Data augmentation is an intuitive solution to increase the diversity of training instances in the machine learning community. Mixup is acknowledged as an effective and efficient mix-based data augmentation method, following a linear alignment assumption that the linear interpolations of features align the corresponding linear interpolations of labels. Unfortunately, this assumption can be violated in many complex scenarios, resulting in augmented instances with noisy labels, especially for regression problems. To solve this problem, we propose an easy-to-implement mixup method, namely DEnosing MIXUP (DE-mixup), which iteratively corrects the noisy response targets by leveraging an auxiliary noise estimation task with mixup deep features. Additionally, we suggest an efficient optimization method with alternating direction method of multipliers. We compare DE-mixup with the existing mixup variants and other prevalent data augmentation methods across benchmark regression datasets. Empirical results indicate the effectiveness of DE-mixup under the in-distribution and out-of-distribution cases.

Evaluating behavioral descriptors remains a major bottleneck in standard and automated Quality-Diversity optimization. Existing methods either rely on expert-crafted descriptors or fully automated ones that tend to lack interpretability and alignment with task objectives, both requiring significant resources and time. This challenge is especially pronounced in swarm systems, where descriptors must represent both individual behaviors and collective dynamics. We address this problem by introducing a Mid-Evolution behavioral descriptor mutation mechanism, where descriptors are modified during a run rather than between full re-training. This enables direct, fair comparison among differing descriptors without costly restarts, and aids in identifying descriptors that align with task objectives. Across three swarm robotics tasks (herding, block collection, and radar coverage) this Mid-Evolution mutation strategy often identifies higher-performing descriptors or eliminates weaker ones more efficiently than retraining. This approach improves evaluation efficiency and enables faster, more practical automated discovery of meaningful behavioral descriptors.

Behavioral heterogeneity yields problem solving benefits in biological collective behavior systems such as insect colonies and human societies and in artificial collective behavior systems such as distributed computer networks and swarm-robotics systems. In this study, we investigate comparative methods for two-step collective behavior evolution designed to encourage the evolution of behavioral diversity in swarm robotic applications. Specifically, we investigate behavioral diversity evolution given pre-evolved behaviors in collective behaviors that are effective across increasingly complex and difficult collective herding task environments. Results indicate that a minimal complement of pre-evolved (lower task-performance) collective herding behaviors was suitable for achieving high task performance across all environments and task difficulty levels. Results support the efficacy of the two-step approach for evolving behaviorally heterogeneous groups in collective behavior tasks that benefit from groups comprising various complementary behaviors.

In biological societies, complex interactions between the behavior and morphology of evolving organisms and their environment have given rise to a wide range of complex and diverse social structures. Similarly, in artificial counterparts such as swarm robotics systems, collective behaviors emerge via the interconnected dynamics of robot morphology (sensory-motor configuration), behavior (controller), and environment (task). Various studies have demonstrated morphological and behavioral diversity enables biological groups to exhibit adaptive, robust, and resilient collective behavior across changing environments. However, in artificial (swarm robotic) systems there is little research on the impact of changing environments on morphological and behavioral (body-brain) diversity in emergent collective behavior, and the benefits of such diversity. This study uses evolutionary collective robotics as an experimental platform to investigate the impact of increasing task environment complexity (collective behavior task difficulty) on the evolution and benefits of morphological and behavioral diversity in robotic swarms. Results indicate that body-brain evolution using coupled behavior and morphology diversity maintenance yields higher behavioral and morphological diversity, which is beneficial for collective behavior task performance across task environments. Results also indicate that such behavioral and morphological diversity maintenance coupled with body-brain evolution produces neuro-morpho complexity that does not increase concomitantly with task complexity.

