Dissertations / Theses on the topic 'Online Model Adaptive'

A distributed online curriculum and courseware (DONC2) development model is proposed in this study. Collaborative courseware development teams which may work in distributed academic or private institutions who need to develop higher quality, reduced cost, on time products are the users of DONC2 development model. The related features from the disciplines of instructional design and software engineering were combined and concepts like usability, especially in terms of formative and summative evaluation, interoperability and reusability were integrated into the model. The research is conducted as a collective case study, including four cases with distinctive characteristics to reveal the several practices in online curriculum and courseware development work. The DONC2 development model was proposed using the results gathered from the investigated cases and a literature survey. The model uses the iterative incremental and agile software development approaches in order to overcome the disadvantages of other linear system development approaches. This enables building releasable products in short time periods with increased quality. Furthermore, continuous communication, evaluation and feedback as well as good project management and readiness to adapt to changes are integrated as the essential characteristics. DONC2 development model is different than previous linear and non-adaptive models in all of these aspects.

Today the majority of the driver assistance systems are rule-basedcontrol systems that help the driver control the truck. But driversare looking for something more personal and exible that can controlthe truck in a human way with their own preferences. Machine learningand articial intelligence can help achieve this aim. In this studyArticial Neural Networks are used to model the driver steering behaviourin the Scania Lane Keeping Assist. Based on this, trajectoryplanning and steering wheel torque response are modelled to t thedriver preference. A model predictive controller can be used to maintainstate limitations and to weigh the two modelled driver preferencestogether. Due to the diculties in obtaining an internal plant modelfor the model predictive controller a variant of a PI-controller is addedfor integral action instead. The articial neural network also containsan online learning feature to further customize the t to the driverpreference over time.
Idag används till största del regelbaserade reglersystem förförarassistanssystem i lastbilar. Men lastbilschaufförer vill ha någotmer personligt och flexibelt, som kan styra lastbilen på ett mänskligtsätt med förarens egna preferenser. Maskininlärning och artificiell intelligenskan hjälpa till för att uppnå detta mål. I denna studie användsartificiella neurala nätverk för att modellera förarens styrbeteende genomScania Lane Keeping Assist. Med användning av detta modellerasförarens preferenser med avseende på placering på vägbanan och momentpåslag på ratten. En modell prediktiv kontroller kan användas föratt begränsa tillstånd och för att väga de två modellerade preferensernamot varann. Eftersom det var mycket svårt att ta fram den internaprocessmodellen som krävdes för regulatorn används istället en variantav en PI-kontroller för att styra lastbilen. De artificiella neuralanätverken kan också tillåtas att lära sig under körning för att anpassasig till förarens preferenser över tid.

Enrollment in online courses is increasing at a much higher rate than enrollment in on campus courses. Initially, online systems were developed by moving course content from in-class courses as is to an online platform. Later, Web 2.0 technology was implemented in order to improve students’ online engagement. These systems considered all students as one homogeneous group and ignored the fact that different students learn in different ways and at different speeds. Later, adaptive online learning systems were developed based on the assumption that if the instructional approach matches the student learning style, student performance and experience will improve. The use of these systems yielded mixed results because there is no agreement on what, how, and when to adapt instructions. The problem is that there is still a lack of empirical evidence about which online learning system’ design is the most effective, efficient, and engaging. There were two goals for this study. The first was to develop a new instructional theory and design model suitable for personalizing and adapting online learning. The first goal was achieved by developing student personalized, adaptive, and comprehensive e-learning spaces instructional theory and design model. This theory is based on finding the best fit among student characteristics, knowledge domain objectives, and technology used in delivering the online course. The second goal was to implement the newly developed theory and design model in an e-learning system prototype. This goal was achieved by developing and internally validating the e-learning system prototype by utilizing a panel of five instructional design experts. The Delphi method was used to solicit input from the expert panel in three rounds of validation. The validation process resulted in the experts’ consensus that the prototype incorporated the instructional theory and design model well and that this instructional theory holds the promise of increasing online learning courses’ effectiveness, efficiency, and student engagement.

