Search

Coordination in a group relies heavily on the type and quality of interactions and communication among individuals. In swarm robotics, communication can make the difference between a heap of isolated robots working independently of each other, and a connected swarm displaying self-organisation. Communication between robots in a swarm can be indirect, for instance through stigmergy whereby robots exploit the sign of previous actions to coordinate, or direct, by means of messages exchanged among robots for the purpose of influencing each other's behaviour. In the latter case, messages can consist either of simple signals, or more structured information, possibly encoding some concept representing features of the environment or of the desired coordination outcome. More complex communication can support more complex self-organising behaviors, deeply impacting on how the swarm tackles the task at hand. In this work, we consider different ways of exploiting communication in the context of a foraging task, in which robots search an open environment for resources to be exploited. Foraging requires abilities such as navigation, exploration and collective decision making. Coordination within a foraging context can lead to higher efficiency in exploiting resources, both in the short or in the long run, by avoiding over-exploitation. Throughout my thesis, foraging is used as a means to study the coupling between different communication processes and the undertaking of a meaningful task by the robots. Specifically, we study three different uses of communication during foraging. Firstly, we focus on simple aggregation of information, and study three parameter-free information processing mechanisms. These result in varying behavior, from the selection of a single resource by the whole swarm to the robots splitting among the resources present. This study is supported by an extensive analysis of navigation and congestion, helping to explain how swarm density can affect the perceived, Doctorat en Sciences de l'ingénieur et technologie, info:eu-repo/semantics/nonPublished

It’s difficult not to be amazed by the intricate process through which a honeybee swarm selects anew nest site. By exploring different options and sharing information with one another, the beescan collectively choose a single site. A key feature of these collective decisions is that they arenot directed by a centralized planner, but instead arise entirely from the interactions of individualswith each other and their environment. Inspired by such natural phenomena, roboticists are nowimplementing similar strategies in robot swarms to enhance collective decision-makingAs a social species, humans also frequently make collective decisions. Many important decisionsfor our societies are not made by a single individual but emerge from extensive verbal discussionsbetween multiple individuals. This dissertation aims to build a conceptual and experimentalbridge linking the study of human collective decision making with the more embodied collectivedecision making of animal and robotic swarms, which mostly take place through movements inphysical space. To do so, we place human participants in the type of embodied decision makingscenarios that swarms usually operate in.We created a virtual environment in which large groups of human participants can interact withone another and perform various tasks that are commonly studied in swarm robotics research. Intwo behavioral experiments, we used this virtual environment to study human behavior duringcollective decision making. In the first experiment, we studied how participants can form acollective perception of their environment by each exploring some part of the environment andthen sharing that information. In a subsequent experiment, we study how a group of participantscan select the best site in their environment while only using movement-based communication.The third study presented in this thesis reports on how human participants can implicitly signaltheir confidence in their movements during decision making in a shared physical, Doctorat en Sciences psychologiques et de l'éducation, info:eu-repo/semantics/nonPublished

In this work, we present an open-source unmanned aerial vehicle (UAV) platform for research in swarm robotics. In swarm robotics, groups of robots collaborate using local interactions and collectively solve tasks beyond an individual robot's capabilities. Individual robots must have onboard processing, communication, and sensing capabilities to autonomously react to their neighbors and immediate environment. Most research involving UAVs in swarm robotics presents only simulation results, while key landmark studies with real UAV swarms have used UAV platforms that were custom-built for the respective study. One important reason for this is that no commercial UAV platform comes pre-equipped with the ability to identify and track the positions and poses of nearby drones using only onboard sensors and computation, and in research platforms, the relevant sensing technologies are currently under development. Our aim is to provide a platform that allows swarm robotics researchers to test their algorithms on real UAVs, without having to develop their own custom-built UAVs or to wait until more advanced sensing technology is ready off-the-shelf. We provide a well-documented, entirely open-source UAV platform - S-drone (Swarm-drone) - to foster and support UAV swarm research in a laboratory environment. The S-drone uses fiducial markers in the environment and cooperative feature-level sensor fusion for inter-robot tracking to track the presence, identity, relative 2D position, and relative 2D orientation of neighboring peers. The S-drone is suitable for a wide range of contexts, supports quad-camera vision-based navigation and a variety of onboard sensing, and is extensible. It is especially suited for swarm robotics research because it can operate using strictly onboard processing and sensing without the need for global positioning systems, motion capture systems, or ground stations for off-board sensing., SCOPUS: ar.j, info:eu-repo/semantics/published

