Evolutionary Robotics

Cooperative coevolutionary algorithms (CCEAs) rely on multiple coevolving populations for the evolution of solutions composed of coadapted components. CCEAs enable, for instance, the evolution of cooperative multiagent systems composed of heterogeneous agents, where each agent is modelled as a component of the solution. Previous works have, however, shown that CCEAs are biased toward stability: the evolutionary process tends to converge prematurely to stable states instead of (near-)optimal solutions. In this study, we show how novelty search can be used to avoid the counterproductive attraction to stable states in coevolution. Novelty search is an evolutionary technique that drives evolution toward behavioural novelty and diversity rather than exclusively pursuing a static objective. We evaluate three novelty-based approaches that rely on, respectively (1) the novelty of the team as a whole, (2) the novelty of the agents’ individual behaviour, and (3) the combination of the two. We...

The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. Printed on acid-free paper. Routledge is an imprint of the Taylor & Francis Group, an informa business ISBN 978-87-93237-03-2 (print) While every effort is made to provide dependable information, the publisher, authors, and editors cannot be held responsible for any errors or omissions.

While central to robotics, biology and cognitive science, the concept of autonomy remains still difficult to make operative in the realm of Alife simulation models of cognitive agents. Its deep significance as a transition concept between life and cognition (a milestone on the Alife route to AI) remains obscured in the intricate relation between metabolic/constructive processes and behavioral adaptive processes in living systems. Within a naturalized and biologically inspired dynamical approach to cognition a definition of behavioral adaptive autonomy is provided: homeostatic maintenance of essential variables under viability constraints through self-modulating behavioral coupling with the environment, hierarchically decoupled from metabolic (constructive) processes. This definition allows for a naturalized notion of behavioral adaptive functionality (that defines a proper level of modelling within Alife), structurally and interactively emergent: the mapping of the agent-environment system's state space trajectories into the viability subspace of the essential variables of the organism.

to the gap between simulation and the real world. Also, it is necessary to model the environment every time a new task is given. Watson et al. 2,3 proposed "embodied evolution (EE)" for solving these issues, in which real robots evolve based on their interactions with the actual environment without simulation on a host computer. Usui and Arita 4 extended the framework by providing each robot with an evolving population of virtual individuals in order to reduce the dependence of the number of the robots and the frequency of encounter with other robots on the speed of evolution. However, these studies brought in new issues. We have to accept robot behavior of a low fi tness, especially in the early generations. Also, an evaluation based on robot behavior needs a longer convergence time in general. This article introduces pre-evaluation into the EE framework in order to restrain the behavior of a robot whose fi tness is estimated to be low, and evaluates the proposed architecture that is applied to a humanoid robot. The mechanism for pre-evaluation is introduced as a co-evolutionary system in the proposed architecture. In other words, both the robot behavior and the evaluation of the robot behavior evolve while interacting with each other. We expect the introduction of pre-evaluation to decrease the number of falls in the early stages of the evolution of biped walking.

Whereas naturally occurring swarms thrive when crowded, physical interactions in robotic swarms are either avoided or carefully controlled, thus limiting their operational density. Here we present a mechanical design rule that allows robots to act in a collision-dominated environment. We introduce the Morphobots -a robotic swarm platform developed to implement embodied computation through a morpho-functional design. By engineering a 3D-printed exoskeleton we encode a re-orientation response to an external body force (such as gravity) or a surface force (such as a collision). We show that the force-orientation response is generic, and can augment existing swarmrobotic platforms (e.g Kilobots) as well as custom robots even 10 times larger. At the individual level, the exoskeleton improves the motility and stability, and also allows to encode two contrasting dynamical behaviors in response to an external force or a collision (including collision with a wall or a movable obstacle, and on a dynamically tilting plane). This force-orientation response adds a mechanical layer to the robot's sense-act cycle at the swarm level, leveraging steric interactions for collective phototaxis when crowded. Enabling collisions also promotes information flow, facilitating online distributed learning. Each robot runs an embedded algorithm that ultimately optimizes collective performance. We identify an effective parameter that controls the force-orientation response and explore its implications in swarms that transition from dilute to crowded. Experimenting with both physical swarms (of up to 64 robots), and simulated swarms (of up to 8192 agents) show that the effect of morphological-computation increases with growing swarm size.

