The machine learning (ML) industry has taken great strides forward and is today facing new challenges. Many more models are developed, used and served within the industry. Datasets that models are trained on, are constantly changing. This demands that modern machine learning processes can handle large number of models, extreme load and support recurring updates in a scalable manner. To handle these challenges, there is a concept called model serving. Model serving is a relatively new concept where more efforts are required to address both conceptual and technical challenges. Existing ML model serving solutions aim to be scalable for the purpose of serving one model at a time. The industry itself requires that the whole ML process, the number of served models and that recurring updates are scalable. That is why this thesis presents an open-source framework for large-scale ML model serving that aims to meet the requirements of today’s ML industry. The presented framework is proven to handle a large-scale ML model serving environment in a scalable way but with some limitations. Results show that the number of parallel requests the framework can handle can be optimized. This would make the solution more efficient in the sense of resource utilization. One avenue for future improvements could be to integrate the developed framework as an application into the open-source machine learning platform STACKn.

The amount of open-source information – that is, data, images, and footage that is openly available to the public - is growing exponentially. With it, so is the number of citizens analysing this data to form open-source intelligence (OSINT). Using the 2022 invasion of Ukraine as a case study, this study highlights the motivations behind the citizen OSINT analysts who are uncovering events on the frontline and verifying significant amounts of data from such events. Through interviews with 10 citizen OSINT analysts – all voluntarily contributing to OSINT in relation to the invasion of Ukraine, as well as other major OSINT projects – this study demonstrates the motivations behind this growing community. The findings reflect a new era of participation and advocacy and are a demonstration of self-determination theory. The findings demonstrate citizens’ sceptic views towards traditional media but also, that with a more analytical approach, with improved transparency and collaboration there is reason to be optimistic about the future of journalism and audience engagement.

Ce projet de recherche-création est axé sur le développement d’un outil de création sonore, qui vise à faciliter l’improvisation de musiques électroacoustiques spatialisées sur de multiples haut-parleurs. L’outil nommé Live 4 Life permet plus spécifiquement de composer des polyrythmies spatialisées, combinées à des textures enveloppantes en temps réel. Actuellement, il est particulièrement difficile de contrôler dynamiquement et simultanément de multiples sons dans l’espace, en raison en partie de la stratification du processus de composition dans plusieurs logiciels et de la complexité à spatialiser individuellement chaque événement sonore. En dotant l’improvisateur d’outils de spatialisation flexibles, celui-ci peut exploiter les dimensions spatiales tout au long du processus de création, de la composition jusqu’à la performance. L’intégration, à même l’outil Live 4 Life, des autres paramètres sonores, comme le rythme ou les vitesses de lecture, permet également de créer de nouveaux liens dans la composition de l’espace. L’outil, implémenté sur la plate-forme en source libre SuperCollider, propose une approche comprenant différentes échelles temporelles, où la spatialisation est déterminée autant au niveau local sur des événements individuels que global sur des flux sonores. Des œuvres ouvertes du point de vue spatial et temporel peuvent ainsi être réalisées à partir de multiples séquences pré-composées, qui sont organisées et modifiées en temps réel. Cet outil de performance spatiale met l’accent sur la construction et la déformation de boucles de paramètres, rythmiques ou spatiales. Il permet, en outre, de questionner le développement spatio-temporel entre les objets sonores eux-mêmes, leurs traitements ou leurs réflexions et ouvre sur des questions plus larges sur les liens qui existent entre l’espace et le temps réel. / The research/creation project is focused on the development of a spatialised sound creation tool, which aims at facilitating the improvisation of electroacoustic music on multiple loudspeakers. More specifically, this tool, which is called Live 4 Life, allows to compose spatialised polyrhythms, combined with enveloping textures in real time. Currently, it is particularly difficult to dynamically and simultaneously control multiple sounds in space, due in part to the layering of the composition process in multiple software programs and the complexity of individually spatialising each sound event. By providing the improviser with flexible spatialisation tools, it allows him to exploit spatial dimensions throughout the creative process, from composition to performance. The integration of other sound parameters, such as rhythm, spectrum or playback speeds, in a global creation tool also allows to create new links in the composition of space. The tool, implemented on the open source platform SuperCollider, offers an approach comprising different time scales, where spatialisation is determined both locally on individual events and globally on sound streams. Spatially and temporally open forms can thus be realised from multiple pre-composed sequences, which are organised and modified in real time. This spatial performance tool emphasises the construction and deformation of parameter loops, rhythmic or spatial. It also makes it possible to question the spatio-temporal development between sound objects themselves, their processing or their reflections, and opens up broader questions on the links that exist between space and real time.

