Social insects allocate their workforce in a decentralised fashion, addressing multiple tasks and... more Social insects allocate their workforce in a decentralised fashion, addressing multiple tasks and responding effectively to environmental changes. This process is fundamental to their ecological success, but the mechanisms behind it are not well understood. While most models focus on internal and individual factors, empirical evidence highlights the importance of ecology and social interactions. To address this gap, we propose a game theoretical model of task allocation. Our main findings are twofold: Firstly, the specialisation emerging from self-organised task allocation can be largely determined by the ecology. Weakly specialised colonies in which all individuals perform more than one task emerge when foraging is cheap; in contrast, harsher environments with high foraging costs lead to strong specialisation in which each individual fully engages in a single task. Secondly, social interactions lead to important differences in dynamic environments. Colonies whose individuals rely on their own experience are predicted to be more flexible when dealing with change than colonies relying on social information. We also find that, counter to intuition, strongly specialised colonies may perform suboptimally, whereas the group performance of weakly specialised colonies approaches optimality. Our simulation results fully agree with the predictions of the mathematical model for the regions where the latter is analytically tractable. Our results are useful in framing relevant and important empirical questions, where ecology and interactions are key elements of hypotheses and predictions.
Social insect colonies distribute their workforce with amazing flexibility across a large array o... more Social insect colonies distribute their workforce with amazing flexibility across a large array of diverse tasks under fluctuating external conditions and internal demands. Deciphering the individual rules of task selection and task performance is at the heart of understanding how colonies can achieve this collective feature. Models play an important role in this endeavor, as they allow us to investigate how the rules of individual behavior give rise to emergent patterns at the colony level. Modulation of individual behavior occurs at many different timescales and to successfully use a model we need to ensure that it applies on the timescale under observation. Here, we focus on short timescales and ask the question whether the most commonly used class of models (response threshold models) adequately describes behavioral modulation on this timescale. We study the fanning behavior of bumblebees on temperature-controlled brood dummies and investigate the effect of (i) stimulus intensity, (ii) repeated task performance, and (iii) task performance feedback. We analyze the timing patterns (rates of task engagement and task disengagement) using survival analysis. Our results show that stimulus intensity does not significantly influence individual task investment at these comparably short timescales. In contrast, repeated task performance and task performance feedback affect individual task investment. We propose an explicitly time-resolved individual-based model and simulate this model to study how patterns of individual task engagement influence task involvement at the group level, finding support for the hypothesis that regulation mechanisms at different timescales can improve performance at the group level in dynamic environments.
A wide range of group-living animals construct tangible infrastructure networks, often of remarka... more A wide range of group-living animals construct tangible infrastructure networks, often of remarkable size and complexity. In ant colonies, infrastructure construction may require tens of thousands of work hours distributed among many thousand individuals. What are the individual behaviours involved in the construction and what level of complexity in inter-individual interaction is required to organize this effort? We investigate this question in one of the most sophisticated trail builders in the animal world: the leafcutter ants, which remove leaf litter, cut through overhangs and shift soil to level the path of trail networks that may cumulatively extend for kilometres. Based on obstruction experiments in the field and the laboratory, we identify and quantify different individual trail clearing behaviours. Via a computational model, we further investigate the presence of recruitment, which—through direct or indirect information transfer between individuals—is one of the main organizing mechanisms of many collective behaviours in ants. We show that large-scale transport networks can emerge purely from the stochastic process of workers encountering obstructions and subsequently engaging in removal behaviour with a fixed probability. In addition to such incidental removal, we describe a dedicated clearing behaviour in which workers remove additional obstructions independent of chance encounters. We show that to explain the dynamics observed in the experiments, no information exchange (e.g. via recruitment) is required, and propose that large-scale infrastructure construction of this type can be achieved without coordination between individuals.
Self-organised collective decision making is one of the core components of swarm intelligence, an... more Self-organised collective decision making is one of the core components of swarm intelligence, and numerous swarm algorithms that are widely used in optimisation and optimal control have been inspired by the biological mechanisms driving it. Beyond the life sciences and bio-inspired engineering, collective decision making is important in a number of other disciplines, most prominently economics and the social sciences. A paradigmatic model system for collective decision making is the foraging behaviour of mass recruiting ant colonies. While this system has been investigated extensively, our knowledge about its function in dynamic environments is still incomplete at best. We show that the mathematical model of mass foraging is really just a specific instance of a very general class of rational group decision making processes. We analyse this general class using an information-theoretic framework, which allows us to abstract from the specific details of a fixed model system. We specifically investigate how noisy communication can enable groups to share information about changes in an environment more efficiently. In the present paper, we show that an optimal noise level exists and that this optimal level depends on the rate of change in the environment. We explain this on the basis of stochastic resonance theory and show why stochastic attractor switching is a suitable base mechanism for adaptive group decision making in dynamic environments.
Self-organized mechanisms are frequently encountered in nature and known to achieve flexible, ada... more Self-organized mechanisms are frequently encountered in nature and known to achieve flexible, adaptive control and decision-making. Noise plays a crucial role in such systems: It can enable a self-organized system to reliably adapt to short-term changes in the environment while maintaining a generally stable behavior. This is fundamental in biological systems because they must strike a delicate balance between stable and flexible behavior. In the present paper we analyse the role of noise in the decision-making of the true slime mold Physarum polycephalum, an important model species for the investigation of computational abilities in simple organisms. We propose a simple biological experiment to investigate the reaction of P. polycephalum to time-variant risk factors and present a stochastic extension of an established mathematical model for P. polycephalum to analyze this experiment. It predicts that—due to the mechanism of stochastic resonance—noise can enable P. polycephalum to correctly assess time-variant risk factors, while the corresponding noise-free system fails to do so. Beyond the study of P. polycephalum we demonstrate that the influence of noise on self-organized decision-making is not tied to a specific organism. Rather it is a general property of the underlying process dynamics, which appears to be universal across a wide range of systems. Our study thus provides further evidence that stochastic resonance is a fundamental component of the decision-making in self-organized macroscopic and microscopic groups and organisms.
Few ant species construct cleared trails. Among those that do, leaf-cutting Atta ants build the m... more Few ant species construct cleared trails. Among those that do, leaf-cutting Atta ants build the most prominent networks, with single colonies clearing debris and obstructions from hundreds of meters of trails annually. Workers on cleared paths move at higher speed than they do over uncleared litter, and one measurement of the time and energetic costs of trail clearance suggests that benefits of trail usage far outweigh the investment costs of trail clearing. The ecological basis of trail clearing remains uncertain, however, because no full account has been made of benefits and costs in common units that allow comparison. We make such an account using a scalable, integrative model of trail investment and foraging energetics. Contrary to assumptions in previous work, we find that trail clearing needs not always be energetically profitable for leaf-cutting ants. Profitability depends on the workforce composition, specifically, on how many ants in a traffic stream act as maintenance workforce to respond to sudden and unpredictable obstructions, such as leaf fall. Such maintenance patrols have not previously been recognized as a cost of trail building. If the patrolling workforce is not too large, the energetic savings from foraging over cleared trails offset the investment and maintenance costs within a few days. Under some conditions, however, amortization can take weeks or months, or trail clearing can become unprofitable altogether. This suggests that Atta colonies must have a mechanism to regulate the intensity of their trail clearing behavior. We explore possible mechanisms and make testable predictions for future research.