The capacity of re-using previously acquired skills can greatly enhance robots' learning speed and behavioral complexity. 'Intrinsically Motivated Reinforcement Learning (IMRL)' is a framework that exploits this idea and proposes to build agents capable of solving several specific tasks by assembling general-purpose building-block behaviors ('skills') previously acquired on the basis of 'intrinsic motivations'. This paper proposes a novel neural-network hierarchical reinforcement-learning architecture which exploits 'evolutionary robotics (ER)' techniques that not only allow tackling important limits of IMRL, as shown in previous papers, but they also allow investigating two other important issues, namely: (1) the optimization of the parameters that regulate the architecture's learning processes; (2) the optimization of the time the architecture dedicates to the acquisition of the skills' repertoire. These two issues are investigated here through a simulated robot engaged in solving compositional path-following navigation tasks. The results obtained indicate that the proposed approach allows obtaining a remarkable improvement of performance of the architecture, while at the same time decreasing the time the system needs to learn the skills, with respect to cases where hand-tuned parameters are used.

This paper describes a new approach for promoting the evolution of relatively complex behaviours in evolutionary robotics, based on the use of noise in simulation. A`homing navigation' behaviour is evolved (in simulation) for the Khepera mobile robot, and it is shown that high noise levels in the simulation promote the evolution of relatively complex behavioural and neural dynamics. It is also demonstrated that simulation noise can actually accelerate arti cial evolution.

In recent years simulation tools for agent-environment interactions have included increasingly complex and physically realistic conditions. These simulations pose challenges for researchers interested in evolutionary robotics because the computational expense of running multiple evaluations can be very high. Here, we address this issue by applying evolutionary techniques to a simplified simulation of a simulation itself. We show this approach to be successful when transferring controllers evolved for example visual tasks from a simplified simulation to a comparatively rich visual simulation.

This paper describes a new approach for promoting the evolution of relatively complex behaviours in evolutionary robotics, based on the use of noise in simulation. A`homing navigation' behaviour is evolved (in simulation) for the Khepera mobile robot, and it is shown that high noise levels in the simulation promote the evolution of relatively complex behavioural and neural dynamics. It is also demonstrated that simulation noise can actually accelerate arti cial evolution.

The search for the neural substrate of vertebrate action selection has focused on structures in the fore-and mid-brain, particularly on the basal ganglia. Yet, the behavioural repertoire of decerebrate and neonatal animals suggests the existence of a relatively self-contained neural substrate for action selection in the brainstem. We propose that the medial reticular formation (mRF) is this substrate's main component, reviewing evidence that the mRF's inputs, outputs, and intrinsic organisation are consistent with the requirements of an action selection system. We argue that the internal architecture of the mRF is composed of interconnected neuron clusters; our quantitative model of this anatomy suggests the mRF's intrinsic circuitry constitutes a small-world network, and may have evolved to reduce axonal wiring. We use computational models to enumerate and illustrate potential configurations of action representation within the internal circuitry of the mRF. We show that each cluster's output could represent activation of an action component; thus, co-activation of a set of these clusters would lead to the coordinated behavioural response observed in the animal. New results are presented that provide evidence for an alternative scheme: inputs to the mRF are organised to contact clusters, but recruit a pattern of reticulo-spinal neurons from across clusters to generate an action. We propose that this reconciles the anatomical structure with behavioural data showing action sequencing is degraded, rather than individual actions lost, as the mRF is progressively lesioned. Finally, we consider the potential integration of the basal ganglia and mRF substrates for selection and suggest they may collectively form a layered/hierarchical control system. * That is not to say their behaviours are reducible to simple sensory-motor reflexes. The locomotive behaviour of C. elegans after the detection of food depends on how well-fed they are (de Bono & Maricq, 2005): the internal milieu plays a role even with a brain of just 302 neurons. Even the idea of a stereotyped sensory-motor reflex in small-scale animals may be particularly misleading. They often display a stochastic motion response to repeated stimuli that stands in contrast to the highly repeatable movements of vertebrates.