In evolutionary robotics, Multi-Level Evolution (MLE) has been demonstrated for effective robot designs using a bottom-up approach, first evolving which materials to use for modular components and then how these components are connected into a functional robot design. This paper evaluates MLE robotic design, as an evolutionary design method on various task (robot ambulation) environments in comparison to human designed robots (pre-designed robot controller-morphology couplings). Results indicate that the MLE method evolves robots that are effective across increasingly difficult (locomotion) task environments, out-performing pre-designed robots, and thus provide further support for the efficacy of MLE as an evolutionary robotic design method. Furthermore, results indicate the MLE method enables the evolution of suitable robotic designs for various environments, where such designs would be non-intuitive and unlikely in conventional robotic design.

In nature, morphological diversity enhances functional diversity, however, there is little swarm (collective) robotics research on the impact of morphological and behavioral (body-brain) diversity that emerges in response to changing environments. This study investigates the impact of increasingly complex task environments on the artificial evolution of body-brain diversity in simulated robot swarms. We investigate whether increasing task environment complexity (collective behavior tasks requiring increasing degrees of cooperative behavior) mandates concurrent increases in behavioral, morphological, or coupled increases in body-brain diversity in robotic swarms. Experiments compared three variants of collective behavior evolution across increasingly complex task environments: two behavioral diversity maintenance variants and body-brain diversity maintenance. Results indicate that body-brain diversity maintenance yielded a significantly higher behavioral and morphological diversity in evolved swarms overall, which was beneficial in the most complex task environment.

This paper investigates a novel method combining Scalable Evolution Strategies (S-ES) and Hierarchical Reinforcement Learning (HRL). S-ES, named for its excellent scalability, was popularised with demonstrated performance comparable to state-of-the-art policy gradient methods. However, S-ES has not been tested in conjunction with HRL methods, which empower temporal abstraction thus allowing agents to tackle more challenging problems. We introduce a novel method merging S-ES and HRL, which creates a highly scalable and efficient (compute time) algorithm. We demonstrate that the proposed method benefits from S-ES's scalability and indifference to delayed rewards. This results in our main contribution: significantly higher learning speed and competitive performance compared to gradient-based HRL methods, across a range of tasks.

The incidence of most diseases varies greatly with seasons, and global climate change is expected to increase its risk. Predictive models that automatically capture trends between climate and diseases are likely to be beneficial in minimizing disease outbreaks. Machine learning (ML) predictive analytic tools have been popularized across many health-care applications, however the optimal task performance of such ML tools largely depends on manual parameter tuning and calibration. Such manual tuning significantly limits the full potential of ML methods, especially for high-dimensional and complex task domains, as typified by real-world health-care application data-sets. Additionally, the inaccessibility of many health-care data-sets compounds innate problems of method comparison, predictive accuracy and the overall advancement of ML based health-care applications. In this study we investigate the impact of Relevance Estimation and Value Calibration, an evolutionary parameter optimization method applied to automate parameter tuning for comparative ML methods (Deep learning and Support Vector Machines) applied to predict daily diarrhoea cases across various geographic regions. Data-augmentation is also used to complement real-world noisy, sparse and incomplete data-sets with synthetic data-sets for training, validation and testing. Results support the efficacy of evolutionary parameter optimization and data synthesis to boost predictive accuracy in the given task, indicating a significant prediction accuracy boost for the deep-learning models across all data-sets.

Autonomous robots are increasingly used in remote and hazardous environments, where damage to sensory-actuator systems cannot be easily repaired. Such robots must therefore have controllers that continue to function effectively given unexpected malfunctions and damage to robot morphology. This study applies the Intelligent Trial and Error (IT&E) algorithm to adapt hexapod robot control to various leg failures and demonstrates the IT&E map-size parameter as a critical parameter in influencing IT&E adaptive task performance. We evaluate robot adaptation for multiple leg failures on two different map-sizes in simulation and validate evolved controllers on a physical hexapod robot. Results demonstrate a trade-off between adapted gait speed and adaptation duration, dependent on adaptation task complexity (leg damage incurred), where map-size is crucial for generating behavioural diversity required for adaptation.

