Dissertations / Theses on the topic 'Machine'

Traditionally, Machine Translation (MT) systems are developed by targeting fluency (i.e. output grammaticality) and adequacy (i.e. semantic equivalence with the source text) criteria that reflect the needs of human end-users. However, recent advancements in Natural Language Processing (NLP) and the introduction of NLP tools in commercial services have opened new opportunities for MT. A particularly relevant one is related to the application of NLP technologies in low-resource language settings, for which the paucity of training data reduces the possibility to train reliable services. In this specific condition, MT can come into play by enabling the so-called “translation-based” workarounds. The idea is simple: first, input texts in the low-resource language are translated into a resource-rich target language; then, the machine-translated text is processed by well-trained NLP tools in the target language; finally, the output of these downstream components is projected back to the source language. This results in a new scenario, in which the end-user of MT technology is no longer a human but another machine. We hypothesize that current MT training approaches are not the optimal ones for this setting, in which the objective is to maximize the performance of a downstream tool fed with machine-translated text rather than human comprehension. Under this hypothesis, this thesis introduces a new research paradigm, which we named “MT for machines”, addressing a number of questions that raise from this novel view of the MT problem. Are there different quality criteria for humans and machines? What makes a good translation from the machine standpoint? What are the trade-offs between the two notions of quality? How to pursue machine-oriented objectives? How to serve different downstream components with a single MT system? How to exploit knowledge transfer to operate in different language settings with a single MT system? Elaborating on these questions, this thesis: i) introduces a novel and challenging MT paradigm, ii) proposes an effective method based on Reinforcement Learning analysing its possible variants, iii) extends the proposed method to multitask and multilingual settings so as to serve different downstream applications and languages with a single MT system, iv) studies the trade-off between machine-oriented and human-oriented criteria, and v) discusses the successful application of the approach in two real-world scenarios.

Thesis: Ph. D., Massachusetts Institute of Technology, School of Architecture and Planning, Program in Media Arts and Sciences, 2017.
Cataloged from PDF version of thesis. Due to the condition of the original material with text runs off the edges of the pages, the reproduction may have unavoidable flaws.
Includes bibliographical references (pages 167-172).
Generative models are venerated as full probabilistic models that randomly generate observable data given a set of latent variables that cannot be directly observed. They can be used to simulate values for variables in the model, allowing analysis by synthesis or model criticism, towards an iterative cycle of model specification, estimation, and critique. However, many datasets represent a combination of several viewpoints - different ways of looking at the same data that leads to various generalizations. For example, a corpus that has data generated by multiple people may be mixtures of several perspectives and can be viewed with different opinions by others. It isn't always possible to represent the viewpoints by clean separation, in advance, of examples representing each perspective and train a separate model for each point of view. In this thesis, we introduce lensing, a mixed-initiative technique to (i) extract lenses or mappings between machine-learned representations and perspectives of human experts, and (2) generate lensed models that afford multiple perspectives of the same dataset. We explore lensing of latent variable model in their configuration, parameter and evidential spaces. We apply lensing to three health applications, namely imbuing the perspectives of experts into latent variable models that analyze adolescent distress and crisis counseling.
by Karthik Dinakar.
Ph. D.

The amount of available data has allowed the field of machine learning to flourish. But with growing data set sizes comes an increase in algorithm execution times. Cluster computing frameworks provide tools for distributing data and processing power on several computer nodes and allows for algorithms to run in feasible time frames when data sets are large. Different cluster computing frameworks come with different trade-offs. In this thesis, the scalability of the execution time of machine learning algorithms running on the Hadoop cluster computing framework is investigated. A recent version of Hadoop and algorithms relevant in industry machine learning, namely K-means, latent Dirichlet allocation and naive Bayes are used in the experiments. This paper provides valuable information to anyone choosing between different cluster computing frameworks. The results show everything from moderate scalability to no scalability at all. These results indicate that Hadoop as a framework may have serious restrictions in how well tasks are actually parallelized. Possible scalability improvements could be achieved by modifying the machine learning library algorithms or by Hadoop parameter tuning.

This thesis is concerned with the application of simulated evolution (SE) to the steady-state parameter identification problem of a simulated and real 3-phase induction machine, over the no-load direct-on-line start period. In the case of the simulated 3-phase induction machine, the Kron's two-axis dynamic mathematical model was used to generate the real and simulated system responses where the induction machine parameters remain constant over the entire range of slip. The model was used in the actual value as well as the per-unit system, and the parameters were estimated using both the genetic algorithm (GA) and the evolutionary programming (EP) from the machine's dynamic response to a direct-on-line start. Two measurement vectors represented the dynamic responses and all the parameter identification processes were subject to five different levels of measurement noise. For the case of the real 3-phase induction machine, the real system responses were generated by the real 3-phase induction machine whilst the simulated system responses were generated by the Kron's model. However, the real induction machine's parameters are not constant over the range of slip, because of the nonlinearities caused by the skin effect and saturation. Therefore, the parameter identification of a real3-phase induction machine, using EP from the machine's dynamic response to a direct-on-line start, was not possible by applying the same methodology used for estimating the parameters of the simulated, constant parameters, 3-phase induction machine.

Nos travaux s’orientent vers les problèmes de réseaux sans fil liés à la coexistence des communications machine-to-machine (M2M) et humain-humain (H2H). On souhaite souligner l'impact mutuel entre les trafics M2M et H2H dans un contexte d’Internet des objets (IoT : Internet Of Things) en particulier lors des catastrophes. Les communications M2M, qui devraient connaître une roissance exponentielle dans un avenir proche, constitueront un facteur important pour affecter tous les réseaux mobiles. On prévoit un grand nombre d'appareils M2M qui entraînera inévitablement des problèmes de saturation et aura des impacts remarquables sur les trafics, les services et les applications M2M et H2H. Pour étudier les influences mutuelles M2M et H2H, nous développons un nouveau modèle markovien à temps continu (CTMC) pour simuler, analyser et mesurer les différentes stratégies d'accès aux réseaux sans fil. Notre modèle nous a permis de contourner certaines limitations des simulateurs professionnels de LTE-A (Long Term Evolution-Advanced) comme SimuLTE en terme d’un nombre massif d'appareils M2M, une flexibilité de certains paramètres ou pour élaborer plus des outils statistiques. Lors d’un sinistre et suite à un énorme nombre de M2M souhaitant accéder aux réseaux sans-fil, nous avons constaté un épuisement rapide de la bande passante allouée dans les réseaux LTE-M (Long Term Evolution for Machines) ou Narrow Band for IoT (NB-IoT). Pour résoudre ce problème, nous proposons une nouvelle approche appelée Adaptive eNodeB (A-eNB) pour les réseaux LTE-M et NB-IoT. Selon nos simulations, l’A-eNB peut résoudre progressivement le problème de surcharge tout en assurant une satisfaisante qualité de service (QoS) pour le trafic H2H. Avec le concept d’A-eNB, un réseau LTE-M pourra adapter ses ressources pour faire face à une augmentation progressive du nombre de connexions M2M accédant au réseau LTE-M / NB-IoT et en même temps réduire l'impact sur le trafic H2H
This Ph.D. work aims to study the Machine-to-Machine (M2M) congestion overload problem and the mutual impact among M2M and Human-to-Human (H2H) traffics in IoT (Internet of Things) environments specifically during disaster events. M2M devices with their expected exponential booming in the near future, will be one of the significant factors to influence all mobile networks. Inevitably, the expected huge number of M2M devices causes saturation problems, and leads to remarkable impacts on both M2M and H2H traffics, services and applications. To study the M2M and H2H mutual influences, we create a new platform model based on Continuous-Time Markov Chain (CTMC) to simulate, analyze and measure radio access strategies due to the limitations of existing Long Term Evolution-Advanced (LTE-A) simulators (i.e, SimuLTE) in term of massive M2M devices, parameter flexibility and statistical tools. Additionally, during disaster events, a fast bandwidth depletion of the limited bandwidth assigned to M2M devices in Long Term Evolution for Machines (LTE-M) and Narrow Band for IoT (NB-IoT) networks is expected due to the high arrival request of M2M device network access. To address this problem, we propose a new approach named Adaptive eNodeB (A-eNB) for both LTE-M and NB-IoT networks. The A-eNB can solve gradually the overload problem, while keeping the H2H traffic Quality of Service (QoS) not to be affected badly. The network adaptation is provided through a dynamic LTE-M resource reservation aiming to increase the number of M2M connections accessing the LTE-M/NB-IoT network and to decrease the impact on H2H traffic

The purpose of this thesis is the design of the software necessary for teleoperation of a fluid handling laboratory. It does not include the implementation of this software. The laboratory for which it is designed is being developed at the University of Arizona, and is a model of the fluid handling laboratory aboard Space Station. The software includes man/machine, machine/machine, and machine/instrument interfaces. The man/machine interface is menu driven and consists of high level commands which are independent of the devices in the laboratory. The machine/machine interface is also device independent. It consists of intermediary commands and maps the commands of the man/machine interface into the low level, device dependent, commands and programs of the machine/instrument interface. Although the software is primarily designed for the model laboratory, the needs of a remotely operated fluid handling laboratory aboard Space Station have been considered.