Present paper analyzes and compares two innovation models in robot design. The first part describes the incremental development process of Teo: a robotic tool for Autistic Spectrum Disorder (ASD) treatment for children. The second part describes the radical innovation model used for Riby: a robotic tool for ASD treatment for adults. At the end of this paper, authors discuss the results of each case and expose the conclusions as a possible new method of product development in social robotics.

Novelty search is an evolutionary approach which promotes phenotypic diversity in a population. Novelty search has been successfully applied to a wide range of domains and a number of variants have been proposed. Here we introduce Multiple Assessment Directed Novelty Search (MADNS), which exploits the notion that a diverse population optimised through phenotypic novelty may contain solutions to multiple conflicting objectives. We show that by utilising the MADNS algorithm, an evolutionary trajectory may be simultaneously directed towards conflicting objectives. We conclude that, through applying MADNS and MC-MADNS, a divergent evolutionary trajectory may be directed to provide simultaneous solutions to multiple conflicting problems in domains with large potential for exploration.

A review is given on the use of evolutionary techniques for the automatic design of adaptive robots. The focus is on methods which use neural networks and have been tested on actual physical robots. The chapter also examines the role of simulation and the use of domain knowledge in the evolutionary process. It concludes with some predictions about future directions in robotics.

Our previous study showed that embodied cultured neural networks and spiking neural networks with spike-timing dependent plasticity (STDP) can learn a behavior as they avoid stimulation from outside. In a sense, the embodied neural network can autonomously change their activity to avoid external stimuli. In this paper, as a result of our experiments using cultured neurons, we find that there is also a second property allowing the network to avoid stimulation: if the network cannot learn to avoid the external stimuli, it tends to decrease the stimulus-evoked spikes as if to ignore the input neurons. We also show such a behavior is reproduced by spiking neural networks with asymmetric-STDP. We consider that these properties can be regarded as autonomous regulation of self and non-self for the network.

We study the dynamic categorization ability of an autonomous agent that distinguishes rectangular and triangular objects. The objects are distributed on a two-dimensional space and the agent is equipped with a recurrent neural network that controls its navigation dynamics. As the agent moves through the environment, it develops neural states which, while not symbolic representations of rectangles or triangles, allow it to distinguish these objects. As a result, it decides to avoid triangles and remain for longer periods of time at rectangles. A significant characteristic of the network is its plasticity, which enables the agent to switch from one navigation mode to another. Diversity of this switching behavior will be discussed.

A current issue in evolutionary robotics involves the coevolution of robot controllers and body morphologies built from modular parts. As part of ongoing research, a model for the evolution of the morphologies and neural network controllers of robots is described. Several robots are evolved for locomotion in simulation built from modules representing cheap, preexisting parts and one is physically built that has comparable behaviour with its original simulated version. The behaviour in simulation of such example robots is ...

Para muchos procesos del mundo real es posible diseñar un controlador difuso que provea buena regularidad usando sólo conocimiento experto. No obstante ello, para lograr un desempeño satisfactorio es necesario hacer uso de métodos más sofisticados. En este trabajo proponemos un modelo basado en sistemas difusos recurrentes, donde el antecedente de las reglas está determinado por una función de pertenencia multidimensional definida en términos de regiones de Voronoi. Las conexiones recurrentes permiten mantener una memoria que guarda información previa. Un algoritmo evolutivo puede evolucionar todas las componentes del sistema difuso recurrente. Además es posible insertar en el sistema conocimiento a priori de forma simple y efectiva. El modelo propuesto es evaluado sobre un problema de control de un robot móvil que debe avanzar evitando obstáculos y siguiendo una trayectoria dirigida por señales luminosas.

