Nous abordons dans cette thèse un problème important en simulation multi-agent pour l'étude des systèmes complexes: celui d'assembler de multiples expertises par une approche multi-niveau. Alors que les approches existantes considèrent habituellement la vue d'un seul expert principal sur le système, nous proposons d'utiliser une approche multi-niveau pour intégrer plusieurs expertises sous la forme d'agents de différents niveaux d'abstraction. Nous montrons qu'il est ainsi possible de rester proche des concepts manipulés par les différents experts (ce qui permet de faciliter le processus de validation dans leurs domaines respectifs) et de combiner les différents niveaux de ces concepts, de manière à ce que chaque expert puisse comprendre les dynamiques des éléments liés à son domaine. Nous proposons le méta-modèle SIMLAB basé sur une représentation unifiée des concepts par des agents pouvant s'influencer les uns les autres dans différents axes et différents niveaux. Ce travail est concrétisé dans le cadre de l'étude de l'activité humaine en relation avec la consommation électrique. Il s'agit là d'un exemple typique de système complexe nécessitant de multiples expertises issues de différents domaines tels que l'ergonomie, l'énergétique, la sociologie, la thermique, ... Dans ce contexte, nous présentons ensuite la mise en oeuvre de notre approche dans la plate-forme SMACH de simulation des comportements humains et nous décrivons un ensemble d'expérimentations illustrant les différentes caractéristiques de notre approche. Nous montrons enfin la capacité de SIMLAB à reproduire et à étendre en simulation une étude réalisée sur le terrain de gestion de la demande énergétique. / The purpose of this work is to tackle a key problem in the study of complex systems when using multi-agent simulation: how to assemble several domains of expertise with a multi-level approach. While existing approaches usually consider the viewpoint of a unique main expert, we propose to use a multi-level model to integrate the multiple domains of expertise embodied in agents located at different abstraction levels. In this work, we show that it is possible to both stay close to the concepts manipulated by the experts (for the sake of the validation process in the domain of each expert) and combine the levels of those concepts. That way, each expert can easily understand the dynamics of the components related to their domain.We present SIMLAB, our meta-model based on a unified representation of the concepts using agents. Each agent can influence the others on different axes and levels. This work is materialised in a study of human activity in relation to electrical consumption. It is a typical example of complex system which requires many domains of expertise such as psychology, energetics, sociology, heat science, … In this context, we present the implementation of our approach in SMACH, a simulation platform of human behaviours. We Then describe several experiments to illustrate the characteristics of our approach. Finally, we show how SIMLAB can reproduce and extend in silico a field study of energy demand management.