With a constant increase of the datasphere so does the need for proper management of this data in order to minimise potential inefficiencies when it comes to the usage of this data. Geodata is no exception to the need of management.  The purpose of this thesis was to investigate the current challenges of geodata management in Swedish companies and public authorities through qualitative analysis by interviews. Geodata- and GIS-users from the public and private sector made up the pool of interviewees and a total of 20 interviews were conducted. Despite a large diversity of daily tasks at hand, from data transfer for a customer in a system change process to updating attributes in NIS tools, all of the interviewees had some degree of challenges with respect to management of geodata. The results showed that the main challenges regarded inadequate or lack of quality geodata and its metadata as well as clarity of the location of these datasets. Scarcity of common understanding of geodata and GIS-systems among colleagues of geodata- and GIS-users causes these colleagues to often deliver incorrect, poorly formatted or low quality geodata and metadata. A large number of geodata file formats also contributes to confusion amongst geodata- and GIS-users and non-users which directly and indirectly causes some inefficiency. It was determined that the challenges of geodata management are highly abundant in the Swedish public and private sector. Furthermore, it was evaluated that the possible solutions would be to simplify geodata with less file formats as well as better and more clear coordination at organisational levels. Educating non-geodata and GIS-users in the workforce as well as in higher educational institutions that have majors related to geodata was also suggested as a possible solution to minimise challenges.

<p dir="ltr">Methodologies for optimizing UAVs' control for varied environmental conditions have become crucial in the recent development for UAV control sector, yet they are lacking. This research focuses on the dynamism of the Gazebo simulator and PX4 Autopilot flight controller, frequently referenced in academic sectors for their versatility in generating close-to-reality digital environments. This thesis proposed an integrated simulation system that ensures realistic wind and gust interactions in the digital world and efficient data extraction by employing an industrial standard control communication protocol called MAVLink with the also the industry standard ground control software QGroundControl, using real and historical weather information from NOAA database. This study also looks into the potential of reinforcement learning, namely the DDPG algorithm, in determining optimal UAV safety distance, trajectory prediction, and mission planning under wind disruption. The overall goal is to enhance UAV stability and safety in various wind-disturbed conditions. Mainly focusing on minimizing potential collision risks in areas such as streets, valleys, tunnels, or really anywhere has winds and obstacles. The ROS network further enhanced these components, streamlining UAV response analysis in simulated conditions. This research presents a machine-learning approach to UAV flight safety and efficiency in dynamic environments by synthesizing an integrated simulation system with reinforcement learning. And the results model has a high accuracy, reaching 91%, 92%, and 97% accuracy on average in prediction of maximum shifting displacement, and left/right shifting displacement, when testing with real wind parameters from KLAF airport. </p>