This paper describes an evolutionary robotics approach to study spatial cognition. Through the analysis of evolving neuroagents, run on behavioral, computational and evolutionary/development level, we show how relevant is considering the dynamic perspective in the observed phenomenon genesis and how useful can be using the evolutionary robotics methodology to address this issue. The experiments described in this short paper use simulated agents led by an artificial neural network and evolved through artificial selection to accomplish some spatial tasks commonly used in animal psychology. Results indicate that behaviors, especially biases, can be understood only taking into account the evolutionary/ development pathway of each agent. Keywords Spatial cognition; Neuro-agents; Artificial life techniques; Neural control system 1. Geometric information overwhelms non-geometric cues (geometric primacy); 2. Non-geometric cues prevail over geometric ones (nongeometric primacy); 3. Integration of both cues. When the first case was observed for the first time in rats studied by Cheng [2] and Margules and Gallistel [3], geometric primacy was explained postulating the existence of a geometric module: geometric and non-geometric information are processed separately, with the geometry module playing a dominant role. Non-geometric primacy and integration of cues do not support the modular hypothesis, that for some years have been critiqued (consider for example, Twyman and Newcombe, [4]). In this debate, some interesting hints come from the studies on humans. Hermer and Spelke's [5] and Gouteux and Spelke' studies [6] have tested adults that proved to

This paper is concerned with the learning of basic behaviors in autonomous robots. In this way, we present a method for the adaptation of basic reactive behaviors implemented as fuzzy controllers applying a genetic algorithm to the evolution of the fuzzy rule system. In this sense, we show our experiments in the evolution of control rules based on symbolic concepts represented as linguistic labels. The rules will be formulated in a fuzzy way and in order to test the rules obtained in each generation of the genetic algorithm a real robot has been used. The individual with the best performance is chosen to generate a new population: the elite strategy. All the new individuals were tested in the same real environment. In conclusion, the individuals of the last generation offer a set of rules that provides better performance than the ones designed by a non-expert designer.

This study introduces an innovative approach to evolving artificial neural networks (ANNs) by incorporating Hox gene-inspired mutations, aiming to enhance the adaptability and efficiency of ANNs in complex problem-solving scenarios. Drawing inspiration from the modular and hierarchical control exhibited by Hox genes in biological development, we propose a novel genetic algorithm that mimics these characteristics to evolve neural network architectures. Our methodology leverages hierarchical genome representations and modular mutation strategies to explore a more diverse and potentially effective genetic landscape. Through a series of experiments, we demonstrate that networks evolved with Hox gene-inspired mutations exhibit improved performance in tasks requiring complex pattern recognition and adaptability, compared to those evolved through traditional evolutionary algorithms. The findings suggest that integrating principles from developmental biology into genetic algorithms can significantly advance the capabilities of ANNs, offering a promising direction for future research in artificial intelligence and computational neuroscience. The implications of this work extend beyond the immediate performance enhancements, highlighting the potential for bio-inspired approaches to contribute to the development of more robust, efficient, and adaptable neural network models.

This paper investigates respiratory odor navigation with minimal evolutionary robots. We introduce a novel agent tasked with locating a chemical source solely through the use of a respiratory sensor, a challenge inspired by the active sampling strategies observed in a wide variety of animals (e.g., sniffing, whisking). Prevailing hypotheses suggest that odor navigation serves predominantly to gate behaviors in response to information gleaned from different sensory modalities. We analyze the agent's behavior and neural dynamics using dynamical systems theory, demonstrating the possibility of strategies that instead rely solely on the information obtained from their respiratory sensor. Our findings reveal that agents can successfully locate chemical sources through the synchronization of breathing rates with motor outputs, mirroring sensorimotor coupling strategies recently identified in the experimental literature. This research contributes to the theoretical understanding of sensory odor navigation and the role of physiology in agent-environment interactions.

Robotics technology has made significant advancements in various fields in industry and society. It is clear how robotics has transformed manufacturing processes and increased productivity. Additionally, navigation robotics has also been impacted by these advancements, with investors now investing in autonomous transportation for both public and private use. This research aims to explore how training scenarios affect the learning process for autonomous navigation tasks. The primary objective is to address whether the initial conditions (learning cases) have a positive or negative impact on the ability to develop general controllers. By examining this research question, the study seeks to provide insights into how to optimize the training process for autonomous navigation tasks, ultimately improving the quality of the controllers that are developed. Through this investigation, the study aims to contribute to the broader goal of advancing the field of autonomous navigation and develop...