Leaf‐cutting ants display regular diel cycles of foraging, but the regulatory mechanisms underlyi... more Leaf‐cutting ants display regular diel cycles of foraging, but the regulatory mechanisms underlying these cycles are not well known. There are, however, some indications in the literature that accumulation of leaf tissue inside a nest dampens recruitment of foragers, thereby providing a negative feedback that can lead to periodic foraging. We investigated two foraging cycles occurring simultaneously in an Atta colombica colony, one involving leaf harvesting and the other exploiting an ephemeral crop of ripe fruit.
Leaf harvesting followed a typical diel pattern of a 10–12 h foraging bout followed by a period of inactivity, while fruit harvesting occurred continuously, but with a regular pre‐dawn dip in activity that marked a 24 h cycle.
Although the results of the present study are drawn from a single field colony, the difference found is consistent with a mechanism of negative feedback regulation acting in parallel on two resources that differ in their rates of distribution and processing, creating cycles of formation and depletion of material caches.
This hypothesis should provoke further interest from students of ant behaviour and some simple manipulative experiments that would begin to test it are outlined. Any role of resource caches in regulating foraging by Atta colonies may have similarities to the logistics of warehouse inventories in human economic activity.
ALife XV Workshop on multidisciplinary applications of evolutionary game theory, 2016
We investigate the effects of social interactions in task al- location using Evolutionary Game Th... more We investigate the effects of social interactions in task al- location using Evolutionary Game Theory (EGT). We propose a simple task-allocation game and study how different learning mechanisms can give rise to specialised and non- specialised colonies under different ecological conditions. By combining agent-based simulations and adaptive dynamics we show that social learning can result in colonies of generalists or specialists, depending on ecological parameters. Agent-based simulations further show that learning dynamics play a crucial role in task allocation. In particular, introspective individual learning readily favours the emergence of specialists, while a process resembling task recruitment favours the emergence of generalists.
Int. Conf. on Bio-inspired Information and Communications Technologies (formerly BIONETICS), 2016
One of the main factors behind the amazing ecological success of social insects is their ability ... more One of the main factors behind the amazing ecological success of social insects is their ability to flexibly allocate the colony's workforce to all the different tasks it has to address. Insights into the self-organised task allocation methods used for this have given rise to the design of an important class of bio-inspired algorithms for network control, industrial optimisation, and other applications. The most widely used class of models for self-organised task allocation, which also forms the core of these algorithms, are response threshold models.
We revisit response threshold models with new experiments using temperature regulation in bumblebee colonies as the model system. We show that standard response threshold models do not fit our experiments and present an alternative behavioural model. This captures a fine-grained, time resolved picture of task engagement, which enables us to investigate task allocation with a different set of statistical methods. Using these we show that our model fits the experiment well and explains its salient aspects.
We compare the effectiveness of our model behaviour with that of response threshold models and demonstrate that it can lead to more efficient task management when demands fluctuate. Our results have the potential to provide a basis for the design of more efficient task allocation algorithms for dynamic environments and to elucidate important biological questions, such as the functional role of inter-individual variation.
We present a unified approach to describing certain types of collective decision making in swarm ... more We present a unified approach to describing certain types of collective decision making in swarm robotics that bridges from a microscopic individual-based description to aggregate properties. Our approach encompasses robot swarm experiments, microscopic and probabilistic macroscopic-discrete simulations as well as an analytic mathematical model. Following up on previous work, we identify the symmetry parameter, a measure of the progress of the swarm towards a decision, as a fundamental integrated swarm property and formulate its time evolution as a continuous-time Markov process. Contrary to previous work, which justified this approach only empirically and a posterior,I, we justify it from first principles and drive hard limits on the parameter regime in which it is applicable
We present a method for mesoscopic, dynamic Monte Carlo simulations of pattern formation in excit... more We present a method for mesoscopic, dynamic Monte Carlo simulations of pattern formation in excitable reaction–diffusion systems. Using a two-level parallelization approach, our simulations cover the whole range of the parameter space, from the noise-dominated low-particle number regime to the quasi-deterministic high-particle number limit. Three qualitatively different case studies are performed that stand exemplary for the wide variety of excitable systems. We present mesoscopic stochastic simulations of the Gray-Scott model, of a simplified model for intracellular Ca2+ oscillations and, for the first time, of the Oregonator model. We achieve simulations with up to 10^10 particles. The software and the model files are freely available and researchers can use the models to reproduce our results or adapt and refine them for further exploration
For many biological applications, a macroscopic (deterministic) treatment of reaction-drift-diffu... more For many biological applications, a macroscopic (deterministic) treatment of reaction-drift-diffusion systems is insufficient. Instead, one has to properly handle the stochastic nature of the problem and generate true sample paths of the underlying probability distribution. Unfortunately, stochastic algorithms are computationally expensive and, in most cases, the large number of participating particles renders the relevant parameter regimes inaccessible. In an attempt to address this problem we present a genuine stochastic, multi-dimensional algorithm that solves the inhomogeneous, non-linear, drift-diffusion problem on a mesoscopic level. Our method improves on existing implementations in being multi-dimensional and handling inhomogeneous drift and diffusion. The algorithm is well suited for implementation on data-parallel hardware architectures such as general-purpose graphics processing units (GPUs). We integrate the method into an operator-splitting approach that decouples chemical reactions from the spatial evolution. We demonstrate the validity and applicability of our algorithm with a comprehensive suite of standard test problems that also serve to quantify the numerical accuracy of the method. We provide a freely available, fully functional GPU implementation. Integration into Inchman, a user-friendly web service, that allows researchers to perform parallel simulations of reaction-drift-diffusion systems on GPU clusters is underway.
Reaction-diffusion systems can be used to model a large variety of complex self-organized phenome... more Reaction-diffusion systems can be used to model a large variety of complex self-organized phenomena occurring in biological, chemical, and social systems. The common macroscopic description of these systems, based on a Fokker-Planck equation (FPE), suffers from major limitations. Most importantly, it fails at low particle densities and it is impossible to incorporate individual-level experimental observations. A microscopic Langevin-type individual-based description can – in principle – address these issues but is challenging and computationally expensive to the point that hardware limitations severely restrict their applicability to models of realistic size.
We present a graphics-processor accelerated stochastic simulation solver that obtains performance gains of up to two orders of magnitude even on workstations. We provide a versatile web-interface allowing researcher to perform complex experiments and parameter studies on a dedicated GPU cluster.
We present a massively parallel stochastic simulation algorithm (SSA) for reaction-diffusion syst... more We present a massively parallel stochastic simulation algorithm (SSA) for reaction-diffusion systems implemented on Graphics Processing Units (GPUs). These are designated chips optimized to process a high number of floating point operations in parallel, rendering them well-suited for a range of scientifichigh-performance computations. Newer GPU generations provide a high-level programming interface which turns them into General-Purpose Graphics Processing Units ( GPGPUs). Our SSA exploits GPGPU architecture to achieve a performance gain of two orders of magnitude over the fastest existing implementations on conventional hardware.