Exoskeletons are wearable devices that enhance or supplement the user’s natural abilities. They have been demonstrated to be efective in alleviating pain, reducing work related injuries, improving working conditions, and can play a pivotal role in improving recovery times and recovery outcomes. Commercial exoskeletons are expensive, specialised and not easily accessible to the average user. This thesis describes the design and manufacturing processes for the OpenXO, an open-source knee exoskeleton. The focus of the thesis is the design and manufacture of the exoskeleton drive system. It implements a cycloidal drive design. Additionally, a method of designing tapered crossed roller bearings was developed that allows for easy integration into elements of the drive system. The open source aspect is further supported by designing the OpenXO around commercially available additive manufacturing technologies. Rapid prototyping and iterative test-based design methods were used in conjunction with empirical testing and validation of both the design and manufacturing methods. Performance validations were conducted on an unpowered exoskeleton. The tests focus on ease of use and comfort. Validation on the exoskeleton and its components was performed at various stages during the design process. The resulting drive design was signifcantly lighter than commercially available solutions. The stator design implemented allows for press ft accuracy between the rotor and stator while allowing for smooth rotation. The fully assembled exoskeleton was tested by 5 individuals. All participants performed tasks to test the usability the exoskeleton in common day to day activities. The participants managed to perform several tasks with ease. However, the exoskeleton was prone to misalignment in specifc circumstances. Gait analysis on a user wearing the exoskeleton shows that the exoskeleton does infuence gait patterns. However, the user does not experience signifcant impact on their perceived range of motion. These tests do not demonstrate the efectiveness of the exoskeleton when it comes to powered assistance. Further work is needed to test and validate the powered assist functionality of the exoskeleton. / Eksoskeletonit ovat puettavia laitteita, jotka parantavat tai täydentävät käyttäjän luonnollisia kykyjä. Niiden on osoitettu olevan tehokkaita kipujen lievittämisessä, työtapaturmien vähentämisessä, työolosuhteiden parantamisessa ja niillä voi olla keskeinen rooli paranemisaikojen ja toipumistulosten parantamisessa. Kaupalliset eksoskeletonit ovat kalliita, erikoistuneita eivätkä tavallisen käyttäjän helposti saatavilla. Tämä opinnäytetyö kuvaa avoimen lähdekoodin polven eksoskeleton OpenXO:n suunnittelu-ja valmistusprosessit. Erityisesti painopiste on sykloidiseen käyttöjärjestelmään perustuvan ulkopuolisen tukirankajärjestelmän suunnittelussa ja valmistuksessa. Lisäksi kartiorullalaakereiden suunnittelumenetelmä kehitettiin siten, että se mahdollistaa helpon integroinnin käyttöjärjestelmän elementteihin. Avoimen lähdekoodin näkökulmaa tuetaan edelleen suunnittelemalla OpenXO kaupallisesti saatavilla olevien lisäaineiden valmistustekniikoiden ympärille. Nopeaa prototyyppiä ja iteratiivisia testipohjaisia suunnittelumenetelmiä käytettiin sekä suunnittelu-että valmistusmenetelmien empiirisen testauksen ja validoinnin yhteydessä. Suorituskyvyn validointi suoritettiin tehottomalla eksoskeletonilla 5 vapaaehtoisen poolissa. Testit suuntautuivat arjen yleisten toimintojen ympärille ja keskittyivät pääasiassa helppokäyttöisyyteen ja käyttömukavuuteen. Eksoskeleton ja sen komponenttien validointi suoritettiin suunnitteluprosessin eri vaiheissa. Tuloksena saatu käyttörakenne oli huomattavasti kevyempi kuin kaupallisesti saatavilla olevat ratkaisut. Toteutettu staattorin rakenne mahdollisti roottorin ja staattorin välisen puristussovituksen tarkkuuden säilyttäen samalla tasaisen pyörimisen. Osallistujat onnistuivat suorittamaan suurimman osan tehtävistä helposti. Eksoskeleton oli kuitenkin altis kohdistusvirheelle tietyissä olosuhteissa. Eksoskeletonia käyttävän käyttäjän kävelyanalyysi osoittaa, että ulkoinen luuranko vaikuttaa kävelykuvioihin, mutta käyttäjä ei koe merkittävää vaikutusta havaittuun liikerataan. Siitä huolimatta tarvitaan lisätyötä eksoskeletonin tehollisen aputoiminnon testaamiseksi ja validoimiseksi. / <p>Presentation conducted online via the Zoom video conferencing platform.</p>