An open question for both natural and artificial evolutionary systems is how, and under what environmental and evolutionary conditions complexity evolves. This study investigates the impact of increasingly complex task environments on the evolution of robot complexity. Specifically, the impact of evolving body-brain couplings on locomotive task performance, where robot evolution was directed by either body-brain exploration (novelty search) or objective-based (fitness function) evolutionary search. Results indicated that novelty search enabled the evolution of increased robot body-brain complexity and efficacy given specific environment conditions. The key contribution is thus the demonstration that body-brain exploration is suitable for evolving robot complexity that enables high fitness robots in specific environments.

The impact of the environment on evolving increasingly complex morphologies (bodies) and controllers (brains) remains an open question in evolutionary biology and has important implications for the evolutionary design of robots. This study uses evolutionary robotics as an experimental platform to evaluate relationships between environment complexity and evolving body-brain complexity given energy costs on evolving complexity. We evolve robot body-brain designs for increasingly complex environments (difficult cooperative transport tasks) in a collective robotic gathering simulation. The impact of complexity costs on body-brain evolution is evaluated across such increasingly complex environments. Results indicate that complexity costs enable the evolution of simpler body-brain designs that are effective in simple environments but yield negligible behavior (task performance) differences in more complex environments.

Recently, there has been increased research on adaptive control systems for vehicles that operate on autonomous vehicle only roads. Specifically, roads without current infrastructure constraints of traffic lights, stop signals at intersections or vehicle lanes. This study investigates controller automation for vehicles that must navigate and coordinate with each other on such autonomous vehicle only roads. We comparatively evaluate fitness-function (objective) versus behavior-based (novelty search) versus hybridized objective-novelty evolutionary search for synthesizing autonomous vehicle coordinated driving behavior. The goal of such evolved coordinated driving behavior is to maximize effective (safe) and efficient (expedient) autonomous vehicle traffic throughput for given roads. Results indicate that while novelty and hybrid search evolved effective and efficient driving behaviors, these behaviors did not generalize to new roads as well as driving behaviors evolved with objective-based search.

The evolutionary cost of morphological complexity in biological populations remains an open question. This study investigates the impact of imposing a cost on morphological complexity given co-adapting behavior-morphology couplings in simulated robots. Specifically, we investigate the environmental and evolutionary conditions for which morphological complexity can be evolved without sacrificing behavioral efficacy. This study evaluates the relationship between between task difficulty (environment complexity) and evolved morphological complexity. We use multi-objective neuro-evolution to evolve robot controller-morphology couplings in task environments of increasing difficulty, where the objectives are to minimize the cost of (morphological) complexity and to maximize behavior quality (task performance) over evolutionary time. Results indicate that imposing a cost of complexity induces the evolution of simpler morphologies with negligible differences in behavior (task performance) across varying task environments. That is, with a cost of complexity, evolution maintained a constant selection pressure for morphological complexity across all environments.

Recently there has been increased research interest in developing adaptive control systems for autonomous vehicles. This study presents a comparative evaluation of two distinct approaches to automated intersection management for a multi-agent system of autonomous vehicles. The first is a centralized heuristic control approach using an extension of the Autonomous Intersection Management (AIM) system. The second is a decentralized neuro-evolution approach that adapts vehicle controllers so as they collectively navigate intersections. This study tests both approaches for controlling groups of autonomous vehicles on a network of interconnected intersections, without the constraints of traffic lights or stop signals. These task environments thus simulate potential future scenarios where vehicles must drive autonomously without specific road infrastructure constraints. The capability of each approach to appropriately handle various types of interconnected intersections, while maintaining an efficient throughput of vehicles and minimizing delay is tested. Results indicate that neuro-evolution is an effective method for automating collective driving behaviors that are robust across a broad range of road networks, where evolved controllers yield comparable task performance or out-perform an AIM controller.