Les décès du cancer de la peau sont majoritairement des mélanomes malins. Il est considéré comme l’un des plus dangereux cancer. A ses débuts, les mélanomes malins sont traités avec des simples biopsies et sont complètement curable. Pour cela, une détection précoce est la meilleure solution pour réduire ses conséquences désastreuses. Imagerie médicale telle que la dermoscopie et les caméras à images standard sont les outils disponibles les plus adaptées pour diagnostiquer précocement les mélanomes. Le diagnostic assisté par ordinateur (CAD) est développé dans le but d’accompagner les radiologistes dans la détection et le diagnostic. Cependant, il y a un besoin d’améliorer la précision de la segmentation et de détection des lésions. Dans ce travail, le modèle de Chan et Vese a été adapté pour segmenter davantage les variations à l’intérieur des lésions avec un résultats très encouragent. La deuxième tâche consiste à extraire des caractéristiques afin de discriminer les mélanomes. Deux méthodes ont été développée, une se basant sur l’irrégularité des bords des lésions et l’autre par la fusion des caractéristiques texturales et structurelles. Les résultats ont montrés de bonnes performances avec une précision de 86.54% et de 86.07%, respectivement. / The vast majority of skin cancer deaths are due to malignant melanoma. It is considered as one of the most dangerous cancers. In its early stages, malignant melanoma is completely curable with a simple biopsy. Therefore, an early detection is the best solution to improve skin cancer prognostic. Medical imaging such as dermoscopy and standard camera images are the most suitable tools available to diagnose melanoma at early stages. To help radiologists in the diagnosis of melanoma cases, there is a strong need to develop computer aided diagnosis (CAD) systems. The accurate segmentation and classification of pigment skin lesions still remains a challenging task due to the various colors and structures developed randomly inside the lesions. The current work focused on two main tasks. Inthe first task, a new approach of the segmentation of skin lesions based on Chan and Vese model is developed. The model is adapted to segment the variations of the pigment inside the lesion and not only the main border. The subjective evaluation, applied on a database of standard camera images, obtained a very encouraging results with 97.62% of true detection rate. In the second main task, two feature extraction methods were developed for the analysis of standard camera and dermoscopy images. The method developed for the standard camera skin cancer images is based on border irregularities, introducing two new concepts, which are valleys and crevasses as first and second level of the border irregularity. The method has been implemented on DermIs and DermQues, two databases of standard camera images, and achieved an accuracy of 86.54% with a sensitivity of 80% and a specificity of 95.45%. The second method consisted of a fusion of structural and textural features. The structural features were extracted from wavelet and curvelet coefficients, while the textural features were obtained from the local binary pattern operator. The method has been implemented on the PH2 database for dermoscopy images with 1000-random sampling cross validation. The obtained results achieved an accuracy, a sensitivity and a specificity of 86:07%, 78.93% and 93.25%. Compared to the existing methods, the proposed methods in this work show very good performances.

Indiana University-Purdue University Indianapolis (IUPUI) / Topology Optimization is a technique that allows for the obtaining structures which maximize the use of the material. This is done by intelligently deciding the binary distribution of solid material and void, in a discretized given space. Several researchers have provided methods to tackle binary topology optimization. New ef- forts are focused on extending the application for multi-phase optimizations. At the industrial level, several components designed are made up of more than one material to reduce weight and production costs. The objective of this work is to implement the algorithm of Hybrid Cellular Automaton for multi-material topology optimiza- tion. The commonly used interpolation rule SIMP, which allows to relate the design variables to the mechanical properties of the material, is replaced by ordered SIMP interpolation function. The multiple volume constraints are applied sequentially, starting with the most elastic material. When a constraint is satisfied, the elements assigned to this material remain passive by a defined number of iterations to promote the convergence of the solution. Examples are shown for static and dynamic loads. The work demonstrates the versatility of algorithms based on control systems to solve problems of multiple phases and transient response fields.

The desire to have enhanced and increased feature sets in embedded applications has contributed to a significant increase in the computational demands of such systems over the years. To support such demand and yet maintain reasonable power/energy budgets, the industry has begun a shift to multi-core architectures even in the embedded systems domain. Embedded real-time applications such as Avionics and Automotive systems are no exception to this trend. Such systems have strict certification requirements of subsets of their functionality, which result in strict temporal constraints on those subsets, while other subsets may have less strict requirements. Migrating such {\em mixed criticality} systems from single-core to multi-core platforms is challenging because application/component isolation and freedom from interference among them must be guaranteed. Safe and efficient, architecture-aware mapping and scheduling of system components (e.g., partitions, tasks, etc. as relevant to a particular domain) on the multiple cores is at the center of any scheme to migrate such systems from single-core to multi-core platforms. In this dissertation, we propose, develop and evaluate a unified framework to automate the mapping and scheduling process with the consideration of several architectural and application level requirements/constraints (e.g., communication and cache conflicts among system components, constraints prohibiting the allocation of certain system components on the same core, etc.)

The use of multi-robot teams in the Robot Operating System (ROS) has encountered difficulty in advancement because of a lack of effective ways for the robots to communicate. Several attempts towards solving this problem have been made, but these approaches have had trouble with either low fault tolerance or high network load. The Gray Transceiver is an interface and communication protocol for inter-robot communication using ROS. The Gray Transceiver leverages multicasting for reduced network load and increased fault tolerance. Results from simulations, high throughput testing, and live multi-robot evaluations are included. The live mult-robot and simulation evaluations show that it functions properly operating across multiple robots while tolerating faults. The high throughput test shows how the Gray Transceiver operates under high load across a several types of conditions.