Objectif : Comparer les dispositifs de délivrance automatisée d’insuline open-source (DDAI-OS) et les DDAI commerciaux hybrides sur le contrôle glycémique, la qualité de vie rapportée, et la sécurité chez des adultes avec diabète de type 1 (DT1). Méthodes : Étude prospective, observationnelle, de non-infériorité, comparative et en vie réelle, incluant 78 adultes canadiens avec un DT1 et utilisant un DDAI ≥ 3 mois. Quatre semaines de mesure continue du glucose ont permis d’évaluer le % temps passé dans la cible de glucose (%TIR, 3,9-10,0 mmol/L). Les indicateurs de qualité de vie ont été évalués par des échelles de mesure validées. Les mesures de sécurité sont le temps passé en hypoglycémie, la survenue d’hypoglycémie sévère ou d’acido-cétose et les problèmes techniques. Résultats : Les participants du groupe DDAI-OS étaient non inférieurs au groupe DDAI commercial sur le %TIR (78,3% [SD 11,0] vs. 71,2% [SD 10,9], différence moyenne 7,2% [95% CI 1,9% à 12,5%], p<0.001), même après ajustement sur plusieurs facteurs confondants. Le groupe DDAI-OS a passé plus de temps en hypoglycémie (<3,9 mmol/L) (3,9% [SD 3,1] vs. 1,8% [SD 1,3], p<0.001) et a rapporté moins de peur de l’hypoglycémie. Aucun épisode d’hypoglycémie sévère ou d’acido-cétose n’a été rapporté, avec un nombre de problèmes techniques similaires entre les deux groupes. Conclusion : Les DDAI-OS hybrides sont sécuritaires et non inférieurs aux DDAI commerciaux hybrides en termes de %TIR chez des adultes vivant avec un DT1 dans des conditions de vie réelle. Nos résultats soutiennent que les DDAI-OS peuvent être considérés pour la gestion du DT1. / Background: Comparison between unregulated open-source (OS) automated insulin delivery (AID) systems and commercial AID (C-AID) systems remains scarce. Objective: Compare both AID systems regarding glucose management, patient-reported outcomes (PROs), and safety among adults with type 1 diabetes (T1D) in real-life conditions. Design: Prospective, observational, non-inferiority, comparative, real-world study. Setting: On-site (a diabetes clinic in Montreal) and online (a T1D registry and social media platforms) across Canada. Participants: 78 adults with T1D, having used an AID system for ≥ 3 months, and living in Canada (26 OS-AID and 52 C-AID users). Measurements: 4-week’s data from a blinded continuous glucose monitor were used to assess effectiveness (primary outcome: 24h time in range % [TIR%], with a non-inferiority margin of 5%). Other outcomes included PRO measures using validated scales. Safety outcomes included time spent in hypoglycemia, severe hypoglycemia, diabetic ketoacidosis (DKA), and technical issues. Results: OS-AIDs were non-inferior to C-AIDs regarding 24h TIR% (78.3% [SD 11.0] vs. 71.2% [SD 10.9], mean difference 7.2% [95% CI 1.9% to 12.5%], p<0.001), even after adjusting for various confounding factors. OS-AIDs spent more time in hypoglycemia (<3.9 mmol/L) than C-AIDs (3.9% [SD 3.1] vs. 1.8% [SD 1.3], p<0.001) and reported less fear of hypoglycemia. No severe hypoglycemia or DKA was reported in either group, with a similar occurrence rate of technical issues between groups. Conclusion: OS-AIDs are safe and non-inferior to C-AIDs for TIR% among adults with T1D in real-world settings. Our findings support that both OS-AID and C-AID systems can be considered for T1D management.

No description available.

The IT sector is in constant change and new IT emerges constantly. This enables organizations to adopt new IT at a correspondingly rapid pace. Nonetheless, it is not certain that the adoption of new IT contributes positively to the organization’s existing application portfolio. Hence, the purpose of this study is to provide insights to how organizations adopt IT and the factors that are important to ensure positive effects from adopting new IT. A case study, consisting of semi-structured interviews, was conducted on a Norwegian governmental organization. The majority of current literature on the field are describing IT adoption for commercial for-profit organizations. Since the case-organization is a governmental organization, this study will also aspire to investigate how that influence IT adoption. The empirical findings of this study show that there is a purpose of having a strategy for adopting new IT to an existing application portfolio. A framework describing the adoption process and factors of importance for successfully adopting new IT is provided. Also, the empirical findings show that being a governmental organization does not influence the adoption of IT significantly. However, the findings show that other aspects of each individual organization can have more significant influence on the adoption process. / IT-sektorn är i ständig förändring och ny IT blir ständigt tillgänglig. Detta gör det möjligt för organisationer att ta sig an ny IT i motsvarande snabba takt. Det är dock inte säkert att implementering av ny IT bidrar positivt till organisationens befintliga applikationsportfölj. Därför är syftet med denna studie att ge insikt om hur organisationer tar sig an IT och de faktorer som är viktiga för att säkerställa positiva effekter från adoption av ny IT. En fallstudie, bestående av semi-strukturerade intervjuer genomfördes på en norsk statlig organisation. Majoriteten av befintlig litteratur inom fältet beskriver IT-adoption för kommersiella vinstdrivande organisationer. Eftersom fallstudieorganisationen är en statlig organisation kommer denna studie också att undersöka hur det påverkar IT-adoption i organisationen. De empiriska resultaten från denna studie visar att det finns ett syfte med en strategi för att adoptera ny IT till en befintlig applikationsportfölj. Ett ramverk som beskriver adoptionsprocessen och faktorer som är viktiga för en framgångsrik adoptionsprocess av ny IT finns framtaget. De empiriska resultaten visar också att det att vara en statlig organisation inte påverkar adoption av IT i särskilt hög grad. Resultaten visar dock att andra aspekter av varje enskild organisation kan ha mer inflytande på adoptionsprocessen.