Self-organized mechanisms are widely used in nature to achieve flexible, adaptive control and dec... more Self-organized mechanisms are widely used in nature to achieve flexible, adaptive control and decision-making. It has recently been shown that noise can play a crucial functional role in such systems. Essentially, noise can enable a self-organized system to reliably adapt to short-term changes in the environment while maintaining a generally stable behavior. This is fundamental in biological systems because they must strike a delicate balance between stable and flexible behavior. We investigate the question how noise influences the decision-making of the true slime mold Physarum polycephalum, an important model species for the investigation of computational abilities in simple organisms. We propose a simple biological experiment to investigate the reaction of P. polycephalum to time-variant risk factors. We present a stochastic extension of an established mathematical model for P. polycephalum to analyze this experiment. It predicts that noise can enable P. polycephalum to correctly assess time-variant risk factors, while the corresponding noise-free system fails to do so. Importantly, our analysis holds interest beyond the study of P. polycephalum. In conjunction with earlier work it demonstrates that the influence of noise on self-organized decision-making is not tied to a specific self-organized mechanism or its physical realization. Rather it is a general property of the underlying process dynamics, which appears to be universal across a wide range of systems. Our study thus provides further evidence that stochastic resonance is a fundamental component of the decision-making in self-organized macroscopic groups and organisms.
Symmetry breaking is commonly found in self-organized collective decision making. It serves an im... more Symmetry breaking is commonly found in self-organized collective decision making. It serves an important functional role, specifically in biological and bio-inspired systems. The analysis of symmetry breaking is thus an important key to understanding self-organized decision making. However, in many systems of practical importance avail-able analytic methods cannot be applied due to the complexity of the scenario and consequentially the model. This applies specifically to self-organization in bio-inspired engineering. We propose a new modelling approach which allows us to formally analyze important properties of such processes. The core idea of our approach is to infer a compact model based on stochastic processes for a one-dimensional symmetry parameter. This enables us to analyze the fundamental properties of even complex collective decision making processes via Fokker–Planck theory. We are able to quantitatively address the effectiveness of symmetry breaking, the stability, the time taken to reach a consensus, and other parameters. This is demonstrated with two examples from swarm robotics
Proceedings of the Royal Society B: Biological Sciences, 2009
Recruitment via pheromone trails by ants is arguably one of the best-studied examples of self-org... more Recruitment via pheromone trails by ants is arguably one of the best-studied examples of self-organization in animal societies. Yet it is still unclear if and how trail recruitment allows a colony to adapt to changes in its foraging environment. We study foraging decisions by colonies of the ant Pheidole megacephala under dynamic conditions. Our experiments show that P. megacephala, unlike many other mass recruiting species, can make a collective decision for the better of two food sources even when the environment changes dynamically. We developed a stochastic differential equation model that explains our data qualitatively and quantitatively. Analysing this model reveals that both deterministic and stochastic effects(noise) work together to allow colonies to efficiently track changes in the environment. Our study thus suggests that a certain level of noise is not a disturbance in self-organized decision-making but rather serves an important functional role.
IEEE International Conference on Self-Adaptive and Self-Organizing Systems, 2008
One of the core aspects that make self-organized systems an interesting engineering paradigm is t... more One of the core aspects that make self-organized systems an interesting engineering paradigm is their potential to behave adaptively. Unravelling the fundamental mechanisms that drive this adaptiveness is of prime importance for understanding and designing such systems. The present paper demonstrates that noise is one of the core ingredients that enable self-organized systems to behave adaptively. This suggests that noise should be taken into account as a constructive component when engineering them. Our study analyses two different but closely related self-organized systems: a man-made system, Ant Colony Optimization algorithms (ACO), and real ant colonies, the natural system that inspired ACO. We demonstrate that the conventionally used mean-field analysis is not a correct description of their behavior in dynamic environments. This can only be achieved by a stochastic analysis that quantitatively takes noise into account. We present such an analysis based on Ito-Diffusions and Fokker-Planck equations and show it to be consistent with experimental data. Real ant colonies and ACO are both controlled by coupled self-limiting feedback loops. Decision making in such systems can be understood as stochastic attractor switching. This is the basis of our analysis. As coupled feedback mechanism are a universal control mechanism found in many types of self-organized systems, we expect our approach to be applicable to a vast array of other natural and man-made self-organized systems.
Social insects allocate their workforce in a decentralised fashion, addressing multiple tasks and... more Social insects allocate their workforce in a decentralised fashion, addressing multiple tasks and responding effectively to environmental changes. This process is fundamental to their ecological success, but the mechanisms behind it are not well understood. While most models focus on internal and individual factors, empirical evidence highlights the importance of ecology and social interactions. To address this gap, we propose a game theoretical model of task allocation. Our main findings are twofold: Firstly, the specialisation emerging from self-organised task allocation can be largely determined by the ecology. Weakly specialised colonies in which all individuals perform more than one task emerge when foraging is cheap; in contrast, harsher environments with high foraging costs lead to strong specialisation in which each individual fully engages in a single task. Secondly, social interactions lead to important differences in dynamic environments. Colonies whose individuals rely on their own experience are predicted to be more flexible when dealing with change than colonies relying on social information. We also find that, counter to intuition, strongly specialised colonies may perform suboptimally, whereas the group performance of weakly specialised colonies approaches optimality. Our simulation results fully agree with the predictions of the mathematical model for the regions where the latter is analytically tractable. Our results are useful in framing relevant and important empirical questions, where ecology and interactions are key elements of hypotheses and predictions.
Social insect colonies distribute their workforce with amazing flexibility across a large array o... more Social insect colonies distribute their workforce with amazing flexibility across a large array of diverse tasks under fluctuating external conditions and internal demands. Deciphering the individual rules of task selection and task performance is at the heart of understanding how colonies can achieve this collective feature. Models play an important role in this endeavor, as they allow us to investigate how the rules of individual behavior give rise to emergent patterns at the colony level. Modulation of individual behavior occurs at many different timescales and to successfully use a model we need to ensure that it applies on the timescale under observation. Here, we focus on short timescales and ask the question whether the most commonly used class of models (response threshold models) adequately describes behavioral modulation on this timescale. We study the fanning behavior of bumblebees on temperature-controlled brood dummies and investigate the effect of (i) stimulus intensity, (ii) repeated task performance, and (iii) task performance feedback. We analyze the timing patterns (rates of task engagement and task disengagement) using survival analysis. Our results show that stimulus intensity does not significantly influence individual task investment at these comparably short timescales. In contrast, repeated task performance and task performance feedback affect individual task investment. We propose an explicitly time-resolved individual-based model and simulate this model to study how patterns of individual task engagement influence task involvement at the group level, finding support for the hypothesis that regulation mechanisms at different timescales can improve performance at the group level in dynamic environments.
