[kand] Sähkötekniikan korkeakoulu / ELEC

Permanent URI for this collectionhttps://aaltodoc.aalto.fi/handle/123456789/13

Browse

Recent Submissions

Now showing 1 - 20 of 3731
  • Kovarianssimatriisin estimointimenetelmien vertailu ja soveltaminen portfolion optimoinnissa
    (2025-05-16) Nissinen, Paavo
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    Portfolionhallinta ja riskinhallinta ovat keskeisiä teemoja rahoitustieteissä. Ei riitä, että sijoittajat ymmärtävät yksittäisten omaisuuserien ominaisuuksia vaan heidän on myös tiedostettava miten eri sijoituskohteet vaikuttavat toisiinsa. Yksi keskeinen työkalu sijoitusten välisten riippuvuuksien arvioimisessa on kovarianssimatriisi. Kovarianssimatriisin estimointi on olennainen osa portfolion optimointia, sillä tämän avulla voidaan laskea optimaalisen minimivarianssiportfolion omaisuuserien painotukset. Tässä kandidaatintyössä vertaillaan kahta eri menetelmää osakeportfolion painojen määrittämisessä. Ensimmäinen menetelmä on perinteinen otoskovarianssimatriisi, joka lasketaan osakkeiden historiallisista tuotoista. Toinen menetelmä hyödyntää faktorimalleihin perustuvaa estimointia, joissa osakkeiden tuottojen välisiä riippuvuuksia selitetään systemaattisilla riskitekijöillä. Faktorimallin avulla kovarianssimatriisin estimaatti muodostetaan regressioanalyysin kautta, jossa mallinnetaan, kuinka herkästi kunkin osakkeen tuotto reagoi valittuihin faktoreihin. Faktorimalliin perustuvan kovarianssimatriisin estimoinnissa hyödynnettiin Fama-Frenchin viiden faktorin mallia. Kandidaatintyön empiirinen analyysi perustuu OMX Helsinki 25-indeksin perusteella valittuihin osakkeisiin, joiden päivittäisiä tuottoja tarkastellaan yhdeksän vuoden ajalta. Molempia kovarianssimatriisin estimointimenetelmiä hyödyntäen laskettiin painotukset kahdelle eri portfoliolle ja niiden suorituskykyä mitattiin portfolion kokonaistuoton, keskihajonnan ja Sharpen luvun avulla. Tulokset osoittavat, että faktorimalliin perustuva kovarianssimatriisi tuotti johdonmukaisesti matalamman volatiliteetin verrattuna otoskovarianssimatriisiin melkein kaikilla tarkastelluilla vuosilla. Lisäksi faktorimallilla saavutettiin parempi Sharpen luku useimpina vuosina, mikä viittaa siihen, että faktorimallia hyödyntävä portfolio johtaisi tehokkaampaan riskikorjattuun tuottoon.
  • Offline Reinforcement Learning for Autonomous Driving Applications
    (2025-05-31) Lund, Tyko
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    This thesis presents the fundamental systems of autonomous vehicles and the decisionmaking challenges that autonomous vehicles face. Online and offline reinforcement learning are presented on a high abstraction level and they are compared in the context of autonomous driving. The thesis provides a comprehensive motivation to explain why offline reinforcement learning could be especially suitable for autonomous driving. After presenting the necessary preliminary information, the thesis compares two different studies that employed offline reinforcement learning using real driving data. The results of the studies are analyzed to provide an overview of the current state of offline reinforcement learning for autonomous driving. Although the studies demonstrated the successful use of real driving data with offline reinforcement learning methods, they highlighted certain issues. The studies showed that offline reinforcement learning algorithms struggle in learning from human demonstrations, and noted that offline reinforcement learning algorithms might face problems learning from real driving data where collisions are rare. The studies found that data augmentation improved learning results. Data augmentation methods are a relevant area for future research. Both studies used a simplified driving scenario with limited domain, and their results do not translate well to the goal of level 5 autonomy. Considering the goal of level 5 autonomy, a relevant direction for future research is the scalability of different offline reinforcement learning methods and the relationship between existing data and required training data.