Indonesia’s energy demand has been growing rapidly driven by increasing population, urbanization, and rapid economic growth. With increasing energy demand, the emissions associated with the energy sector continue to increase. With the gradual increase in demand and dominant share of fossil fuels in the energy mix, implementing the energy efficiency measures is crucial for Indonesia to achieve its energy and climate goals. From the policy perspective, National Energy plan of Indonesia aims to achieve higher levels of energy efficiency to reduce the overall energy intensity. Indonesia also has commitments to reduce greenhouse gas emissions and achieve SDG targets. This report reviews the current status of energy demand and energy efficiency in Indonesia and evaluates the potential of implementing energy efficiency measures and fuel switching options to achieve future low carbon energy future. Long term energy model of Indonesia is modelled using the open-source modelling tool OSeMOSYS. Different scenarios have been developed to investigate the outcome of implementing energy efficiency and fuel switching measures in the Residential, Commercial, and Transportation sectors. The results are presented in terms of reduction in total final energy use, greenhouse gas emissions, and local air pollution. Cost-Benefit analysis of the applied measures present their financial feasibility. With the deployment of efficient appliances, up to 30% electricity savings can be achieved in the residential and commercial sector. Vehicle electrification can contribute towards reduction in annual energy use by 48% by the end of modelling period. Measures in the residential and commercial sector directly contribute towards emission reductions. Vehicle electrification does not show proportionate reduction in emissions compared to energy use reduction due to high carbon intensity of the electricity grid. However, significant reduction in local air pollutants can be achieved. Cost benefit analysis shows that deployment of efficient appliances is financially feasible with maximum 2 years of payback period. On the other hand, successful deployment of electric vehicles will require tangible support from government due to its high price premium compared to conventional vehicles. Energy efficiency measures and fuel switching also contribute substantially to achieving Sustainable Development Goal 7.3. In conclusion, this study presents a set of technically and economically feasible energy system development options for Indonesia. From the modelling perspective, this study identifies ways to implement demand side management measures in the energy supply modelling system OSeMOSYS. / Indonesiens energibehov har ökat snabbt drivet av ökande befolkning, urbanisering och snabbekonomisk tillväxt. Med ökande energibehov fortsätter utsläppen i energisektorn att öka. Medden gradvisa ökningen i efterfrågan och den dominerande andelen fossila bränslen ienergimixen är genomförandet av energieffektivitetsåtgärderna avgörande för att Indonesienska uppnå sina energi- och klimatmål. Ur politiskt perspektiv syftar Indonesiens nationella energiplan till att uppnå högre nivåer av energieffektivitet för att minska den totala energiintensiteten. Indonesien har också åtaganden att minska utsläppen av växthusgaser och uppnå SDG-mål. Denna rapport granskar den aktuella statusen för efterfrågan på energi och energieffektivitet i Indonesien och utvärderar potentialen för att genomföra energieffektivitetsåtgärder och alternativ för bränsleomkoppling för att uppnå framtida energiförbrukning med låg koldioxid. Indonesiens långsiktiga energimodell modelleras med hjälp av open-sourcemodelleringsverktyget OSeMOSYS. Olika scenarier har utvecklats för att undersöka resultatet av genomförande av energieffektivitet och bränsleomkopplingsåtgärder inom bostads-, kommersiellt och transportsektorn. Resultaten presenteras i termer av minskning av den totalaslutliga energiförbrukningen, växthusgasutsläpp och lokal luftföroreningar. Kostnadsnyttoanalys av de tillämpade åtgärderna utgör deras ekonomiska genomförbarhet. Med användning av effektiva apparater kan upp till 30% elbesparing uppnås i bostads- och affärssektorn. Fordonselektrifiering kan bidra till minskning av den årliga energiförbrukningen med 48% i slutet av modelleringsperioden. Åtgärder inom bostads- och kommersiell sektor bidrar direkt till utsläppsminskningar. Fordonselektrifiering visar inte proportionell minskning av utsläpp jämfört med energiförbrukningen på grund av hög kolintensitet i elnätet. Emellertid kan en betydande minskning av lokala luftföroreningar uppnås. Kostnads för delningsanalys visar att distribution av effektiva apparater är ekonomiskt möjlig med maximalt 2 års återbetalningsperiod. Å andra sidan kommer framgångsrik distribution av elfordon att kräva konkret stöd från regeringen på grund av dess höga prispremie jämfört med konventionella fordon. Energi effektivitetsåtgärder och bränsleomkoppling bidrar också väsentligt till att uppnå mål för hållbar utveckling 7.3. Sammanfattningsvis presenterar denna studie en uppsättning tekniska och ekonomiskt genomförbara energisystemutvecklingsalternativ för Indonesien. Från modelleringsperspektivet identifierar denna studie sätt att implementera hanteringsåtgärder på efterfrågesidan i modelleringssystemet för energiförsörjning OSeMOSYS.