Previous research has demonstrated that computational models of Gene Regulatory Networks (GRNs) can adapt so as to increase their evolvability, where evolvability is defined as a populations responsiveness to environmental change. In such previous work, phenotypes have been represented as bit strings formed by concatenating the activations of the GRN after simulation. This research is an extension where previous results supporting the evolvability of GRNs are replicated, however, the phenotype space is enriched with time and space dynamics with an evolutionary robotics task environment. It was found that a GRN encoding used in the evolution of a way-point navigation behavior in a fluctuating environment results in (robot controller) populations becoming significantly more responsive (evolvable) over time. This is as compared to a direct encoding of controllers which was unable to improve its evolvability in the same task environment.

Robustness and evolvability have traditionally been seen as conflicting properties of evolutionary systems, due to the fact that selection requires heritable variation on which to operate. Various recent studies have demonstrated that organisms evolving in environments fluctuating non-randomly become better at adapting to these fluctuations, that is, increase their evolvability. It has been suggested that this is due to the emergence of biases in the mutational neighborhoods of genotypes. This paper examines a potential consequence of these observations, that a large bias in certain areas of genotype space will lead to increased robustness in corresponding phenotypes. The evolution of boolean networks, which bear similarity to models of gene regulatory networks, is simulated in environments which fluctuate between task targets. It was found that an increase in evolvability is concomitant with the emergence of highly robust genotypes, where evolvability was defined as the populations adaptability. Analysis of the genotype space elucidated that evolution finds regions containing robust genotypes coding for one of the target phenotypes, where these regions overlap or are situated in close proximity. Results indicate that genotype space topology impacts the relationship between robustness and evolvability, where the separation of robust regions coding for the various targets was detrimental to evolvability.

Novelty Search (NS) has been proposed as an alternative search approach for black-box optimization methods where the fitness function is replaced and only novel solutions are searched for. NS has been demonstrated as advantageous when the fitness landscape is highly deceptive and misdirects the search process towards local optima. In this research we test the efficacy of NS in comparison to a purely objective based approach and a hybrid approach that combines NS and a fitness function in combination with two behavior characterization schemes. The task is non-Markovian double-pole balancing. Results indicate that the success of NS strongly depends upon the behavior characterization scheme used, given that NS performed the best under one scheme and relatively poorly under the other scheme.

This paper presents a study on the impact of different robot sensory configurations (morphologies) in simulated robot teams that must accomplish a collective (cooperative) behavior task. The study's objective was to investigate if effective collective behaviors could be efficiently evolved given minimal morphological complexity of individual robots in an homogenous team. A range of sensory configurations are tested in company with evolved controllers for a collective construction task. Results indicate that a minimal sensory configuration yields the highest task performance, and increasing the complexity of the sensory configuration does not yield an increased task performance.

Crowd simulations have become increasingly popular in films over the last decade, appearing in large crowd shots of many big name block-buster films. An important requirement for crowd simulations in films is that they should be directable both at a high and low level. As agent-based techniques allow for low-level directability and more believable crowds, they are typically used in this field. However, due to the bottom-up nature of these techniques, to achieve high level directability, agent-level parameters must be adjusted until the desired crowd behavior emerges. As manually adjusting parameters is a time consuming and tedious process, this paper investigates a method for automating this, using Neuro-Evolution. To this end, the Conventional Neuro-Evolution (CNE), Covariance Matrix Adaptation Evolutionary Strategy (CMA-ES), Neuro-Evolution of Augmenting Topologies (NEAT), and Enforced Sub Populations (ESP) algorithms are compared across a variety of representative crowd simulation scenarios. Overall, it was found that CMA-ES generally performs the best across the selected simulations.

Crowd simulations are a set techniques used to control groups of agents and are exemplified by scenes from movies such as The Lord of the Rings and Inception. A problem which all crowd simulation techniques suffer from is the balance between control of the crowd behaviour and the autonomy of the agents. One possible solution to this problem is to use Neuro-Evolution (NE) to evolve the agent models so that the agents behave realistically and the emergent crowd behaviour is controllable. Since this is not an application area which has been investigated much, it is unknown which NE parameters and operators work well. This paper attempts to address this by comparing the performance of a set of crossover operators with a range of probabilities in three simulations: Car Racing, Mouse Bridge Crossing, and a War-Robot Battle. Overall it was found that Laplace crossover worked the best across all our simulations.

