Evolvable Hardware (EHW) is a promising area in electronics today. Evolutionary Algorithms (EA), together with a circuit simulation tool or real hardware, automatically designs a circuit for a given problem. The circuits evolved may have unconventional designs and be less dependent on the personal knowledge of a designer. Nowadays, EA are represented by Genetic Algorithms (GA), Genetic Programming (GP) and Evolutionary Strategy (ES). While GA is definitely the most popular tool, GP has rapidly developed in recent years and is notable by its outstanding results. However, to date the use of ES for analogue circuit synthesis has been limited to a few applications. This work is devoted to exploring the potential of ES to create novel analogue designs. The narrative of the thesis starts with a framework of an ES-based system generating simple circuits, such as low pass filters. Then it continues with a step-by-step progression to increasingly sophisticated designs that require additional strength from the system. Finally, it describes the modernization of the system using novel techniques that enable the synthesis of complex multi-pin circuits that are newly evolved. It has been discovered that ES has strong power to synthesize analogue circuits. The circuits evolved in the first part of the thesis exceed similar results made previously using other techniques in a component economy, in the better functioning of the evolved circuits and in the computing power spent to reach the results. The target circuits for evolution in the second half are chosen by the author to challenge the capability of the developed system. By functioning, they do not belong to the conventional analogue domain but to applications that are usually adopted by digital circuits. To solve the design tasks, the system has been gradually developed to support the ability of evolving increasingly complex circuits. As a final result, a state-of-the-art ES-based system has been developed that possesses a novel mutation paradigm, with an ability to create, store and reuse substructures, to adapt the mutation, selection parameters and population size, utilize automatic incremental evolution and use the power of parallel computing. It has been discovered that with the ability to synthesis the most up-to-date multi-pin complex analogue circuits that have ever been automatically synthesized before, the system is capable of synthesizing circuits that are problematic for conventional design with application domains that lay beyond the conventional application domain for analogue circuits.

This thesis is divided into two parts. The first part considers and develops some of the statistics used in genetic programming (GP) while the second uses those statistics to study and develop a form of incremental evolution and an early termination heuristic for GP. The first part looks in detail at success proportion, Koza's minimum computational effort, and a measure we rename "success effort". We describe and develop methods to produce confidence intervals for these measures as well as confidence intervals for the difference and ratio of these measures. The second part studies Jackson's fitness-based incremental evolution. If the number of fitness evaluations are considered (rather than the number of generations) then we find some potential benefit through reduction in the effort required to find a solution. We then automate the incremental evolution method and show a statistically significant improvement compared to GP with automatically defined functions (ADFs). The success effort measure is shown to have the critical advantage over Koza's measure as it has the ability to include a decreasing cost of failure. We capitalise on this advantage by demonstrating an early termination heuristic that again offers a statistically significant advantage.

<p>The aim of the main work in the thesis is to investigate Incremental Evolution methods for designing a suitable behavior arbitration mechanism for behavior-based (BB) robot controllers for autonomous mobile robots performing tasks of higher complexity. The challenge of designing effective controllers for autonomous mobile robots has been intensely studied for few decades. Control Theory studies the fundamental control principles of robotic systems. However, the technological progress allows, and the needs of advanced manufacturing, service, entertainment, educational, and mission tasks require features beyond the scope of the standard functionality and basic control. Artificial Intelligence has traditionally looked upon the problem of designing robotics systems from the high-level and top-down perspective: given a working robotic device, how can it be equipped with an intelligent controller. Later approaches advocated for better robustness, modifiability, and control due to a bottom-up layered incremental controller and robot building (Behavior-Based Robotics, BBR). Still, the complexity of programming such system often requires manual work of engineers. Automatic methods might lead to systems that perform task on demand without the need of expert robot programmer. In addition, a robot programmer cannot predict all the possible situations in the robotic applications. Automatic programming methods may provide flexibility and adaptability of the robotic products with respect to the task performed. One possible approach to automatic design of robot controllers is Evolutionary Robotics (ER). Most of the experiments performed in the field of ER have achieved successful learning of target task, while the tasks were of limited complexity. This work is a marriage of incremental idea from the BBR and automatic programming of controllers using ER. Incremental Evolution allows automatic programming of robots for more complex tasks by providing a gentle and easy-to understand support by expertknowledge — division of the target task into sub-tasks. We analyze different types of incrementality, devise new controller architecture, implement an original simulator compatible with hardware, and test it with various incremental evolution tasks for real robots. We build up our experimental field through studies of experimental and educational robotics systems, evolutionary design, distributed computation that provides the required processing power, and robotics applications. University research is tightly coupled with education. Combining the robotics research with educational applications is both a useful consequence as well as a way of satisfying the necessary condition of the need of underlying application domain where the research work can both reflect and base itself.</p>

