This thesis investigates the novel idea of using a computer to create and optimise conceptual designs of a range of differently-shaped three-dimensional solid objects from scratch. An extensive literature review evaluates all related areas of research and reveals that no such system exists. The development of a generic evolutionary design system, using a genetic algorithm (GA) as its core, is then presented. The thesis describes a number of significant advances necessitated by the development of this system. Firstly, a new low-parameter spatial-partitioning representation of solid objects is introduced, which allows a wide range of solid objects to be appropriately defined and easily manipulated by a GA. Secondly, multiobjective optimisation is investigated to allow users to define design problems without fine-tuning large numbers of weights. As a result of this, the new concepts of acceptability, range-independence and importance are introduced and a new multiobjective ranking method is identified as being most appropriate. Thirdly, variable-length chromosomes in GAs are addressed, to allow the number of primitive shapes that define a design to be variable. This problem is overcome by the use of a new hierarchical crossover operator, which uses the new concept of a semantic hierarchy to reference chromosomes. Additionally, the thesis describes how the performance of the GA is improved by using an explicit mapping stage between genotypes and phenotypes, steady-state reproduction with preferential selection, and a new lifespan limiter. A library of modular evaluation software is also presented, which allows a user to define new design problems quickly and easily by picking combinations of modules to guide the evolution of designs. Finally, the feasibility of the generic evolutionary design of solid objects is demonstrated by presenting the successful evolution of both conventional and unconventional designs for fifteen different solid-object design tasks, e.g. tables, heatsinks, penta-prisms, boat hulls, aerodynamic cars.

In the neuronal circuits of natural and artificial agents, memory is usually implemented with recurrent connections, since recurrence allows past agent state to affect the present, on-going behavior. Here, an interesting question arises in the context of evolution: how reactive agents could have evolved into cognitive ones with internalized memory? This study strives to find an answer to the question by simulating neuroevolution on artificial neural networks, with the hypothesis that internalization of external material interaction can be a plausible evolutionary path leading to a fully internalized memory system. A series of computational experiments were performed to gradually verify the above hypothesis. The first experiment demonstrated the possibility that external materials can be used as memory-aids for a memoryless reactive artificial agents in a simple 1-dimensional environment. Here, the reactive artificial agents used environmental markers as memory references to be successful in the ball-catching task that requires memory. Motivated by the result of the first experiment, an extended experiment was conducted to tackle a more complex memory problem using the same principle of external material interaction. This time, the reactive artificial agents are tasked to remember the locations of food items and the nest in a 2-dimensional environment. Such path-following behavior is a trivial foraging strategy of various lower animals such as ants and fish. The final experiment was designed to show the evolution of internal recurrence. In this experiment, I showed the evolutionary advantage of external material interaction by comparing the results from neural network topology evolution algorithms with and without the material interaction mechanism. The result confirmed that the agents with external material interaction learned to solve the memory task faster and more accurately. The results of the experiments provide insights on the possible evolutionary route to an internalized memory. The use of external material interaction can help reactive artificial agents to go beyond the functionality restricted by their simple network structure. Moreover, it allows much faster convergence with higher accuracy than the topological evolution of the artificial agents. These results suggest one plausible evolutionary path from reactive, through external material interaction, to recurrent structure.

In this thesis, we use evolutionary robotics techniques to automatically design and synthesise behaviour for groups of simulated and real robots. Our contribution will be on the design of non-trivial individual and collective behaviour; decisions about solitary or social behaviour will be temporal and they will be interdependent with communicative acts. In particular, we study time-based decision-making in a social context: how the experiences of robots unfold in time and how these experiences influence their interaction with the rest of the group. We propose three experiments based on non-trivial real-world cooperative scenarios. First, we study social cooperative categorisation; signalling and communication evolve in a task where the cooperation among robots is not a priori required. The communication and categorisation skills of the robots are co-evolved from scratch, and the emerging time-dependent individual and social behaviour are successfully tested on real robots. Second, we show on real hardware evidence of the success of evolved neuro-controllers when controlling two autonomous robots that have to grip each other (autonomously self-assemble). Our experiment constitutes the first fully evolved approach on such a task that requires sophisticated and fine sensory-motor coordination, and it highlights the minimal conditions to achieve assembly in autonomous robots by reducing the assumptions a priori made by the experimenter to a functional minimum. Third, we present the first work in the literature to deal with the design of homogeneous control mechanisms for morphologically heterogeneous robots, that is, robots that do not share the same hardware characteristics. We show how artificial evolution designs individual behaviours and communication protocols that allow the cooperation between robots of different types, by using dynamical neural networks that specialise on-line, depending on the nature of the morphology of each robot. The experiments briefly described above contribute to the advancement of the state of the art in evolving neuro-controllers for collective robotics both from an application-oriented, engineering point of view, as well as from a more theoretical point of view.