Agriculture is becoming more dependent on computerized solutions to make thefarmer’s job easier. The big step that many companies are working towards is fullyautonomous vehicles that work the fields. To that end, the equipment fitted to saidvehicles must also adapt and become autonomous. Making this equipment autonomoustakes many incremental steps, one of which is developing an accurate and reliable tanklevel measurement system. In this thesis, a system for tank level measurement in a seedplanting machine is evaluated. Traditional systems use load cells to measure the weightof the tank however, these types of systems are expensive to build and cumbersome torepair. They also add a lot of weight to the equipment which increases the fuel consump-tion of the tractor. Thus, this thesis investigates the use of radar sensors together witha number of Machine Learning algorithms. Fourteen radar sensors are fitted to a tankat different positions, data is collected, and a preprocessing method is developed. Then,the data is used to test the following Machine Learning algorithms: Bagged RegressionTrees (BG), Random Forest Regression (RF), Boosted Regression Trees (BRT), LinearRegression (LR), Linear Support Vector Machine (L-SVM), Multi-Layer Perceptron Re-gressor (MLPR). The model with the best 5-fold crossvalidation scores was Random For-est, closely followed by Boosted Regression Trees. A robustness test, using 5 previouslyunseen scenarios, revealed that the Boosted Regression Trees model was the most robust.The radar position analysis showed that 6 sensors together with the MLPR model gavethe best RMSE scores.In conclusion, the models performed well on this type of system which shows thatthey might be a competitive alternative to load cell based systems.

In questo lavoro di tesi è stato analizzato l'avvento dell'industria 4.0 all'interno dell' industria nel settore packaging. In particolare, è stata discussa l'importanza della diagnostica predittiva e sono stati analizzati e testati diversi approcci per la determinazione di modelli descrittivi del problema a partire dai dati. Inoltre, sono state applicate le principali tecniche di Machine Learning in modo da classificare i dati analizzati nelle varie classi di appartenenza.

Self-healing systems promise operating cost reductions in large-scale computing environments through the automated detection of, and recovery from, faults. However, at present there appears to be little known empirical evidence comparing the different approaches, or demonstrations that such implementations reduce costs. This thesis compares previous and current self-healing approaches before demonstrating a new, unsupervised approach that combines artificial neural networks with performance tests to perform fault identification in an automated fashion, i.e. the correct and accurate determination of which computer features are associated with a given performance test failure. Several key contributions are made in the course of this research including an analysis of the different types of self-healing approaches based on their contextual use, a baseline for future comparisons between self-healing frameworks that use artificial neural networks, and a successful, automated fault identification in cloud infrastructure, and more specifically virtual machines. This approach uses three established machine learning techniques: Naïve Bayes, Baum-Welch, and Contrastive Divergence Learning. The latter demonstrates minimisation of human-interaction beyond previous implementations by producing a list in decreasing order of likelihood of potential root causes (i.e. fault hypotheses) which brings the state of the art one step closer toward fully self-healing systems. This thesis also examines the impact of that different types of faults have on their respective identification. This helps to understand the validity of the data being presented, and how the field is progressing, whilst examining the differences in impact to identification between emulated thread crashes and errant user changes – a contribution believed to be unique to this research. Lastly, future research avenues and conclusions in automated fault identification are described along with lessons learned throughout this endeavor. This includes the progression of artificial neural networks, how learning algorithms are being developed and understood, and possibilities for automatically generating feature locality data.

Nella presente tesi di laurea viene preso in considerazione l’algoritmo di classificazione Support Vector Machine. Piu` in particolare si considera la sua formulazione come problema di ottimizazione Mixed Integer Program per la classificazione binaria super- visionata di un set di dati.

Machine-to-Machine (M2M) communications is an emerging communication paradigm that provides ubiquitous connectivity between devices along with an ability to communicate autonomously without human intervention. M2M communications acts as an enabling technology for the practical realization of Internet-of-Things (IoT). However, M2M communications differs from conventional Human-to-Human (H2H) communications due to its unique features such as massive number of connected devices, small data transmissions, little or no mobility, requirements of high energy efficiency and reliability, etc. These features create various challenges for existing communication networks which are primarily optimized for H2H communications. Therefore, novel solutions are required to meet the key requirements of M2M communications. In addition, enhancements are required at different layers of the protocol stack to support co-existence of M2M devices and H2H users. The main objective of this research is to investigate the challenges of M2M communications in two broad types of M2M networks; capillary M2M and cellular M2M networks. The primary focus is on developing novel solutions, algorithms, and protocol enhancements for successfully enabling M2M communications. Since cognitive radio technology is very promising for M2M communications, special emphasis is on capillary M2M networks with cognitive radio based Physical layer. Besides, the focus is also on exploring new frontiers in M2M communications. This thesis covers different aspects of M2M communications. Considering the motivation for cognitive M2M and service requirements of M2M devices, two cognitive MAC protocols have been proposed. The first protocol is centralized in nature and utilizes a specialized frame structure for co-existence with the primary network as well as handling different Quality-of-Service (QoS) requirements of M2M devices. The second protocol is a distributed cognitive MAC protocol, which is specially designed to provide high energy efficiency and reliability for M2M devices operating in challenging wireless environments. Both protocols explicitly account for the peculiarities of cognitive radio environments. The protocols have been evaluated using analytical modeling and simulation studies. Recently IETF has standardized a specially designed routing protocol for capillary M2M networks, known as RPL (Routing for Low Power and Lossy Networks). RPL is emerging as the de facto routing protocol for many M2M applications including the smart grid. On the other hand, the application of cognitive radio for smart grid communication is under active investigation in the research community. Hence, it is important to investigate the applicability and adaptation of RPL in cognitive radio environments. In this regard, an enhanced RPL based routing protocol has been proposed for cognitive radio enabled smart grid networks. The enhanced protocol provides novel modifications to RPL for protecting the primary users along with meeting the utility requirements of the secondary network. An important challenge in LTE-based cellular networks with M2M communications is the uplink radio resource management as available resources are shared between M2M devices and H2H users, having different and often conflicting QoS requirements. Apart from this, energy efficiency requirements become critically important. Further, the specific constraints of Single Carrier Frequency Division Multiple Access (SC-FDMA) complicate the resource allocation problem. In this respect, an energy efficient resource allocation algorithm for the uplink of LTE networks with M2M/H2H co-existence under statistical QoS guarantees has been developed, that is based on canonical duality theory. The proposed algorithm outperforms classical algorithms in terms of energy efficiency while satisfying the QoS requirements of M2M devices and H2H users. A new frontier in M2M communications is the nano-M2M communications, which is envisioned to create the Internet-of-Nano-Things (IoNT). Molecular communication (MC) is a promising communication technique for nano-M2M communications. In literature, no model for error performance of MC exists. Therefore, an error performance model has been developed that explicitly accounts for noise and interference effects. Since relaying and network coding based solutions are gaining popularity for nano-M2M networks, the error performance of a network coded molecular nano-M2M network has been evaluated as well. Finally, the thesis is concluded based on the overall picture of the research conducted. In addition, some directions for future work are included as well.

Mestrado em Engenharia de Computadores e Telemática
Although the area of Machine-to-Machine communications and, consequently, the Internet of Things, have undergone a great improvement regarding interoperability, there is still no ’de facto’ solution proposal to achieve large scale, even global, interoperability. As a first step, this work provides a theoretical analysis of proposals relevant to the area, mainly analysing how they achieve some essential requirements for the Internet of Things, such as scalability, heterogeneity and management. Later, focusing in ETSI’s M2M standard, is first given a high-level description of its vision, approach and architecture, and then, finally, from a more practical point of view, is also presented and tested a functional implementation of an ETSI M2M compliant gateway, which provides an empirical evaluation of the standard.
Apesar de a área das comunicações Máquina-a-Máquina e, consequentemente, a Internet das Coisas, terem sofrido uma grande melhoria relativamente à interoperabilidade, ainda não existe nenhuma solução considerada "dominante" que permita atingir uma interoperabilidade em larga escala, até mesmo global. Desta forma, numa primeira instância este trabalho visa fornecer uma análise teórica de propostas relevantes para a área, onde se analisa maioritariamente como é que essas propostas atingem alguns requisitos essenciais para a Internet das Coisas, como a escalabilidade, heterogeneidade e gestão. Posteriormente, focando-se no standard ETSI M2M, é dado em primeiro lugar uma descrição de alto nível da sua visão, abordagem e arquitectura, e depois, finalmente, de um ponto de vista prático, é ainda apresentada e testada uma implementação funcional de uma gateway condescendente com o standard, o que fornece uma avaliação mais empírica do mesmo.