This paper deals with the reactive control of an autonomous robot which move safely in a crowded real world unknown environment and to reach specified target by avoiding static as well as dynamic obstacle. The inputs to the proposed neurocontroller consist of left, right, and front obstacle distance to its locations and target angle between a robot and a specified target being acquired by an array of sensors. A four layer neural networks is used to design and develop the neurocontroller to solve the path and time optimization problem of mobile robots which deals the with cognitive tasks such as learning, adaptation, generalization and optimization. Back propagation method is used to trained the network. This paper analyzes the kinematical modeling of mobile robots as well as the design of control systems for the autonomous motion of the robot. The training of the nets and the control performances analysis have been done in a real experimental setup. The simulation results are compared with experimental results which are satisfactory and shows a very good agreement.

In this paper, the performance of the utility function method for behavioral organization is investigated in the framework of a simple guard robot. In order to achieve the best possible results, it was found that high-order polynomials should be used for the utility functions, even though the use of such polynomials, involving many terms, increases the running time needed for the evolutionary algorithm to find good solutions.

The possibility to use competitive evolutionary algorithms to generate long-term progress is normally prevented by the convergence on limit cycle dynamics in which the evolving agents keep progressing against their current competitors by periodically rediscovering solutions adopted previously over and over again. This leads to local but not to global progress, i.e. progress against all possible competitors. We propose a new competitive algorithm that produces long-term global progress by identifying and by filtering out opportunistic variations, i.e. variations leading to progress against current competitors and retrogression against other competitors. The efficacy of the method is validated on the co-evolution of predator and prey robots, a classic problem that has been used in other related researches. The accumulation of global progress over many generations leads to effective solutions that involve the production of rather articulated behaviors. The complexity of the behavior displayed by the evolving robots increases across generations although progresses in performance are not always accompanied by behavior complexification.

The possibility of using competitive evolutionary algorithms to generate long-term progress is normally prevented by the convergence on limit cycle dynamics in which the evolving agents keep progressing against their current competitors by periodically rediscovering solutions adopted previously. This leads to local but not to global progress (i.e., progress against all possible competitors). We propose a new competitive algorithm that produces long-term global progress by identifying and filtering out opportunistic variations, that is, variations leading to progress against current competitors and retrogression against other competitors. The efficacy of the method is validated on the coevolution of predator and prey robots, a classic problem that has been used in related researches. The accumulation of global progress over many generations leads to effective solutions that involve the production of articulated behaviors. The complexity of the behavior displayed by the evolving robot...

Embodied Evolution (EE)" is a methodology in evolutionary robotics, in which, without simulations on a host computer, real robots evolve based on the interactions with actual environment. However, we had to accept robot behavior with low fitness especially in the early generations when adopting the EE framework. We introduced pre-evaluation into the EE framework so as to restrain robot behavior whose fitness is predicted to be low. This paper reports on the introduction of pre-evaluation into the Embodied-Evolution framework for a biped robot in order to reduce the risk of falling.

This paper describes the analysis of a crawl gait multi-objective optimization system that combines bio-inspired Central Patterns Generators (CPGs) and a multi-objective evolutionary algorithm. In order to optimize the crawl gait, a multiobjective problem, an optimization system based on NSGA-II allows to find a set of non-dominated solutions that correspond to different motor solutions The experimental results highlight the effectiveness of this multi-objective approach.

I would like to thank my advisers, Professor Susana Vieira and Professor Sancho Oliveira, for all the guidance and supervision given. A special thanks to Professor Anders Christensen that also advise and ensure the quality of this dissertation, being always available and with willingness to help. I am very grateful to all for providing me such great learning experience. I also would like to acknowledge IT-Instituto de Telecomunicações and FCT-Foundation of Science and Technology, which partly supported this study. To my family, in particular, my mother and father, for all the love and support, and for how much they care about my education. To my grandmother , which taught me the value of hard work ("when something has to be done, it deserves to be well done"). To my friends, for their encouragement and understanding the various skipped night-outs. And to my love, César Costa. There are no words to thank him enough, for how much he helped during the process, for enduring with me so many weekends at home, for being always there when needed.

