hardware accelerators have recently been designed to provide the computational resources that AI and machine learning tools need. These edge accelerators have high-performance hardware which helps maintain the precision needed to accomplish this mission. Furthermore, this classification dilemma that investigates channel interdependencies using either depth-wise or group-wise convolutional features, has benefited from the inclusion of Bottleneck modules. Because of its increasing use in portable applications, the classic inverted residual block, a well-known architecture technique, has gotten more recognition. This work takes it a step forward by introducing a design method for porting CNNs to low-resource embedded systems, essentially bridging the difference between deep learning models and embedded edge systems. To achieve these goals, we use closer computing strategies to reduce the computer's computational load and memory usage while retaining excellent deployment efficiency. This thesis work introduces HBONext, a mutated version of Harmonious Bottlenecks (DHbneck) combined with a Flipped version of Inverted Residual (FIR), which outperforms the current HBONet architecture in terms of accuracy and model size miniaturization. Unlike the current definition of inverted residual, this FIR block performs identity mapping and spatial transformation at its higher dimensions. The HBO solution, on the other hand, focuses on two orthogonal dimensions: spatial (H/W) contraction-expansion and later channel (C) expansion-contraction, which are both organized in a bilaterally symmetric manner. HBONext is one of those versions that was designed specifically for embedded and mobile applications. In this research work, we also show how to use NXP Bluebox 2.0 to build a real-time HBONext image classifier. The integration of the model into this hardware has been a big hit owing to the limited model size of 3 MB. The model was trained and validated using CIFAR10 dataset, which performed exceptionally well due to its smaller size and higher accuracy. The validation accuracy of the baseline HBONet architecture is 80.97%, and the model is 22 MB in size. The proposed architecture HBONext variants, on the other hand, gave a higher validation accuracy of 89.70% and a model size of 3.00 MB measured using the number of parameters. The performance metrics of HBONext architecture and its various variants are compared in the following chapters.
45
Joshi, Sanket Ramesh. "HBONext: An Efficient Dnn for Light Edge Embedded Devices." Thesis, 2021. http://dx.doi.org/10.7912/C2/17.
Full textAbstract:
Indiana University-Purdue University Indianapolis (IUPUI)Every year the most effective Deep learning models, CNN architectures are showcased based on their compatibility and performance on the embedded edge hardware, especially for applications like image classification. These deep learning models necessitate a significant amount of computation and memory, so they can only be used on high-performance computing systems like CPUs or GPUs. However, they often struggle to fulfill portable specifications due to resource, energy, and real-time constraints. Hardware accelerators have recently been designed to provide the computational resources that AI and machine learning tools need. These edge accelerators have high-performance hardware which helps maintain the precision needed to accomplish this mission. Furthermore, this classification dilemma that investigates channel interdependencies using either depth-wise or group-wise convolutional features, has benefited from the inclusion of Bottleneck modules. Because of its increasing use in portable applications, the classic inverted residual block, a well-known architecture technique, has gotten more recognition. This work takes it a step forward by introducing a design method for porting CNNs to lowresource embedded systems, essentially bridging the difference between deep learning models and embedded edge systems. To achieve these goals, we use closer computing strategies to reduce the computer’s computational load and memory usage while retaining excellent deployment efficiency. This thesis work introduces HBONext, a mutated version of Harmonious Bottlenecks (DHbneck) combined with a Flipped version of Inverted Residual (FIR), which outperforms the current HBONet architecture in terms of accuracy and model size miniaturization. Unlike the current definition of inverted residual, this FIR block performs identity mapping and spatial transformation at its higher dimensions. The HBO solution, on the other hand, focuses on two orthogonal dimensions: spatial (H/W) contraction-expansion and later channel (C) expansion-contraction, which are both organized in a bilaterally symmetric manner. HBONext is one of those versions that was designed specifically for embedded and mobile applications. In this research work, we also show how to use NXP Bluebox 2.0 to build a real-time HBONext image classifier. The integration of the model into this hardware has been a big hit owing to the limited model size of 3 MB. The model was trained and validated using CIFAR10 dataset, which performed exceptionally well due to its smaller size and higher accuracy. The validation accuracy of the baseline HBONet architecture is 80.97%, and the model is 22 MB in size. The proposed architecture HBONext variants, on the other hand, gave a higher validation accuracy of 89.70% and a model size of 3.00 MB measured using the number of parameters. The performance metrics of HBONext architecture and its various variants are compared in the following chapters.