Data Ductus is a Swedish IT-consultant company, their customer base ranging from small startups to large scale cooperations. The company has steadily grown since the 80s and has established offices in both Sweden and the US. With the help of machine learning, this project will present a possible solution to the errors caused by the human factor in the logistic business.A way of preprocessing data before applying it to a machine learning algorithm, as well as a couple of algorithms to use will be presented.
Data Ductus är ett svenskt IT-konsultbolag, deras kundbas sträcker sig från små startups till stora redan etablerade företag. Företaget har stadigt växt sedan 80-talet och har etablerat kontor både i Sverige och i USA. Med hjälp av maskininlärning kommer detta projket att presentera en möjlig lösning på de fel som kan uppstå inom logistikverksamheten, orsakade av den mänskliga faktorn.Ett sätt att förbehandla data innan den tillämpas på en maskininlärning algoritm, liksom ett par algoritmer för användning kommer att presenteras.

De nos jours, les exigences de productivité et de maîtrise des coûts ont incité les industriels à développer de nouvelles machines, et avec elles, de nouveaux enjeux sont apparus : souplesse de la structure, vibration, effets dynamiques non-négligeables, etc. Pourtant, leur mise en œuvre est toujours issue de méthodes employées pour les machines conventionnelles. Ces travaux s’intéressent donc à la définition de stratégies globales englobant la prise en compte de la structure utilisée et de la tâche à réaliser, appliquée à deux contextes d’illustration. Dans le contexte de l’usinage, nous proposons un réglage des machines basé sur le modèle comportemental de la structure qui ne nécessite que peu de modifications manuelles et permettant un gain de temps pour la mise en œuvre. Une nouvelle loi de commande en couple calculé est également proposé, elle permet de réduire les phénomènes vibratoires lors de phases dynamiquement exigeantes. Dans le contexte des essais mécaniques, l’objectif est de montrer la faisabilité de l’utilisation de machines parallèles à 6 degrés de liberté dans le cadre d’essais dont la gestion des conditions aux limites est critique. Nous proposons une instrumentation et un schéma de commande qui permettent de respecter les consignes avec une erreur maximale de l’ordre de 0.40μm, même dans le cas d’éprouvettes très rigide (en béton par exemple)
Nowadays, the requirements in productivity and costs mastering have forced the industrial manufacturers to develop new kind of mechanisms. Thus, the complexity of the machine-tools structures and machining processes has increased and new challenges have emerged : flexible structure, vibration, non-negligible dynamic effects, etc ... However, their implementation still comes from methods used for conventional machines. These works are thus about defining overall strategies including consideration of the kind of structure used and the task to realise. Two illustrative contexts are used. In the context of machining, we propose a generic tuning method based on kinematic and dynamic model of machine-tools structure that requires only a few manual modifications, in order to save time for implementation. A new computed torque control law is proposed, it reduces vibration phenomena in dynamical demanding phases. In the context of the mechanical tests, the objective is to demonstrate the feasibility of using parallel machines with 6 degrees of freedom in the context of mechanical tests, whereas the boundary conditions are perfectly controlled. We propose an instrumentation and control scheme that is able to perform mechanical tests with a maximum error of about 0.40 mu m, even in the case of very rigid specimen (concrete for example)