In this paper, a new coevolutive method, called Uniform Coevolution, is introduced to learn weights of a neural network controller in autonomous robots. An evolutionary strategy is used to learn high-performance reactive behavior for navigation and collisions avoidance. The introduction of coevolutive over evolutionary strategies allows evolving the environment, to learn a general behavior able to solve the problem in different environments. Using a traditional evolutionary strategy method, without coevolution, the learning process obtains a specialized behavior. All the behaviors obtained, with/without coevolution have been tested in a set of environments and the capability of generalization is shown for each learned behavior. A simulator based on a mini-robot Khepera has been used to learn each behavior. The results show that Uniform Coevolution obtains better generalized solutions to examples-based problems.

In this article, a co‐evolutive method is used to evolve neural controllers for general obstacle‐avoidance of a Braitenberg vehicle. During a first evolutionary process, Evolution Strategies were applied to generate neural controllers; the generality of the obtained behaviors was quite poor. During a second evolutionary process, a new co‐evolutive method, called Uniform Co‐evolution, is introduced to co‐evolve both the controllers and the environment. A comparison of both methods shows that the co‐evolutive approach improves the generality of controllers. © 2002 Wiley Periodicals, Inc.

In recent years, information theory has come into the focus of researchers interested in the sensorimotor dynamics of both robots and living beings. One root for these approaches is the idea that living beings are information processing systems and that the optimization of these processes should be an evolutionary advantage. Apart from these more fundamental questions, there is much interest recently in the question how a robot can be equipped with an internal drive for innovation or curiosity that may serve as a drive for an open-ended, self-determined development of the robot. The success of these approaches depends essentially on the choice of a convenient measure for the information. This article studies in some detail the use of the predictive information (PI), also called excess entropy or effective measure complexity, of the sensorimotor process. The PI of a process quantifies the total information of past experience that can be used for predicting future events. However, the application of information theoretic measures in robotics mostly is restricted to the case of a finite, discrete state-action space. This article aims at applying the PI in the dynamical systems approach to robot control. We study linear systems as a first step and derive exact results for the PI together with explicit learning rules for the parameters of the controller. Interestingly, these learning rules are of Hebbian nature and local in the sense that the synaptic update is given by the product of activities available directly at the pertinent synaptic ports. The general findings are exemplified by a number of case studies. In particular, in a two-dimensional system, designed at mimicking embodied systems with latent oscillatory locomotion patterns, it is shown that maximizing the PI means to recognize and amplify the latent modes of the robotic system. This and many other examples show that the learning rules derived from the maximum PI principle are a versatile tool for the self-organization of behavior in complex robotic systems.

I describe the simplest living organism, the cell, as a symbol-matter system-an observable case of how a natural representation using a wordprocessing format constrains the real-time behavior of a material organism. My purpose is not to model cognition or language at the brain level, which cells have achieved only after some three billion years of evolution. Rather, I use this primeval embodiment of a symbol-matter system as an exercise in mental hygiene for cognitive scientists to test, and hopefully improve, the clarity of their fundamental explanatory concepts. I focus on the assumptions of information processing and direct perception approaches as specific examples of inadequate theories for understanding even simple symbol-matter systems. Based on the genetic organization of cells, I propose a semantic closure principle that requires both the physical constraints of direct perception and the syntactic constraints of information processing for explanation of symbol-matter systems. This exercise is motivated by the belief that if we expect to get anywhere with the mind-body problem at the brain level, then our concepts must at least be adequate in scope and precision to explain the symbol-matter relation in single cells where it all started.

