Rapporten tar upp hur dagens true wireless-teknik fungerar och hur den kan implementeras i separata högtalare. Den förklarar de fundamentala komponenterna som behövs för att konstruera en fungerande trådlös högtalare, sammanfattar överföringstekniken Bluetooth och går in på radarsystem samt tekniken som används i Acconeers radarkomponenter XC112/XR112. Rapporten förklarar specifika komponenter som hörlurarna Urbanista Stockholm och hur dess funktioner kan påverkas från externa källor. Dokumentering i hur installationen samt implementeringen av Raspberry Pi och radarsensorerna går till tas även upp i rapporten. I arbetet har geststyrning implementerats i ena högtalaren med hjälp av radarsensorer för att kontrollera funktionerna som hörlurarna har inbyggt. Hörlurarna kommunicerar med en uppspelningsenhet via Bluetooth, vanligtvis en mobiltelefon. De funktioner som används är att byta låt, ändra volym och pausa/spela musiken. Eftersom de olika hörlurarna kontrollerar olika funktioner har en 433Hz radiosändare används för att kunna skicka över specifika kommandon till andra högtalaren. Resultatet av arbetet visar att true wireless-tekniken kan implementeras väl i separata högtalare, med hjälp av Acconeers radarsensorer kunde olika gester registreras och kommandon skickas till hörlurarna. För att gester ska registreras krävs specifika avstånd och rätt hastighet på svepet med handen. Kommunikationen med radiosändaren blev instabil och behövdes både placeras rätt och justeras för att fungera. När väl kommunikationen med radiomodulerna fungerade så utfördes kommandona på den andra hörluren korrekt. / This report explains how today's true wireless technology works and how it could be implemented in separate speakers. It explains the fundamental components needed to design a working wireless speaker, summarizes the Bluetooth transmission technology and goes into radar systems as well as the technology used in Acconer's radar components XC112/XR112. Furthermore, the report goes into specific components such as the Urbanista Stockholm headphones and how its functions can be controlled from external sources. Documentation on how the installation as well as implementation of the Raspberry Pi and the radar sensors works is also included in the report. With the work, gesture control has been implemented in one of the speakers with the help of radar sensors to control the functions that the headphones have built-in. The headphones communicate with a playback device via Bluetooth, usually a mobile phone. The functions are changing songs, adjusting volume and pausing/playing the music. Because the different headphones control different functions, a 433Hz radio transmitter has been used to be able to transmit specific commands to the other speaker. The results of the work show that true wireless technology can be well implemented in separate speakers, with help of Acconeer’s radar sensors, different gestures could be registered and commands sent to the headphones. In order for the gestures to be registered, specific distances and the correct speed of the sweep are required. The communication between the radio transmitter and receiver became unstable and needed to be placed correctly and adjusted to work. Once the communication worked the commands were executed correctly by the other headphone.

In recent years, companies of all sizes have become increasingly dependent on customer user data and processing it using machine learning (ML) methods. These methods do, however, require the raw user data to be stored locally on a server or cloud service, raising privacy concerns. Hence, the purpose of this paper is to analyze a new alternative ML method, called federated learning (FL). FL allows the data to remain on each respective device while still being able to create a global model by averaging local models on each client device. The analysis in this report is based on two different types of simulations. The first is simulations in a virtual environment where a larger number of devices can be included, while the second is simulations on a physical testbed of Raspberry Pi (RPI) single-board computers. Different parameters are changed and altered to find the optimal performance, accuracy, and loss of computations in each case. The results of all simulations show that fewer clients and more training epochs increase the accuracy when using independent and identically distributed (IID) data. However, when using non-IID data, the accuracy is not dependent on the number of epochs, and it becomes chaotic when decreasing the number of clients which are sampled each round. Furthermore, the tests on the RPIs show results which agree with the virtual simulation. / På den senaste tiden har företag blivit allt mer beroende av ku rs användardata och har börjat använda maskininlärningsmodeller för att processera datan. För att skapa dessa modeller behövs att användardata lagras lokalt på en server eller en molntjänst, vilket kan leda till integritetsproblematik. Syftet med denna rapport är därför att analysera en ny alternativ metod, vid namn ”federated learning” (FL). Denna metod möjliggör skapandet av en global modell samtidigt som användardata förblir kvar på varje klients enhet. Detta görs genom att den globala modellen bestäms genom att beräkna medelvärdet av samtliga enheters lokala modeller. Analysen av metoden görs baserat på två olika typer av simuleringar. Den första görs i en virtuell miljö för att kunna inkluderastörre mängder klientenheter medan den andra typen görs på en fysisk testbädd som består av enkortsdatorerna Raspberry Pi (RPI). Olika parametrar justeras och ändras för att finna modellens optimala prestanda och nogrannhet. Resultaten av simuleringarna visar att färre klienter och flera träningsepoker ökar noggrannheten när oberoende och likafördelad (på engelska förkortat till IID) data används. Däremot påvisas att noggrannheten inte är beroende av antalet epoker när icke-IID data nyttjas. Noggrannheten blir däremot kaotisk när antalet klienter som används för att träna på varje runda minskas. Utöver observeras det även att testresultaten från RPI enheterna stämmer överens med resultatet från simuleringarna. / Kandidatexjobb i elektroteknik 2022, KTH, Stockholm