This thesis aims at contributing to overcome two of the main challenges for the deployment of M2M networks in data collection scenarios for the Internet of Things: the management of massive numbers of end-devices that attempt to get access to the channel; and the need to extend the network lifetime. In order to solve these challenges, two complementary strategies are considered. Firstly, the thesis focuses on the design, analysis and performance evaluation of MAC protocols that can handle abrupt transitions in the traffic load and minimize the energy consumption devoted to communications. And secondly, the use of energy harvesting (EH) is considered in order to provide the network with unlimited lifetime. To this end, the second part of the thesis focuses on the design and analysis of EH-aware MAC protocols. While the Frame Slotted-ALOHA (FSA) protocol has been traditionally adopted in star topology networks for data collection, results show that FSA-based protocols lack of scalability and present synchronization problems as the network density increases. Indeed, the frame length of FSA must be adjusted to the number of contenders, which may be complex to attain in dense networks with large and dynamic number of end-devices. In order to overcome these issues, a tree splitting-based random access protocol, referred to as Low Power Contention Tree-based Access (LP-CTA), is proposed in the first part of this thesis. In LP-CTA, the frame length can be very short and fixed, which facilitates synchronization and provides better network scalability than FSA. While LP-CTA uses data slots for contention, it is possible to use short access requests in minislots, where collisions are resolved using tree splitting, and avoid the contention in data. Since these minislots can be much shorter than the duration of a data packet, the performance can be improved. The Low Power Distributed Queuing (LP-DQ) protocol proposed in this thesis is based on this idea. LP-DQ combines tree splitting with the logic of two distributed queues that manage the contention resolution and the collision-free data transmission. Results show that LP-DQ outperforms LP-CTA and FSA in terms of delay and energy efficiency, and it relaxes the need to know the size of the network and adapts smoothly to any change in the number of end-devices. The approach of LP-DQ is convenient when the messages transmitted by each end-device fit in one single slot, however, if the end-devices generate long messages that have to be fragmented, it is better to add a reservation mechanism in order to boost the performance. In this sense, the LPR-DQ protocol is proposed as an extension of LP-DQ where the concept of reservation is integrated to allow the end-devices reserve as many collision-free slots as needed. The second part of the thesis is devoted to the integration of the MAC layer with the use of energy harvesting. The variability and fluctuations of the harvested energy is considered for the design of EH-aware MAC protocols and three performance metrics are proposed: the probability of delivery, the data delivery ratio and the time efficiency. Previous works on data collection networks with EH focus on DFSA. In this thesis, the EH-CTA protocol is proposed as an adaptation of LP-CTA that takes the energy harvesting process into account. Results show that EH-CTA outperforms DFSA if the energy threshold for an end-device to become active is properly configured. In addition, while DFSA needs to adapt the frame length dynamically, EH-CTA uses a fixed frame length, thus facilitating scalability and synchronization. Finally, the EH-RDFSA and EH-DQ protocols are proposed for scenarios where data must be fragmented. EH-RDFSA is a combination of RFSA and DFSA, and EH-DQ is an extension of LPR-DQ.
Esta tesis contribuye a resolver dos de los retos para el despliegue de redes M2M en escenarios de recolección de datos para el Internet de las Cosas: la gestión del acceso al canal de un número masivo de dispositivos; y la necesidad de extender la vida de la red. Para resolverlos se consideran dos estrategias complementarias. En primer lugar, se centra en el diseño, el análisis y la evaluación de protocolos MAC que pueden manejar transiciones abruptas de tráfico y reducen el consumo de energía. Y en segundo lugar, se considera el uso de mecanismos de captura de energía (Energy Harvesters, EH) para ofrecer un tiempo de vida ilimitado de la red. Con este fin, la segunda parte de la tesis se centra en el diseño y el análisis de protocolos MAC de tipo "EH-aware". Mientras que Frame Slotted-ALOHA (FSA) ha sido tradicionalmente adoptado en aplicaciones de recolección de datos, los resultados muestran que FSA presenta problemas de escalabilidad y sincronización cuando aumenta la densidad de la red. De hecho, la longitud de trama de FSA se debe ajustar según sea el número de dispositivos, lo cual puede ser difícil de estimar en redes con un número elevado y dinámico de dispositivos. Para superar estos problemas, en esta tesis se propone un protocolo de acceso aleatorio basado en "tree-splitting" denominado Low Power Contention Tree-based Access (LP-CTA). En LP-CTA, la longitud de trama puede ser corta y constante, lo cual facilita la sincronización y proporciona mejor escalabilidad. Mientras que LP-CTA utiliza paquetes de datos para la contienda, es posible utilizar solicitudes de acceso en mini-slots, donde las colisiones se resuelven utilizando "tree-splitting", y evitar la contención en los datos. Dado que estos mini-slots pueden ser mucho más cortos que la duración de un slot de datos, el rendimiento de LP-CTA puede ser mejorado. El protocolo Low Power Distributed Queuing (LP-DQ) propuesto en esta tesis se basa en esta idea. LP-DQ combina "tree-splitting" con la lógica de dos colas distribuidas que gestionan la resolución de la contienda en la solicitud de acceso y la transmisión de datos libre de colisiones. Los resultados demuestran que LP-DQ mejora LP-CTA y FSA en términos de retardo y eficiencia energética, LP-DQ no requiere conocer el tamaño de la red y se adapta sin problemas a cualquier cambio en el número de dispositivos. LP-DQ es conveniente cuando los mensajes transmitidos por cada dispositivo caben en un único slot de datos, sin embargo, si los dispositivos generan mensajes largos que requieren fragmentación, es mejor añadir un mecanismo de reserva para aumentar el rendimiento. En este sentido, el protocolo LPR-DQ se propone como una extensión de LP-DQ que incluye un mecanismo de reserva para permitir que cada dispositivo reserve el número de slots de datos según sea el número de fragmentos por mensaje. La segunda parte de la tesis está dedicada a la integración de la capa MAC con el uso de "Energy Harvesters". La variabilidad y las fluctuaciones de la energía capturada se consideran para el diseño de protocolos MAC de tipo "EH-aware" y se proponen tres métricas de rendimiento: la probabilidad de entrega, el "Data Delivery Ratio" y la eficiencia temporal. Los trabajos previos en redes de recolección de datos con EH se centran principalmente en DFSA. En esta tesis, el protocolo EH-CTA se propone como una adaptación de LP-CTA que tiene en cuenta el proceso de captura de energía. Los resultados muestran que EH-CTA supera DFSA si el umbral de energía para que un dispositivo se active está configurado correctamente. Además, mientras que en DFSA se necesita adaptar la longitud de trama de forma dinámica, EH-CTA utiliza una longitud de trama fija, facilitando así la escalabilidad y la sincronización. Por último, se proponen los protocolos EH-RDFSA y EH-DQ para escenarios en los que los datos deben ser fragmentados. EH-RDFSA es una combinación de RFSA y DFSA, y EH-DQ es una extensión de LPR-DQ.
Aquesta tesi contribueix a resoldre dos dels reptes per al desplegament de xarxes M2M en escenaris de recol·lecció de dades per a l'Internet de les Coses: la gestió de l'accés al canal d'un nombre massiu de dispositius; i la necessitat d'extendre la vida de la xarxa. Per resoldre'ls es consideren dues estratègies complementàries. En primer lloc, es centra en el disseny, l'anàlisi i l'avaluació de protocols MAC que poden manegar transicions abruptes de trànsit i redueixen el consum d'energia. I en segon lloc, es considera l'ús de mecanismes de captura d'energia (Energy Harvesters, EH) per a oferir un temps de vida il·limitat de la xarxa. Amb aquesta finalitat, la segona part de la tesi es centra en el disseny i l'anàlisi de protocols MAC de tipus "EH-aware".Mentre que Frame Slotted-ALOHA (FSA) ha estat tradicionalment adoptat en aplicacions de recol·lecció de dades, els resultats mostren que FSA presenta problemes d'escalabilitat i sincronització quan augmenta la densitat de la xarxa. De fet, la longitud de trama de FSA s'ha d'ajustar segons sigui el nombre de dispositius, la qual cosa pot ser difícil d'estimar en xarxes amb un nombre elevat i dinàmic de dispositius. Per superar aquests problemes, en aquesta tesi es proposa un protocol d'accés aleatori basat en "tree-splitting" denominat Low Power Contention Tree-based Access (LP-CTA). En LP-CTA, la longitud de trama pot ser curta i constant, la qual cosa facilita la sincronització i proporciona millor escalabilitat.Mentre que LP-CTA utilitza paquets de dades per a la contenció, és possible utilitzar sol·licituds d'accés a mini-slots, on les col·lisions es resolen utilitzant "tree-splitting", i evitar la contenció a les dades. Atès que aquests mini-slots poden ser molt més curts que la durada d'un slot de dades, el rendiment de LP-CTA pot ser millorat. El protocol Low Power Distributed Queuing (LP-DQ) proposat en aquesta tesi es basa en aquesta idea. LP-DQ combina "tree-splitting" amb la lògica de dues cues distribuïdes que gestionen la resolució de la contenció en la sol·licitud d'accés i la transmissió de dades lliure de col·lisions. Els resultats demostren que LP-DQ millora LP-CTA i FSA en termes de retard i eficiència energètica, LP-DQ no requereix conèixer la mida de la xarxa i s'adapta sense problemes a qualsevol canvi en el nombre de dispositius.LP-DQ és convenient quan els missatges transmesos per cada dispositiu caben en un únic slot de dades, però, si els dispositius generen missatges llargs que requereixen fragmentació, és millor afegir un mecanisme de reserva per augmentar el rendiment. En aquest sentit, el protocol LPR-DQ es proposa com una extensió de LP-DQ que inclou un mecanisme de reserva per a permetre que cada dispositiu reservi el nombre de slots de dades segons sigui el nombre de fragments per missatge.La segona part de la tesi està dedicada a la integració de la capa MAC amb l'ús de "Energy Harvesters". La variabilitat i les fluctuacions de l'energia capturada es consideren per al disseny de protocols MAC de tipus "EH-aware" i es proposen tres mètriques de rendiment: la probabilitat d'entrega, el "Data Delivery Ratio" i l'eficiència temporal.Els treballs previs en xarxes de recol·lecció de dades amb EH se centren principalment en DFSA. En aquesta tesi, el protocol EH-CTA es proposa com una adaptació de LP-CTA que té en compte el procés de captura d'energia. Els resultats mostren que EH-CTA supera DFSA si el llindar d'energia perquè un dispositiu s'activi s'ajusta correctament. A més, mentre que a DFSA es necessita adaptar la longitud de trama de forma dinàmica, EH-CTA utilitza una longitud de trama fixa, facilitant així l'escalabilitat i la sincronització. Finalment, es proposen els protocols EH-RDFSA i EH-DQ per a escenaris en els quals les dades han de ser fragmentades. EH-RDFSA és una combinació de RFSA i DFSA, i EH-DQ és una extensió de LPR-DQ.

Nowadays, the great diffusion of advanced devices, such as smart-phones, has shown that there is a growing trend to rely on new technologies to generate and/or support progress; the society is clearly ready to trust on next-generation communication systems to face today’s concerns on economic and social fields. The reason for this sociological change is represented by the fact that the technologies have been open to all users, even if the latter do not necessarily have a specific knowledge in this field, and therefore the introduction of new user-friendly applications has now appeared as a business opportunity and a key factor to increase the general cohesion among all citizens. Within the actors of this technological evolution, wireless machine-to-machine (M2M) networks are becoming of great importance. These wireless networks are made up of interconnected low-power devices that are able to provide a great variety of services with little or even no user intervention. Examples of these services can be fleet management, fire detection, utilities consumption (water and energy distribution, etc.) or patients monitoring. However, since any arising technology goes together with its security threats, which have to be faced, further studies are necessary to secure wireless M2M technology. In this context, main threats are those related to attacks to the services availability and to the privacy of both the subscribers’ and the services providers’ data. Taking into account the often limited resources of the M2M devices at the hardware level, ensuring the availability and privacy requirements in the range of M2M applications while minimizing the waste of valuable resources is even more challenging. Based on the above facts, this Ph. D. thesis is aimed at providing efficient security solutions for wireless M2M networks that effectively reduce energy consumption of the network while not affecting the overall security services of the system. With this goal, we first propose a coherent taxonomy of M2M network that allows us to identify which security topics deserve special attention and which entities or specific services are particularly threatened. Second, we define an efficient, secure-data aggregation scheme that is able to increase the network lifetime by optimizing the energy consumption of the devices. Third, we propose a novel physical authenticator or frame checker that minimizes the communication costs in wireless channels and that successfully faces exhaustion attacks. Fourth, we study specific aspects of typical key management schemes to provide a novel protocol which ensures the distribution of secret keys for all the cryptographic methods used in this system. Fifth, we describe the collaboration with the WAVE2M community in order to define a proper frame format actually able to support the necessary security services, including the ones that we have already proposed; WAVE2M was funded to promote the global use of an emerging wireless communication technology for ultra-low and long-range services. And finally sixth, we provide with an accurate analysis of privacy solutions that actually fit M2M-networks services’ requirements. All the analyses along this thesis are corroborated by simulations that confirm significant improvements in terms of efficiency while supporting the necessary security requirements for M2M networks.