Cette thèse s'intéresse à l'adaptation des lois de pilotage d'un avion de transport civil aux différentes incertitudes qui peuvent affecter sa dynamique. Le procédé de pilotage adaptatif est censé fonctionner en temps réel à bord de l'avion afin d'optimiser la performance boucle fermée en fonction des conditions dans lesquelles il évolue. Les incertitudes peuvent être liées à la méconnaissance des conditions de vol (par exemple la vitesse et l'altitude), à des non-linéarités aérodynamiques inconnues ou encore à la méconnaissance du pilote aux commandes. Les procédés adaptatifs qui répondent à ces problèmes se doivent d'être performants sur l'ensemble du domaine opérationnel de l'avion en présence de perturbations réalistes. D'autres contraintes spécifiques peuvent être ajoutées en fonction du contexte (par exemple des charges limites, la stabilité aéroélastique, etc.). Plusieurs méthodes adaptatives sont testées afin d'adapter le système aux larges incertitudes qui le composent. Elles associent en général un estimateur en ligne (aussi appelé loi de mise-à-jour) à une loi de commande structurée. La synthèse de ces deux éléments peut être réalisée simultanément pour les méthodes adaptatives dites " directes ", comme par exemple le Model Reference Adaptative Control qui utilise la stabilité au sens de Lyapounov. Mais cette synthèse peut aussi être découplée pour les méthodes adaptatives dites "indirectes", ce qui offre un large choix de techniques pour chaque élément (comme les Moindres Carrés pour l'estimation de paramètres physiques incertains et la synthèse sous forme LFR pour le correcteur). Le choix de la méthode dépend fortement du contexte applicatif et des nombreuses contraintes associées. Trois applications sont au cœur de ce mémoire. Elles traitent de l'ajustement de lois de guidage à un modèle pilote inconnu, du contrôle longitudinal de non-linéarités de l'avion, et de la mise au point de lois longitudinale et latérale de pilotage manuel qui s'adaptent à des conditions de vol inconnues. Des méthodes avancées d'analyse linéaire et non-linéaire (dérivées de la µ-analyse et d'algorithmes d'optimisation) sont aussi mises en place pour valider ces systèmes sophistiqués adaptatifs en temps réel. D'une façon générale, les méthodes adaptatives indirectes ont donné le plus de satisfaction. Leur performance est aussi bonne que celle des méthodes directes, mais le fait qu'elles estiment en ligne des paramètres physiques facilite la surveillance temps réel du procédé adaptatif et sa validation
This thesis deals with adapting flight control laws of a civil transport aircraft to various incertainties which can affect its behaviour. The adaptive flight control system is supposed to run in real time onboard the airplane so that its closed-loop performance is optimized with respect to the current conditions. These incertainties may be linked to unknown flight conditions (e.g. unknown airspeed and altitude), or unknown aerodynamics non-linearities or even unknown behaviour of the pilot in command. The adaptive schemes that are derived to answer these problems must be valid on the whole flight envelope with realistic disturbances but other additional contraints may exist depending on the context (e.g. loads limits, aeroelastic stability, etc.). To accommodate for large uncertainties on the system, adaptive methods are investigated. They usually combine an online estimator (also called an update law) with a structured flight control law. The synthesis of both elements may be simultaneous on 'direct' adaptive methods, e.g. on Model Reference Adaptive Control, using Lyapunov's stability theory. But it can also be decoupled on 'indirect' adaptive methods, giving a full spectrum of techniques for both elements (such as Least-Squares for estimating unknown physical parameters and the LFR framework for designing controllers). The choice of a specific method really depends on the application context and the related constraints.Three applications are the core of this report. They deal with adjusting guidance law to the pilot's unknown behaviour, controlling a longitudinal non-linearity, and providing manual longitudinal and lateral flight control laws which adapt to unknown flight conditions. Advanced linear and non-linear analysis techniques (based on µ-analysis or on optimization algorithms) are also applied to validated these sophisticated real-time adaptive systems. Results showed that indirect adaptive schemes were generally the most satisfactory. Their performance is similar to the one of direct schemes but as indirect methods provide physical parameter estimates, real-time monitoring and offline validation seem quite easier

This project was a step forward in improving the voltage profile of traditional low voltage distribution networks with high photovoltaic generation or high peak demand. As a practical and economical solution, the developed methods use a Dynamic Voltage Restorer or DVR, which is a series voltage compensator, for continuous and communication-less power quality enhancement. The placement of DVR in the network is optimised in order to minimise its power rating and cost. In addition, new approaches were developed for grid synchronisation and control of DVR which are integrated with the voltage quality improvement algorithm for stable operation.