<i>Phytophthora infestans</i>, the oomycete pathogen responsible for late blight of potato and tomato, is regarded as the biggest threat to global potato production and is thought to cost the industry around £6 billion annually. Traditionally, fungicides have been used to control the disease, but this is both economically and environmentally costly, as multiple chemical applications may be required during a single growing season. <i>P. infestans</i> has rapidly overcome genetic resistances introduced into cultivated potato from wild species. This provides the rationale for developing artificial resistance genes to create durable resistance to late blight disease.<i>Phytophthora</i> species secrete essential effectors into plant cells that target critical host cellular mechanisms to promote disease. One such <i>P. infestans</i> effector is AVR3a^KI which is recognised by the potato R3a protein, a member of the CC-NB-LRR type resistance gene family. However, the closely related virulent form, AVR3a^EM, which is homozygous in more than 70% of wild <i>P. infestans</i> isolates, evades this recognition. Domain swapping experiments have revealed that the LRR domain of R3a is involved in recognition of AVR3a^KI, as the CC-NB domain of an R3a-paralog which does not mediate recognition of AVR3a^KI, is able to induce a HR when combined with the LRR of wild-type R3a. However, a chimeric protein consisting of the CC-NB domain of a more distantly-related homolog of R3a and the LRR of domain of R3a, is unable to recognise AVR3a^KI, suggesting that function is achieved only when the different domains of an R protein are attuned to recognition and signalling. Gain-of-function variants of <i>R3a</i> (<i>R3a*</i>), engineered by an iterative process of error-prone PCR, DNA fragmentation, re-assembly of the leucine rich repeat (LRR)-encoding region of <i>R3a</i>, are able to recognise both forms of AVR3a. This gain-of-recognition is accompanied by a gain-of-mechanism, as shown by a cellular re-localisation from the cytoplasm to prevacuolar compartments upon perception of recognised effector forms. However, R3a* variants do not confer resistance to AVR3a^EM-carrying isolates of <i>P. infestans</i>.Future efforts will target the NB-ARC domain of R3a, in a bid to fine-tune the intra-cellular signalling of gain-of-recognition R3a* variants. It is hoped that a shuffled <i>R3a*</i> gene, capable of conferring resistance to <i>P. infestans</i> isolates harbouring AVR3a^EM, will provide durable late blight resistance when deployed in the field in combination with other mechanistically different R proteins.

An ambitious long-term goal for neuroevolution, which studies how artificial evolutionary processes can be driven to produce brain-like structures, is to evolve neurocontrollers with a high density of neurons and connections that can adapt and learn from past experience. Yet while neuroevolution has produced successful results in a variety of domains, the scale of natural brains remains far beyond reach. In this dissertation two extensions to the recently introduced Hypercube-based NeuroEvolution of Augmenting Topologies (HyperNEAT) approach are presented that are a step towards more brain-like artificial neural networks (ANNs). First, HyperNEAT is extended to evolve plastic ANNs that can learn from past experience. This new approach, called adaptive HyperNEAT, allows not only patterns of weights across the connectivity of an ANN to be generated by a function of its geometry, but also patterns of arbitrary local learning rules. Second, evolvable-substrate HyperNEAT (ES-HyperNEAT) is introduced, which relieves the user from deciding where the hidden nodes should be placed in a geometry that is potentially infinitely dense. This approach not only can evolve the location of every neuron in the network, but also can represent regions of varying density, which means resolution can increase holistically over evolution. The combined approach, adaptive ES-HyperNEAT, unifies for the first time in neuroevolution the abilities to indirectly encode connectivity through geometry, generate patterns of heterogeneous plasticity, and simultaneously encode the density and placement of nodes in space. The dissertation culminates in a major application domain that takes a step towards the general goal of adaptive neurocontrollers for legged locomotion.

<p>This master thesis investigates the principal capability of artificial evolution to produce tool use behavior in adaptive agents, excluding the application of life-time learning or adaptation mechanisms. Tool use is one aspect of complex behavior that is expected from autonomous agents acting in real-world environments. In order to achieve tool use behavior an agent needs to identify environmental objects as potential tools before it can use the tools in a problem-solving task. Up to now research in robotics has focused on life-time learning mechanisms in order to achieve this. However, these techniques impose great demands on resources, e.g. in terms of memory or computational power. All of them have shown limited results with respect to a general adaptivity. One might argue that even nature does not present any kind of omni-adaptive agent. While humans seem to be a good example of natural agents that master an impressive variety of life conditions and environments (at least from a human perspective, other examples are spectacular survivability observations of octopuses, scorpions or various viruses) even the most advanced engineering approaches can hardly compete with the simplest life-forms in terms of adaptation. This thesis tries to contribute to engineering approaches by promoting the application of artificial evolution as a complementing element with the presentation of successful pioneering experiments. The results of these experiments show that artificial evolution is indeed capable to render tool use behavior at different levels of complexity and shows that the application of artificial evolution might be a good complement to life-time approaches in order to create agents that are able to implicitly extract concepts and display tool use behavior. The author believes that off-loading at least parts of the concept retrieval process to artificial evolution will reduce resource efforts at life-time when creating autonomous agents with complex behavior such as tool use. This might be a first step towards the vision of a higher level of autonomy and adaptability. Moreover, it shows the demand for an experimental verification of commonly accepted limits between qualities of learned and evolved tool use capabilities.</p>