A wide range of group-living animals construct tangible infrastructure networks, often of remarka... more A wide range of group-living animals construct tangible infrastructure networks, often of remarkable size and complexity. In ant colonies, infrastructure construction may require tens of thousands of work hours distributed among many thousand individuals. What are the individual behaviours involved in the construction and what level of complexity in inter-individual interaction is required to organize this effort? We investigate this question in one of the most sophisticated trail builders in the animal world: the leafcutter ants, which remove leaf litter, cut through overhangs and shift soil to level the path of trail networks that may cumulatively extend for kilometres. Based on obstruction experiments in the field and the laboratory, we identify and quantify different individual trail clearing behaviours. Via a computational model, we further investigate the presence of recruitment, which—through direct or indirect information transfer between individuals—is one of the main organizing mechanisms of many collective behaviours in ants. We show that large-scale transport networks can emerge purely from the stochastic process of workers encountering obstructions and subsequently engaging in removal behaviour with a fixed probability. In addition to such incidental removal, we describe a dedicated clearing behaviour in which workers remove additional obstructions independent of chance encounters. We show that to explain the dynamics observed in the experiments, no information exchange (e.g. via recruitment) is required, and propose that large-scale infrastructure construction of this type can be achieved without coordination between individuals.
Self-organised collective decision making is one of the core components of swarm intelligence, an... more Self-organised collective decision making is one of the core components of swarm intelligence, and numerous swarm algorithms that are widely used in optimisation and optimal control have been inspired by the biological mechanisms driving it. Beyond the life sciences and bio-inspired engineering, collective decision making is important in a number of other disciplines, most prominently economics and the social sciences. A paradigmatic model system for collective decision making is the foraging behaviour of mass recruiting ant colonies. While this system has been investigated extensively, our knowledge about its function in dynamic environments is still incomplete at best. We show that the mathematical model of mass foraging is really just a specific instance of a very general class of rational group decision making processes. We analyse this general class using an information-theoretic framework, which allows us to abstract from the specific details of a fixed model system. We specifically investigate how noisy communication can enable groups to share information about changes in an environment more efficiently. In the present paper, we show that an optimal noise level exists and that this optimal level depends on the rate of change in the environment. We explain this on the basis of stochastic resonance theory and show why stochastic attractor switching is a suitable base mechanism for adaptive group decision making in dynamic environments.
Self-organized mechanisms are frequently encountered in nature and known to achieve flexible, ada... more Self-organized mechanisms are frequently encountered in nature and known to achieve flexible, adaptive control and decision-making. Noise plays a crucial role in such systems: It can enable a self-organized system to reliably adapt to short-term changes in the environment while maintaining a generally stable behavior. This is fundamental in biological systems because they must strike a delicate balance between stable and flexible behavior. In the present paper we analyse the role of noise in the decision-making of the true slime mold Physarum polycephalum, an important model species for the investigation of computational abilities in simple organisms. We propose a simple biological experiment to investigate the reaction of P. polycephalum to time-variant risk factors and present a stochastic extension of an established mathematical model for P. polycephalum to analyze this experiment. It predicts that—due to the mechanism of stochastic resonance—noise can enable P. polycephalum to correctly assess time-variant risk factors, while the corresponding noise-free system fails to do so. Beyond the study of P. polycephalum we demonstrate that the influence of noise on self-organized decision-making is not tied to a specific organism. Rather it is a general property of the underlying process dynamics, which appears to be universal across a wide range of systems. Our study thus provides further evidence that stochastic resonance is a fundamental component of the decision-making in self-organized macroscopic and microscopic groups and organisms.
Few ant species construct cleared trails. Among those that do, leaf-cutting Atta ants build the m... more Few ant species construct cleared trails. Among those that do, leaf-cutting Atta ants build the most prominent networks, with single colonies clearing debris and obstructions from hundreds of meters of trails annually. Workers on cleared paths move at higher speed than they do over uncleared litter, and one measurement of the time and energetic costs of trail clearance suggests that benefits of trail usage far outweigh the investment costs of trail clearing. The ecological basis of trail clearing remains uncertain, however, because no full account has been made of benefits and costs in common units that allow comparison. We make such an account using a scalable, integrative model of trail investment and foraging energetics. Contrary to assumptions in previous work, we find that trail clearing needs not always be energetically profitable for leaf-cutting ants. Profitability depends on the workforce composition, specifically, on how many ants in a traffic stream act as maintenance workforce to respond to sudden and unpredictable obstructions, such as leaf fall. Such maintenance patrols have not previously been recognized as a cost of trail building. If the patrolling workforce is not too large, the energetic savings from foraging over cleared trails offset the investment and maintenance costs within a few days. Under some conditions, however, amortization can take weeks or months, or trail clearing can become unprofitable altogether. This suggests that Atta colonies must have a mechanism to regulate the intensity of their trail clearing behavior. We explore possible mechanisms and make testable predictions for future research.
Leaf‐cutting ants display regular diel cycles of foraging, but the regulatory mechanisms underlyi... more Leaf‐cutting ants display regular diel cycles of foraging, but the regulatory mechanisms underlying these cycles are not well known. There are, however, some indications in the literature that accumulation of leaf tissue inside a nest dampens recruitment of foragers, thereby providing a negative feedback that can lead to periodic foraging. We investigated two foraging cycles occurring simultaneously in an Atta colombica colony, one involving leaf harvesting and the other exploiting an ephemeral crop of ripe fruit.
Leaf harvesting followed a typical diel pattern of a 10–12 h foraging bout followed by a period of inactivity, while fruit harvesting occurred continuously, but with a regular pre‐dawn dip in activity that marked a 24 h cycle.
Although the results of the present study are drawn from a single field colony, the difference found is consistent with a mechanism of negative feedback regulation acting in parallel on two resources that differ in their rates of distribution and processing, creating cycles of formation and depletion of material caches.
This hypothesis should provoke further interest from students of ant behaviour and some simple manipulative experiments that would begin to test it are outlined. Any role of resource caches in regulating foraging by Atta colonies may have similarities to the logistics of warehouse inventories in human economic activity.
ALife XV Workshop on multidisciplinary applications of evolutionary game theory, 2016
We investigate the effects of social interactions in task al- location using Evolutionary Game Th... more We investigate the effects of social interactions in task al- location using Evolutionary Game Theory (EGT). We propose a simple task-allocation game and study how different learning mechanisms can give rise to specialised and non- specialised colonies under different ecological conditions. By combining agent-based simulations and adaptive dynamics we show that social learning can result in colonies of generalists or specialists, depending on ecological parameters. Agent-based simulations further show that learning dynamics play a crucial role in task allocation. In particular, introspective individual learning readily favours the emergence of specialists, while a process resembling task recruitment favours the emergence of generalists.
Int. Conf. on Bio-inspired Information and Communications Technologies (formerly BIONETICS), 2016
One of the main factors behind the amazing ecological success of social insects is their ability ... more One of the main factors behind the amazing ecological success of social insects is their ability to flexibly allocate the colony's workforce to all the different tasks it has to address. Insights into the self-organised task allocation methods used for this have given rise to the design of an important class of bio-inspired algorithms for network control, industrial optimisation, and other applications. The most widely used class of models for self-organised task allocation, which also forms the core of these algorithms, are response threshold models.
