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.<br>

<div>Driven by breakthroughs in mobile and IoT devices, on-device computation becomes promising. Meanwhile, there is a growing concern over its security: it faces many threats</div><div>in the wild, while not supervised by security experts; the computation is highly likely to touch users’ privacy-sensitive information. Towards trustworthy on-device computation, we present novel system designs focusing on two key applications: stream analytics, and machine learning training and inference.</div><div><br></div><div>First, we introduce Streambox-TZ (SBT), a secure stream analytics engine for ARM-based edge platforms. SBT contributes a data plane that isolates only analytics’ data and</div><div>computation in a trusted execution environment (TEE). By design, SBT achieves a minimal trusted computing base (TCB) inside TEE, incurring modest security overhead.</div><div><br></div><div>Second, we design a minimal GPU software stack (50KB), called GPURip. GPURip allows developers to record GPU computation ahead of time, which will be replayed later</div><div>on client devices. In doing so, GPURip excludes the original GPU stack from run time eliminating its wide attack surface and exploitable vulnerabilities.</div><div><br></div><div>Finally, we propose CoDry, a novel approach for TEE to record GPU computation remotely. CoDry provides an online GPU recording in a safe and practical way; it hosts GPU stacks in the cloud that collaboratively perform a dryrun with client GPU models. To overcome frequent interactions over a wireless connection, CoDry implements a suite of key optimizations.</div>

Beach cusps, formed during a storm are observed. The storm acted for three hours together with high water conditions. These beach cusps exhibited a quasiuniform wavelength of 8 m. The measured topography after the storm, the calculated drift velocity of the incident wave and the synchronous edge wave are similar in scale and shape. The over one wavelength measured grain size is also correlated to the topography. A nonlinear hydro-numerical model is used to investigate the reaction of edge waves on alongshore change in bottom topography. Edge waves are greatly amplified over beach cusps. The relative amplification of edge waves on beach cusps is more than 17 times that without a change in longshore topography. Amplification is slightly greater for edge waves travelling in the same direction as the longshore current during the storm. An almost random positioning of sediment starts a feedback loop, which allows edge waves to grow and build short periodic cusps.

Habitat fragmentation, and the resulting increase in edge habitat, has important effects on birds, including the increased probability of nest predation, changes in habitat structure, and the increased presence of non-native plant species. It is critical to understand the effects of fragmentation at all stages of the avian life cycle, including the often overlooked postfledging period. Because much of juvenile mortality occurs during the immediate postfledging period, and juvenile mortality contributes substantially to population dynamics, it is necessary to understand if fledgling survival is reduced in edge habitats and if fledglings' survival is influenced by their habitat use. During 2008 and 2009 I radio-tracked 52 fledgling Spotted Towhees (Pipilo maculatus) during the 30-day postfledging period in a 24-ha urban park near Portland, Oregon. Thirty-six fledglings (69%) survived the 27-day tracking period (an estimated 62.1% survived the entire 30-day postfledging period). At least 9 of 16 predation events were attributable to domestic cats (Felis domesticus) or Western Screech-owls (Megascops kennicottii). Although fledglings were more likely to be found near edges than the park interior, fledglings located closer to park edges had a higher probability of dying. However, I found that towhee nests were more likely to be found near edges, nests near edges produced more fledglings, and nestlings near edges were heavier. I used a STELLA-based stochastic model of nest success and fledgling survival to show that the benefits initially gained by nesting near edges were reversed during the postfledging period. The number of fledglings per nest that survived to the end of the 30-day postfledging period was significantly lower near edges than in the park interior. This apparent preference for nesting near edges, paired with higher fledgling mortality near edges, is consistent with the idea that edges are ecological traps. Fledgling habitat was significantly more structurally dense and had a greater abundance of non-native plant species, particularly Himalayan Blackberry (Rubus armeniacus), than nest habitat. Towhees avoided English Ivy (Hedera helix) for both nesting and care of fledglings. However, fledgling survival was not related to vegetation characteristics, which suggests that increased fledgling mortality near edges was a direct result of increased predator abundance or predation near edges, and was not an artifact of changes in habitat near edges. My results help to establish that fledgling survival and the unique habitat requirements of fledglings should be considered along with nest success and nest habitat when examining the effects of habitat fragmentation on bird populations. More broadly, this study has important implications for conservation, as it exemplifies how phenomena such as ecological traps created by anthropogenic changes in the environment can be overlooked if only one life history stage is studied.