  • Esteiden kiertäminen automaattitrukeilla
    (2025-05-30) Kössi, Antti-Pekka
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    Teknologioita paikallaan pysyvien esteiden kiertämiseen on olemassa, mutta monet nykyisistä automaattitrukeista (AGV, Automated Guided Vehicle) eivät niitä hyödynnä. Siksi tässä tutkielmassa pyritään perehtymään tarjolla oleviin algoritmeihin sekä löytämään niiden käyttöön ottamisen kannalta oleellisia tekijöitä. Tutkielma toteutettiin kirjallisuuskatsauksena, jossa perehdyttiin esteiden kiertämiseen liittyvään reitinsuunnitteluun yleisesti sekä esitellään tarkemmin yleisesti käytetyistä metodeista Dynamic Window Approach (DWA), Timed Elastic Band (TEB) sekä Model Predictive Control (MPC). Tutkielmassa käsiteltiin myös automaattitrukkistandardia ISO 3691-4:2023, josta tunnistettiin esteiden kiertämisen kannalta oleellisia turvallisuusvaatimuksia. Lisäksi tutkielmassa tarkasteltiin, miten automaattitrukkien paikannus- ja ohjausjärjestelmät liittyvät esteiden kiertämiseen. Teknologioiden käytäntöön viemisen tueksi esitettiin pääpiirteinen suunnitelma ja esteenkiertoalgoritmin valinnassa käytettäviä vertailukriteereitä.
  • Event Triggered State Estimation for Constrained Nonlinear Systems
    (2025-05-30) Kokkonen, Kimmo
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    Työ tarkastelee järjestelmiä, joissa laitteita hallitaan keskuksesta käsin ja laitteet kommunikoivat keskuksen kanssa yhteistä verkkoa pitkin. Nämä laitteet mittaavat reaalimaailman suureita, ja mittausten perusteella keskus estimoi mitattavia suureita ja näin hallitsee laitteita reaaliajassa. Työ keskittyy erityisesti sellaisiin järjestelmiin, joissa itse laitteet sekä verkko ovat rajoitettuja, mikä tarkoittaa, että laitteilla on rajoitettu virrankäyttö ja laskentatehokkuus, ja verkolla rajoitettu kaistanleveys. Tämä rajoitteisuus tarkoittaa, että lähetettyä dataa täytyy saada vähennettyä, jotta mahdollistetaan verkon tasapainoinen toimintakyky. Datan vähentäminen verkossa nostaa esiin kaksi ongelmaa. Ensinnäkin, kuinka lähetettyä dataa voidaan vähentää laitteen päässä ja miten itse keskus mukautuu tähän rajoitettuun dataan. Työ keskittyy event-trigger (ET) -strategiaan, mikä mahdollistaa datan tehokkaan vähentämisen säilyttäen kuitenkin systeemin approksimoinnin tarkkuuden keskuksen päässä. Työn painopisteenä on erityisesti epälineaariset systeemit, koska todellisen elämän mallit ovat aina epälineaarisia. Työn ensimmäisessä osiossa keskitytään näytteenoton strategioihin, joilla datan lähetystä vähennetään laitteiden sensoreissa. Näistä esimerkkejä ovat Send-on-Delta (SOD) ja Innovation-based triggering (IBT). Työ osoittaa, että IBT on sopeutuvin ja intuitiivisin vaihtoehto. Työn toinen osa keskittyy filttereihin, eli laitteen tilaa estimoiviin ratkaisuihin. Työ esittelee ensiksi lineaarisen Kalman-suotimen, edeten kohti epälineaarista tapausta. Työssä osoitetaan, että ET-strategian implementointi murtaa Kalman-suotimen gaussisen olettamuksen, eivätkä näin ollen Kalman-suotimeen perustuvat ratkaisut voi taata maksimaalista tarkkuutta systeemin estimointiin. Osoittautuu, että näytteenottometodin valinta riippuu järjestelmän rajoitteista. IBT:tä ei voida valita sellaisille järjestelmille, joissa sekä sensorin virrankulutus ja verkon kaistanleveys ovat rajoitettuja. Optimaalisen epälineaarisen filtterin osalta työ tarjoaa ratkaisuksi ET particle filter (ETPF) -mallin, ja osoittaa ettei sillä ole epälineaarisia eikä gaussisia rajoituksia. ETPF:n osalta rajoittavana tekijänä on laskentateho. Jos järjestelmän rakenne on sellainen, että estimointi tapahtuu laitteissa itsessään ilman keskusta, laskentateho ei riitä ETPF:n implementointiin. Mitattavien todellisen elämän suureiden estimointi käyttäen ET-menetelmiä riippuu siis systeemin rakenteesta ja sen rajoitteista. Analysoimalla nämä oikein voidaan valita tehokkain järjestelmälle sopiva menetelmä. Tietyissä tilanteissa on valittava, mitkä ominaisuudet estimaattorin suorituskyvylle ovat tärkeimmät, ja mistä voidaan tinkiä.