This master thesis investigates the principal capability of artificial evolution to produce tool use behavior in adaptive agents, excluding the application of life-time learning or adaptation mechanisms. Tool use is one aspect of complex behavior that is expected from autonomous agents acting in real-world environments. In order to achieve tool use behavior an agent needs to identify environmental objects as potential tools before it can use the tools in a problem-solving task. Up to now research in robotics has focused on life-time learning mechanisms in order to achieve this. However, these techniques impose great demands on resources, e.g. in terms of memory or computational power. All of them have shown limited results with respect to a general adaptivity. One might argue that even nature does not present any kind of omni-adaptive agent. While humans seem to be a good example of natural agents that master an impressive variety of life conditions and environments (at least from a human perspective, other examples are spectacular survivability observations of octopuses, scorpions or various viruses) even the most advanced engineering approaches can hardly compete with the simplest life-forms in terms of adaptation. This thesis tries to contribute to engineering approaches by promoting the application of artificial evolution as a complementing element with the presentation of successful pioneering experiments. The results of these experiments show that artificial evolution is indeed capable to render tool use behavior at different levels of complexity and shows that the application of artificial evolution might be a good complement to life-time approaches in order to create agents that are able to implicitly extract concepts and display tool use behavior. The author believes that off-loading at least parts of the concept retrieval process to artificial evolution will reduce resource efforts at life-time when creating autonomous agents with complex behavior such as tool use. This might be a first step towards the vision of a higher level of autonomy and adaptability. Moreover, it shows the demand for an experimental verification of commonly accepted limits between qualities of learned and evolved tool use capabilities.

No description available.

Over the last 100 years sexual selection has advanced into a vast field of theoretical and empirical research. While Darwin’s idea of female preference being an integral mechanism of sexual selection is no longer debated, our understanding of female preference is still very limited. For example, we know little about the genetic variation in female preference, and the costs of preference over and above the costs of mating with particular male phenotypes. Additionally, while costs of mate choice are well documented, the benefits of mate choice and their implications are still debated. For example, controversy exists over the inevitability of good gene benefits and their capability to promote adaptive sexual selection. Furthermore, the adaptiveness of sexual selection itself is debated. Our understanding of the traits involved in mate choice is also far from complete. Here I investigated aspects of sexual selection in Drosophila simulans, employing a range of behavioural approaches along with artificial selection and environmental manipulations. The findings presented here indicate that female preference can evolve when directly selected on, and that preference itself is not particularly costly. There was also no conclusive evidence for the good genes benefits of mate choice in D. simulans. These benefits are considered crucial in promoting the adaptiveness of sexual selection, and although we found sexual selection to be adaptive under some test conditions it was not adaptive in other conditions. Our investigations into traits involved in mate choice established sex-specific genetic variation in cuticular hydrocarbons and the genetic architecture of this trait was found to sex-specific evolution of cuticular hydrocarbons under natural and sexual selection. Additionally, we found that a secondary sexual character, the sex combs was positively allometric – just like most signalling and weapon traits, and there was no association between trait fluctuating asymmetry and trait size. These findings collectively indicate that sexual selection in D. simulans is consistent with classical models of this process.

Le but de ce doctorat est de déterminer dans quelle mesure les algorithmes issus de l’intelligence artificielle, principalement les Algorithmes Evolutionnaires et la Programmation Génétique, pourraient aider les algorithmes de l’automatique classique afin de permettre aux engins autonomes de disposer de capacités bien supérieures, et ce dans les domaines de l’identification, de la planification de trajectoire, du pilotage et de la navigation.De nouveaux algorithmes ont été développés, dans les domaines de l’identification, de la planification de trajectoire, de la navigation et du contrôle, et ont été testés sur des systèmes de simulation et des aéronefs du monde réel (Oktokopter du ST2I, Bebop.Drone de la société Parrot, Twin Otter et F-16 de la NASA) de manière à évaluer les apports de ces nouvelles approches par rapport à l’état de l’art.La plupart de ces nouvelles approches ont permis d’obtenir de très bons résultats comparés à l’état de l’art, notamment dans le domaine de l’identification et de la commande, et un approfondissement des travaux devraient être engagé afin de développer le potentiel applicatifs de certains algorithmes. / The object of this PhD has consisted in elaborating evolutionary computing algorithms to find interesting solutions to important problems in several domains of automation science, applied to aircrafts mission conduction and to understand what could be the advantages of using such approaches, compared to the state-of-the-art, in terms of efficiency, robustness, and effort of implementation.New algorithms have been developed, in Identification, Path planning, Navigation and Control and have been tested on simulation and on real world platforms (AR.Drone 3.0 UAV (Parrot), Oktokopter UAV, Twin Otter and military fighter F-16 (NASA LaRC)), to assess the performances improvements, given by the new proposed approaches.Most of these new approaches provide very interesting results; and research work (on control by evolutionary algorithms, identification by genetic programming and relative navigation) should be engaged to plan potential applications in different real world technologies.