46
Huang, Jian-Xun, and 黃建勛. "Three-Tier Capacity and Traffic Optimization for Core, Edges, and Devices in Mobile Edge Computing." Thesis, 2017. http://ndltd.ncl.edu.tw/handle/29663301878031255532.
Full textAbstract:
碩士國立臺灣科技大學
資訊管理系
105
5G wireless communications systems has been proposed under the dramatical growth demand. The vision of 5G provides extremely low latency, and extends from 4G. In order to process more and more complexitive application in local mobile devices, so the local capabilities would be larger and larger. Due to that, MEC (mobile edge computing) has been proposed. Our work will solve two problems: (1) provide a costless model for operators, and decide the number of edges and the capacity of all servers in MEC (2) how to allocate the traffic. In order to achieve that, our research is aimed to compute the optimization about capacity and traffic allocation on homogeneous hierarchical architecture at three-tier MEC network. Our objective is constructing a network on MEC architecture with minimum total computational capacity beyond the latency constraint of 5G. We develope an algorithm, named as “two-phase iterative optimization (TPIO)”, to solve our two problems simultaneously. TPIO uses queueing theory to calculate the delay of traffic. the optimal network setting under ultra-low latency by the interactive adjustment of the two factor－capacity and traffic. In the final result, we know that three-tier architecture can save about 42% total cost than two-tier. In the percentage of traffic satisfying latency constraint, the total cost of QoS in 90% would have extra 34% cost than QoS in 50%. Due to the aforementioned, our work can find out the optimal static network setting and point out the influence of “edge tier” to overall network.
47
Papík, Ondřej. "Komunikace v prostředí tzv. mobile edge-cloud." Master's thesis, 2018. http://www.nusl.cz/ntk/nusl-387378.
Full textAbstract:
Edge-cloud brings the computation power as close to the clients as possible. This reduces latencies and overall computation time in the cloud. Thanks to the mobile nature of clients we must be able to migrate tasks among different servers. The goal of this thesis is to examine possible problems in communication and propose the architecture of framework. Our framework uses gRPC and is written as module to it. It is platform independent, uses reliable communication and focuses on easy usage. We provide implementation of this framework with some example uses. 1
48
(11185971), Md Naim Miah. "VOICE COMMAND RECOGNITION WITH DEEP NEURAL NETWORK ON EDGE DEVICES." Thesis, 2021.