Mozambique is to fifty percent covered by forest, most of which belongs to the biodiverse miombo woodlands. The last decades, Mozambique has been suffering from rapid deforestation. The once continuous forest cover has turned into a mosaic of forest patches, farmland, settlements etc. The remaining forest patches are in many cases very isolated. These forests have distinct edges towards the neighbouring land, which means that the edge zones have different environmental conditions (more light, higher temperatures etc.) and tree species composition than the interior. In order to examine how the forests of Mozambique are affected by edge effects, the highly isolated Nhamacoa forest was studied. An additional aim of the project was to make a floristic inventory of the forest in order to further assess its conservation status. Specimens were collected, pressed and photographed for identification. Edge effects were studied in plots at the edge and in the interior of the forest in a paired design. Trees inside the plots were identified, counted and measured (dbh (diameter at breast height) and height) to search for differences in species richness, diversity, biomass and height-to-dbh ratios. Environmental parameters (air temp., soil temp., light and slope) were also measured. In total, 76 species of trees were sampled and 44 (35 in the interior and 32 at the edge) of these were found inside the plots. The interior plots harboured significantly more individuals and species of trees than their paired edge plots. Additionally, biomass and height-to-dbh ratios were higher in the interior plots than in the edge plots. These differences strongly suggest that the Nhamacoa forest is affected by edge effects, although none of the measured environmental parameters could explain why. That the Nhamacoa forest is affected by edge effects goes in line with the research hypothesis and shows that it is important to maintain large and intact pieces of forest in order to preserve the Mozambican miombo forests. For future studies, additional environmental parameters (wind speed, humidity etc.) could be examined in order to better explain the presence of edge effects in the Nhamacoa forest.

Peatlands provide multiple ecosystem services, including extensive carbon sequestration and storage, yet many peatlands have been degraded or destroyed. Peatlands carbon storage capacity is connected to inherently low decomposition rates, causing the build-up of organic matter. Reasons suggested for this include waterlogged conditions that reduce the amount of available oxygen for the decomposer community, a low pH that inhibits bacterial decomposition, and colder temperatures lowering metabolic rates. This study focused on edge effects on decomposition in the transition zone between Sphagnum bogs and surrounding forest, with the expectation that decomposition is lowest in the bog and highest in the forest, but with a mix of factors causing intermediate decomposition rates near the bog edge. Transitional decomposition rates were measured across six bogs in central Sweden during the summer of 2021, using the tea bag index. Three 20 meter transects, each containing seven pairs of tea bags, were buried across the margins of each bog, centred at the edge of the Sphagnum moss. Soil moisture content, pH, and plant composition was also recorded at each burial site, and with 12 temperature loggers placed evenly among four of the bogs. The results confirmed the hypothesis regarding edge effects, and soil moisture was the dominant factor affecting the decomposition rate, having a strong significant negative correlation. The interaction of pH was significant but with an unexpected negative correlation, most likely due to the low pH in the surrounding forest. Temperature displayed no significance, and plants indicative of low decomposition rates included Vaccinium oxycoccos, Drosera rotundifolia, and especially Sphagnum species. The conclusions are that there is an edge effect on decomposition, and maintaining, or restoring, the hydrology of a peatland is the most important factor for continued carbon storage, with a rough estimation of an areas decomposition rate possible to estimate based on its vegetation. / Torvmarker bidrar med flera ekosystemtjänster, däribland utbrett koldioxidupptag och kollagring, trots detta har många torvmarker idag antingen skadats eller förstörts. Torvmarkers förmåga att lagra kol beror på en låg nedbrytningstakt, vilket gör att organiskt material ackumuleras; möjliga förklaringar som föreslagits till detta är vattensjuk mark som minskar syretillgången för nedbrytarsamhället, ett surt pH som missgynnar bakteriell nedbrytning, samt lägre temperaturer vilket sänker metabolismen generellt. Den här studien fokuserade på hur kanteffekter påverkar nedbrytningen i övergångszonen mellan vitmossemyrar och skog, med förväntningen att nedbrytningen är lägst ute i myren och högst i skogen, men att flera olika faktorer orsakar en intermediär nedbrytningshastighet i myrkanten. Nedbrytningshastigheten i 6 myrars övergångszoner, belägna i centrala Sverige, uppskattades under sommaren 2021 med hjälp av tepåseindexet (tea bag index). Tre 20 meters transekter med 7 par tepåsar vardera begravdes vid varje myr, med mitten vid kanten på vitmossan. Mätningar gjordes vid varje tepåsepar gällande markfuktighet, pH, och växtsamhälle, samt vid 4 myrar så placerades totalt 12 temperaturloggrar. Resultaten bekräftar hypotesen om kanteffekter och markfuktighet framträdde som den dominerande faktorn, med en signifikant negativ effekt på nedbrytningshastigheten. Effekten av pH var också signifikant men med en oväntad negativ korrelation, förmodligen på grund av den sura jorden i omgivande tallskog (Pinus sylvestris). Temperatur hade inte någon signifikant effekt, och växter som indikerade låg nedbrytning var Vaccinium oxycoccos, Drosera rotundifolia, samt framförallt Sphagnum spp.. Slutsatsen är att det finns en kanteffekt och att bevara, eller återskapa, hydrologin av en vitmossemyr är den viktigaste faktorn för fortsatt kolinlagring, med en grov skattning av ett områdes nedbrytningshastighet möjlig att basera på markernas växtlighet.

