ED Sciences Physiques et de l'Ingénieur
Lossless data compression : application to vibration analysis of aeronautical equipment and radar imagery.
by Guillaume COTTIN (Laboratoire de l'Intégration du Matériau au Système)
The defense will take place at 14h00 - Amphithéâtre Jean Paul Dom, IMS, Bâtiment A31 - 351 avenue de la libération, 33400, Talence
in front of the jury composed of
- Franck CAZAURANG - Professeur des universités - Université de Bordeaux - Directeur de these
- Mohamed-Chaker LARABI - Professeur des universités - Université de Poitiers - Rapporteur
- Yannis POUSSET - Professeur des universités - Université de Poitiers - Rapporteur
- Magalie THOMASSIN - Maîtresse de conférences - Université de Lorraine - Examinateur
- Jérôme SARACCO - Professeur des universités - Bordeaux INP - CoDirecteur de these
- Marc DONIAS - Maître de conférences - Bordeaux INP - Examinateur
Lossless data compression is a current issue affecting many areas of application. Faced with ever-increasing volumes of data, embedded systems must be able to store, transmit, and process this information while operating with limited computing resources. Complete data integrity is also essential for analysis and decision-making. Lossless compression ensures the integrity of these signals and is necessary in many sectors. This thesis therefore offers an in-depth study of lossless compression applied to these two types of data. The first part deals with vibration signals based on dynamic partitioning guided by Shannon entropy and adaptive Huffman coding. This methodology enables online compression that is efficient and compatible with the constraints of embedded systems. The second part extends this work to two-dimensional data: radar imaging. For this work, two approaches are explored. One vectorizes the image following an analysis of its intensity vector, which then allows the previous one-dimensional processing to be used according to the selected reading direction. The other is based on a decomposition into bit planes, exploiting the informational characteristics of the most significant bits to generate a compact representation of the image. These methods aim to optimize processing, storage, and transmission while maintaining the integrity of the radar information. This thesis therefore proposes original and effective lossless compression solutions that are suitable for heterogeneous data while meeting the critical requirements of embedded environments. It offers an innovative solution for managing growing volumes of vibration and radar data, contributing to both predictive maintenance and the operational efficiency of avionics systems.
Solving vehicle routing problems with drones in urban logistics: approach based on exact optimization methods and heuristics
by Sylvain LICHAU (Laboratoire de l'Intégration du Matériau au Système)
The defense will take place at 14h00 - Amphithéâtre E Université de Bordeaux, Bâtiment A29, 351 cours de la Libération, 33400 Talence
in front of the jury composed of
- Rémy DUPAS - Professeur - Université de Bordeaux - Directeur de these
- Eduardo UCHOA - Full professor - Universidade Federal Fluminense - CoDirecteur de these
- Francesca GUERRIERO - Full professor - University of Calabria - Rapporteur
- Roberto BALDACCI - Full professor - Hamad Bin Khalifa University - Rapporteur
- François CLAUTIAUX - Professeur - Université de Bordeaux - Examinateur
- Caroline PRODHON - Professeure - Université de Technologie de Troyes - Examinateur
Drones represent a promising complement to delivery logistics. They offer a fundamentally different performance profile than conventional trucks. They have lower operating costs and higher speeds, at the expense of reduced range and capacity. In scenarios where these limitations are manageable, drones can fully replace traditional delivery trucks. However, in most practical applications, these complementary vehicle types should be strategically combined to leverage their respective operational strengths. This thesis mainly proposes exact mathematical optimization methods for routing problems with drones. The first chapter gives an overview of the literature on routing problems with drones, in the field of operations research, with an emphasis on exact optimization methods. We highlight the variety of problems addressed by researchers and the need to increase the size of the instances solved to optimality. The second chapter briefly recalls the concepts of mathematical programming required for understanding branch-cut-and-price algorithms. We also describe the branch-cut-and-price algorithm used in this thesis, including state-of-the-art components and the labeling algorithm used to solve the pricing problems. The third chapter addresses the Two-Echelon Vehicle Routing Problem with Drones. We propose a new set-partitioning model, in which partial routes corresponding to drone movements are enumerated using a dynamic program. We propose an adaptation of the well-known rounded capacity cuts, specifically tailored to address this problem. In addition, we propose a heuristic branch-cut-and-price based on the exact algorithm. Computational experiments show that the exact algorithm significantly increases the size of optimally solved instances. To further evaluate the model, the algorithm is then extended to address the Park-and-Loop Routing Problem. The fourth chapter addresses the Truck-based Drone Delivery Routing Problem with Time Windows. We develop an exact branch-cut-and-price algorithm and a labeling algorithm to solve the pricing subproblem. We compare two alternative pricing models defined on distinct graphs and show that they are non-dominated: the most effective choice depends on instance characteristics. Computational experiments exhibit the efficiency of the proposed approaches. We also propose new challenging instances with increased drone capacity. The fifth chapter studies a variant of the Vehicle Routing Problem with Time Windows, which integrates two operationally relevant features: flexible departure times from the depot and strict limits on route duration. Although more generic, this problem can be assimilated to a drone routing problem in which the route duration limits the drone's range. We develop an exact branch-cut-and-price algorithm, featuring a bidirectional labeling algorithm as its primary component for solving the complex pricing subproblem. We also show how the related problem of minimizing total route duration can be treated. Computational experiments confirm the algorithm's performance.