  • Audio watermarking using deep learning
    (2025-05-30) Mikkola, Samuel
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    This literature study explores deep learning-based audio watermarking techniques in response to growing concerns over synthetic media authenticity. Traditional methods often lack robustness and scalability. This study evaluates five state-of-the-art models: RobustDNN, WavMark, MaskMark, AudioSeal, and SilentCipher based on architec- ture, robustness features, watermark localization, training and loss functions. Results show that while no method excels universally, SilentCipher demonstrates superior robustness and real-time performance, whereas MaskMark offers the highest capacity. AudioSeal balances imperceptibility and detection capabilities. The comparative anal- ysis highlights trade-offs among methods and suggests future directions for improving adversarial resilience, real-time deployment, and generalization across diverse audio environments.
  • Exploring Gaps in Multi-objective Reinforcement Learning Theory and Algorithms
    (2025-05-30) Cajander, Niclas
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    Multi-objective reinforcement learning (MORL) addresses decision-making problems where multiple, possibly conflicting objectives must be optimized simultaneously. Unlike single-objective reinforcement learning, MORL employs vector-valued rewards to capture the trade-offs between objectives, necessitating advanced theoretical frameworks and algorithms. This thesis explores the gaps in MORL theory and algorithms for learning problems, focusing on fully observable, single-task, single agent Multi-objective Markov Decision Processes (MOMDPs) and the challenges posed by different utility functions and optimality criteria. The thesis examines key theoretical aspects, including solution sets such as the Pareto Front (PF) and Convex Coverage Set (CCS), and their dependence on utility functions and policy types. The thesis discusses single-policy algorithms and highlights their limitations, particularly for the understudiedESRcriterion, whichis crucialforscenarios requiring optimization over single episodes, such as medical treatments. The thesis also reviews multi-policy algorithms, categorizing them into outer-loop and inner-loop methods, and discusses their applicability to real-world problems. Crucial aspects include sample-efficiency, scalability with respect to the number of objectives, and scalability of state-action spaces. It is clear that the ESR criterion is still understudied, as very few algorithms have been proposed. Moreover, as no solution set has been defined for ESR, there only exist single-policy algorithms for the criterion. There exist many multi-policy methods with clear strengths, but they are often limited by some other factor.