No description available.

Rapporten behandlar problemen med drift- och underhåll på vindkraftverk och hur man kan minska behovet av dessa genom övervakning av vindkraftverkens tillstånd. Målet med rapporten är att undersöka om det går att förutse när och hur ett vindkraftverk kommer behöva underhållas, samt beskriva de metoder som används för att förhindra slitage av turbinbladen. Forskningsfrågor som behandlas inkluderar tillståndsövervakning av vindturbiner och vilka metoder som används för att undvika slitage på turbinbladen. Arbetet kommer att utföras som en litteraturstudie eftersom det inte är möjligt med fältstudier av vindkraftverk. Viktiga frågor som besvaras är vilka faktorer som orsakar erosion och hur det leder till minskad effektivitet och produktionskapacitet hos vindkraftsparker. Vidare tas det upp hur underhåll utförs på vindkraftverken idag och hur det kan komma att ske i framtiden, med fokus på tillståndsövervakande system och olika skyddslösningar. Även skillnaden mellan landbaserad och havsbaserad vindkraft tas upp, samt olika stadier av erosionsutbredningen på turbinbladen och hur de påverkar energiproduktionen. Slutligen diskuteras olika lösningar för att minska sannolikheten för erosion, såsom att minska maxhastigheten under otjänligt väder. För havsbaserade vindkraftverk finns det svårigheter vad gäller byggande och underhåll, men fördelar som bättre vindförhållanden och möjlighet att bygga på djupare vatten. Flytande vindkraftverk används för att möjliggöra placering på djupare vatten och längre bort från land. Framtida vindkraftverk kommer ha större turbiner och högre torn, samtidigt som minskningar av energikostnaderna för både land- och havsbaserad vindkraft är trolig. SeaTwirl har en revolutionerande design där bladet roterar genom vertikalplanet, vilket minskar effekterna av erosionsbaserade skador som regndroppar och partiklar kan påföra. Dessa flytande vindkraftverk som är tänkta att placeras på havsytan har en relativt låg kostnad, lägre underhållsbehov och lägre materialåtgång, vilket gör dem konkurrenskraftiga. Framtida utveckling finns i form av exempelvis SR Energy som börjat dela ut "vindbonus" till de som bor nära vindkraftsparker. Ytterligare ett exempel är Modvion som använder trälaminat istället för stålkonstruktioner, vilket möjliggör enklare konstruktion, kräver färre inspektioner och material kan vid avveckling återanvändas vid husbyggen. / The report addresses the issues of operation and maintenance of wind turbines and how their need can be reduced through monitoring the condition of wind turbines. The aim of the report is to investigate whether it is possible to predict when and how a wind turbine will need to be maintained, as well as describe the methods used to prevent wear of the turbine blades. Research questions that are addressed include condition monitoring of wind turbines and the methods used to avoid wear on turbine blades. The work will be carried out as a literature study since it is not possible to do field studies of wind turbines. Important questions that are answered include what factors cause erosion and how it leads to reduced efficiency and production capacity of wind farms. Furthermore, it is discussed how maintenance is currently carried out on wind turbines and how it may be done in the future, with a focus on condition monitoring systems and various protective solutions. Also, the difference between land-based and offshore wind power is discussed, as well as different stages of blade erosion and how they affect energy production. Finally, different solutions to reduce the likelihood of erosion are discussed, such as reducing the maximum speed during inclement weather. For offshore wind turbines, there are difficulties with construction and maintenance, but advantages such as better wind conditions and the ability to build in deeper water. Floating wind turbines are used to enable placement in deeper water and farther from shore. Future wind turbines will have larger turbines and higher towers, while reducing energy costs for both land and offshore wind power is likely. SeaTwirl has a revolutionary design where the blade rotates through the vertical plane, which reduces the effects of erosion-based damage that raindrops and particles can cause. These floating wind turbines that are intended to be placed on the ocean surface have a relatively low cost, lower maintenance needs, and lower material consumption, making them competitive. Future developments include, for example, SR Energy, which has started to distribute "wind bonuses" to those who live near wind farms. Another example is Modvion, which uses wooden laminate instead of steel structures, enabling simpler construction, requiring fewer inspections, and allowing material reuse in building construction.