The advent of smarter vehicles coupled with declining costs for solid-state light detection and ranging (LiDAR) sensors unlocks the potential of cheaper devices that experience less mechanical wear while providing smaller form factors. The early state of the market makes evaluating the limited supply of available sensors interesting. This thesis investigated the possibility of creating a wireless distance measurement system, at relatively low cost, using a commercially available solid-state LiDAR sensor. A complete prototype was designed using a LeddarTech VU8 sensor. The system was fully integrated with a power supply PCB and a 3D printed enclosure. Distance data could be wirelessly transferred via Bluetooth using a Raspberry Pi 4 B. A graphical user interface (GUI) was developed to display distance data and to interact with the system in real-time. The system intends to provide a foundation for future work.

The dissertation aims to expand access through a low-cost technological innovation system S-MLS to learners in underdeveloped areas with difficulties in accessing education. Technology is advancing rapidly. However, many parts of the world need access to educational advances, which are hindered due to war, political situations, and low literacy and income. A qualitative phenomenological approach explores the lived experience using the solar-powered computing and learning management system (LMS) to support the development of educational access in underrepresented societies, developing countries, and rural areas where access to proper classroom education is non-existent. Proof of concept is used with a group of students in a rural area, a developing country, and within an underrepresented population to check the feasibility of using the equipment in a real-world setting. A technology acceptance model would be used to identify the user's perceived interest and user acceptance. The community of inquiry theory would find the first-hand experience and point of view of the learner. The student group interviews would be through semi-structured interviews. Observations, surveys, video/audio recordings, and artifacts would be gathered for further analysis. The data collected would be analyzed using interpretative phenomenology analysis (IPA), close examination, and management of development themes through thoughts, observations, and reflections on the technological experience and future research and implementations provided. The projected finding would be to check that a solar-powered Raspberry Pi system with MoodleBox operating system that runs Moodle (Modular Object-Oriented Dynamic Learning Environment) LMS would be feasible to provide learning underdeveloped areas to enhance education.

Die Masterarbeit stellt die Entwicklung und Anwendung des interdisziplinären High-Fidelity-Prototyps, PiaPi dar. Basierend auf der 'Research through Design' Methode kombiniert der Prototyp LEGO, Musik und Licht, um die Mensch-Computer-Interaktion zu fördern. Eine qualitative Studie zeigt, dass der Prototyp Testpersonen dazu motiviert, kreativ zu interagieren, Neugier weckt und dabei Berührungsängste abbaut. Zur wissenschaftlichen Einordnung wurden Elemente der HCI mit Forschung zu Spaß und Spiel kombiniert. Besonderes Augenmerk liegt auf dem potenziellen Einsatz des PiaPi als Projekt für den fächerübergreifenden Unterricht an Schulen. Dafür wurden in der Entwicklung verschiedene Aspekte der Pädagogik und Didaktik wie des Informatikunterrichts, Ganztagsangeboten und des fächerübergreifenden Unterrichts berücksichtigt. Im Rahmen der Arbeit wurde mit dem PiaPi ein Prototyp entwickelt, der für die Weiterentwicklung zum Projekt im Rahmen des Informatikunterrichts an Schulen nachweislich geeignet ist.:Inhalt Abkürzungsverzeichnis iv Abbildungsverzeichnis v Tabellenverzeichnis vii Anhang viii 1. Einleitung 1 1.1 Idee und Motivation 1 1.2 Ziel der Arbeit 2 1.3 Forschungsstand 3 1.4 Aufbau der Arbeit 4 2. Wissenschaftliche Grundlagen 5 2.1 Technische und methodische Grundlagen 5 2.1.1 Research through Design 6 2.1.2 Prototypen 8 2.1.3 Mini-Computer Raspberry Pi 12 2.2 Theoretische Grundlagen 14 2.2.1 Spiel und Vergnügen in der Mensch-Computer-Interaktion 15 2.2.2 Visualisierung von Ton 18 2.2.3 Informatik und Schule 23 Informatik an (Grund-)Schulen 23 Fächerübergreifender Unterricht 26 Ganztagsangebote und Ganztagsschulen 27 2.3 Verwandte Projekte 29 2.3.1 Der LEGO-Zug im Museum 29 2.3.2 Das LEGO-Mikroskop - “Hands on” by Yuksel Temiz 33 3. Konzept und Umsetzung des Prototypen 36 3.1 Idee und Entwurf 36 3.1.1 Die Idee 36 3.1.2 Variante A 37 3.1.3 Variante B 41 3.2 Elektrotechnische Umsetzung 43 3.2.1 Aufbau und Funktionen 43 3.2.2 Bauteile des PiaPi 45 Der Raspberry Pi 45 Die Tasten 48 Die LEDs 51 Die IMU 52 3.3 Entwicklung und Implementierung der Software 53 3.3.1 Überblick des Software-Ablaufs 53 3.3.2 Initialisierung und Klassen 54 3.3.3 Vom Tastendruck zum Ton 57 3.3.4 Es werde Licht 59 4. Methodik der Nutzer:innenstudie 60 4.1. Usability Testing 60 4.2 Qualitative Interviews 63 5. Nutzer:innenstudie zum PiaPi 67 5.1 Testung und qualitatives Interview 67 5.2 Auswertung der Studie 71 5.2.1 Interaktion mit dem PiaPi 72 5.2.2 Emotionen der Tester:innen 73 5.2.3 Bewertung der Designbesonderheiten 75 5.2.4 Anwendungsvorschläge und Bewertung der Konzeptidee 76 5.3. Herausforderungen und Erkenntnisse 78 6. Fazit und Ausblick 81 Literaturverzeichnis 86 Selbstständigkeitserklärung 91