  • Matkapuhelinverkkojen aiheuttama hiilijalanjälki ja sen vähentäminen
    (2025-05-08) Hildén, Santeri
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    Tutkitaan matkapuhelinverkkojen aiheuttamaa hiilijalanjälkeä ja keinoja sen vähentämiseen. Pyritään selvittämään suurimmat päästölähteet ja arvioimaan teknologisia ja rakenteellisia ratkaisuja, joilla verkkojen ympäristövaikutuksia voidaan pienentää. Työ toteutettiin kirjallisuuskatsauksena, jossa hyödynnettiin useita kansainvälisiä tutkimuksia ja raportteja. Pääasialliseksi teemaksi asetettiin tukiasemien energiankulutus, jonka todettiin kattavan jopa 80 % koko matkapuhelinverkkojen energiankulutuksesta. Huomattiin myös, että tukiasemien valmistus, kuljetus ja huolto aiheuttavat merkittäviä, mutta vähemmän tutkittuja päästöjä. Energiankulutuksen vähentämiseksi on kehitetty useita ratkaisuja, kuten tukiasemien dynaaminen lepotila säätely, älykkäät virransäästöalgoritmit ja energiatehokkaat radioteknologiat. Näillä voidaan saavuttaa suuria säästöjä energiankulutuksessa. Uusiutuvien energialähteiden, kuten aurinkoenergian, käyttö nähdään lisämahdollisuutena päästöjen vähentämiseen erityisesti kehittyvillä alueilla, mutta nykytilanteessa niiden käyttöönottoa ei pidetä laajasti taloudellisesti kannattavana ilman ulkoisia ohjausmekanismeja, kuten hiiliverotusta. Tulevaisuuden teknologioiden, kuten 6G:n ja generatiivisen tekoälyn, avulla mahdollistetaan entistä tarkempi verkonhallinta ja energiankäytön optimointi. Erityisesti tekoälypohjaisten järjestelmien, kuten DETA:n, avulla voidaan merkittävästi vähentää verkkojen päästöjä. Voidaan todeta, että teknologisia keinoja päästöjen vähentämiseen on olemassa, mutta niiden vaikutuksen todettiin riippuvan taloudellisista kannustimista ja poliittisesta ohjauksesta. Kestävän mobiiliverkon saavuttaminen edellyttää kokonaisvaltaista lähestymistapaa, jossa yhdistetään teknologia, talous ja ympäristöpolitiikka.
  • A systematic review of attention management in multi-device environments
    (2025-05-30) Holopainen, Samuli
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    As devices, such as mobile phones, laptops and tablets, have become more common in our day-to-day lives, environments consisting of multiple devices have emerged as increasingly relevant. These environments allow for improved functionality and computing, however, their level of interaction continuity varies. Recent studies have examined the effects of these environments on attention, highlighting their challenges, and examining different interaction techniques to mitigate the negative effects on attention. This systematic review synthesizes the current studies on attention management in multi-device environments with its primary objectives as providing a structured, comprehensive synthesis of the key challenges identified in existing literature and evaluating the strategies proposed to address them. The review was conducted following the PRISMA 2020 statement. For the purpose of this thesis, a search across 7 different databases was conducted to identify relevant studies. To be included the studies had to be focused on managing user’s attention in multi-device environments, addressing issues, such as notifications, device-switching and interruptions. From the database search, 17 studies meeting the inclusion criteria were included, and by searching through their reference lists, an additional 3 were identified. Two of these studies were later excluded from the synthesis, the first since it was not in English and the second for its narrowness. The synthesis highlighted two different key challenges for attention management: notifications and device-switching. These challenges were addressed by various strategies, such as notification management systems, feedforward and feedback mechanisms, and seamless switching strategies. While several approaches were identified as capable in reducing the negative effects on attention, the usability of these strategies in real-world applications remains limited, primarily due to the diversity of devices and user contexts. Overall, the findings suggest that managing attention in multi-device environments requires adaptive and personalized solutions that account for the complexity of different usage patterns. Future research could further investigate how personalization can improve task continuity in device switching.