The aim of the main work in the thesis is to investigate Incremental Evolution methods for designing a suitable behavior arbitration mechanism for behavior-based (BB) robot controllers for autonomous mobile robots performing tasks of higher complexity. The challenge of designing effective controllers for autonomous mobile robots has been intensely studied for few decades. Control Theory studies the fundamental control principles of robotic systems. However, the technological progress allows, and the needs of advanced manufacturing, service, entertainment, educational, and mission tasks require features beyond the scope of the standard functionality and basic control. Artificial Intelligence has traditionally looked upon the problem of designing robotics systems from the high-level and top-down perspective: given a working robotic device, how can it be equipped with an intelligent controller. Later approaches advocated for better robustness, modifiability, and control due to a bottom-up layered incremental controller and robot building (Behavior-Based Robotics, BBR). Still, the complexity of programming such system often requires manual work of engineers. Automatic methods might lead to systems that perform task on demand without the need of expert robot programmer. In addition, a robot programmer cannot predict all the possible situations in the robotic applications. Automatic programming methods may provide flexibility and adaptability of the robotic products with respect to the task performed. One possible approach to automatic design of robot controllers is Evolutionary Robotics (ER). Most of the experiments performed in the field of ER have achieved successful learning of target task, while the tasks were of limited complexity. This work is a marriage of incremental idea from the BBR and automatic programming of controllers using ER. Incremental Evolution allows automatic programming of robots for more complex tasks by providing a gentle and easy-to understand support by expertknowledge — division of the target task into sub-tasks. We analyze different types of incrementality, devise new controller architecture, implement an original simulator compatible with hardware, and test it with various incremental evolution tasks for real robots. We build up our experimental field through studies of experimental and educational robotics systems, evolutionary design, distributed computation that provides the required processing power, and robotics applications. University research is tightly coupled with education. Combining the robotics research with educational applications is both a useful consequence as well as a way of satisfying the necessary condition of the need of underlying application domain where the research work can both reflect and base itself.

Le doctorat s'inscrit dans le cadre d'une bourse CIFRE et d'un partenariat entre l'ENSTA Bretagne, l'IRISA et Thales Air Systems. Les préoccupations de ce dernier, et plus particulièrement de l'équipe de rattachement, sont de réaliser des systèmes à logiciels prépondérants embarqués. La complexité de ces systèmes et les besoins de compétitivité associés font émerger la notion de "Model-Based Product Lines(MBPLs)". Celles-ci tendent à réaliser une synergie de l'abstraction de l'Ingénierie Dirigée par les Modèles (IDM) et de la capacité de gestion de la capitalisation et réutilisation des Lignes de Produits (LdPs). La nature irrévocablement dynamique des systèmes réels induit une évolution permanente des LdPs afin de répondre aux nouvelles exigences des clients et pour refléter les changements des artefacts internes de la LdP. L'objectif de cette thèse est unique, maîtriser des incréments d'évolution d'une ligne de produits de systèmes complexes, les contributions pour y parvenir sont duales. La thèse est que 1) une variabilité multidimensionnelle ainsi qu'une modélisation relationnelle est requise dans le cadre de lignes de produits de systèmes complexes pour en améliorer la compréhension et en faciliter l'évolution (proposition d'un cadre générique de décomposition de la modélisation et d'un langage (DSML) nommé PLiMoS, dédié à l'expression relationnelle et intentionnelle dans les MBPLs), et que 2) les efforts de spécialisation lors de la dérivation d'un produit ainsi que l'évolution de la LdP doivent être guidé par une architecture conceptuelle (introduction de motifs architecturaux autour de PLiMoS et du patron ABCDE) et capitalisés dans un processus outillé semi-automatisé d'évolution incrémentale des lignes de produits par extension. / The PhD (CIFRE fundings) was supported by a partnership between three actors: ENSTA Bretagne, IRISA and Thales Air Systems. The latter's concerns, and more precisely the ones from the affiliation team, are to build embedded software-intensive systems. The complexity of these systems, combined to the need of competitivity, reveal the notion of Model-Based Product Lines (MBPLs). They make a synergy of the capabilities of modeling and product line approaches, and enable more efficient solutions for modularization with the distinction of abstraction levels and separation of concerns. Besides, the dynamic nature of real-world systems induces that product line models need to evolve continually to meet new customer requirements and to reflect changes in product line artifacts. The aim of the thesis is to handle the increments of evolution of complex systems product lines, the contributions to achieve it are twofolds. The thesis claims that i) a multidimensional variability and a relational modeling are required within a complex system product line in order to enhance comprehension and ease the PL evolution (Conceptual model modularization framework and PliMoS Domain Specific Modeling Language proposition; the language is dedicated to relational and intentional expressions in MBPLs), and that ii) specialization efforts during product derivation have to be guided by a conceptual architecture (architectural patterns on top of PLiMoS, e.g.~ABCDE) and capitalized within a semi-automatic tooled process allowing the incremental PL evolution by extension.