Natural lifeforms specialise to their environmental niches across many levels; from low-level features such as DNA and proteins, through to higher-level artefacts including eyes, limbs, and overarching body plans. We propose Multi-Level Evolution (MLE), a bottom-up automatic process that designs robots across multiple levels and niches them to tasks and environmental conditions. MLE concurrently explores constituent molecular and material 'building blocks', as well as their possible assemblies into specialised morphological and sensorimotor configurations. MLE provides a route to fully harness a recent explosion in available candidate materials and ongoing advances in rapid manufacturing processes. We outline a feasible MLE architecture that realises this vision, highlight the main roadblocks and how they may be overcome, and show robotic applications to which MLE is particularly suited. By forming a research agenda to stimulate discussion between researchers in related fields, we hope to inspire the pursuit of multi-level robotic design all the way from material to machine. Perspective article-Howard, David, et al. "Evolving embodied intelligence from materials to machines." Nature Machine Intelligence 1.1 (2019): 12.

This paper describes two implementations of a potential field sharing multi-robot system which we term as pessimistic and optimistic. Unlike other multirobot systems in which coordination is designed explicitly, it is an emergent property of our system. The robots perform no reasoning and are purely reactive in nature. We extend our previous work in simulated search and rescue where there was only one target to the search for multiple targets. As in our previous work the sharing systems with six or more robots outperformed the equivalent non-sharing system. We conclude that potential field sharing has a positive impact on robots involved in a search and rescue problem.

Robotic swarms and mobile sensor networks are used for environmental monitoring in various domains and areas of operation. Especially in otherwise inaccessible environments, decentralized robotic swarms can be advantageous due to their high spatial resolution of measurements and resilience to failure of individuals in the swarm. However, such robotic swarms might need to be able to compensate misplacement during deployment or adapt to dynamical changes in the environment. Reaching a collective decision in a swarm with limited communication abilities without a central entity serving as decision-maker can be a challenging task. Here, we present the CIMAX algorithm for collective decision-making for maximizing the information gathered by the swarm as a whole. Agents negotiate based on their individual sensor readings and ultimately make a decision for collectively moving in a particular direction so that the swarm as a whole increases the amount of relevant measurements and thus accessible information. We use both simulation and real robotic experiments for presenting, testing, and validating our algorithm. CIMAX is designed to be used in underwater swarm robots for troubleshooting an oxygen depletion phenomenon known as ''anoxia.''

Inspired by natural swarms, numerous control schemes enabling robotic swarms, mobile sensor networks and other multi-agent systems to exhibit various self-organized behaviors have been suggested. In this work, we present a Wave Oriented Swarm Programming Paradigm (WOSPP) enabling the control of swarms with minimalistic communication bandwidth in a simple manner, yet allowing the emergence of diverse complex behaviors and autonomy of the swarm. Communication in the proposed paradigm is based on "ping"-signals inspired by strategies for communication and self organization of slime mold (dictyostelium discoideum) and fireflies (lampyridae). Signals propagate as information-waves throughout the swarm. We show that even with 1-bit bandwidth communication between agents suffices for the design of a substantial set of behaviors in the domain of essential behaviors of a collective. Ultimately, the reader will be enabled to develop and design a control scheme for individual swarms. Keywords Swarm control • Programming paradigm • Bio-inspired • Selforganization • Agent based modeling • Mobile sensor network * these authors contributed equally to this work; order of their listed names was chosen randomly

Nature showcases swarms of animals performing various complex tasks efficiently where capabilities of individuals alone in the swarm are often quite limited. Swarm intelligence is observed when agents in the swarm follow simple rules which enable the swarm to perform certain complex tasks. This decentralized approach of nature has inspired the artificial intelligence community to apply this approach to engineered systems. Such systems are said to have no single point of failure and thus tend be more resilient. The aim of this paper is to put this notion of resilience to the test and quantify the robustness of two swarm algorithms, namely "swarmtaxis" and "FSTaxis". The first simulation results of the effects of introducing an impairment in agent-to-agent interactions in these two swarm algorithms are presented in this paper. While the FSTaxis algorithm shows a much higher resilience to agent-to-agent communication failure, both the FSTaxis and swarmtaxis algorithms are found to have a non-zero tolerance towards such failures.