info:eu-repo/semantics/published

A metaheuristic is a collection of algorithmic concepts that can be used to define heuristic methods applicable to a wide set of optimization problems for which exact/analytical approaches are either limited or impractical. In other words, a metaheuristic can be considered a general algorithmic framework that can be easily adapted to different optimization problems. In this article, we discuss the two main approaches used to create new metaheuristics: manual design, which is based on the designer’s “intuition” and often involves looking for inspiration in other fields of knowledge, and automatic design, which seeks to reduce human involvement in the design process by harnessing recent advances in automatic algorithm configuration methods. In this context, we discuss the trend of manually designed “novel” metaphor-based metaheuristics inspired by natural, artificial, and even supernatural behaviors. In recent years, this trend has been strongly criticized due to the uselessness of new metaphors in devising truly novel algorithms and the confusion such metaheuristics have created in the literature. We then present automatic design as a powerful alternative to manual design that has the potential to render the “novel” metaphor-based metaheuristics trend obsolete. Finally, we examine several fundamental aspects of the field of metaheuristics and offer suggestions for improving them., info:eu-repo/semantics/published

In collective perception, agents sample spatial data and use the samples to agree on some estimate. In this paper, we identify the sources of statistical uncertainty that occur in collective perception and note that improving the accuracy of fully decentralized approaches, beyond a certain threshold, might be intractable. We propose self-organizing hierarchy as an approach to improve accuracy in collective perception by reducing or eliminating some of the sources of uncertainty. Using self-organizing hierarchy, aspects of centralization and decentralization can be combined: robots can understand their relative positions system-wide and fuse their information at one point, without requiring, e.g. a fully connected or static communication network. In this way, multi-sensor fusion techniques that were designed for fully centralized systems can be applied to a self-organized system for the first time, without losing the key practical benefits of decentralization. We implement simple proof-of-concept fusion in a self-organizing hierarchy approach and test it against three fully decentralized benchmark approaches. We test the perceptual accuracy of the approaches for absolute conditions that are uniform time-invariant, time-varying, and spatially nonuniform with high heterogeneity, as well as the scalability and fault tolerance of their accuracy. We show that, under our tested conditions, the self-organizing hierarchy approach is generally more accurate, more consistent, and faster than the other approaches and also that its accuracy is more scalable and comparably fault-tolerant. Under spatially nonuniform conditions, our results indicate that the four approaches are comparable in terms of similarity to the reference samples. In future work, extending these results to additional methods, such as collective probability distribution fitting, is likely to be much more straightforward in the self-organizing hierarchy approach than in the decentralized approaches., info:eu-repo/semantics/published

Self-organized groups of robots have generally coordinated their behaviors using quite simple social interactions. Although simple interactions are sufficient for some group behaviors, future research needs to investigate more elaborate forms of coordination, such as social cognition, to progress towards real deployments. In this perspective, we define social cognition among robots as the combination of social inference, social learning, social influence, and knowledge transfer, and propose that these abilities can be established in robots by building underlying mechanisms based on behaviors observed in humans. We review key social processes observed in humans that could inspire valuable capabilities in robots and propose that relevant insights from human social cognition can be obtained by studying human-controlled avatars in virtual environments that have the correct balance of embodiment and constraints. Such environments need to allow participants to engage in embodied social behaviors, for instance through situatedness and bodily involvement, but, at the same time, need to artificially constrain humans to the operational conditions of robots, for instance in terms of perception and communication. We illustrate our proposed experimental method with example setups in a multi-user virtual environment., SCOPUS: ar.j, info:eu-repo/semantics/published