Two important physical aspects that determine the performance of a running train are the total running resistance that acts on the whole train moving forward, and the available adhesion (utilizable wheel-rail-friction) for propulsion and breaking. Using the measured and available signals, online identification of the current running resistance and available adhesion and also prediction of future values for a distance ahead of the train, is desired. With the aim to enhance the precision of those calculations, this thesis investigates the potential of online identification and prediction utilizing the Extended Kalman Filter. The conclusions are that problems with observability and sensitivity arise, which result in a need for sophisticated methods to numerically derive the acceleration from the velocity signal. The smoothing spline approximation is shown to provide the best results for this numerical differentiation. Sensitivity and its need for high accuracy, especially in the acceleration signal, results in a demand of higher sample frequency. A desire for other profound ways of collecting further information, or to enhance the models, arises with possibilities of future work in the field.
Två viktiga fysikaliska aspekter som bestämmer prestandan för ett tåg i drift är det totala gångmotståndet som verkar på hela tåget, samt den tillgängliga adhesionen (användbara hjul-räl-friktionen) för framdrivning och bromsning. Från de tillgängliga signalerna önskas identifiering, samt prediktering, av dessa två storheter, under drift. Med målet att förbättra precisionen av dessa skattningar undersöker detta examensarbete potentialen av skattning och prediktering av gångmotstånd och adhesion med hjälp av Extended KalmanFiltering. Slutsatsen är att problem med observerbarhet och känslighet uppstår, vilket resulterar i ett behov av sofistikerade metoder att numeriskt beräkna acceleration från en hastighetssignal. Metoden smoothing spline approximation visar sig ge de bästa resultaten för denna numeriska derivering. Känsligheten och dess medförda krav på hög precision, speciellt på accelerationssignalen, resulterar i ett behov av högre samplingsfrekvens. Ett behov av andra adekvata metoder att tillföra ytterligare information, eller att förbättra modellerna, ger upphov till möjliga framtida utredningar inom området.

Purpose: The purpose of this paper to fill the research gap of lack ofcross-country studies in Germany, Hungary and Sweden in thecontext of online and mobile payment adaption and usage. Research Questions: The authors of this study aim at answering the followingquestion: What are the main similarities and differences in onlineand mobile payment acceptance in Germany, Hungary andSweden in the age group of 18-35?Theoretical Framework:The theoretical framework presents prevalent theories aboutfactors which are connected to online and mobilepayment. In addition, the Technical Acceptance Model wasused as a base for the conceptual framework developed by theauthors.Methodology: This study includes both qualitative and quantitative primarydata. The conducted survey resulted in 1016 answers; out of these869 answers were eligible. In addition, 150 companies wereasked and 30 qualitative interviews (10 per each country) wereconducted.Analysis: The analysis aimed at investigating how the acceptance of theusage of online and mobile payment differs among Germany,Hungary Sweden. Thereby, the following factors wereinvestigated: (1) perceived ease of use, (2) perceived usefulness,(3) perceived security, (4) perceived risk, (5) social influence, (6)information asymmetry and (7) intention to use. Conclusion: The findings of this cross-country study showed that Germany,Hungary and Sweden show a relatively similar level ofacceptance in online payment. In contrast, the acceptance in thecontext of mobile payment differed greatly. Whereas Swedenshowed a high level of acceptance, Germany and Hungaryshowed a relatively moderate level of acceptance.