There are many forms of speed control for an autonomous ground vehicle currently in development. Most use a simple PID controller to achieve a speed specified by a higher-level motion planning algorithm. Simple controllers may not provide a desired acceleration profile for a ground vehicle. Also, without extensive tuning the PID controller may cause excessive speed overshoot and oscillation. This paper examines an approach that was designed to allow a greater degree of control while reducing the computing load on the motion planning software. The SpAM+PI (Speed-based Acceleration Map + Proportional Integral controller) algorithm outlined in this paper uses three inputs: current velocity, desired velocity and desired maximum acceleration, to determine throttle and brake commands that will allow the vehicle to achieve its correct speed. Because this algorithm resides on an external controller it does not add to the computational load of the motion planning computer. Also, with only two inputs that are needed only when there is a change in desired speed or maximum desired acceleration, network traffic between the computers can be greatly reduced. The algorithm uses splines to smoothly plan a speed profile from the vehicle's current speed to its desired speed. It then uses a lookup table to determine the correct pedal position (throttle or brake) using the current vehicle speed and a desired instantaneous acceleration that was determined in the splining step of the algorithm. Once the pedal position is determined a PI controller is used to minimize error in the system. The SpAM+PI approach is a novel approach to the speed control of an autonomous vehicle. This academic experiment is tested using Odin, Team Victor Tango's entry into the 2007 DARPA Urban Challenge which won 3rd place and a $500,000 prize. The evaluation of the algorithm exposed both strengths and weaknesses that guide the next step in the development of a speed control algorithm. / Master of Science

This thesis presents scaling methods for sizing a prototype micro prismatic revolute (PR) manipulator actuated by permanent magnet (PM) direct current (d.c.) gearmotors. Dimensional analysis was the principle tool used in this investigation, and addressed the problems of scaling a trajectory planner, control law, and gearmotors that exhibit internal nonlinear friction. Similitude methods were used to develop a scaleable two degree-of-freedom trajectory planner from a third order polynomial. Scaling laws were developed from Buckingham's Pi theorem to facilitate the selection process of gearmotors. Nondimensional, nonlinear, differential equations were developed to describe viscous, Coulomb and static friction in comparative PM d.c. motors. From the insights gained through dimensional analysis, a scaleable controller based on the computed torque method was developed and implemented with a cubic trajectory planner. Model and prototype PR manipulator systems were simulated using a hybrid Matlab/Simulink simulation scheme. Experimental systems were constructed with dissimilar model and prototype motors. Control was provided by an AT class PC equipped with 12-bit A/D, D/A cards operating at a sample rate of 100 Hz. The control algorithm was written in Borland 3.1 C for DOS. Results from the experimental testing showed excellent agreement between the test and simulated data and verified the viability of the scaling laws. The techniques presented in this thesis are expected to be applicable to any application that involves scaling PM d.c. micro gearmotors that have significant internal friction terms. These simple, practical tools should be especially beneficial to designers of micro robotic systems. / Master of Science