Which off-line fully-automatic optimization-based design method produces control software that will yield the best performance once executed on a swarm of physical robots? This fundamental question cannot currently be answered due to the lack of two elements: i) an appropriate experimental protocol for the evaluation and comparison of fully-automatic design methods, and ii) a procedure that reliably predicts the real-world performance of control software. This dissertation focuses on addressing this void.The literature on optimization-based design of robot swarms suffers from the absence of a clearly established state of the art. It is in fact, for the most part, a collection of feasibility studies, and little effort has been devoted to the systematic empirical evaluation and comparison of the methods or ideas proposed. Recent papers have formally characterized two approaches to the optimization-based design that were entangled in the literature: the fully-automatic and the semi-automatic approaches. In light of this novel categorization, we show that the experimental protocols employed so far for the evaluation and comparison of design methods do not respect the tenets of fully-automatic design, and we propose one that does.One of the most challenging issues when designing control software off-line on the basis of a simulation model is the reality gap: the unavoidable discrepancies between the design model and reality. It is generally understood that, because of the reality gap, the design model overestimates the performance that control software eventually yields when executed on physical robots. As a result, conducting expensive and time consuming tests on physical robots is mandatory to reliably assess control software. We introduce the concept of pseudo-reality: a simulation model, different from the one used in the design, whose purpose is to evaluate control software. With this concept, we show via a series of experiments that the reality-gap problem is to be, Doctorat en Sciences de l'ingénieur et technologie, info:eu-repo/semantics/nonPublished

The aim of this dissertation is to investigate the role of optimization in the automatic modular design of control software for robot swarms. One of the main challenges in swarm robotics is to design the behavior of the individual robots so that a desired collective mission can be performed. Optimization-based design methods utilize an optimization algorithm to search for well-performing instances of control software. In optimization-based design, past research has mainly focused on proving the feasibility of optimization-based design methods for given missions. With this approach, researchers could tackle a wide range of missions. However, only a few works compare the role of the components of any chosen optimization-based design method. In particular, very little attention has been devoted to the optimization algorithm, arguably the central element in optimization-based design. In the context of my research, I focused on automatic modular design, an optimization-based design approach that combines modules into higher-level control architectures. Automatic modular design has shown to produce control software that not onlyperforms well in simulation but that also transfers well into reality.In this dissertation, I present a study of different types of optimization algorithms: local-search, model-free racing, and model-based. I defined three automatic modular design methods and compared them against state-of-the-art methods from the literature. I assessed and compared these design methods in experiments for several missions, both in simulation and on real robots. In particular, I showed that, while the choice of the optimization algorithm has an impact on the performance of the generated control software, it appears to not compromise the ability to cross the reality gap satisfactorily.The work presented in this dissertation represents a first step towards systematically investigating the role of optimization in optimization-based design. More work is still needed to, Doctorat en Sciences de l'ingénieur et technologie, info:eu-repo/semantics/nonPublished

A metaheuristic is a general algorithmic framework that can be easily adapted to many different optimization problems for which exact/analytical approaches are either limited or impractical. There are two main approaches used to create new metaheuristic implementations: manual design, which is based on the designer’s "intuition"—often involving looking for inspiration in other fields of knowledge—and automatic design, which seeks to remove human intervention from the design process by harnessing recent advances in automatic algorithm configuration methods and machine learning. Compared to manual design, which is typically time-consuming and error-prone, automatic design is often much more efficient and has advantages, such as removing biases from the creation process and increasing the flexibility with which new metaheuristic designs can be explored. Yet, to this day, the vast majority of metaheuristics proposed in the literature are still created by algorithm designers who manually define the different components that make up the implementation based on their knowledge (empirical, theoretical or intuitive) and/or by considering new sources of inspiration. The main topic of this thesis is how to improve the way we create metaheuristics by replacing manual design with automatic design. We focus specially on those cases where the use of manual design involves taking inspiration from natural and even supernatural behaviors, since this approach has been used intensively in the last two decades, spawning hundreds of so-called "novel" metaphor-based metaheuristics. Unfortunately, for the vast majority of them, it is still unknown what is their really novelty apart from the use of the new metaphors. In order to put an end to this situation and, indeed, to try to remedy it, we advocate for replacing "intuition" and "new sources of inspiration" with automated design methods. To demonstrate the feasibility of this approach and its advantages, we develop a metaheuristic softwa, Doctorat en Sciences de l'ingénieur et technologie, info:eu-repo/semantics/nonPublished