This paper proposes a new Neuro-Evolution (NE) method for automated controller design in agent-based systems. The method is Generational Neuro-Evolution (GeNE), and is comparatively evaluated with established NE methods in a multi-agent predator-prey task. This study is part of an ongoing research goal to derive efficient (minimising convergence time to optimal solutions) and scalable (effective for increasing numbers of agents) controller design methods for adapting agents in neuro-evolutionary multi-agent systems. Dissimilar to comparative NE methods, GeNE employs tiered selection and evaluation as its generational fitness evaluation mechanism and, furthermore, re-initializes the population each generation. Results indicate that GeNE is an appropriate controller design method for achieving efficient and scalable behavior in a multi-agent predator-prey task, where the goal was for multiple predator agents to collectively capture a prey agent. GeNE outperforms comparative NE methods in terms of efficiency (minimising the number of genotype evaluations to attain optimal task performance).

This paper evaluates two Neuro-Evolution (NE) methods to adapt controllers in simulated robot teams. The first method evolves controllers with fixed topologies and adapts team size as a function of task complexity. The second method evolves controller topology as a function of task complexity, but keeps team sizes constant. These methods are: Collective Neuro-Evolution 2 (CONE-2), and Neuro-Evolution for Augmenting Topologies (NEAT). CONE-2 and NEAT are comparatively tested in a collective construction task. The goal is to ascertain the most appropriate controller evolution method for adapting teams to solve a collective construction task, with varying cooperative behavior requirements. Results indicate that CONE-2 is most effective at adapting controllers as the complexity of the task increases. In environments where multiple forms of cooperative behavior are required, CONE-2 evolves teams with a higher average task performance. CONE-2 is demonstrated as being effective at evolving behavioral heterogeneity in teams, which results in a higher team fitness, comparative to NEAT evolved teams, in environments that require cooperation.

This paper reports upon two adaptive approaches for deriving words in an artificial language simulation. The efficacy of a Particle Swarm Optimization (PSO) method versus an Artificial Evolution (AE) method was examined for the purpose of adapting communication between agents. The objective of the study was for agents to derive a common (shared) lexicon for talking about food resources in the simulation environment. In the simulation, communication was essential for agent survival and as such facilitated lexicon adaptation. Results indicated that PSO was effective at adapting agents to quickly converge to a common lexicon, where, on average, one word for each food type was derived. AE required more method iterations to converge to a common lexicon that contained, on average, multiple words for each food type. However, there was greater word diversity in the lexicon converged upon by AE evolved agents, compared to that converged upon by PSO adapted agents.

This paper describes user-supervised Evolutionary Algorithm (EA) experiments that investigate the evolution of a sensible fictional dialogue. A user-supervised EA was used given the difficulty of defining a fitness function for evolving art tasks. Two EAs were tested for the task of evolving dialogue given an English word population. The EAs required user-assigned fitness values to be given as input with varying degrees of frequency during the evolutionary process. The success of the EAs were comparatively evaluated with respect to two-point recombination and a novel complement gene scan operator. Task performance was evaluated according to average fitness, word and genotype diversity, and the number of words used in the fittest evolved dialogue. Results indicated that for both EAs, complement gene scan was more effective for evolving complex, sensible and grammatically correct dialogue, comparative to sentences evolved by the EAs using two-point recombination.

This paper presents a study that compares the efficacy of Neuro-Evolution (NE) versus Particle Swarm Optimization (PSO) for evolving Artificial Neural Network (ANN) controllers in an unsupervised adaptation process. The research objective is to ascertain which adaptive method is most appropriate for deriving agent behaviors in a competitive co-evolution pursuit-evasion task. This task requires one predator agent to capture one prey agent in a simulation where behavior adaptation is guided by an arms race of competitive co-evolution. Results indicate that NE was overall more effective at deriving pursuit and evasion behaviors according to the task performance measures defined for this study.