PONTIFÍCIA UNIVERSIDADE CATÓLICA DO RIO DE JANEIRO
COORDENAÇÃO DE APERFEIÇOAMENTO DO PESSOAL DE ENSINO SUPERIOR
PROGRAMA DE EXCELENCIA ACADEMICA
A organização lógica de data centers recai principalmente na questão estratégica de distribuir os serviços nos equipamentos de forma que os custos operacionais sejam os menores possíveis. Além desses custos, devem ser considerados outros aspectos que envolvem a interdependência de seus serviços internos e a distribuição entre suas localidades, visando assim melhorar a qualidade de seu produto aos seus clientes. Este trabalho explora o problema de atribuição de processos a máquinas do desafio ROADEF de 2012 pelos métodos de programação inteira e geração de colunas. Apresenta estratégias para lidar com as dificuldades numéricas encontradas. Na geração de colunas, analisa técnicas para acelerar a convergência, por meio de resolver o mestre restrito após cada variável, geração prévia de colunas e estabilização das variávies duais. Ao final do trabalho, são comparados os resultados obtidos com os melhores resultados oficiais.
A data center logic organization lies mainly by the strategic decision on how distribute services between machines, so the operational costs should be the smallest as possible. Beside those costs, must also consider the interdependence of their own services, the distribution between their localities, to improve the quality of their product to their customers. This work explores the challenge ROADEF 2012 machine assignment problem by the means of integer programming and column generation. Shows strategies to address numeric issues. At column generation, it analyzes techniques to speed up the convergence, by solving after each variable adiction, a previous generation of columns and stabilization of duals variables. At the end of the work, it compares the results obtained are compared with the best official results.

In the field of machine translation there are various systems available which have different strengths and weaknesses. This thesis investigates the combination of two systems, a rule based one and a statistical one, to see if such a hybrid system can provide higher quality translations. The classification approach was taken, where a support vector machine is used to choose which sentences from each of the two systems result in the best translation. To label the sentences from the collected data a new method of simulated annealing was applied and compared to previously tried heuristics. The results show that a hybrid system has an increased average BLEU score of 6.10% or 1.86 points over the single best system, and that using the labels created through simulated annealing, over heuristic rules, gives a significant improvement in classifier performance.

Cette thèse s'insère dans un projet de recherches visant l'informatique vectorielle, tant du point de vue programmation que du point de vue architecture des machines. Les aspects programmation du projet étudient une alternative à l'approche classique utilisant un langage scalaire associé à un vectoriseur. Nous nous intéressons à la définition d'outils autorisant l'expression d'algorithmes vectoriels de manière explicite, de manière indépendante de toute cible matérielle particulière. Cette approche résout les problèmes de portabilité d'une manière simple, les programmes ne faisant aucune référence à l'architecture sous-jacente. Pour que les programmes puissent être compilés d'une manière optimale, le langage donne la possibilité au programmeur d'exprimer la connaissance dont il dispose sur son algorithme, ainsi que sur les données traitées. Ces informations n'ont plus à être extraites des sources par un outil d'analyse (vectoriseur dans l'approche classique). La compilation repose sur l'utilisation d'une couche virtuelle (machine abstraite MAD et son langage Devil). Celle-ci est décrite dans ce document. A côté de ces aspects langages, nous nous intéressons aux aspects matériels des machines vectorielles. La définition dans notre langage de vecteurs originaux (autres que de simples tableaux) nous a mené à étudier la prise en compte des vecteurs au niveau matériel. Aussi, nous avons défini une mémoire contenant des vecteurs en tant que tels et y autorisant l'accès. Les définitions de cette mémoire et d'un processeur associé sont données. Enfin, en marge de ces travaux, face aux problèmes dus aux conflits d'accès aux données dans la mémoire des supercalculateurs, nous introduisons un modèle d'exécution des opérations vectorielles pipelinées dénommé traitement vectoriel désordonné (tvd). Dans ce modèle un conflit mémoire ne bloque pas nécessairement le port mémoire, et donc les pipelines qui y sont chaînés. Nous montrons l’intérêt de ce modèle selon le taux de conflits. Par ailleurs, nous montrons que ce modèle est implantable assez facilement sur un processeur de type Cray

Cette thèse s’intéresse à l’analyse de structures de machines électriques non conventionnelles destinées à la production d’électricité par des groupes électrogènes. Les topologies recherchées doivent utiliser moins de matières actives et/ou simplifier la procédure de fabrication par rapport aux machines actuelles.Une des structures est dédiée au système d’excitation de la machine. Elle possède un stator à griffes avec un bobinage toroïdal. Son dimensionnement est réalisé à l’aide d’un modèle de calcul par éléments finis en 3D, qui a été validé expérimentalement. Cette structure permet d’obtenir une réduction significative de la quantité de cuivre dans le bobinage.Les deux autres structures traitées concernent l’alternateur principal. Une machine synchro-réluctante à barrières deflux et à rotor bobiné a été proposée. L’objectif est d’augmenter la densité de couple grâce au couple de saillance. Elle est dimensionnée et comparée avec la structure conventionnelle. Ce cas met en évidence les limitations des structures proches de la structure actuelle.La dernière structure est une nouvelle topologie de machine électrique. Elle combine les caractéristiques des machines à flux axial et des machines à griffes, permettant de simplifier le système d’excitation. Une approche de modélisation originale a été développée pour l’analyse de cette topologie 3D en vue de son dimensionnement par optimisation. Le système d’excitation a un rôle important dans les performances de cette machine.Cette thèse constate aussi que l’évolution future des matériaux magnétiques pourrait tirer meilleur profit des structures non-conventionnelles avec des trajets de flux tridimensionnels
The PhD project aims to analyse nonconventional structures of electric machines for electric power generation with diesel gensets. The researched topologies should use less active materials and/or simplify the manufacturing process compared to today’s machines.One of the structures is intended for the excitation system of the electric machine. It is composed of a claw-pole stator and a toroidal winding. It is designed using a 3D Finite Element model, previously validated by a prototype. This structure allows a significant reduction of the quantity of copper of the field winding.The other two structures under study are proposed for the main generator. A synchronous-reluctance machine with flux-barriers and a field winding in the rotor is analysed. The objective is to increase the torque density by increasing the reluctance torque. The designed machine is compared to the conventional structure. This case shows the limitations of structures similar to the current structure.The last structure is a new topology of electric machine. It combines the characteristics of claw-pole machines andaxial-flux machines. An original modelling approach is developed to analyse this 3D structure for its design based on an optimization algorithm. The study shows that the excitation system has a key role in its performances.This work also notes that the future evolution of magnetic materials should benefit the use of non-conventional structures with 3D flux paths

"In terms of what they are capable of, it seems to me, when you have the distance narrowing between humans and machines in the sense that if we are becoming more machine-like, it's easier to see the machine as more human-like. I don't want to be overly dramatic about it, but I think more and more people wonder, is this living or are we just going through the motions? What's happening? Is everything being leached out of life? Are the whole texture and values and everything kind of draining away? Well, that would take many other lectures, but it's not so much the actual advance of the technology: If machines can be human, humans can be machines. The truly scary point is the narrowing of the distance between the two".In John Zerza’s talk ‘Against Technology’ at Stanford University, he observed that when you have the distance between humans and machines narrowing then in a sense we are becoming more machine-like, and it’s easier to see the machine as more human like. These views are similar to the views I have been considering for some time in my art practice.My research paper attempts to chart the relationship between my art practice and personal and global circumstances as an international student from Beijing studying at an art school in Melbourne.Living Machines also finds expressions of these ideas in the theories of Michel Focault, Gilles Deleuze, Jacques Lacan and Harold Pinter. The artists I have been investigating include Marcel Duchamp, Ai WeiWei, Ang Lee, Tom Friedman and A Constructed World .The informal nature of the writing attempts to articulate my philosophical stance taken in the studio-based research. My studio research practice comprises collaboration and installations where I construct objects from found materials, and use video, animation, and performance to explore material and spatial equivalences to the concept of body as machine.

Thesis (Ph. D.)--Massachusetts Institute of Technology, School of Architecture and Planning, Program in Media Arts and Sciences, 2007.
Includes bibliographical references (p. [137]-146).
Machine Therapy is a new practice combining art, design, psychoanalysis, and engineering work in ways that access and reveal the vital, though often unnoticed, relevance of people's interactions and relationships with machines. Machine Therapy will be illustrated through the construction of several systems including re-appropriated domestic devices such as Blendie, wearble apparatuses such as ScreamBody, and body-signal-based companion machines - Umo, Amo, and Omo - that function through visceral interactions including breathing and non-verbal sounds. These systems will be used to explore themes of human-machine relations in terms of visceral, cathartic, and reflexive expressions and communications. This work incorporates elements from my technical research in digital signal processing, machine learning, mechanical engineering, and sensor design. Combining these areas of research and practice, I have been able to help manifest new objects and relationships that are unique in some aspects while maintaining quotidian familiarity in other aspects. These apparatuses enable unusual explorations of what we interact with when we interact with machines. I hypothesize that the answer will turn out to be much more than the machine itself, and will include our sense of self, agency in the interpersonal and political world, and our shared psychological, emotional, cultural, and perceptual approaches to the world. The importance of the parapractic elements and also the therapeutic properties of the Machine Therapy machines will be evaluated in studies of participants' interactive engagements with the machines as well as their affective responses to the machines.
Kelly Dobson.
Ph.D.