Through cooperation, robot swarms can perform tasks or solve problems that a single robot from the swarm could not perform/solve by itself. However, it has been shown that a single Byzantine robot (such as a malfunctioning or malicious robot) can disrupt the coordination strategy of the entire swarm. Therefore, a versatile swarm robotics framework that addresses security issues in inter-robot communication and coordination is urgently needed. Here, we show that security issues can be addressed by setting up a token economy between the robots. To create and maintain the token economy, we used blockchain technology, originally developed for the digital currency Bitcoin. The robots were given crypto tokens that allowed them to participate in the swarm’s security-critical activities. The token economy was regulated via a smart contract that decided how to distribute crypto tokens among the robots depending on their contributions. We designed the smart contract so that Byzantine robots soon ran out of crypto tokens and could therefore no longer influence the rest of the swarm. In experiments with up to 24 physical robots, we demonstrated that our smart contract approach worked: The robots could maintain blockchain networks, and a blockchain-based token economy could be used to neutralize the destructive actions of Byzantine robots in a collective-sensing scenario. In experiments with more than 100 simulated robots, we studied the scalability and long-term behavior of our approach. The obtained results demonstrate the feasibility and viability of blockchain-based swarm robotics., SCOPUS: ar.j, info:eu-repo/semantics/published

Hierarchical frameworks - a special class of directed frameworks with a layer-by-layer architecture - can be an effective mechanism to coordinate robot swarms. Their effectiveness was recently demonstrated by the mergeable nervous systems paradigm (Mathews et al. 2017), in which a robot swarm can switch dynamically between distributed and centralized control depending on the task, using self-organized hierarchical frameworks. New theoretical foundations are required to use this paradigm for formation control of large swarms. In particular, the systematic and mathematically analyzable organization and reorganization of hierarchical frameworks in a robot swarm is still an open problem. Although methods for framework construction and formation maintenance via rigidity theory exist in the literature, they do not address cases of hierarchy in a robot swarm. In this article, we extend bearing rigidity to directed topologies and extend the Henneberg constructions to generate self-organized hierarchical frameworks with bearing rigidity. We investigate three-key self-reconfiguration problems: 1) framework merging; 2) robot departure; and 3) framework splitting. We also derive the mathematical conditions of these problems and then develop algorithms that preserve rigidity and hierarchy using only local information. Our approach can be used for formation control generally, as in principle it can be coupled with any control law that makes use of bearing rigidity. To demonstrate and validate our proposed hierarchical frameworks and methods, we apply them to four scenarios of reactive formation control using an example control law., SCOPUS: ar.j, info:eu-repo/semantics/published

In swarm robotics, the design problem is one of the most pressing issues of today's research. Robots in a swarm must be autonomous and are usually individually programmed. However, the collective behavior of the swarm, which is expected to emerge from the interactions between the robots and between the robots and their environment, is best described at the level of the swarm. A great deal of effort has been devoted in an attempt to find suitable solutions to create control software for individual robots so that they achieve a particular collective behavior. One solution is automatic design, in which control software is generated automatically by an optimization process that configures a certain control software architecture. Traditionally, automatic design is accomplished through neuroevolution, that is, training a neural network in a simulated environment using an evolutionary algorithm, directly on the mission to tackle. Neuroevolutionary approaches have shown to be very effective, and work quite well when control software is designed and assessed in a simulated environment. However, they have been shown to be susceptible to the reality gap, the intrinsic and unavoidable difference between reality and the simulation environment used to generate the control software. In recent years, another approach has been proposed: modular design. In modular design, the control software is composed of modules that can be created once and for all and re-used for multiple missions. Restricting the space of possible behaviors, by limiting the control software to a combination of existing modules, effectively increases the robustness to the reality gap of modular methods in comparison to neuroevolutionary ones. However, modular methods require human expertise in swarm robotics in order to create the modules, making them more challenging to implement than neuroevolutionary methods. The objective of this thesis is to combine the advantages of neuroevolution and modular methods: the e, Doctorat en Sciences de l'ingénieur et technologie, info:eu-repo/semantics/nonPublished