This paper is a preliminary study of the types of collective behavior tasks that are best solved by neuro-evolution (NE). This research tests a hypothesis that for a multi-rover task, the best approach (for deriving effective collective behaviors) is to evolve complete artificial neural network (ANN) controllers, and then combine controller behaviors in a collective behavior context. Such methods are called multi-agent conventional neuro-evolution (multi-agent CNE). This is opposed to methods such as enforced sub-populations (ESP) which evolves individual neurons and then combines them to form complete ANN controllers. Single and multi-agent CNE and ESP approaches to evolving collective behavior solutions are tested comparatively in the multi-rover task. The multi-rover task requires that teams of rovers (controllers) cooperate in order to detect features of interest in a virtual environment. Results indicate that a multi-agent CNE approach derives rover teams with a higher task performance and genotype diversity, comparative to ESP.

In this paper we apply three Neuro-Evolution (NE) methods as controller design approaches in a collective behavior task. These NE methods are Enforced Sub-Populations, Multi-Agent Enforced Sub-Populations, and Collective Neuro- Evolution. In the collective behavior task, teams of simulated robots search an unexplored area for objects that are to be used in a collective construction task. Results indicate that the Collective Neuro-Evolution method, a cooperative co-evolutionary approach that allows for regulated recombination between genotype populations is appropriate for deriving artificial neural network controllers in a set of increasingly difficult collective behavior task scenarios.

We compare the efficacy of the Enforced Sub-Populations (ESP) and Collective Neuro-Evolution (CONE) methods for designing behavioral specialization in a multi-rover collective behavior task. These methods are tested for Artificial Neural Network (ANN) controller design in an extension of the multi-rover task, where behavioral specialization is known to benefit task performance. The task is for multiple simulated autonomous vehicles (rovers) to maximize the detection of points of interest (red rocks) in a virtual environment. The task requires rovers to collectively sense such points of interest in order for them to be detected. Results indicate that the CONE method facilitates a level of specialization appropriate for achieving a significantly higher task performance, comparative to rover teams evolved by the ESP method.

This paper introduces the collective neuro evolution (CONE) method, and compares its efficacy for designing specialization, with a conventional neuro-evolution (NE) method. Specialization was defined at both the individual agent, and at the agent group level. The CONE method was tested comparatively with the conventional NE method in an extension of the multi-rover task domain, where specialization exhibited at both the individual and group level is known to benefit task performance. In the multi-rover domain, the task was for many agents (rovers) to maximize the detection and evaluation of points of interest in a simulated environment, and to communicate gathered information to a base station. The goal of the rover group was to maximize a global evaluation function that measured performance (fitness) of the group. Results indicate that the CONE method was appropriate for facilitating specialization at both the individual and agent group levels, where as, the conventional NE method succeeded only in facilitating individual specialization. As a consequence of emergent specialization derived at both the individual and group levels, rover groups evolved by the CONE method were able to achieve a significantly higher task performance, comparative to groups evolved by the conventional NE method.

Neuro-Evolution (NE) methods have been successfully applied to the rover task domain [1], [2]. However, extending this domain to include the notion of using NE to facilitate emergent specialization, in order to increase task performance, has not yet been investigated. We introduce the Collective Neuro Evolution (CONE) method, and compares its efficacy for designing specialization, with a conventional NE method. CONE and conventional NE were applied to an extension of the multi-rover task [1], [2] for the purpose of designing collective behavior. This task requires solutions for controlling groups of simulated autonomous vehicles (rovers) that seek to maximize the number of points of interest discovered in an unexplored environment (global evaluation function). Rovers operated in a discrete simulation environment, and used complementary sensors and actuators so as to maximize the global evaluation function.