No description available.

Multi-task learning (MTL) is a rapidly growing field in the world of autonomous vehicles, particularly in the area of computer vision. Autonomous vehicles are heavily reliant on computer vision technology for tasks such as object detection, object segmentation, and object tracking. The complexity of sensor data and the multiple tasks involved in autonomous driving can make it challenging to design effective systems. MTL addresses these challenges by training a single model to perform multiple tasks simultaneously, utilizing shared representations to learn common concepts between a group of related tasks, and improving data efficiency. In this thesis, we proposed a scalable MTL system for object detection that can be used to construct any MTL network with different scales and shapes. The proposed system is an extension to the state-of-art algorithm called Mask RCNN. It is designed to overcome the limitations of learning multiple objects in multi-label learning. To demonstrate the effectiveness of the proposed system, we built three different networks using it and evaluated their performance on the state-of-the-art BDD100k dataset. Our experimental results demonstrate that the proposed MTL networks outperform a base single-task network, Mask RCNN, in terms of mean average precision at 50 (mAP50). Specifically, the proposed MTL networks achieved a mAP50 of 66%, while the base network only achieved 53%. Furthermore, we also conducted comparisons between the proposed MTL networks to determine the most efficient way to group tasks together in order to create an optimal MTL network for object detection on the BDD100k dataset.

A new high-resolution spatial discretization scheme is presented for use within the volume-ofluid (VOF) method. This scheme is an adaptation of the previously published boundary gradient maximization (BGM) scheme, with an extension for time-dependent solutions. The scheme does not explicitly reconstruct the interface, but rather resolves the values of the volume fraction in the area of the interface. The scheme is upwind biased to provide numerical stability, and the face values are limited to meet boundedness criteria and prevent variable overshoot. Unlike most other high-resolution schemes found in the literature, the proposed scheme does not use upwind-biased and downwind-biased “switching” to maintain stability. This thesis presents a number of test cases including 2-D and 3-D cases on both structured and unstructured grids. The results display the method’s ability to predict good shape of the volume fraction interface and resolve the volume fraction discontinuity.

This thesis is concentrated on parallelizing a sequential algorithm for finding k best non-intersecting local sequence alignments. In this thesis, the DNA local sequence alignment and the related problems are formally defined and efficient algorithms for solving these problems are presented. The problem have important applications in molecular biology. Based on the analysis of the characteristics of the local sequence alignment problem and a multi-transputer system, the problem was partitioned into subproblems and nicely mapped onto the transputer nodes. Then, an efficient parallel program is designed and implemented. By comparing the outputs of the sequential program and the parallel program, the performance of the parallel program is estimated. An average speedup of 6.3 is achieved on a 8-node configuration and an average speed-up of 11 is achieved on a 16-node configuration. / Thesis / Master of Engineering (ME)

The main contribution of this thesis is a new method of image compression based on a recently developed adaptive transform called Mixtures of Principal Components (MPC). Our multi-resolution extension of MPC-called Multi-Resolution Mixtures of Principal Components (MR-MPC) compresses and decompresses images in stages. The first stage processes the original images at very low resolution and is followed by stages that process the encoding errors of the previous stages at incrementally higher resolutions. To evaluate our multi-resolution extension of MPC we compared it with MPC and with the excellent performing wavelet based scheme called SPIHT. Fifty chest radiographs were compressed and compared to originals in two ways. First, Peak Signal to Noise Ratio (PSNR) and five distortion factors from a perceptual distortion measure called PQS were used to demonstrate that our multi-resolution extension of MPC can achieve rate distortion performance that is 220% to 720% better than MPC and much closer to that of SPIHT. And second, in a study involving 724 radiologists' evaluations of compressed chest radiographs, we found that the impact of MR-MPC and SPIHT at 25:1, 50:1, 75:1 on subjective image quality scores was less than the difference of opinion between four radiologists. / Thesis / Master of Science (MS)