Recent developments in AI is changing our world. It already governs our digital life. In my thesis I take the position that AI involvement in the field of architecture is inevitable, and indeed already here. AI is neither something we can simply accept, nor wholly ignore. Rather, we should try to understand and work with it. These algorithms should not be seen as mere tools with predictable, repeatable outcomes, they are something more complex. I’ve explored the world of AI by means of teaching a machine to design diverse, typologically similar objects: residential doorways from Stockholm. By instructing the machine to read and recreate these objects it has learned to design objects similar to them. While the machine does not know what it has designed, it has nevertheless reinterpreted the residential gate, thus offering an opportunity to glimpse into to the “mind” of AI, a world equally as unknown as omnipresent.

This thesis consists of a collection of poems. By virtue of its content and arrangement, the collection ruminates on and attempts to work through the problem of corporeality and bodily experience: the anxieties surrounding illness, mortality, and the physicality of contemporary life. This collection explores the tension inherent in the mind/body duality and, rather than prescribing solutions, offers multiple avenues and perspectives through which to view bodily experience, as well as how that experience affects an individual’s identity, agency, and sense of self.

Can I create my own design companion? My own design AI? How far do I go using the machine? What are the poetics of machine learning? This thesis is about exploring art and artificial intelligence, specifically machine learning which is the study of computer algorithms that improve through experience. The core thing of what machine learning does is to find patterns in data to then use those patterns to in some way predict the future.  I define a machine which works and generates images according to the given rules. The rules are set in time and in data. The decision, however, as in all creative processes, is up to the creator (in this the architect) so it is as much a part of the creation as the setting up of the data. The method is a mix of my own personality and imagination and the impersonal machine (my computer).  With me during the process, I found inspiration from other creators working with machines in different experimental ways that diverge from the original purpose of their machine/tool. The project is an investigation of contemporary technologies where I try to understand my tool through a series of experiments.

This document examines the work in my MFA thesis exhibition. The objects in that installation address specific socio-aesthetic sites of class and power. The personal and cultural narratives examined here further explain the objects’ symbolic potential.

The PIPPIN machine describes two simulations which are intended to help students understand the compile and execute process of a simple computer. The first simulator takes a simple mathematical expression as input and then translates it into an assembly language. The second simulator will execute an assembly language program.

From a general discussion concerning the shortcomings of the received view of scientific theories and scientific explanation I conclude that metaphorical thinking, until quite recently restricted to literary analysis, may play a significant role not only in the way theories are conceived, but also in the way that meaning is ascribed to the concepts used in science. The analysis of the literal realm of 'machine' considers three aspects that could appear in the metaphorical assimilation of organisms to machines: the contrivance aspect, which is the 'hardware'; the fact that machines exhibit purpose; finally, the integrated aspect of the machine (its harmony). The study is devoted only to the first aspect. I offer a narrative of pivotal ideas about the workings of the biological individual, from the clockworks of the early mechanicists to modern biochemistry, not just as a succession of discoveries but also as alleged accomplishments of the 'machine metaphor' revealing its scope. Some recognized milestones in the history of ideas about the inner workings of organisms are surveyed: the proposals of the early mechanicists during the Scientific Revolution, Lavoisier's view of respiration as combustion, Liebig's description of the human body as a chemical machine and the suggestion that the chemical accomplishments in living beings are the result of myriads of fermentation-like processes. I devote special attention to the problem of the direct conversion of chemical energy into mechanical energy using the evolution of ideas about muscular contraction as the main example. During the period 1900-1930 the study of colloidal behavior was considered to be the right path for unraveling most of the mysteries of vital processes. I carefully describe this work particularly the proposed models for muscular contraction and enzymatic action. The dismissal of this colloidal approach after the acceptance of the existence of those particular kinds of macromolecules that exist in living organisms marks the entrance of our modern approach. One of the remarkable features of the modern approach is the incessant elaboration of the idea of 'molecular machine'. I conclude with a discussion of the problem how literally can this metaphor be taken?
Graduate and Postdoctoral Studies
Graduate

Cellular network based Machine-to-Machine (M2M) communications have been growing rapidly in recent years, being used in a wide range of services such as security, metering, health, remote control, tracking and so on. A critical issue that needs to be considered in M2M communications is the energy efficiency, typically the machines are powered by batteries of low capacity and it is important to optimize the way the power is consumed. In search of better M2M systems, we propose a context-aware framework for M2M communications so the machine type communication (MTC) devices dynamically adapt their settings depending on a series of characteristics such as data reporting mode and quality of service (QoS) features so higher energy efficient is achieved, extending the operating lifetime of the M2M network. Simulations were performed with four commonly used M2M applications:home security, telehealth, climate and smart metering, achieving considerable energy savings and operating lifetime extension on the network. Thus, it is shown that contexts play an important role on the energy efficiency of a M2M system.

Attacks against mobile systems have escalated over the past decade. There have been increases of fraud, platform attacks, and malware. The Internet of Things (IoT) offers a new attack vector for Cybercriminals. M2M contributes to the growing number of devices that use wireless systems for Internet connection. As new applications and platforms are created, old vulnerabilities are transferred to next-generation systems. There is a research gap that exists between the current approaches for security framework development and the understanding of how these new technologies are different and how they are similar. This gap exists because system designers, security architects, and users are not fully aware of security risks and how next-generation devices can jeopardize safety and personal privacy. Current techniques, for developing security requirements, do not adequately consider the use of new technologies, and this weakens countermeasure implementations. These techniques rely on security frameworks for requirements development. These frameworks lack a method for identifying next generation security concerns and processes for comparing, contrasting and evaluating non-human device security protections. This research presents a solution for this problem by offering a novel security framework that is focused on the study of the “functions and capabilities” of M2M devices and improves the systems development life cycle for the overall IoT ecosystem.

Thesis: M.B.A., Massachusetts Institute of Technology, Sloan School of Management, 2019, In conjunction with the Leaders for Global Operations Program at MIT
Thesis: S.M., Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science, 2019, In conjunction with the Leaders for Global Operations Program at MIT
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 70-71).
As the popularity of online retail expands, world-class electronic commerce (e-commerce) businesses are increasingly adopting collaborative robotics and Internet of Things (IoT) technologies to enhance fulfillment efficiency and operational advantage. E-commerce giants like Alibaba and Amazon are known to have smart warehouses staffed by both machines and human operators. The robotics systems specialize in transporting and maneuvering heavy shelves of goods to and from operators. Operators are left to higher-level cognitive tasks needed to process goods such as identification and complex manipulation of individual objects. Achieving high system throughput in these systems require harmonized interaction between humans and machines. The robotics systems must minimize time that operators are waiting for new work (idle time) and operators need to minimize time processing items (takt time). Over time, these systems will naturally generate extensive amounts of data. Our research provides insights into both using this data to design a machine-learning (ML) model of takt time, as well as exploring methods of interpreting insights from such a model. We start by presenting our iterative approach to developing a ML model that predicts the average takt of a group of operators at hourly intervals. Our final XGBoost model reached an out-of-sample performance of 4.01% mean absolute percent error (MAPE) using over 250,000 hours of historic data across multiple warehouses around the world. Our research will share methods to cross-examine and interpret the relationships learned by the model for business value. This can allow organizations to effectively quantify system trade-offs as well as identify root-causes of takt performance deviations. Finally, we will discuss the implications of our empirical findings.
by Anjian Wu.
M.B.A.
S.M.
M.B.A. Massachusetts Institute of Technology, Sloan School of Management
S.M. Massachusetts Institute of Technology, Department of Electrical Engineering and Computer Science

The process of forest harvesting is highly mechanized in most industrialized countries, with felling and processing of trees performed by technologically advanced forestry machines. However, the maneuvering of the vehicles through the forest as well as the control of the on-board hydraulic boom crane is currently performed through continuous manual operation. This complicates the introduction of further incremental productivity improvements to the machines, as the operator becomes a bottleneck in the process. A suggested solution strategy is to enhance the production capacity by increasing the level of automation. At the same time, the working environment for the operator can be improved by a reduced workload, provided that the human-machine interaction is adapted to the new automated functionality. The objectives of this thesis are 1) to describe and analyze the current logging process and to locate areas of improvements that can be implemented in current machines, and 2) to investigate future methods and concepts that possibly require changes in work methods as well as in the machine design and technology. The thesis describes the development and integration of several algorithmic methods and the implementation of corresponding software solutions, adapted to the forestry machine context. Following data recording and analysis of the current work tasks of machine operators, trajectory planning and execution for a specific category of forwarder crane motions has been identified as an important first step for short term automation. Using the method of path-constrained trajectory planning, automated crane motions were demonstrated to potentially provide a substantial improvement from motions performed by experienced human operators. An extension of this method was developed to automate some selected motions even for existing sensorless machines. Evaluation suggests that this method is feasible for a reasonable deviation of initial conditions. Another important aspect of partial automation is the human-machine interaction. For this specific application a simple and intuitive interaction method for accessing automated crane motions was suggested, based on head tracking of the operator. A preliminary interaction model derived from user experiments yielded promising results for forming the basis of a target selection method, particularly when combined with some traded control strategy. Further, a modular software platform was implemented, integrating several important components into a framework for designing and testing future interaction concepts. Specifically, this system was used to investigate concepts of teleoperation and virtual environment feedback. Results from user tests show that visual information provided by a virtual environment can be advantageous compared to traditional video feedback with regards to both objective and subjective evaluation criteria.