Multibody dynamic analysis (MDA) has become part of the standard toolkit used to reconstruct the biomechanics of extinct animals. However, its use is currently almost exclusively limited to steady state activities such as walking and running at constant velocity. If we want to reconstruct the full range of activities that a given morphology can achieve then we must be able to reconstruct non-steady-state activities such as starting, stopping, and turning. In this paper we demonstrate how we can borrow techniques from the robotics literature to produce gait controllers that allow us to generate non-steady-state gaits in a biologically realistic quadrupedal simulation of a chimpanzee. We use a novel proportional-derivative (PD) reach controller that can accommodate both the non-linear contraction dynamics of Hill-type muscles and the large numbers of both single-joint and two-joint muscles to allow us to define the trajectory of the distal limb segment. With defined autopodial traject...

At the University of Bielefeld a new bio-inspired, hexapod robot system called HECTOR has been developed and is currently set up. To benefit from bioinspired control approaches it is fundamental to identify the most important body aspects in biological examples and to transfer body features and control approaches as pairs to the technical system. According to this, the main functional characteristics of HECTOR as presented in this paper are the elasticity in the self-contained leg joint-drives with integrated sensory processing capabilities, actuated body joints and in addition a lean bus system for onboard communication.

Service robots will play an increasing and more important role in the society in the next years. One of the main challenges is to endow robots with enough autonomy to operate on real environments. To reach that goal, the design of controllers to solve simple tasks must be automatized. Engineers look for learning algorithms that are general, robust, require low expertise knowledge, and generate controllers that can run on the real robot without any tuning stage. In this paper, a framework to learn behaviors (controllers) in mobile robotics, fulfilling the previous requirements, has been used. The framework is based on two modules: dataset generation and a data-driven evolutionary-based learning algorithm to obtain fuzzy controllers. Nevertheless, the design of a fuzzy controller still requires the selection of the type of learning algorithm, and also to choose the value of some design parameters. In this paper we present an exhaustive study on a set of evolutionary-based data-driven learning algorithms, for learning fuzzy controllers in mobile robotics, that cover a wide range of the accuracy/interpretability trade-off. The study has also evaluated the influence of the values of all the design parameters over accuracy and interpretability. The objective is to analyze the performance of the different algorithms for the design of behaviors in mobile robotics, and to extract some general rules that can help in the process to design new behaviors. The analysis comprises two different behaviors (wall-following and moving object following) and more than 450 tests, both in simulation and on a Pioneer II AT robot. Results have shown very good performances in complex and realistic conditions for the different combinations of algorithms and parameters.

Greetings, and thank you for publishing with SAGE. We have prepared this page proof for your review. Please review the entire proof for accuracy and provide necessary corrections by digitally marking this PDF using Adobe Reader. Click "Comment" in the upper right corner of Adobe Reader to access the markup tools as follows: For textual edits, please use the "Annotations" tools. Please refrain from using the two tools crossed out below, as data loss can occur when using these tools. For formatting requests, questions, or other complicated changes, please insert a comment using "Drawing Markups."

The standard approach in evolutionary robotics is to evolve neural networks for control by encoding the parameters of the network in the genome. By contrast, we have evolved a neural controller based on biological principles from molecular and ...

Evolution gave rise to human and animal intelligence here on Earth. We argue that the path to developing artificial human-like-intelligence will pass through mimicking the evolutionary process in a nature-like simulation. In Nature, there are two processes driving the development of the brain: evolution and learning. Evolution acts slowly, across generations, and amongst other things, it defines what agents learn by changing their internal reward function. Learning acts fast, across one's lifetime, and it quickly updates agents' policy to maximise pleasure and minimise pain. The reward function is slowly aligned with the fitness function by evolution, however, as agents evolve the environment and its fitness function also change, increasing the misalignment between reward and fitness. It is computationally expensive to replicate these two processes in simulation. This work proposes Evolution via Evolutionary Reward (EvER) which allows learning to single-handedly drive the search for policies with increasingly evolutionary fitness by ensuring the alignment of the reward function with the fitness function. In this search, EvER makes use of the whole state-action trajectories that agents go through their lifetime. In contrast, current evolutionary algorithms discard this information and consequently limit their potential efficiency at tackling sequential decision problems. We test our algorithm in two simple bio-inspired environments and show its superiority at generating more capable agents at surviving and reproducing their genes when compared with a state-of-the-art evolutionary algorithm.