The history of Software-Defined Radios (SDRs) epitomizes innovation in wireless communication. Initially serving military needs, SDRs swiftly transitioned to civilian applications, revolutionizing communication. This thesis explores SDR applications such as Spectrum Scanning Systems, Contraband Cellphone Detection, and Human Activity Recognition via Wi-Fi signals. SDRs empower Spectrum Scanning Systems to monitor and analyze radio frequencies, optimizing spectrum allocation for seamless wireless communication. In Contraband Cellphone Detection, SDRs identify unauthorized signals in restricted areas, bolstering security efforts by thwarting illicit cellphone usage. Human Activity Recognition utilizes Raspberry Pi 3B+ to track movement patterns via Wi-Fi signals, offering insights across various sectors. Additionally, the thesis conducts a comparative analysis of Wi-Fi-based Human Activity Recognition and Radar for accuracy assessment. SDRs continue to drive innovation, enhancing wireless communication and security in diverse domains, from defense to healthcare and beyond.

Constrained optimization is pivotal in deep learning for enforcing predefined operational, ethical, and regulatory constraints in models. This thesis explores iterative gradient-based methods for solving constrained optimization. These methods are preferred for their adaptability in the non-convex, stochastic, and high-dimensional learning landscape of neural networks. Lagrangian methods, which integrate constraints directly into the objective function using Lagrange multipliers, are particularly effective. This formulation facilitates a unified approach to handle both the model's performance and its adherence to constraints simultaneously. However, the min-max problems inherent in Lagrangian frameworks often suffer from unstable oscillatory dynamics. Solving them using traditional gradient descent-ascent methods potentially leads to suboptimal convergence and prolonged training times. To mitigate these issues, this thesis introduces the νPI algorithm, a novel algorithm that integrates Proportional-Integral (PI) control theory into the optimization of Lagrange multipliers. The νPI algorithm is designed to stabilize the optimization process by adjusting the updates to the Lagrange multipliers based on the current and past constraint violations, effectively dampening the oscillations typically observed with gradient descent-ascent methods. Experimental validations across various deep learning applications demonstrate that νPI not only stabilizes the learning process but also accelerates convergence to optimal or near-optimal solutions, outperforming traditional methods in terms of reliability and speed. Through its application, νPI proves to be a robust tool in managing the dynamics of constrained optimization, enhancing the capability of deep learning models to meet stringent external constraints. / L’optimisation contrainte est essentielle dans l’apprentissage profond pour faire respecter les contraintes opérationnelles, éthiques et réglementaires prédéfinies dans les modèles. Cette thèse explore les méthodes itératives basées sur le gradient pour résoudre l’optimisation contrainte. Ces méthodes sont privilégiées pour leur adaptabilité dans le paysage d’apprentissage non convexe, stochastique et de haute dimension des réseaux neuronaux. Les méthodes lagrangiennes, qui intègrent directement les contraintes dans la fonction objectif à l’aide de multiplicateurs de Lagrange, sont particulièrement efficaces. Cette formulation facilite une approche unifiée pour gérer simultanément la performance du modèle et son adhérence aux contraintes. Cependant, les problèmes min-max inhérents aux cadres lagrangiens souffrent souvent de dynamiques oscillatoires instables. Leur résolution à l’aide des méthodes traditionnelles de descente de gradient-ascent peut potentiellement conduire à une convergence sous-optimale et à des temps de formation prolongés. Pour atténuer ces problèmes, cette thèse introduit l’algorithme νPI, un nouvel algorithme qui intègre la théorie de contrôle Proportionnel-Intégral (PI) dans l’optimisation des multiplicateurs de Lagrange. L’algorithme νPI est conçu pour stabiliser le processus d’optimisation en ajustant les mises à jour des multiplicateurs de Lagrange en fonction des violations de contraintes actuelles et passées, atténuant efficacement les oscillations typiquement observées avec les méthodes de descente de gradient-ascent. Les validations expérimentales à travers diverses applications d’apprentissage profond démontrent que νPI stabilise non seulement le processus d’apprentissage mais accélère également la convergence vers des solutions optimales ou quasi-optimales, surpassant les méthodes traditionnelles en termes de fiabilité et de vitesse. Grâce à son application, νPI se révèle être un outil robuste pour gérer les dynamiques de l’optimisation contrainte, améliorant la capacité des modèles d’apprentissage profond à répondre à des contraintes externes strictes.

University of Macau / Faculty of Social Sciences and Humanities / Department of Chinese