Production equipment accounts for a large portion of the energy use from industry. But so far there has been no standardized way of requiring energy efficiency when purchasing a new machine. Scania is therefore implementing energy efficiency requirements in their purchasing process for production equipment. As a part of this, there needs to be a way of validating that the requirements have been fulfilled. This study aims to find how requirements on energy efficiency in production equipment can be validated in a user friendly and time efficient way. Firstly, the energy efficiency requirements set by Scania and by regulations are mapped. Then these requirements are clearly defined to enable a validation. Two component-level measurements of one machine tool and one industrial washing machine are analyzed. And then a cost analysis is conducted to determine the timespan that can be said to be time efficient for a validation procedure. The results from this are used to develop a validation method and an interactive protocol to make the validation more user friendly. This method is then tested through a simulated validation. The method proposed consists of two parts, an inspection and a measurement. The inspection is purely visual and validates the requirements on efficiency class for electrical motors and pumps, as well as requirements of specific equipment. The measurement is performed by running the machine through four different machine states in eight steps and validates requirements on when energy is used, and how much is used. The proposed method validates all energy efficiency requirements set by Scania for machine tools and industrial washing machines. It can be performed in a timespan that is far shorter than what is cost efficient. The proposed method can validate requirements on the energy use from any electrical components, compressed air use, and visually confirm that required equipment is present and some of its properties based on labelling. It will also be able to validate any new requirements on the energy use of electrical components, meaning it can easily be applied to other types of production equipment.
Produktionsutrustning står för en stor andel av energianvändningen inom industrin. Men än så länge finns det inget standardiserat sätt att kravställa energieffektivitet vid inköp av nya maskiner. Scania har därför börjat implementera krav på energieffektivitet i deras inköpsprocess för produktionsutrustning. Som en del av detta behövs ett sätt att validera att de ställda kraven också uppfylls. Denna studie undersöker hur krav på energieffektivitet kan valideras på ett användarvänligt och tidseffektivt sätt. Först kartläggs de energieffektivitetskrav som ställs av Scania och lagstiftning. Dessa krav definieras sedan så tydligt som möjligt för att möjliggöra en validering. Två mätningar av energianvändning på komponentnivå på en bearbetningsmaskin och en industriell tvättmaskin analyseras. Sedan utförs en kostnadsanalys för att avgöra ett tidsspann som kan sägas vara tidseffektivt för en valideringsprocess. Resultaten från detta används sedan för att utveckla en valideringsmetod och ett interaktivt protokoll. Denna metoden testas sedan genom en simulerad validering. Den föreslagna metoden består av två delar, en inspektion och en mätning. Inspektionen är endast visuell och validerar kraven på effektivitetsklass på motorer och pumpar, samt krav på specifik utrustning. Mätningen utförs genom att köra maskinen genom fyra olika maskinlägen i åtta steg och validerar krav på när energi används, och hur mycket som används. Den föreslagna metoden validerar alla krav på energieffektivitet som Scania ställer på bearbetningsmaskiner och industriella tvättmaskiner. Den kan utföras under ett tidsspann som är mycket kortare än gränsen för vad som är kostnadseffektivt. Den föreslagna metoden kan validera krav på energianvändning från elektriska komponenter, tryckluftsanvändning, och visuellt bekräfta att kravställd utrustning är på plats och vissa egenskaper baserat på märkningen. Metoden kommer också att kunna validera alla nya krav på energianvändning från elektriska komponenter, vilket innebär att den enkelt kan appliceras på andra typer av produktionsutrustning.

This thesis introduces the development of an integrated system for the design of layouts for special purpose machines (SPMs). SPMs are capable of performing several machining operations (such as drilling, milling, and tapping) at the same time. They consist of elements that can be arranged in different layouts. Whilst this is a unique feature that makes SPMs modular, a high level of knowledge and experience is required to rearrange the SPM elements in different configurations, and also to select appropriate SPM elements when product demand changes and new layouts are required. In this research, an integrated system for SPM layout design was developed by considering the following components: an expert system tool, an assembly modelling approach for SPM layouts, an artificial intelligence tool, and a CAD design environment. SolidWorks was used as the 3D CAD environment. VisiRule was used as the expert system tool to make decisions about the selection of SPM elements. An assembly modelling approach was developed with an SPM database using a linked list structure and assembly relationships graph. A case-based reasoning (CBR) approach was developed and applied to automate the selection of SPM layouts. These components were integrated using application programing interface (API) features and Visual Basic programming language. The outcome of the application of the novel approach that was developed in this thesis is reducing the steps for the assembly process of the SPM elements and reducing the time for designing SPM layouts. As a result, only one step is required to assemble any two SPM elements and the time for the selection process of SPM layouts is reduced by approximately 75% compared to the traditional processes. The integrated system developed in this thesis will help engineers in design and manufacturing fields to design SPM layouts in a more time-effective manner.