Gegenwärtige Infrastrukturen für verteilte Dienste und Anwendungen, insbesondere das Internet, entwickeln sich zunehmend zu mobilen verteilten Systemen. Durch die Integration drahtloser Netze, mobiler bzw. dedizierter Endgeräte und nicht zuletzt durch die Mobilität der Benutzer steigt die Heterogenität und Dynamik der Systeme hinsichtlich der eingesetzten Endgeräte, Kommunikationstechnologien sowie Benutzeranforderungen und Anwendungssituationen. Diese Eigenschaften sind mobilen Systemen inhärent und bleiben trotz der fortschreitenden Entwicklung der Technologien bestehen. Daraus resultieren spezifische Anforderungen an Anwendungen und Dienste, denen insbesondere die Softwareentwicklung Rechnung tragen muss. In der vorliegenden Arbeit wird die Adaptivität von Softwaresystemen als wesentlicher Lösungsansatz für mobile verteilte Infrastrukturen thematisiert. Dazu werden wesentliche Mechanismen zur Adaption sowie der Überschneidungsbereich von Adaptionsmechanismen, "Context-Awareness" und Softwareentwicklung untersucht. Ziel ist es, Erkenntnisse über Basismechanismen und Grundprinzipien der Adaption zu gewinnen und diese zur systematischen Entwicklung adaptiver Anwendungen auszunutzen. Aus der Analyse des State-of-the-Art werden als erstes wichtiges Ergebnis der Arbeit wesentliche Basismechanismen zur Adaption identifiziert, umfassend klassifiziert und hinsichtlich eines Einsatzes in mobilen verteilten Infrastrukturen bewertet. Auf dieser Grundlage wird ein Meta-Modell zur systematischen Entwicklung adaptiver Anwendungen erarbeitet. Dieses erlaubt die Beschreibung adaptiver Anwendungen durch die Komposition von Basismechanismen zur Struktur- und Parameteradaption. Die Steuerung der Adaption durch Kontext und Meta-Informationen kann explizit beschrieben werden. Das Meta-Modell kann Entwickler beim Entwurf adaptiver Anwendungen unterstützen, stellt aber auch einen Ausgangspunkt für deren Analyse und Validierung sowie zur Kodegenerierung dar. Durch die explizite Beschreibung der verwendeten Adaptionsmechanismen und deren Abhängigkeiten von Kontext können Anwendungsmodelle außerdem zur Dokumentation verwendet werden. Im Rahmen der Validierung konnte die Integrierbarkeit der Basismechanismen und die flexible Anwendbarkeit des Modells zur systematischen Entwicklung adaptiver Anwendungen nachgewiesen werden.

碩士
國立清華大學
資訊工程學系
93
In this thesis, we propose a statistical model-based contour tracking method based on the Condensation framework. The models include a novel contour prediction model and two statistical object models. The object models consist of the grayscale histogram and contour shape PCA models computed from the previous tracking results. With the incremental singular value decomposition (SVD) technique, these three models are learned and updated very efficiently during tracking. We show that the proposed shape prediction model performs better than the affine predictor though experiments. Experimental results show the proposed contour tracking algorithm is very stable in tracking human heads on real videos with object scaling, rotation, partial occlusion, and illumination changes.

abstract: Nearly 60% of the world population uses a mobile phone, which is typically powered by a system-on-chip (SoC). While the mobile platform capabilities range widely, responsiveness, long battery life and reliability are common design concerns that are crucial to remain competitive. Consequently, state-of-the-art mobile platforms have become highly heterogeneous by combining a powerful SoC with numerous other resources, including display, memory, power management IC, battery and wireless modems. Furthermore, the SoC itself is a heterogeneous resource that integrates many processing elements, such as CPU cores, GPU, video, image, and audio processors. Therefore, CPU cores do not dominate the platform power consumption under many application scenarios. Competitive performance requires higher operating frequency, and leads to larger power consumption. In turn, power consumption increases the junction and skin temperatures, which have adverse effects on the device reliability and user experience. As a result, allocating the power budget among the major platform resources and temperature control have become fundamental consideration for mobile platforms. Dynamic thermal and power management algorithms address this problem by putting a subset of the processing elements or shared resources to sleep states, or throttling their frequencies. However, an adhoc approach could easily cripple the performance, if it slows down the performance-critical processing element. Furthermore, mobile platforms run a wide range of applications with time varying workload characteristics, unlike early generations, which supported only limited functionality. As a result, there is a need for adaptive power and performance management approaches that consider the platform as a whole, rather than focusing on a subset. Towards this need, our specific contributions include (a) a framework to dynamically select the Pareto-optimal frequency and active cores for the heterogeneous CPUs, such as ARM big.Little architecture, (b) a dynamic power budgeting approach for allocating optimal power consumption to the CPU and GPU using performance sensitivity models for each PE, (c) an adaptive GPU frame time sensitivity prediction model to aid power management algorithms, and (d) an online learning algorithm that constructs adaptive run-time models for non-stationary workloads.
Dissertation/Thesis
Doctoral Dissertation Electrical Engineering 2018

碩士
亞洲大學
資訊工程學系碩士班
96
The study was to develop a system based on the probability reasoning of Bayesian Network and OT theory into the Moon unit. This system was composed of On-line Adaptive Diagnostic Test System and Computerized Adaptive Remedial Instruction. The Moon unit was Nature and Life Technology course in Grade four at primary school. After taking the test, the unpracticed competence indicators of students were diagnosed individually, and then correspondent remedial instruction Flash animators were presented to them to learn. This system tried to induce the error types that students made and to validate the effect of the system. The results were as follows. 1. The stability of execution interface increased up to 96% at Computerised Adaptive Diagnostic Test System. The number of items tested by students in the Computerized Adaptive Diagnostic Test System is 35 averagely. This system can save 10 items averagely, and the test-taking time is also saved simultaneously. 2. After the Adaptive Remedial Instruction, Student’s grades increased twenty. The progress of students is significant. Therefore, the Online Adaptive Learning System proposed in this study can factually test and remedy students’ abilities individually.