  • Data-Driven Feedback Linearization of Nonlinear Systems
    (2025-05-30) Ahlstedt, Zakarias
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    The nonlinear dynamics of many real-world systems pose significant challenges in control engineering. Consequently, these systems are often linearized to simplify analysis and control design. The feedback linearization technique is a common choice, allowing linear control tools to be used with nonlinear systems by mathematical cancellation of nonlinear terms. Linearizing real-world systems, however, is often problematic due to challenges associated with modeling of complex systems. In recent years, data-driven feedback linearization techniques have emerged as a promising solution, leveraging system data in the place of a system model. This thesis provides an overview of data-driven feedback linearization, examining the underlying motivations for linearization, specifically feedback linearization, and the data-driven alternative. The main focus is on reviewing state-of-the-art data-driven feedback linearization techniques through a literature review and case studies to examine the performance of different methodologies and sampling strategies. The case studies show that data-driven methods provide a compelling alternative to the model-based solution, analyzing prediction and domain performance. A comparison of different input signals underscores the importance of persistently exciting inputs. Moreover, the scalability issues for higher-dimensional systems are shown to be significant.
  • Autonomous Vehicle Sensor Weather Performance
    (2025-05-30) Pesonen, Kari
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    Reliable perception in adverse weather remains a major challenge for autonomous vehicles. This literature review examines how rain, snow, fog, and other conditions affect the performance of sensor types, such as LiDAR, radar, cameras, and ultrasonic sensors. Each sensor type provides strengths, but adverse weather conditions amplify their weaknesses and degrade overall performance. To address these limitations, the review highlights sensor fusion, where multiple sensor types are combined to improve robustness. The thesis emphasizes real-world experiments and studies evaluating sensor types and variants in varying conditions. The findings show that no single sensor type is sufficient in all environments, and that both the selection of sensors and data processing methods significantly impact fusion performance. Understanding the role of sensor selection is essential for safe and reliable autonomous navigation in weather conditions.
  • A Comparative Study of SVM and Random Forest Classifiers in Sentiment Analysis
    (2025-05-09) Kyöstiö, Kalle
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    Machine learning models have long been applied to sentiment analysis, which aims to determine the underlying sentiment expressed in data. Usually, the aim is to classify the text as negative or positive. This two-class classification problem is known as binary classification. Sentiment analysis can be used in customer feedback analysis or identifying trends, for example. This thesis starts with a rigorous explanation of Support Vector Machines (SVM) and Random Forest (RF) classifiers. After, small experiment where these models are used to classify movie reviews is also conducted to compare the models' performance. Support vector machines, like their name indicates, use support vectors to fit an optimal hyperplane between classes. Maximizing the margin between classes creates an optimization problem. Additionally, the thesis discusses the kernel trick, which enables the use of nonlinear data with SVMs. Random forests, on the other hand, use an ensemble of decision trees, which work by splitting data based on specific features to make the data more homogeneous. The decision trees are trained on bootstrapped datasets that decorrelate the trees, and the final prediction is then decided by a majority vote. The results indicate that SVMs performed modestly better in sentiment analysis, and their training time was significantly lower. The results also show the importance of data preprocessing, as modifications of preprocessing steps can have considerable effects on the performance of both models. However, the study on the hyperparameters of the models remained limited, which could be an avenue for future research.
  • Causality in Intuitive Physics
    (2025-05-30) Sarnes, Kaarle
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    Humans have an impressive ability to understand physical systems and perceive the interactions that occur within them. This ability is called intuitive physics, and it consists of several intuitive theories about the world. It has been suggested that many cognitive processes operate by combining intuitive theories with causal reasoning. Studying how people understand their physical environment may offer solutions for developing more intelligent AI that can also operate in physical environments. This bachelor9s thesis is a literature review that explores the role of causality in intuitive physics. This thesis focuses in particular on the counterfactual simulation model and the Intuitive Physics Engine. According to the counterfactual simulation model, causation is based on the execution of counterfactual (i.e., alternative) simulations. The Intuitive Physics Engine is a theory suggesting that people simulate physical situations in their minds in a way similar to how physics simulations are run on computers. The conclusion of this thesis is that by combining the counterfactual simulation model with the intuitive physics engine, humans are able to accurately perceive causal relationships in their environment. However, people do not always use this combined model; they also hold incorrect beliefs about physical situations and often rely on imprecise, heuristic methods. Causality is part of human physical reasoning, but it is not the only method people use to understand their surroundings. The human approach to understanding the physical world could be a valuable starting point for developing AI, as intuitive physics may offer a good compromise between accuracy and computational demand.