In the last decades, the vibration analysis has been exploited for monitoring many mechanical systems for industrial applications. Although several works demonstrated how the vibration based diagnostics may reach satisfactory results, the nowadays industrial scenario is deeply changing, driven by the fundamental need of time and cost reduction. In this direction, the academic research has to focus on the improvement of the computational efficiency for the signal processing techniques applied in the mechanical diagnostics field. In the same way, the industrial word requires an increasing attention to the predictive maintenance for estimating the system failure avoiding unnecessary machine downtimes for maintenance operations. In this contest, in the recent years the research activity has been moved to the development of prognostic models for the prediction of the remaining useful life. However, it is important to keep in mind how the two fields are strictly connected, being the diagnostics the base on which build the effectiveness of each prognostic model. On these grounds, this thesis has been focused on these two different but linked areas for the detection and prediction of possible failures inside rotating machines in the industrial framework. The first part of the thesis focuses on the development of a blind deconvolution indicator based on the cyclostationary theory for the fault identification in rotating machines. The novel criterion aims to decrease the computational cost of the blind deconvolution through the exploitation of the Fourier-Bessel series expansion due to its modulated nature more comparable with the fault related vibration pattern. The proposed indicator is extensively compared to the other cyclostationary one based on the classic Fourier transform, taking into account both synthesized and real vibration signals. The comparison proves the improvement given by the proposed criterion in terms of number of operations required by the blind deconvolution algorithm as well as its diagnostic capability also for noisy measured signals. The originality of this part regards the combination of cyclostationarity and Fourier-Bessel transform that leads to the definition of a novel blind deconvolution criterion that keeps the diagnostic effectiveness of cyclostationarity reducing the computational cost in order to meet the industrial requirements. The second part regards the definition of a novel prognostic model from the family of the hidden Markov models constructed on a generalized Gaussian distribution. The target of the proposed method is a better fitting quality of the data distribution in the last damaging phase. In fact, the fault appearance and evolution reflects on a modification of the observation distribution within the states and consequently a generalized density function allows the changes on the distribution form through the values of some model parameters. The proposed method is compared in terms of fitting quality and state sequence prediction to the classic Gaussian based hidden Markov model through the analysis of several run to failure tests performed on rolling element bearings and more complex systems. The novelty of this part regards the definition of a new iterative algorithm for the estimation of the generalized Gaussian model parameters starting from the observations on the physical system for both monovariate and multivariate distributions. Furthermore, the strictly connection between diagnostics and prognostics is demonstrated through the analysis of a not monotonically increasing damaging process proving how the selection of a suitable indicator enables the correct health state estimation.
Negli ultimi decenni, l’analisi vibrazionale è stata sfruttata per il monitoraggio di molti sistemi meccanici per applicazioni industriali. Nonostante molte pubblicazioni abbiano dimostrato come la diagnostica vibrazionale possa raggiungere risultati soddisfacenti, lo scenario industriale odierno è in profondo cambiamento, guidato dalla necessità di ridurre tempi e costi produttivi. In questa direzione, la ricerca deve concentrarsi sul miglioramento dell’efficienza computazionale delle tecniche di analisi del segnale applicate a fini diagnostici. Allo stesso modo, il mondo industriale richiede una sempre maggior attenzione per la manutenzione predittiva, al fine di stimare l’effettivo danneggiamento del sistema evitando così inutili fermi macchina per operazioni manutentive. In tale ambito, negli ultimi anni l’attività di ricerca si sta spostando verso lo sviluppo di modelli prognostici finalizzati alla stima della vita utile residua dei componenti. Tuttavia, è importante ricordare come i due ambiti siano strettamente connessi, essendo la diagnostica la base su cui fondare l’efficacia di ciascun modello prognostico. Su questa base, questa tesi è stata incentrata su queste due diverse, ma tra loro connesse, aree al fine di identificare e predire possibile cause di cedimento su macchine rotanti per applicazioni industriali. La prima parte della tesi è concentrata sullo sviluppo di un nuovo indicatore di blind deconvolution per l’identificazione di difetti su organi rotanti sulla base della teoria ciclostazionaria. Il criterio presentato vuole andare a ridurre il costo computazionale richiesto dalla blind deconvolution tramite l’utilizzo della serie di Fourier-Bessel grazie alla sua natura modulata, maggiormente affine alla tipica firma vibratoria del difetto. L’indicatore proposto viene accuratamente confrontato con il suo analogo basato sulla classica serie di Fourier considerando sia segnali simulati che segnali di vibrazione reali. Il confronto vuole dimostrare il miglioramento fornito dal nuovo criterio in termini sia di minor numero di operazioni richieste dall’algoritmo che di efficacia diagnostica anche in condizioni di segnale molto rumoroso. Il contributo innovativo di questa parte riguarda la combinazione di ciclostazionarietà e serie di Furier-Bessel che porta alla definizione di un nuovo criterio di blind deconvolution in grado di mantenere l’efficacia diagnostica della ciclostazionarietà ma con un minor tempo computazionale per venire incontro alle richieste del mondo industriale. La second parte riguarda la definizione di un nuovo modello prognostico, appartenente alla famiglia degli hidden Markov models, costruito partendo da una distribuzione Gaussiana generalizzata. L’obbiettivo del metodo proposto è una miglior riproduzione della reale distribuzione dei dati, in particolar modo negli ultimi stadi del danneggiamento. Infatti, la comparsa e l’evoluzione del difetto comporta una modifica della distribuzione delle osservazioni fra i diversi stati. Di conseguenza, una densità di probabilità generalizzata permette la modificazione della forma della distribuzione tramite diversi valori dei parametri del modello. Il metodo proposto viene confrontato con il classico hidden Markov model di base Gaussiana in termini di qualità di riproduzione della distribuzione e predizione della sequenza di stati tramite l’analisi di alcuni test di rottura su cuscinetti volventi e sistemi complessi. L’innovatività di questa parte è data dalla definizione di un algoritmo iterativo per la stima dei parametri del modello nell’ipotesi di distribuzione Gaussiana generalizzata, sia nel caso monovariato che multivariato, partendo dalle osservazioni sul sistema fisico in esame.

La tradautomaticitat és el fenomen lingüístic que fa que les traduccions automàtiques sonin a màquina. Aquesta tesi introdueix el concepte de tradautomaticitat com un objecte de recerca i presenta un mètode d'avaluació que consisteix en determinar si la traducció és pròpia d'una màquina en comptes de determinar la seva semblança amb una traducció humana, com en els mètodes d'avaluació actuals. El mètode avalua la qualitat d'una traducció amb una mètrica, la MTS (Machine Translationness Score). Aquesta mètrica és conseqüent amb la percepció de la tradautomaticitat de la gent corrent. La MTS correlaciona bé amb les valoracions de qualitat dels avaluadors humans. A més, la nostra proposta permet realitzar avaluacions de baix cost perquè no necessiten de recursos que són cars d'obtenir (traduccions de referència, corpus d'entrenament, etc.). El criteri de tradautomaticitat té aplicacions que van més enllà de l'avaluació de traduccions automàtiques (detecció de plagi, detecció de publicacions no supervisades a Internet, etc.).
La tradautomacidad es el fenómeno lingüístico que hace que las traducciones automáticas suenen a máquina. Esta tesis introduce el concepto de tradautomaticidad como un objeto de investigación y presenta un método de evaluación que consiste en determinar si la traducción es propia de una máquina en vez de determinar su parecido a una traducción humana, como en los métodos de evaluación actuales. El método evalúa la calidad de una traducción con una métrica, la MTS (Machine Translationness Score). Esta métrica es consecuente con la percepción de la tradautomaticidad de la gente corriente. La MTS correlaciona bien con las valoraciones de calidad de evaluadores humanos. Además, nuestra propuesta permite realizar evaluaciones de bajo coste porque no requieren de recursos que son caros de obtener (traducciones de referencia, corpus de entrenamiento, etc.). El criterio de tradautomaticidad tiene aplicaciones que van más allá de la evaluación de traducciones automáticas (detección de plagio, detección de publicaciones no supervisadas en Internet, etc.).
Machine translationness (MTness) is the linguistic phenomena that make machine translations distinguishable from human translations. This thesis introduces MTness as a research object and presents an MT evaluation method based on determining whether the translation is machinelike instead of determining its humanlikeness as in current evaluation approaches. The method rates the MTness of a translation with a metric, the MTS (Machine Translationness Score). The MTS calculation is in accordance with the results of an experimental study on machine translation perception by common people. MTS proved to correlate well with human ratings on translation quality. Besides, our approach allows the performance of cheap evaluations since expensive resources (e.g. reference translations, training corpora) are not needed. Machine translationness ratings can be applied for other uses beyond machine translation evaluation (plagiarism and other forms of cheating, detection of unsupervised MT documents published on the Web, etc.).

Thesis: Ph. D., Massachusetts Institute of Technology, Department of Aeronautics and Astronautics, 2015.
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 135-143).
I envision a system that enables successful collaborations between humans and machine learning models by harnessing the relative strength to accomplish what neither can do alone. Machine learning techniques and humans have skills that complement each other - machine learning techniques are good at computation on data at the lowest level of granularity, whereas people are better at abstracting knowledge from their experience, and transferring the knowledge across domains. The goal of this thesis is to develop a framework for human-in-the-loop machine learning that enables people to interact effectively with machine learning models to make better decisions, without requiring in-depth knowledge about machine learning techniques. Many of us interact with machine learning systems everyday. Systems that mine data for product recommendations, for example, are ubiquitous. However these systems compute their output without end-user involvement, and there are typically no life or death consequences in the case the machine learning result is not acceptable to the user. In contrast, domains where decisions can have serious consequences (e.g., emergency response panning, medical decision-making), require the incorporation of human experts' domain knowledge. These systems also must be transparent to earn experts' trust and be adopted in their workflow. The challenge addressed in this thesis is that traditional machine learning systems are not designed to extract domain experts' knowledge from natural workflow, or to provide pathways for the human domain expert to directly interact with the algorithm to interject their knowledge or to better understand the system output. For machine learning systems to make a real-world impact in these important domains, these systems must be able to communicate with highly skilled human experts to leverage their judgment and expertise, and share useful information or patterns from the data. In this thesis, I bridge this gap by building human-in-the-loop machine learning models and systems that compute and communicate machine learning results in ways that are compatible with the human decision-making process, and that can readily incorporate human experts' domain knowledge. I start by building a machine learning model that infers human teams' planning decisions from the structured form of natural language of team meetings. I show that the model can infer a human teams' final plan with 86% accuracy on average. I then design an interpretable machine learning model then "makes sense to humans" by exploring and communicating patterns and structure in data to support human decision-making. Through human subject experiments, I show that this interpretable machine learning model offers statistically significant quantitative improvements in interpretability while preserving clustering performance. Finally, I design a machine learning model that supports transparent interaction with humans without requiring that a user has expert knowledge of machine learning technique. I build a human-in-the-loop machine learning system that incorporates human feedback and communicates its internal states to humans, using an intuitive medium for interaction with the machine learning model. I demonstrate the application of this model for an educational domain in which teachers cluster programming assignments to streamline the grading process.
by Been Kim.
Ph. D.