Find full textAbstract:
Interconnected devices are becoming attractive solutions to integrate physical parameters and making them more accessible for further analysis. Edge devices, located at the end of the physical world, measure and transfer data to the remote server using either wired or wireless communication. The exploding number of sensors, being used in the Internet of Things (IoT), medical fields, or industry, are demanding huge bandwidth and computational capabilities in the cloud, to be processed by Artificial Neural Networks (ANNs) – especially, processing audio, video and images from hundreds of edge devices. Additionally, continuous transmission of information to the remote server not only hampers privacy but also increases latency and takes more power. Deep Neural Network (DNN) is proving to be very effective for cognitive tasks, such as speech recognition, object detection, etc., and attracting researchers to apply it in edge devices. Microcontrollers and single-board computers are the most commonly used types of edge devices. These have gone through significant advancements over the years and capable of performing more sophisticated computations, making it a reasonable choice to implement DNN. In this thesis, a DNN model is trained and implemented for Keyword Spotting (KWS) on two types of edge devices: a bare-metal embedded device (microcontroller) and a robot car. The unnecessary components and noise of audio samples are removed, and speech features are extracted using Mel-Frequency Cepstral Co-efficient (MFCC). In the bare-metal microcontroller platform, these features are efficiently extracted using Digital Signal Processing (DSP) library, which makes the calculation much faster. A Depth wise Separable Convolutional Neural Network (DSCNN) based model is proposed and trained with an accuracy of about 91% with only 721 thousand trainable parameters. After implementing the DNN on the microcontroller, the converted model takes only 11.52 Kbyte (2.16%) RAM and 169.63 Kbyte (8.48%) Flash of the test device. It needs to perform 287,673 Multiply-and-Accumulate (MACC) operations and takes about 7ms to execute the model. This trained model is also implemented on the robot car, Jetbot, and designed a voice-controlled robotic vehicle. This robot accepts few selected voice commands-such as “go”, “stop”, etc. and executes accordingly with reasonable accuracy. The Jetbot takes about 15ms to execute the KWS. Thus, this study demonstrates the implementation of Neural Network based KWS on two different types of edge devices: a bare-metal embedded device without any Operating System (OS) and a robot car running on embedded Linux OS. It also shows the feasibility of bare-metal offline KWS implementation for autonomous systems, particularly autonomous vehicles.
49
Chung, ChiBing, and 鍾其斌. "Three-dimensional Analyses of High Voltage Semiconductor Devices Edge Termination Structures." Thesis, 2013. http://ndltd.ncl.edu.tw/handle/36244494762548187158.
Full textAbstract:
碩士東海大學
電機工程學系
101
When the traditional element used in high pressure, the edge termination of structure was often the limiting element voltage capability, and even early crash. To solve this problem, we propose super junction structure to improve its main features to mitigate the gathered electric field distribution and be homogenized. In this thesis, power components are located in a circular shape and laid-out in round shape to avoid the corner and edge early breakdown. In this paper a round super junction structure is proposed to improve the breakdown voltage. SJ structure can reduce the surface electric field, to relieve the density of the electric power lines, so that the breakdown voltage of the device can be increased. Through simulation 3D structure design element, the software simulates and analyzes the electrical components of results. The results show that the super-junction structure can effectively improve the component breakdown voltage, relieved with dense surface electric field and achieved RESURF effect. So the electric field distribution is more flat. Ultimately derive empirical formula surplus surface, so that subsequent component design reference basis.
50
"Study of Knowledge Transfer Techniques For Deep Learning on Edge Devices." Master's thesis, 2018. http://hdl.handle.net/2286/R.I.49325.
Full textAbstract:
abstract: With the emergence of edge computing paradigm, many applications such as image recognition and augmented reality require to perform machine learning (ML) and artificial intelligence (AI) tasks on edge devices. Most AI and ML models are large and computational heavy, whereas edge devices are usually equipped with limited computational and storage resources. Such models can be compressed and reduced in order to be placed on edge devices, but they may loose their capability and may not generalize and perform well compared to large models. Recent works used knowledge transfer techniques to transfer information from a large network (termed teacher) to a small one (termed student) in order to improve the performance of the latter. This approach seems to be promising for learning on edge devices, but a thorough investigation on its effectiveness is lacking.
The purpose of this work is to provide an extensive study on the performance (both in terms of accuracy and convergence speed) of knowledge transfer, considering different student-teacher architectures, datasets and different techniques for transferring knowledge from teacher to student.
A good performance improvement is obtained by transferring knowledge from both the intermediate layers and last layer of the teacher to a shallower student. But other architectures and transfer techniques do not fare so well and some of them even lead to negative performance impact. For example, a smaller and shorter network, trained with knowledge transfer on Caltech 101 achieved a significant improvement of 7.36\% in the accuracy and converges 16 times faster compared to the same network trained without knowledge transfer. On the other hand, smaller network which is thinner than the teacher network performed worse with an accuracy drop of 9.48\% on Caltech 101, even with utilization of knowledge transfer.Dissertation/Thesis
Masters Thesis Computer Science 2018