  • Colour in Human-Centred Interface Design — A Systematic Literature Review
    (2025-05-16) Niemi, Helmi
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    Colour is an essential aspect of user interface (UI) design. UI designers should be aware of the psychological components of colour and how colour can be used to create user friendly interfaces. Research about effects and use of colour in UI contexts is scattered and wide-ranging. A literature review is necessary for compiling effective practices and finding research gaps in the field. The objective of this research is to collect and organise information about the use of colour and its effects on users in the fields of UI/UX design and Human-Computer Interaction through literature in the fields in the time frame of 2010-2025. This period was chosen to capture the widespread adoption of smartphones. The research is limited to mobile and web interfaces, due to their common use in everyday digital interactions, making their colour design especially important. This thesis is a systematic literature review, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Relevant literature was sourced from the Google Scholar and Scopus databases. The quality of the articles was assessed with the Publication Forum (JUFO) classification. Publications that did not meet the baseline quality criteria were excluded. The contributing studies came from three continents and nine countries, with the highest number of studies coming from China (10) followed by the United States (7). As key findings, six thematic syntheses emerged, each addressing the relationship between colour and one of the following: emotion, attention, associations, performance, culture, and accessibility. The results indicated that colour significantly affect users’ emotions, focus, and performance. Individuals of the same cultural background tend to have similar colour interpretations, while notable cultural differences in preference and associations exist. Low-saturation and low-contrast colour schemes promote visual comfort and task performance, while high-saturation and high-contrast colours like red attract attention. Some colour-emotion and colour-concept associations are consistently stronger than others. Colour blindness also affects these colour associations. Future research should further examine how culture and colour blindness affect colour perception.
  • Lagrangian and Hamiltonian neural networks in model learning
    (2025-05-30) Oja, Maria
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    This thesis researches physics-informed machine learning models, Lagrangian neural networks (LNNs), and Hamiltonian neural networks (HNNs) in model learning. Physics-informed neural networks (PINNs) offer a solution where physical laws are embedded into neural networks. This helps to avoid mathematically preventable errors, and often requires less data. However, models such as LNN and HNN can be challenging to expand to non-conservative systems because they assume the total energy to be conserved. This thesis explores LNN- and HNN- based modeling approaches for conservative pendulum systems and a non-conservative damped mass-spring. The reviewed implementations include a trained model using video input, two robotic systems with different motion constraints, and current- and pressure-driven systems. This thesis aims to answer the following three research questions: How are LNNs and HNNs implemented in model learning? How can they be expanded to model non-conservative systems? What are the possible industrial applications? The examination is executed as a literary review. The results indicate that in nearly conservative systems, HNNs demonstrate superior predictive performance. LNNs are relatively simple to expand to non-conservative systems through generalized formulations. However, integrating port-Hamiltonian physics, HNNs can also be extended to model non-conservative systems. Overall, in most examined cases, PINNs demonstrate more accurate performance compared to data-driven methods. Potential future applications include robotics, computer vision, and other autonomous systems, like ones to do with current- and pressure dynamics.
  • Trust in Speech-Based Technologies in relation to risk
    (2025-05-30) Nyström, Sofie
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    As speech-based technologies become increasingly integrated into daily life, trust plays a critical role in their adoption and effective use. This thesis examines how trust in speech-based technologies differs based on the risk level of the scenarios. A systematic literature review identifies key factors influencing trust, including time, explanations, content credibility, perceived ease of use, perceived usefulness, privacy and security concerns, and anthropomorphic design features. The findings suggest that while some trust factors, such as content credibility and repeated exposure, consistently impact trust regardless of risk, others, such as privacy concerns, vary in importance depending on the perceived risk of the interaction. The study also highlights that users may rely less on trust when using speech-based technologies in low-risk contexts, whereas higher trust is essential for adoption in high-risk scenarios. Finally, the thesis discusses strategies for improving trust through ethical alignment, user education, and personalized system design. These insights contribute to the development of more trustworthy and user-centred speech-based technologies.