We revisit response threshold models with new experiments using temperature regulation in bumblebee colonies as the model system. We show that standard response threshold models do not fit our experiments and present an alternative behavioural model. This captures a fine-grained, time resolved picture of task engagement, which enables us to investigate task allocation with a different set of statistical methods. Using these we show that our model fits the experiment well and explains its salient aspects.
We compare the effectiveness of our model behaviour with that of response threshold models and demonstrate that it can lead to more efficient task management when demands fluctuate. Our results have the potential to provide a basis for the design of more efficient task allocation algorithms for dynamic environments and to elucidate important biological questions, such as the functional role of inter-individual variation.
We present a unified approach to describing certain types of collective decision making in swarm ... more We present a unified approach to describing certain types of collective decision making in swarm robotics that bridges from a microscopic individual-based description to aggregate properties. Our approach encompasses robot swarm experiments, microscopic and probabilistic macroscopic-discrete simulations as well as an analytic mathematical model. Following up on previous work, we identify the symmetry parameter, a measure of the progress of the swarm towards a decision, as a fundamental integrated swarm property and formulate its time evolution as a continuous-time Markov process. Contrary to previous work, which justified this approach only empirically and a posterior,I, we justify it from first principles and drive hard limits on the parameter regime in which it is applicable
We present a method for mesoscopic, dynamic Monte Carlo simulations of pattern formation in excit... more We present a method for mesoscopic, dynamic Monte Carlo simulations of pattern formation in excitable reaction–diffusion systems. Using a two-level parallelization approach, our simulations cover the whole range of the parameter space, from the noise-dominated low-particle number regime to the quasi-deterministic high-particle number limit. Three qualitatively different case studies are performed that stand exemplary for the wide variety of excitable systems. We present mesoscopic stochastic simulations of the Gray-Scott model, of a simplified model for intracellular Ca2+ oscillations and, for the first time, of the Oregonator model. We achieve simulations with up to 10^10 particles. The software and the model files are freely available and researchers can use the models to reproduce our results or adapt and refine them for further exploration
For many biological applications, a macroscopic (deterministic) treatment of reaction-drift-diffu... more For many biological applications, a macroscopic (deterministic) treatment of reaction-drift-diffusion systems is insufficient. Instead, one has to properly handle the stochastic nature of the problem and generate true sample paths of the underlying probability distribution. Unfortunately, stochastic algorithms are computationally expensive and, in most cases, the large number of participating particles renders the relevant parameter regimes inaccessible. In an attempt to address this problem we present a genuine stochastic, multi-dimensional algorithm that solves the inhomogeneous, non-linear, drift-diffusion problem on a mesoscopic level. Our method improves on existing implementations in being multi-dimensional and handling inhomogeneous drift and diffusion. The algorithm is well suited for implementation on data-parallel hardware architectures such as general-purpose graphics processing units (GPUs). We integrate the method into an operator-splitting approach that decouples chemical reactions from the spatial evolution. We demonstrate the validity and applicability of our algorithm with a comprehensive suite of standard test problems that also serve to quantify the numerical accuracy of the method. We provide a freely available, fully functional GPU implementation. Integration into Inchman, a user-friendly web service, that allows researchers to perform parallel simulations of reaction-drift-diffusion systems on GPU clusters is underway.
Reaction-diffusion systems can be used to model a large variety of complex self-organized phenome... more Reaction-diffusion systems can be used to model a large variety of complex self-organized phenomena occurring in biological, chemical, and social systems. The common macroscopic description of these systems, based on a Fokker-Planck equation (FPE), suffers from major limitations. Most importantly, it fails at low particle densities and it is impossible to incorporate individual-level experimental observations. A microscopic Langevin-type individual-based description can – in principle – address these issues but is challenging and computationally expensive to the point that hardware limitations severely restrict their applicability to models of realistic size.
We present a graphics-processor accelerated stochastic simulation solver that obtains performance gains of up to two orders of magnitude even on workstations. We provide a versatile web-interface allowing researcher to perform complex experiments and parameter studies on a dedicated GPU cluster.
We present a massively parallel stochastic simulation algorithm (SSA) for reaction-diffusion syst... more We present a massively parallel stochastic simulation algorithm (SSA) for reaction-diffusion systems implemented on Graphics Processing Units (GPUs). These are designated chips optimized to process a high number of floating point operations in parallel, rendering them well-suited for a range of scientifichigh-performance computations. Newer GPU generations provide a high-level programming interface which turns them into General-Purpose Graphics Processing Units ( GPGPUs). Our SSA exploits GPGPU architecture to achieve a performance gain of two orders of magnitude over the fastest existing implementations on conventional hardware.
Self-organized mechanisms are widely used in nature to achieve flexible, adaptive control and dec... more Self-organized mechanisms are widely used in nature to achieve flexible, adaptive control and decision-making. It has recently been shown that noise can play a crucial functional role in such systems. Essentially, noise can enable a self-organized system to reliably adapt to short-term changes in the environment while maintaining a generally stable behavior. This is fundamental in biological systems because they must strike a delicate balance between stable and flexible behavior. We investigate the question how noise influences the decision-making of the true slime mold Physarum polycephalum, an important model species for the investigation of computational abilities in simple organisms. We propose a simple biological experiment to investigate the reaction of P. polycephalum to time-variant risk factors. We present a stochastic extension of an established mathematical model for P. polycephalum to analyze this experiment. It predicts that noise can enable P. polycephalum to correctly assess time-variant risk factors, while the corresponding noise-free system fails to do so. Importantly, our analysis holds interest beyond the study of P. polycephalum. In conjunction with earlier work it demonstrates that the influence of noise on self-organized decision-making is not tied to a specific self-organized mechanism or its physical realization. Rather it is a general property of the underlying process dynamics, which appears to be universal across a wide range of systems. Our study thus provides further evidence that stochastic resonance is a fundamental component of the decision-making in self-organized macroscopic groups and organisms.
Symmetry breaking is commonly found in self-organized collective decision making. It serves an im... more Symmetry breaking is commonly found in self-organized collective decision making. It serves an important functional role, specifically in biological and bio-inspired systems. The analysis of symmetry breaking is thus an important key to understanding self-organized decision making. However, in many systems of practical importance avail-able analytic methods cannot be applied due to the complexity of the scenario and consequentially the model. This applies specifically to self-organization in bio-inspired engineering. We propose a new modelling approach which allows us to formally analyze important properties of such processes. The core idea of our approach is to infer a compact model based on stochastic processes for a one-dimensional symmetry parameter. This enables us to analyze the fundamental properties of even complex collective decision making processes via Fokker–Planck theory. We are able to quantitatively address the effectiveness of symmetry breaking, the stability, the time taken to reach a consensus, and other parameters. This is demonstrated with two examples from swarm robotics
Proceedings of the Royal Society B: Biological Sciences, 2009
Recruitment via pheromone trails by ants is arguably one of the best-studied examples of self-org... more Recruitment via pheromone trails by ants is arguably one of the best-studied examples of self-organization in animal societies. Yet it is still unclear if and how trail recruitment allows a colony to adapt to changes in its foraging environment. We study foraging decisions by colonies of the ant Pheidole megacephala under dynamic conditions. Our experiments show that P. megacephala, unlike many other mass recruiting species, can make a collective decision for the better of two food sources even when the environment changes dynamically. We developed a stochastic differential equation model that explains our data qualitatively and quantitatively. Analysing this model reveals that both deterministic and stochastic effects(noise) work together to allow colonies to efficiently track changes in the environment. Our study thus suggests that a certain level of noise is not a disturbance in self-organized decision-making but rather serves an important functional role.