In this research, a neuro-evolution method called collective neuro-evolution (CONE), is introduced for the design of neural controllers (agents) operating in collective behavior task domains. The efficacy of the CONE method for facilitating emergent behavioral specialization for the benefit of increasing task performance is tested in a pursuit-evasion and collective gathering task. For a comparative study, a conventional neuro-evolution method was applied to the same tasks. In both tasks, the CONE method derived behavioral specialization in groups of agents resulting in higher task performances, where as the conventional neuro-evolution method was unable to derive specialization resulting in comparatively lower task performances

In this paper we study a population of individuals in a simulated artificial environment. These individuals have a "body" as well as a "mind", i.e., some of their features effect their "physical" properties, like speed and strength, while other features influence their "mental" preferences and choices in interacting with the environment and other agents. We compare two approaches to adapting the minds of individuals. In approach 1, the bodies and the minds develop through evolution, while in approach 2 only the bodies evolve and the minds are adapted by lifetime-learning. The results indicate that the evolutionary approach is able to sustain larger and more stable agent populations as well as maintain a higher degree of individual success compared to the lifetime learning approach. Furthermore, quite unexpectedly, the method used for mental development has a strong effect on the development of the physical features within the very same environment: The individuals' bodies evolve to completely different segments of the physical feature space under the two regimes.

The research goal was to engineer agent collectives that most effectively accomplish a cooperative gathering task. In view of this, we compared reproduction schemes for the artificial evolution of agent controller parameters for a cooperative minesweeping task. Agents utilized cooperative behavior to improve task performance in a simulated environment where different types of mines with different fitness rewards were randomly distributed. We compared the evolution of agent controller parameters with respect to temporal and spatial dimensions of agent reproduction schemes. The first dimension concerned agents reproducing only once at the end of their lifetime or multiple times during their lifetime. The second dimension concerned agents reproducing only with agents in adjacent positions (locally restricted) or with agents located anywhere else in the environment (panmictic). Results indicated that the single reproduction at the end of an agent's lifetime and the locally restricted reproduction schemes afforded the agent collective a higher level of performance in its cooperative gathering task

This review presents a review of prevalent results within research pertaining to emergent cooperation in biologically inspired artificial social systems. Results reviewed maintain particular reference to biologically inspired design principles, given that current mathematical and empirical tools have provided only a partial insight into elucidating mechanisms responsible for emergent cooperation, and then only in systems of an abstract nature. This review aims to provide an overview of important and disparate research contributions that investigate utilization of biologically inspired concepts such as emergence, evolution, and self-organization as a means of attaining cooperation in artificial social systems. An introduction and overview of emergent cooperation in artificial life is presented, followed by a survey of emergent cooperation in swarm-based systems, the pursuit-evasion domain, and RoboCup soccer. The final section draws conclusions regarding future directions of emergent cooperation as a problem-solving methodology that is potentially applicable in a wide range of problem domains. Within each of these sections and their respective themes of research, the mechanisms deemed to be responsible for emergent cooperation are elucidated and their key limitations highlighted. The review concludes that current studies in emergent cooperative behavior are limited by a lack of situated and embodied approaches, and by the research infancy of current biologically inspired design approaches. Despite these limiting factors, emergent cooperation maintains considerable future potential in a wide variety of application domains where systems composed of many interacting components must cooperatively perform unanticipated global tasks.

This research concerns the comparison of different artificial evolution approaches to the design of cooperative behavior in a team of simulated mobile robots co-evolved against a second team. The first and second approaches, termed: single pool and plasticity, are characterized by robots that share a single genotype, though the plasticity approach includes a learning mechanism. The third approach, termed: multiple pools, is characterized by robots that use different genotypes. The application domain is a pursuit-evasion game in which a team of three robots termed: pursuers, collectively work to immobilize one of the three robots of the other team, termed: evaders. Results indicate that the multiple pools approach applied within a competitive co-evolution process yields superior performance comparative to the other approaches. Specifically, the co-evolutionary process allows the multiple pools approach to 'bootstrap' complementary behavioral roles, facilitating the evolution of a stable cooperative pursuit strategy.