  • Industrial Human-Robot Collaboration
    (2025-05-30) Hakkarainen, Verneri
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    Human-robot collaboration can be a powerful tool when applied in industrial settings, such as manufacturing. The robot co-operates with the human on the same task and uses intention prediction and communication methods to make the collaboration effective. The robot can benefit from the same social cues that humans use when collaborating with other people. The robot predicts human actions by using human movement modeling, gaze, gestures and surroundings as a guide. Using human intention prediction techniques increases fluency and makes collaboration seem more natural. Additionally, effective human-robot collaboration requires communication between participants. The robot can use speech or visual communication like augmented reality, or many communication methods simultaneously. Effective communication enables the human and the robot to discuss crucial task information like a task completion plan that the robot can generate. The robot can create human centered task plans that reduce human physical effort and emphasize human skills. This thesis is a literature review that examines human modeling and intention prediction methods and communication methods. This thesis also reviews human-robot collaboration applied in industrial settings. The review found that human robot collaboration offers many benefits in industrial applications by combining human creativity and flexibility with robot efficiency. The robot can increase the throughput of manufacturing by decreasing work task completion times and reducing human fatigue. A major difficulty in human-robot collaboration is implementing a safe collaboration robot to use near humans. The study of this field is gaining interest, and many studies indicate that human-robot collaboration offers benefits in industrial applications, making the potential look promising.
  • Liitteettömät koodit ja ¾-konjektuuri
    (2025-05-30) Honda, Aiki
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    Tietokone käsittelee symboleja, binäärisinä lukuina. Symbolien yleisyys kuitenkin vaihtelee. Antamalla useammin ilmestyville symboleille lyhyemmät ja harvemmille pitemmät bittijonot, joita kutsutaan koodisanoiksi, symboleista muodostetusta bittijonosta tulee mahdollisesti lyhyempi. Koodisanojen vaihteleva pituus tekee dekoodauksesta monimutkaisempaa, mutta etuliitteetön koodi ratkaisee tämän; mikään koodisana ei ole minkään toisen koodisanan etuliite. Kyseessä on välitön koodi, jossa dekoodaus tapahtuu välittömästi jotakin symbolia vastaavan koodisanan esiinnyttyä, koska se ei voi olla minkään muun koodisanan etuliite. Etuliitteettömässä koodissa kun data jostain kohtaa korruptoituu, niitä seuraava ehjä data saattaa menettää yksiselitteisyyttä. Tarvitaan liitteetöntä koodia, jossa mikään koodisana ei ole minkään toisen koodisanan etuliite eikä jälkiliite. Tämä ei ole yhtä tehokas kuin etuliitteetön koodi, mutta kestää paremmin datakorruptiota. Liitteettömästä koodista ja sen matemaattisista ominaisuuksista tunnetaan suhteellisen vähän. Tämä kirjoitus käsittelee liitteettömän koodin ominaisuuksia, siihen liittyviä ratkaisemattomia ongelmia kuten ¾-konjektuuria, sekä aiheen ymmärtämiseen tarvittavia taustatietoja ja peruskäsitteitä.
  • Localization of Autonomous Vehicles
    (2025-05-30) Lee, Benjamin
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    Autonomous vehicles (AVs) are equipped with a range of sensors and methodologies to facilitate the perception of their place within an environment. The combination of different sensor types, or sensor fusion, is a pivotal aspect of how AVs localize. The utilization of these sensors in localization can depend on the requirements of the AV as well as the environment it works in. This thesis is a literature review of various sensor fusion and localization methods used in AV localization. The objective of this study is to explore the various sensor fusion and localization methods employed in different AV environments. The review of this thesis is divided into three different categories corresponding to a distinct AV environment. The three categories of environments are: urban, unstructured and controlled. AVs working in urban environments were observed to utilize almost exclusively LIDAR and cameras as their main sensor. This was due to the necessity for high precision localization. Map matching and SLAM were utilized as the localization methods. In unstructured environments, robust multi-sensor fusion was utilized due to the low-feature qualities of unstructured environments. The relative simplicity of controlled environments led to the adoption of simpler localization methods and sensors. Similarly, generally higher localization accuracy was achieved. Finally, LIDAR and camera dominance in high-precision localization was discovered to be a main theme across all categories. Similarly SLAM and map matching were regarded as standard localization methods. ML was utilized to enhance localization. Additionally, GNSS, IMU, radar and other odometry were utilized as auxiliary sensors in the localization process.
