In distributed systems, dependency among different computations of an application leads to a problem of deciding the locations of computations. Spatial requirements of a computation can be expressed in terms of spatial relationships with other computations. This research presents programming abstractions and language constructs which can be used for specifying spatial coordination requirements for distributed computations. A spatial coordination middleware has been implemented for satisfying spatial coordination requirements of systems implemented using the Actor model of concurrent computation. Our approach abstracts spatial requirements of concurrent computations and provides key programming primitives for specifying these requirements. We have also implemented a number of higher level spatial coordination primitives which can be translated into the basic primitives. Spatial requirements can be specified using these primitives and then the runtime system converts them into a constraint satisfaction problem and satisfies them. Our approach reduces the programming complexity and provides a middleware which separates spatial requirements from functional code and enables the application programmer to change spatial requirements at runtime without effecting application's functionality. We have identified some of the high level primitives and provided a mechanism to develop high level primitives on top of the basic primitives. This thesis presents the rationale, design, implementation, and evaluation of spatial coordination. By comparing programs written with and without our spatial coordination primitives, we show how spatial coordination enables a programmer to specify spatial requirements declaratively and simplify the programming task. Experimental results demonstrate the performance of the approach, as the number of constraints increases.

Les humains sont continuellement exposés à des stimulations pour lesquelles leur système perceptivo-cognitif doit créer des représentations mnésiques. Tout en créant un code interne de composantes, ce système doit être en mesure de reconnaître, d'identifier, et de discriminer ces objets lors de prochaines occurrences. Ce processus s'effectue par la création et la mise à jour d'une mémoire épisodique d'exemplaires à dimensionnalité réduite. De plus, le système cognitif doit regrouper les objets similaires en catégories, tout en adaptant le contenu de la mémoire suite à l'ajout d'informations produit par la rencontre de nouveaux objets. Ces processus de niveau « objet » et « catégorie » s'effectuent de façon séparée, par le biais de deux mémoires. Jusqu'à maintenant, aucun modèle formel satisfaisant n'était en mesure de rendre compte de cette variété de comportements humains sans sacrifier la simplicité et l'élégance du système initial pour simuler l'un d'eux. Le modèle FEBAM (pour Feature-Extracting Bidirectional Associative Memory) a été créé dans le but de répondre à cette incapacité de beaucoup de modèles existants à effectuer des tâches cognitives et perceptuelles à l'aide d'un codage interne créé de façon autonome, comme le font les humains. Basé sur une architecture neuronale associative bidirectionnelle, FEBAM peut reproduire les comportements d'autres réseaux de neurones artificiels dont les processus dynamiques sont basés sur l'extraction de composantes, la création de bassins d'attracteurs, ou encore le partitionnement de données (« clustering »), et ce, en utilisant une seule architecture, règle de transmission et procédure d'apprentissage. Dans la présente thèse, il sera montré qu'avec un nombre minimal de principes définitoires, le modèle pourra effectuer des tâches telles que la création autonome d'un code interne de composantes, le développement autonome d'une mémoire d'exemplaires parfaits, ainsi que l'identification et la catégorisation autonomes. Il sera aussi montré, grâce à la proposition d'un mécanisme itératif de croissance de l'architecture, que les catégories créées par le réseau peuvent être réorganisées suite à la présentation de nouvelles informations perceptuelles au système. On montrera également que FEBAM préserve les capacités d'une mémoire autoassociative récurrente (dont il est inspiré), tout en améliorant certains des comportements de cette dernière. Le modèle FEBAM sera également étendu au cas supervisé. Dans ce cas, le modèle FEBAM-RA (RA pour Response Association), grâce à un module supplémentaire, associera les représentations internes des stimuli à leur identité ou à leur appartenance catégorielle prédéfinies. Cette extension se fera sans avoir à ajouter des principes définitoires: ainsi, on utilisera ici la même règle d'apprentissage, la même règle de transmission, et une généralisation de l'architecture de FEBAM. Grâce à cet ajout, le modèle sera en mesure de reproduire de façon qualitative l'effet de la pré-exposition perceptuelle sur la rapidité de l'apprentissage identificatif supervisé, ainsi que l'effet de difficulté de la tâche lorsque l'on compare l'identification et la catégorisation supervisées (dans une situation de tâches simultanées). La contribution principale de cette thèse repose donc dans la parcimonie des principes utilisés. En effet, grâce à un nombre minimal de postulats définitoires, on modélisera donc des processus de traitement d'objets et de catégories, et ce, de façon autonome ou supervisée. Ce projet de recherche constituant la première étape de développement de l'approche FEBAM, quelques améliorations à l'approche de base seront proposées. ______________________________________________________________________________ MOTS-CLÉS DE L’AUTEUR : Modélisation cognitive, Réseaux de neurones artificiels, Extraction de composantes, Catégorisation, Identification.

In distributed systems, dependency among different computations of an application leads to a problem of deciding the locations of computations. Spatial requirements of a computation can be expressed in terms of spatial relationships with other computations. This research presents programming abstractions and language constructs which can be used for specifying spatial coordination requirements for distributed computations. A spatial coordination middleware has been implemented for satisfying spatial coordination requirements of systems implemented using the Actor model of concurrent computation. Our approach abstracts spatial requirements of concurrent computations and provides key programming primitives for specifying these requirements. We have also implemented a number of higher level spatial coordination primitives which can be translated into the basic primitives. Spatial requirements can be specified using these primitives and then the runtime system converts them into a constraint satisfaction problem and satisfies them. Our approach reduces the programming complexity and provides a middleware which separates spatial requirements from functional code and enables the application programmer to change spatial requirements at runtime without effecting application's functionality. We have identified some of the high level primitives and provided a mechanism to develop high level primitives on top of the basic primitives. This thesis presents the rationale, design, implementation, and evaluation of spatial coordination. By comparing programs written with and without our spatial coordination primitives, we show how spatial coordination enables a programmer to specify spatial requirements declaratively and simplify the programming task. Experimental results demonstrate the performance of the approach, as the number of constraints increases.