IEEE International Conference on Self-Adaptive and Self-Organizing Systems, 2008
One of the core aspects that make self-organized systems an interesting engineering paradigm is t... more One of the core aspects that make self-organized systems an interesting engineering paradigm is their potential to behave adaptively. Unravelling the fundamental mechanisms that drive this adaptiveness is of prime importance for understanding and designing such systems. The present paper demonstrates that noise is one of the core ingredients that enable self-organized systems to behave adaptively. This suggests that noise should be taken into account as a constructive component when engineering them. Our study analyses two different but closely related self-organized systems: a man-made system, Ant Colony Optimization algorithms (ACO), and real ant colonies, the natural system that inspired ACO. We demonstrate that the conventionally used mean-field analysis is not a correct description of their behavior in dynamic environments. This can only be achieved by a stochastic analysis that quantitatively takes noise into account. We present such an analysis based on Ito-Diffusions and Fokker-Planck equations and show it to be consistent with experimental data. Real ant colonies and ACO are both controlled by coupled self-limiting feedback loops. Decision making in such systems can be understood as stochastic attractor switching. This is the basis of our analysis. As coupled feedback mechanism are a universal control mechanism found in many types of self-organized systems, we expect our approach to be applicable to a vast array of other natural and man-made self-organized systems.
Sorting and clustering methods inspired by the behavior of real ants are among the earliest metho... more Sorting and clustering methods inspired by the behavior of real ants are among the earliest methods in ant-based meta-heuristics. We revisit these methods in the context of a concrete application and introduce some modifications that yield significant improvements in terms of both quality and efficiency. Firstly, we re-examine their capability to simultaneously perform a combination of clustering and multi-dimensional scaling. In contrast to the assumptions made in earlier literature, our results suggest that these algorithms perform scaling only to a very limited degree. We show how to improve on this by some modifications of the algorithm and a hybridization with a simple pre-processing phase. Secondly, we discuss how the time-complexity of these algorithms can be improved. The improved algorithms are used as the core mechanism in a visual document retrieval system for world-wide web searches.
Clustering with swarm-based algorithms is emerging as an alternative to more conventional cluster... more Clustering with swarm-based algorithms is emerging as an alternative to more conventional clustering methods, such as hierarchical clustering and k-means. Ant-based clustering stands out as the most widely used group of swarm-based clustering algorithms. Broadly speaking, there are two main types of ant-based clustering: the first group of methods directly mimics the clustering behavior observed in real ant colonies. The second group is less directly inspired by nature: the clustering task is reformulated as an optimization task and general purpose ant-based optimization heuristics are utilized to find good or near-optimal clusterings. This papers reviews both approaches and places these methods in the wider context of general swarm-based clustering approaches.
The paper presents self-organizing graphs, a novel approach to graph layout based on a competitiv... more The paper presents self-organizing graphs, a novel approach to graph layout based on a competitive learning algorithm. This method is an extension of selforganization strategies known from unsupervised neural networks, namely from Kohonen's self-organizing map. Its main advantage is that it is very flexibly adaptable to arbitrary types of visualization spaces, for it is explicitly parameterized by a metric model of the layout space. Yet the method consumes comparatively little computational resources and does not need any heavy-duty preprocessing. Unlike with other stochastic layout algorithms, not even the costly repeated evaluation of an objective function is required. To our knowledge this is the first connectionist approach to graph layout. The paper presents applications to 2D-layout as well as to 3D-layout and to layout in arbitrary metric spaces, such as networks on spherical surfaces. 1 Introduction Automatic layout techniques are a crucial component for any application ...
Ant Colony Optimisation algorithms perform competitively with other meta-heuristics for many type... more Ant Colony Optimisation algorithms perform competitively with other meta-heuristics for many types of optimisation problems, but unfortunately their performance does not always degrade gracefully when the problem contains hard constraints. Many industrially relevant problems, such as fleet routing, rostering and timetabling, are typically subject to hard constraints. A complementary technique for solving combinatorial optimisation problems is Constraint Programming (CP). CP
ABSTRACT Ant Colony Optimization (ACO) is a recent stochastic meta-heuristic inspired by the fora... more ABSTRACT Ant Colony Optimization (ACO) is a recent stochastic meta-heuristic inspired by the foraging behaviour of real ants. As for all meta-heuristics the balance between learning based on previous solutions (in-tensification) and exploration of the search space (diversification) is of crucial importance. The present paper explores a novel approach to di-versity control in ACO. The common idea of most diversity control mech-anisms is to avoid or slow down full convergence. We suggest to instead use a fast converging search algorithm that is artificially confined to the critical phase of its convergence dynamics. We also analyze the influence of an ACO parameter that does not seem to have received sufficient at-tention in the ACO literature: α, the exponent on the pheromone level in the probabilistic choice function. Our studies suggest that α does not only qualitatively determine diversity and convergence behaviour, but also that a variable α can be used to render ant algorithms more ro-bust. Based on these ideas we construct an Algorithm ccAS for which we present some encouraging results on standard benchmarks.
International Journal of Production Economics, 2013
Abstract We consider a scheduling problem arising in the mining industry. Ore from several mining... more Abstract We consider a scheduling problem arising in the mining industry. Ore from several mining sites must be transferred to ports to be loaded on ships in a timely manner. In doing so, several constraints must be met which involve transporting the ore and deadlines. These deadlines are two-fold; there is a preferred deadline by which the ships should be loaded and there is a final deadline by which time the ships must be loaded. Corresponding to the two types of deadlines, each task is associated with a soft and hard due time. The ...
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2009
A recent line of research concerns the integration of ant colony optimization and constraint prog... more A recent line of research concerns the integration of ant colony optimization and constraint programming. Hereby, constraint programming is used for eliminating parts of the search tree during the solution construction of ant colony optimization. In the context of a single ...
Many real-world optimization problems can be modelled as combinatorial optimization problems. Oft... more Many real-world optimization problems can be modelled as combinatorial optimization problems. Often, these problems are characterized by their large size and the presence of multiple, conflicting objectives. Despite progress in solving multi-objective combinatorial optimization problems exactly, the large size often means that heuristics are required for their solution in acceptable time. Since the middle of the nineties the trend is
IECON 2011 - 37th Annual Conference of the IEEE Industrial Electronics Society, 2011
Abstract Direct plugin types of battery charging methods are commonly used for powering mobile ob... more Abstract Direct plugin types of battery charging methods are commonly used for powering mobile objects like swarm robots. With a large number of robots that need to be charged frequently, direct plugging in or manual changing of batteries can be very labor intensive ...
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2009
The travelling salesman problem with time windows is a difficult optimization problem that appear... more The travelling salesman problem with time windows is a difficult optimization problem that appears, for example, in logistics. Among the possible objective functions we chose the optimization of the makespan. For solving this problem we propose a so-called Beam-...