  • Benefits of transferring from IEC 61131-3 toIEC 61499
    (3025-05-30) Aalto, Jaakko
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    Automation trends are shifting towards distributed control systems to enable more cost-effective production via system reconfigurability and reusability. Shorter product life cycles and a higher product variety are driving the shift. This thesis studies two prominent standards in the field of industrial automation, IEC 61131-3 and IEC 61499, and focuses on methods and benefits of transitioning between them. The newer standard IEC 61499 was created as a framework for designing and implementing distributed control systems. However it has not been adopted by the automation industry as its predecessor IEC 61131-3 dominates the industry, extensively deployed across both small- and large-scale systems. Despite its preva- lence, IEC 61131-3 is argued to be suboptimal for large and complex systems. Therefore researching transition methods to IEC 61499 and the resulting benefits is of significant relevance. The thesis comprises of a literature review and an experimental part in which a simple IEC 61131-3 system is redesigned in IEC 61499. The system comprises of three programmable logic computers and mostly pneumatic actuators. The programs are implemented in Step7 as sequential flow charts. The experiment uses IEC 61499 composite function blocks to redesign the IEC 61131-3 programs according to IEC 61499. The redesigned IEC 61499 function blocks are simulated to ensure similar behavior and successful transformation. The programs are compared in order to evaluate achieved benefits, which include clear separation of control logic and hardware, event-driven execution and higher system level abstraction. The experimented redesign approach requires further development as it is com- pletely manual. However, it proves redesigning systems is possible with the proposed approach. The achieved benefits require further studies on large-scale industrial systems to evaluate scalability and cost-effectiveness of the transition process.
  • Vendor Agnostic Module Type Package Integration
    (2025-05-29) Järvinen, Taavi
    Sähkötekniikan korkeakoulu | Bachelor's thesis
    The process automation industry is constantly driven by the increasing demand for a faster time-to-market, flexible production and the ability to deliver customized products. These demands are forcing a shift in the design and operation of process automation: several companies are abandoning traditional, monolithic configurations for a move toward a modular process architecture. The Module Type Package (MTP) concept provides a standardized option for modular integration of process equipment into centralized systems. By ensuring consistent formatting of module descriptions in data exchange, the Module Type Package concept enables vendor agnostic module integration and provides plug-and-produce functionalities. The purpose of this thesis is to examine and evaluate the Module Type Package concept. Using relevant literature, it analyzes the Module Type Package concept from technical, organizational and economic perspectives. By clarifying both the advantages and limitations of the concept, this thesis aims to contribute to widespread deployment of the Module Type Package concept. The study found that the Module Type Package concept provides five key benefits. These benefits are vendor agnostic module integration, plug-and-produce features, shorter time-to-market, reduced costs, and flexible production. These allow companies to combine and quickly deploy devices from different manufacturers, enabling rapid adaptation to customer demands. Limitations to widespread adoption of the Module Type Package remain. The limitations discussed by this thesis are inconsistent conformance to Module Type Package guidelines, security risks in communication protocols and organizational consideration linked to displacement. Additionally, this thesis proposes a virtual method for evaluating the suitability of the Module Type Package standards before physical deployment. The study found that successful implementation depends on full compliance with the standard, standardized communication through OPC UA, and virtual testing to determine suitability of the Module Type Package to processes. Future development of the Module Type Package concept depends heavily on cooperation of vendors.