Multi-agent systems (MAS) have been of interest to many researchers during the last decades. This thesis focuses on multi-robot systems (MRS) and programmable matter as two types of MAS. Regarding MRS, the focus is on the 'mergeable nervous system' (MNS) concept which allows the robots to connect to one another and establish a communication network through self-organization and then use the network to temporarily report sensing events and cede authority to a single robot in the system. Here, in a collective perception scenario, we experimentally evaluate the performance of an MNS enabled approach and compare it with that of several decentralized benchmark approaches. We show that an MNS-enabled approach is high-performing, fault-tolerant, and scalable, so it is an appropriate approach for MRS. As a goal of the thesis, using an MNS-enabled approach, we present for the first time a comprehensive comparison of control architectures in multi-robot systems, which includes a comparison of accuracy, efficiency, speed, energy consumption, scalability, and fault tolerance. Our comparisons provide designers of multi-robot systems with a better understanding for selecting the best-performing control depending on the system’s objectives. Additionally, as a separate goal, we design a high-level leader-based programmable matter, which can perform some basic primitive operations in a grid environment, and construct it using lower-level organisms. We design and implement deterministic algorithms for "curl" operation of this high-level matter, an instance of shape formation problem. We prove the correctness of the presented algorithms, analytically determine their complexity, and experimentally evaluate their performance., Doctorat en Sciences de l'ingénieur et technologie, info:eu-repo/semantics/nonPublished

Taking inspiration from natural behaviors to devise new optimization algorithms has played an important role in the history of the field of metaheuristics (Sörensen et al. 2017). Unfortunately, in the last two decades we have been witnessing a new trend by which dozens of metaphor-based metaheuristics based on the most diverse possible set of natural, artificial, social, and sometimes even supernatural phenomena and behaviors are proposed, without a clear motivation beyond the desire of their authors to publish their papers. Despite several attempts (Sörensen 2015; Sörensen et al. 2019; Weyland 2010; Piotrowski et al. 2014; Fong et al. 2016; Camacho Villalón et al. 2019; Camacho Villalón et al. 2020; Swan et al. 2015; Lones 2020; Tzanetos and Dounias 2020) to put an end to the flood of increasingly outlandish, “novel” metaphor-centered metaheuristics, there is still a steady inflow of such papers submitted, and often accepted for publication. For many reasons, we believe this is detrimental to the field of metaheuristics. This letter describes some of the negative effects of publishing such papers in the literature and proposes a necessary action to try to put a limit to this highly undesirable phenomenon.

Multi-agent systems (MAS) have been of interest to many researchers during the last decades. This thesis focuses on multi-robot systems (MRS) and programmable matter as two types of MAS. Regarding MRS, the focus is on the ’mergeable nervous systems’ (MNS) concept which allows the robots to connect to one another and establish a communication network through self-organization and then use the network to temporarily report sensing events and cede authority to a single robot in the system. Here, in a collective perception scenario, we experimentally evaluate the performance of an MNS-enabled approach and compare it with that of several decentralized benchmark approaches. We show that an MNS enabled approach is high-performing, fault-tolerant, and scalable, so it is an appropriate approach for MRS. As a goal of the thesis, using an MNS-enabled approach, we present for the first time a comprehensive comparison of control architectures in multi-robot systems, which includes a comparison of accuracy, efficiency, speed, energy consumption, scalability, and fault tolerance. Our comparisons provide designers of multi robot systems with a better understanding for selecting the best-performing control depending on the system’s objectives. Additionally, as a separate goal, we design a high-level leader-based programmable matter, which can perform some basic primitive operations in a grid environment, and construct it using lower-level organisms. We design and implement deterministic algorithms for “curl” operation of this high-level matter, an instance of shape formation problem. We prove the correctness of the presented algorithms, analytically determine their complexity, and experimentally evaluate their performance., Doctorat en Sciences de l'ingénieur et technologie, info:eu-repo/semantics/nonPublished