Bakgrunden till projektet är problem med kanterosion av turbinblad inom vindkraftverks- industrin. Det är ett fenomen som uppstår när turbinblad roterar i höga hastigheter och träffar partiklar, främst vattendroppar i regn. Denna erosion skadar turbinbladen, vilket både minskar vindkraftverkens effektivitet och sprider partiklar i den lokala miljön. För att både undersöka detta fenomen och ge möjlighet att utvärdera olika materials motståndskraft mot erosion ska en testrigg tas fram. Utöver detta ska testriggen även möjliggöra uppsamling av partiklar för vidare forskning kring deras effekt på miljön. Som utgångspunkt används en tribometer med rotationsmekanism från ett föregående maskinkonstruktionsprojekt på KTH. Ombyggnationen av denna avgränsas till att endast genomföras digitalt med hjälp av CAD för att hålla mängden arbete till en rimlig nivå. Förutom CAD har arbetet även inkluderat kravspecifikationer, beräkningar i MATLAB, FEA-analyser och kostnadskalkyler. Resultatet är en digitalt styrd testrigg med tillhörande komponentlista och instruktioner för tillverkning och genomförande. Inköpskostnad för ombyggnationen uppskattas till 28 000 kr. Riggen för en cylindrisk provbit genom ett artificiellt regn i hög hastighet, vilket resulterar i en accelererad nötningsprocess. Under testets gång dokumenteras erosionen visuellt med hjälp av en kamera och efter testet kan mängden förlorat material mätas i vikt och partiklar samlas upp. Flera andra parametrar dokumenteras automatiskt under testets gång för att ge en mer detaljerad bild av processen och data för undersökningar av repeterbarhet. Slutligen konstateras att testriggen uppnår alla krav ställda på den förutom att den inte har en nödbroms. Avsaknaden av nödbroms diskuteras och det leder till slutsatsen att testriggen, även utan nödbroms, är fullt fungerande, enkel och säker att använda. / The background for this project is a problem in the wind turbine industry, namely leading edge erosion of turbine blades. This occurs when wind turbine blades rotate at high speeds and collide with particles, mainly water drops in rain. This erosion damages the turbine blades, reducing the efficiency of the wind turbines and releasing particles into the local environment. To investigate this phenomenon and evaluate the durability of different materials, a test rig is to be developed. In addition, the test rig will enable the collection of particles to facilitate further research into their environmental impact. A tribometer with a rotation mechanism from a previous project at KTH serves as the starting point. The reconstruction of this tribometer is limited to a digital implementation using CAD to keep the amount of work at an appropriate level. In addition to CAD, the work has also included requirements specifications, calculations in MATLAB, FEA, and cost estimates. The result is a digitally controlled test rig with an accompanying component list and instructions for manufacturing and implementation. The estimated purchase cost for the reconstruction is 28,000 SEK. The rig tests a cylindrical sample by propelling it at high speed through artificial rain, resulting in an accelerated wear process. The erosion is visually documented using a camera during the test. Afterwards the amount of lost material can be measured by weight and the particles collected. Several other parameters are automatically recorded during the test to provide a more detailed picture of the process and data for investigations into repeatability. Finally, it is concluded that the test rig meets all its requirements except for the absence of an emergency brake. The absence of an emergency brake is discussed, leading to the conclusion that despite missing an emergency brake, the test rig is fully functional, easy to use, and safe.