In this thesis the action learning and generation problem on a humanoid robot is studied. Our aim is to realize action learning, generation and recognition in one system and our inspiration source is the mirror neuron hypothesis which suggests that action learning, generation and recognition share the same neural circuitry. Dynamic Movement Primitives, an efficient action learning and generation approach, are modified in order to fulfill this aim. The system we developed (1) can learn from multiple demonstrations, (2) can generalize to different conditions, (3) generates actions in a closed-loop and online fashion and (4) can be used for online action recognition. These claims are supported by experiments and the applicability of the developed system in real world is demonstrated through implementing it on a humanoid robot.

Visualization is the process of making something perceptible to the mind or imagination. The techniques for producing visual imagery of volumetric data have advanced immensely during the last decades to a point where each produced image can include an overwhelming amount of information. An increasingly viable solution to the limitations of the human sense of visual perception is to make use of not only vision, but also additional senses. This thesis presents recent work on the development of principles and algorithms for generating representations of volumetric data through the sense of touch for the purpose of visualization. The primary idea introduced in this work is the concept of yielding constraints, that can be used to provide a continuous set of shapes as a representation of features of interest in various types of volumetric data. Some of the earlier identified standard human exploratory procedures can then be used which enables natural, intuitive and effective interaction with the data. The yielding constraints concept is introduced, and an algorithm based on haptic primitives is described, which forms a powerful yet versatile implementation of the yielding constraints. These methods are also extended to handle time-varying, moving and low quality data. A framework for multimodal visualization has been built on the presented methods, and this is used to demonstrate the applicability and versatility of the work through several example applications taken from different areas.

Robotic motion planning aspires to match the ease and efficiency with which humans move through and interact with their environment. Yet state of the art robotic planners fall short of human abilities; they are slower in computation, and the results are often of lower quality. One stumbling block in traditional motion planning is that points and paths are often considered in isolation. Many planners fail to recognize that substantial shared information exists among path alternatives. Exploitation of the geometric and topological relationships among path alternatives can therefore lead to increased efficiency and competency. These benefits include: better-informed path sampling, dramatically faster collision checking, and a deeper understanding of the trade-offs in path selection. In path sampling, the principle of locality is introduced as a basis for constructing an adaptive, probabilistic, geometric model to influence the selection of paths for collision test. Recognizing that collision testing consumes a sizable majority of planning time and that only collision-free paths provide value in selecting a path to execute on the robot, this model provides a significant increase in efficiency by circumventing collision testing paths that can be predicted to collide with obstacles. In the area of collision testing, an equivalence relation termed local path equivalence, is employed to discover when the work of testing a path has been previously performed. The swept volumes of adjoining path alternatives frequently overlap, implying that a continuum of intermediate paths exists as well. By recognizing such neighboring paths with related shapes and outcomes, up to 90% of paths may be tested implicitly in experiments, bypassing the traditional, expensive collision test and delivering a net 300% boost in collision test performance. Local path equivalence may also be applied to the path selection problem in order to recognize higher-level navigation options and make smarter choices. This thesis presents theoretical and experimental results in each of these three areas, as well as inspiration on the connections to how humans reason about moving through spaces.

Universiẗat, Diss., 2004--München.

What is the difference between intention and action? To start answering this complex question, we have created a serious game that allows us to capture a large quantity of experimental data and study human behavior. In the game, users catch flies, presented to the left or to the right of the screen, by dragging the tongue of a frog across a touchscreen monitor. The movement of interest has a predefined starting point (the frog) and necessarily transits through a via-point (a narrow corridor) before it proceeds to the chosen left/right direction. Meanwhile, the game collects data about the movement performed by the player. This work is focused on the analysis of such movements. We try to find criteria that will allow us to predict (as early as possible) the direction (left/right) chosen by the player. This is done by analyzing kinematic information (e.g. trajectory, velocity profile, etc.). Also, processing such data according to the dynamical movement primitives approach, allows us to find further criteria that support a classification of human movement. Our preliminary results show that individually considered, participants tend to create and use stereotypical behaviors that can be used to formulate predictions about the subjects’ intention to reach in one direction or the other, early after the onset of the movement.

Wireless sensor networks have the potential for becoming a huge market. Ericsson predicts 50 billion devices interconnected to the Internet by the year 2020. Before that, the devices must be made to be able to withstand years of usage without having to change power source as that would be too costly. These devices are typically small, inexpensive and severally resource constrained. Communication is mainly wireless, and the wireless transceiver on the node is typically the most power hungry component. Therefore, reducing the usage of radio is key to long lifetime. In this thesis I identify four problems with the conventional broadcast primitive. Based on those problems, I implement a new communication primitive. This primitive is called Politecast. I evaluate politecast in three case studies: the Steal the Light toy example, a Neighbor Discovery simulation and a full two-month deployment of the Lega system in the art gallery Liljevalchs. With the evaluations, Politecast is shown to be able to massively reduce the amount of traffic being transmitted and thus reducing congestion and increasing application performance. It also prolongs node lifetime by reducing the overhearing by waking up neighbors.

This thesis deals with issues of secure authentication and key management in the Internet of Things. It describes basic protocols used in IoT, cryptographic primitives, communication technologies in IoT and end elements. It also includes a measuring the performance of cryptographic primitives on Raspberry Pi and selecting the appropriate LPWAN simulation technology. The conclusion of the work is devoted to the simulation of a LoRaWAN network