info:eu-repo/semantics/published

Programming robot swarms is hard because system requirements are formulated at the swarm level (i.e. globally) while control rules need to be coded at the individual robot level (i.e. locally). Connecting global to local levels or vice versa through mathematical modeling to predict the system behavior is generally assumed to be the grand challenge of swarm robotics. We propose to approach this problem by programming directly at the swarm level. Key to this solution is the use of heterogeneous swarms that combine appropriate subsets of agents whose hard-coded agent behaviors have known global effects. Our novel global-to-local design methodology allows to compose heterogeneous swarms for the example application of self-organized task allocation. We define a large but finite number of local agent controllers and focus on the global dynamics of behaviorally heterogeneous swarms. The user inputs the desired global task allocation for the swarm as a stationary probability distribution of agents allocated over tasks. We provide a generic method that implements the desired swarm behavior by mathematically deriving appropriate compositions of heterogeneous swarms that approximate these global user requirements. We investigate our methodology over several task allocation scenarios and validate our results with multiagent simulations. The proposed global-to-local design methodology is not limited to task allocation problems and can pave the way to formal approaches to design other swarm behaviors., info:eu-repo/semantics/published

info:eu-repo/semantics/published

Most of our experiences, as well as our intuition, are usually built on a linear understanding of systems and processes. Complex systems in general, and more specifically swarm robotics in this context, leverage non-linear effects to self-organize and to ensure that ‘more is different’. In previous work, the non-linear and therefore counter-intuitive effect of ‘less is more’ was shown for a site-selection swarm scenario. Although it seems intuitive that being able to communicate over longer distances should be beneficial, swarms were found to sometimes profit from communication limitations. Here, we build on this work and show the same effect for the collective perception scenario in a dynamic environment. We also find an additional effect that we call ‘slower is faster’: in certain situations, swarms benefit from sampling their environment less frequently. Our findings are supported by an intensive empirical approach and a mean-field model. All our experimental work is based on simulations using the ARGoS simulator extended with a simulator of the smart environment for the Kilobot robot called Kilogrid., SCOPUS: cp.k, info:eu-repo/semantics/published, ANTS (Conference :Swarm intelligence) & Dorigo M. (2022). Swarm intelligence :13th international conference ants 2022 málaga spain november 2-4 2022 proceedings. Springer. https://doi.org/10.1007/978-3-031-20176-9

We study how robot swarms can achieve a consensus on the best among a set of n possible options available in the environment. While the robots rely on local communication with one another, follow simple rules, and make estimates of the option’s qualities subject to measurement errors, the swarm as a whole is able to make accurate collective decisions. We compare the performance of two prominent decision-making algorithms that are based, respectively, on the direct-switching and the cross-inhibition models, both of which are well-suited for simplistic robots. Most studies used these models to let robots achieve consensus by solely relying on social interactions and ignored the aspect of enabling robots to self-source information from the environment. However, in order to select the best option, we deem sampling environmental information crucial for the successful performance of the task. Through robot-swarm simulations, we show that swarms programmed with the direct-switching model are only able to make consensus decisions in asymmetric environments where options have different quality values. Instead, using cross-inhibition, the robot swarm can also break decision deadlocks and reach a consensus in symmetric environments with equal quality options. We investigate the mechanistic causes of such differences and we find that the time the robots spend in a state of indecision is a key parameter to break the symmetry. This research highlights the importance of considering both social and environmental information when studying collective decision-making., SCOPUS: cp.k, info:eu-repo/semantics/published, ANTS (Conference :Swarm intelligence) & Dorigo M. (2022). Swarm intelligence :13th international conference ants 2022 málaga spain november 2-4 2022 proceedings. Springer. https://doi.org/10.1007/978-3-031-20176-9