Den snabbt växande utvecklingen av AI-baserade applikationer och den stora mängden data dessa applikationer behandlar ställer ökade krav på prestanda och optimering av datorsystemen. För att tillfredsställa de växande datorbehoven används hårdvaruacceleratorer som förbättrar databehandlingshastigheten genom att avlasta den befintliga utrustningen genom att hjälpa till med uppgifter och komplexa beräkningar. De befintliga lösningarna som används i dagsläget är kostsamma och MT-FoU på Umeå Universitetssjukhus efterfrågar därför en alternativ lösning i form av att kombinera mindre integrerande acceleratorer på ett större PCIe-kort. I detta examensarbete designas och implementeras en AI-accelerator bestående av fyra Google Coral Dual Edge TPU M.2 på ett 16x PCIe-kort. Arbetet genomfördes på MT-FoU och målet med examensarbetet var att undersöka om den tilltänkta konstruktionen kan förbättra prestandan hos AI-baserade system och fungera som ett billigare alternativ i verksamheten. Schemaritning och PCB-design utfördes i KiCad och information om gränssnitt och komponenter hämtades främst från tillverkares hemsidor och datablad. Kretsen består i huvudsak av fyra stycken M.2 E key kontaktdon, en 16port/16lane packetswitch och en 16x PCIe-anslutning. Switchen delar upp banorna från PCIe porten så att Edge TPU’erna kan anslutas parallellt i M.2 kontakterna. Edge TPU’erna använder pipelineparallellism för att fördela arbetsuppgifter på varje TPU så att större, mer komplexa program kan exekveras. Vid monteringen av kretskortet uppstod problem med fastlödningen av vissa komponenter. För att undvika att dessa problem uppstår och möjliggöra avlägsnandet av dessa felkällor bör montering istället beställas av fabrik där lödrobot finns tillgängligt. På grund av att tiden för kursen tog slut hann en sådan beställning inte göras och evaluering av den framtagna designen var därför inte möjlig att genomföra. Den design som togs fram var dock betydligt billigare än de existerande lösningarna och med pipelineparallellism förväntas designen kunna utföra komplexa beräkningar och därmed förbättra prestandan i befintliga system. / The rapidly growing development of AI-based applications and the large amount of data these applications process place increased demands on the performance and optimization of conventional computer systems. To satisfy these growing computing requirements, hardware accelerators are used to improve the data processing speed by offloading the existing equipment by executing models and complex calculations. The existing solutions currently used are costly and MT-R&amp;D at Umeå University Hospital is therefore requesting an alternative solution by combining smaller integrating accelerators on a larger PCIe card. In this thesis, an AI accelerator using four Google Coral Dual Edge TPU M.2 on a 16x PCIe card is designed and implemented. The work was carried out at MT-R&amp;D and the goal of the thesis was to investigate whether the intended design can improve the performance of AI-based systems and serve as a cheaper alternative in the institution. Schematic and PCB were designed in KiCad and information on interfaces and components was obtained from manufacturers' websites and data sheets. The circuit’s main components are four M.2 E key connectors, a 16port/16lane packet switch and a 16x PCIe connection. The switch divides the lanes from the PCIe port so that the Edge TPUs can be connected in parallel in the M.2 connectors. The Edge TPUs use pipeline parallelism to distribute models across each TPU so that larger, more complex programs can be executed. When assembling the circuit board, problems arose with the soldering of certain components. In order to avoid these sources of error, assembly should instead be ordered from a factory where a soldering robot is available. Due to the fact that the time for the course ran out, such an order could not be placed and evaluation of the design was therefore not possible to carry out. However, the design that was produced was significantly cheaper than the existing solutions and by using pipeline parallelism, the design is expected to be able to perform complex calculations and thus improve the performance of existing systems.