碩士
亞洲大學
資訊工程學系碩士班
96
This study aims to explore the capability evaluations, sub-skills and error types in the unit of Four Fundamental Arithmetic Operations of Polynomial in the mathematics field of eighth-graders in the junior high school. There are 3 capability evaluations, 23 sub-skills and 37 error types included. By means of the probability reasoning of Bayesian Network, the research aims to construct a computerized adaptive learning mode system of this unit, which includes computerized adaptive diagnostic testing and computerized adaptive remedial instruction, so that students could be assessed, diagnosed and receive remedial instruction. The results are as follows: 　　First, by means of the dynamics threshold of Bayesian Network, the distinguishing rate of error types on average is 94%; the distinguishing rate of sub-skills on average is 91.8%; the distinguishing rate of capability evaluation is 92.1% and the diagnosis rate of the whole Bayesian Network on average is 93.3%, which show the diagnosis accuracy of Bayesian Network in this unit. 　　Second, in the aspects of computerized adaptive diagnostic testing, the rate of reducing test items is 15.8%; the prediction accuracy of the whole Bayesian Network on average is 96.5%; the prediction accuracy of the response of testing items is 97.7%. 　　Last, in the aspects of computerized adaptive remedial instruction, the progress rate of post-test is 87.5%, and the effect of students on different levels is obvious. The progress rate of error types is 37.4%; the progress rate of sub-skills is 33%. Adaptive remedial instruction upgraded students’ achievement and reduced the error types they made and enhanced their sub-skills, so the achievement is significant.

Among student modelling researches, domain-independent student models have usually been a rarity. They are valued because of reusability and economy. The demand on domain-independent student models is further increased by the need to stay competitive in the so-called knowledge economy nowadays and the widespread practice of lifelong learning. On the other hand, the popularity of student-oriented pedagogy triggers the need to provide cognitive support in virtual learning environments which in turn requires student models that create cognitive profiles of students. This study offers an innovative student modelling approach called cognitive trait model (CTM) to address both the needs mentioned above. CTM is a domain-independent and persistent student model that goes beyond traditional concept of student model. It is capable of taking the role of a learning companion who knows about the cognitive traits of the student and can supply this information when the student first starts using a new learning system. The behaviour of the students in the learning systems can then be used to update CTM. Three cognitive traits are included in the CTM in this study, they are working memory capacity, inductive reasoning ability and divergent associative learning. For the three cognitive traits, their domain-independence and persistence are studied and defined, their characteristics are examined, and behaviour patterns that can be used to indicate them are extracted. In this study, a learning system is developed to gather behaviour data of students. Several web-based psychometric tools are also developed to gather the psychometric data about the three cognitive traits of students. In the evaluations, Cognitive trait modelling is then applied on the behaviour data and the results are compared with the psychometric data. The findings prove the effectiveness of CTM and reveal important insights about the three cognitive traits.

User Modeling and Machine Learning for User Modeling have both become important research topics and key techniques in recent adaptive systems. One of the most intriguing problems in the `information age´ is how to filter relevant information from the huge amount of available data. This problem is tackled by using models of the user´s interest in order to increase precision and discriminate interesting information from un-interesting data. However, any user modeling approach suffers from several major drawbacks: User models built by the system need to be inspectable and understandable by the user himself. Secondly, users in general are not willing to give feedback concerning user satisfaction by the delivered results. Without any evidence for the user´s interest, it is hard to induce a hypothetical user model at all. Finally, most current systems do not draw a line of distinction between domain knowledge and user model which makes the adequacy of a user model hard to determine. This thesis presents the novel approach of conceptual user models. Conceptual user models are easy to inspect and understand and allow for the system to explain its actions to the user. It is shown, that ILP can be applied for the task of inducing user models from feedback, and a method for using mutual feedback for sample enlargement is introduced. Results are evaluated independently of domain knowledge within a clear machine learning problem definition. The whole concept presented is realized in a meta web search engine called OySTER.