Genetic and Evolutionary Computation Conference, GECCO'11, 2011
ABSTRACT Hybrid methods for solving combinatorial optimization problems have become increasingly ... more ABSTRACT Hybrid methods for solving combinatorial optimization problems have become increasingly popular recently. The present paper is concerned with hybrids of ant colony optimization and constraint programming which are typically useful for problems with hard constraints. However, the original algorithm suffered from large CPU time requirements. It was shown that such an integration can be made efficient via a further hybridization with beam search resulting in CP-Beam-ACO. The original work suggested this in the context of job scheduling. We show here that this algorithm type is also effective on another problem class, namely the car sequencing. We consider an optimization version, where we aim to optimize the utilization rates across the sequence. Car sequencing is a notoriously difficult problem, because it is difficult to obtain good bounds via relaxations. We show that stochastic sampling provides superior results to well known lower bounds for this problem when combined with CP-Beam-ACO.
2005 IEEE Congress on Evolutionary Computation, 2005
... 2 Ant Colony Optimization ... An alternative to these widely-used ways of handling hard const... more ... 2 Ant Colony Optimization ... An alternative to these widely-used ways of handling hard constraints in constructive meta-heuristics is to incor-porate some form of lookahead into the construction phase so that (almost ... The survey [9] notes that as yet only very simple lookahead ...
ABSTRACT Many applications of swarm robotics require autonomous navigation in unknown environment... more ABSTRACT Many applications of swarm robotics require autonomous navigation in unknown environments. We describe a new collective navigation strategy based on diffusion limited aggregation and bacterial foraging behaviour. Both methods are suitable for typical swarm robots as they require only minimal sensory and control capabilities. We demonstrate the usefulness of the strategy with a swarm that is capable of autonomously finding charging stations and show that the collective search can be significantly more effective than individual-based search.
2008 IEEE Congress on Evolutionary Computation, CEC 2008, 2008
ABSTRACT This paper investigates the use of hybrid meta-heuristics based on ant colony optimizati... more ABSTRACT This paper investigates the use of hybrid meta-heuristics based on ant colony optimization (ACO) for the strip packing problem. Here, a fixed set of rectangular items of fixed sizes have to be placed on a strip of fixed width and infinite height without overlaps and with the objective to minimize the height used. We analyze a commonly used basic placement heuristic (BLF) by itself and in a number of hybrid combinations with ACO. We compare versions that learn item order only, item rotation only, both independently, and rotations conditionally upon placement order. Our analysis shows that integrating a learning meta-heuristic provides a significant performance advantage over using the basic placement heuristic by itself. The experiments confirm that even just learning a placement order alone can provide significant performance improvements. Interestingly, learning item rotations provides at best a marginal advantage. The best hybrid algorithm presented in this paper significantly outperforms previously reported strip packing meta-heuristics.
... string so that part of the recognition task has to be performed by this translation rather th... more ... string so that part of the recognition task has to be performed by this translation rather than ... As pointed out in [26], hyper-edge replacement systems are identical to plex grammars ... a and P are arbitrary edge-and node-labelled graphs, E is the em-bedding function which specifies ...
... 46556, USA. Aaron Sloman A. Sloman@ cs. bham. ac. uk ... computational insights. The author o... more ... 46556, USA. Aaron Sloman A. Sloman@ cs. bham. ac. uk ... computational insights. The author of Chapter 1, Aaron Sloman, has a special place in the de-bate on diagrammatic reasoning within the artificial intelligence community. Sloman's ...
Proceedings IEEE Workshop on Visual Languages, 2000
ABSTRACT this paper has been published in: Proceedings of the 1992 IEEE Workshop on Visual Langua... more ABSTRACT this paper has been published in: Proceedings of the 1992 IEEE Workshop on Visual Languages, Seattle/WA, September 1992.
For applications which generate diagrammatic represen- tations automatic layout techniques are a ... more For applications which generate diagrammatic represen- tations automatic layout techniques are a crucial compo- nent. Since graph-like network diagrams are among the most commonly used and most important types of diagram- matic displays, layout techniques for graphs have been ex- tensively studied. However, a problem with current graph layout methods which are capable of producing satisfactory results for a wide
ABSTRACT Though visual access to spatial database systems has attracted much attention in recent ... more ABSTRACT Though visual access to spatial database systems has attracted much attention in recent years, there have only few deductive visual languages for spatial information systems been proposed. One reason for this may be that most of the spatial visual languages stick to the classical metaphor of map manipulation. We argue that more powerful metaphors are needed to facilitate the development of pictorial languages for spatial deduction that are readily comprehensible and formally manageable at the same time. This paper introduces a fully deductive pictorial language for spatial information based on sketching. 1 Introduction and Motivation In the last decade spatial information systems in general and geographical information systems in particular have become an ever growing field of research. In recent years visual interfaces to spatial information systems have attracted an increasing amount of interest. Several proposals have been made and working systems have been implemented. Some of th...
ABSTRACT In the near future, visual and diagrammatic human-computer interfaces, such as those of ... more ABSTRACT In the near future, visual and diagrammatic human-computer interfaces, such as those of UML-based CASE tools, will be required to offer a much more intelligent behavior than just editing. Yet there is very little formal support and there are almost no tools available for the construction of intelligent diagrammatic environments. The present paper introduces a constraint-based formalism for the specification and implementation of complex diagrammatic environments. We start from grammar-based definitions of diagrammatic languages and show how a constraint solver for diagram recognition and interpretation can automatically be constructed from such grammars. In a second step, the capabilities of these solvers are extended by allowing to axiomatise formal diagrammatic systems so that they can be regarded as a new constraint domain. The ultimate aim of this schema is to establish a constraint logic language for diagrammatic reasoning applications.
ABSTRACT In Chapter 2 we saw that visual language specification methods come in a variety of form... more ABSTRACT In Chapter 2 we saw that visual language specification methods come in a variety of forms, making the systematic comparison of different methods and the abstract classification of visual languages difficult. The fundamental role of the Chomsky hierarchy ...
ABSTRACT This paper discusses picture logic, a visual language for the specification of diagrams ... more ABSTRACT This paper discusses picture logic, a visual language for the specification of diagrams and diagram transformations. Formal specification techniques for diagrammatic or visual languages have previously mainly been targeted towards static diagrammatic languages. For reasoning about certain types of diagrams, however, formalizing a notion of change is inevitable. This is particularly true of visual mathematical notations whose evaluation rules or consequence relations correspond to visual or graphical transformations. The paper presents constraint-based extensions of picture logic which render it suitable for the specification of such diagram notations and the required transformations. Diagrammatic Reasoning and Formalizations In computational diagrammatic reasoning, we can distinguish between reasoning about diagrams and reasoning with diagrams. We can either use non-visual computational methods to reason about diagrams or we can use computational reasoning methods tha...
ABSTRACT A key component of computational diagrammatic reasoning is the automated interpretation ... more ABSTRACT A key component of computational diagrammatic reasoning is the automated interpretation of diagram notations. One common and successful approach to this is based on attributed multiset grammars. The disadvantages of grammars are, however, that they do not allow ready integration of semantic information and that the underlying theory is not strongly developed. Therefore, embeddings of grammars into first-order logic have been investigated. Unfortunately, these are unsatisfactory: Either they are complex and unnatural or else, because of the monotonicity of classical first-order logic, cannot handle diagrammatic reasoning. We investigate the use of two non-standard logics, namely linear logic and situation theory, for the formalization of diagram interpretation and reasoning. The chief advantage of linear logic is that it is a resource-oriented logic, which renders the embedding of grammars straightforward. Situation theory, on the other hand, has been designed for capturing the semantics of natural language and offers powerful methods for modelling more complex aspects of language, such as incomplete views of the world. The paper illustrates embeddings of grammar-based interpretation into both formalisms and also discusses their integration.
... 936. Full Text via CrossRef. LNCS 785S. Conrad, On certification of specifications for TROLL... more ... 936. Full Text via CrossRef. LNCS 785S. Conrad, On certification of specifications for TROLL light objects. In: H. Ehrig and F. Orejas, Editors, Proc. 9th Workshop on Abstract Data Types 4th Compass Workshop (ADT'92), Springer (1994), pp. 158172. ...
We investigate the effects of social interactions in task al- location using Evolutionary Game Th... more We investigate the effects of social interactions in task al- location using Evolutionary Game Theory (EGT). We propose a simple task-allocation game and study how different learning mechanisms can give rise to specialised and non- specialised colonies under different ecological conditions. By combining agent-based simulations and adaptive dynamics we show that social learning can result in colonies of generalists or specialists, depending on ecological parameters. Agent-based simulations further show that learning dynamics play a crucial role in task allocation. In particular, introspective individual learning readily favours the emergence of specialists, while a process resembling task recruitment favours the emergence of generalists.
Division of labour, or the differentiation of the individuals in a collective across tasks, is a ... more Division of labour, or the differentiation of the individuals in a collective across tasks, is a fundamental aspect of social organisations, such as social insect colonies. It allows for efficient resource use and improves the chances of survival for the entire collective. The emergence of large inactive groups of individuals in insect colonies sometimes referred to as laziness, has been a puzzling and hotly debated division-of-labour phenomenon in recent years that is counter to the intuitive notion of effectiveness. It has previously been shown that inactivity can be explained as a by-product of social learning without the need to invoke an adaptive function. While highlighting an interesting and important possibility, this explanation is limited because it is not yet clear whether the relevant aspects of colony life are governed by social learning. In this paper, we explore the two fundamental types of behavioural adaptation that can lead to a division of labour, individual learn...
Significant efforts are being invested to bring the classification and recognition powers of desk... more Significant efforts are being invested to bring the classification and recognition powers of desktop and cloud systems directly to edge devices. The main challenge for deep learning on the edge is to handle extreme resource constraints(memory, CPU speed and lack of GPU support). We present an edge solution for audio classification that achieves close to state-of-the-art performance on ESC-50, the same benchmark used to assess large, non resource-constrained networks. Importantly, we do not specifically engineer the network for edge devices. Rather, we present a universal pipeline that converts a large deep convolutional neural network (CNN) automatically via compression and quantization into a network suitable for resource-impoverished edge devices. We first introduce a new sound classification architecture, ACDNet, that produces above state-of-the-art accuracy on both ESC-10 and ESC-50 which are 96.75% and 87.05% respectively. We then compress ACDNet using a novel network-independe...
Deep Learning has celebrated resounding successes in many application areas of relevance to the I... more Deep Learning has celebrated resounding successes in many application areas of relevance to the Internet-of-Things, for example, computer vision and machine listening. To fully harness the power of deep leaning for the IoT, these technologies must ultimately be brought directly to the edge. The obvious challenge is that deep learning techniques can only be implemented on strictly resource-constrained edge devices if the models are radically downsized. This task relies on different model compression techniques, such as network pruning, quantization and the recent advancement of XNOR-Net. This paper examines the suitability of these techniques for audio classification in microcontrollers. We present an XNOR-Net for end-to-end raw audio classification and a comprehensive empirical study comparing this approach with pruning-and-quantization methods. We show that raw audio classification with XNOR yields comparable performance to regular full precision networks for small numbers of class...
Self-organized mechanisms are frequently encountered in nature and known to achieve flexible, ada... more Self-organized mechanisms are frequently encountered in nature and known to achieve flexible, adaptive control and decision-making. Noise plays a crucial role in such systems: It can enable a self-organized system to reliably adapt to short-term changes in the environment while maintaining a generally stable behavior. This is fundamental in biological systems because they must strike a delicate balance between stable and flexible behavior. In the present paper we analyse the role of noise in the decision-making of the true slime mold Physarum polycephalum, an important model species for the investigation of computational abilities in simple organisms. We propose a simple biological experiment to investigate the reaction of P. polycephalum to time-variant risk factors and present a stochastic extension of an established mathematical model for P. polycephalum to analyze this experiment. It predicts that-due to the mechanism of stochastic resonance-noise can enable P. polycephalum to c...
Ant Colony Optimization (ACO) is a recent stochastic metaheuristic inspired by the foraging behav... more Ant Colony Optimization (ACO) is a recent stochastic metaheuristic inspired by the foraging behaviour of real ants. As for all metaheuristics the balance between learning based on previous solutions (intensification) and exploration of the search space (diversification) is of crucial importance. The present paper explores a novel approach to diversity control in ACO. The common idea of most diversity control mechanisms is to avoid or slow down full convergence. We suggest to instead use a fast converging search algorithm that is artificially confined to the critical phase of its convergence dynamics. We also analyze the influence of an ACO parameter that does not seem to have received sufficient attention in the ACO literature: α, the exponent on the pheromone level in the probabilistic choice function. Our studies suggest that α does not only qualitatively determine diversity and convergence behaviour, but also that a variable α can be used to render ant algorithms more robust. Bas...
Even in application domains where diagrammatic notations are a natural element of discourse and t... more Even in application domains where diagrammatic notations are a natural element of discourse and their meaning is well-understood, their usage in the human computer interface is still very limited. Mainly this is because of the difficulties encountered when imple menting a diagrammatic interface. Despite the fact that a number of editor generators or compilers are available, only a handful of research prototypes sup port the high-level specification of syntax and inter pretation of diagrams. These issues in the context of incremental interactive systems are addressed by the RecopJa editor generator that we are demonstrating. Recopla is a Java-based editor generator used to construct interactive graphical front-ends (so-called Recopla instances) that are specialized for particular types of diagrams, like e.g. an editor for electronic cir cuit diagrams. Recopla lets the user specify appear ance, syntax and semantics of a diagram language and compiles this into a Java implementati...
This paper discusses picture logic, a visual language for the specification of diagrams and diagr... more This paper discusses picture logic, a visual language for the specification of diagrams and diagram transformations. Formal specification techniques for diagrammatic or visual languages have previously mainly been targeted towards static diagrammatic languages. For reasoning about certain types of diagrams, however, formalizing a notion of change is inevitable. This is particularly true of visual mathematical notations whose evaluation rules or consequence relations correspond to visual or graphical transformations. The paper presents constraint-based extensions of picture logic which render it suitable for the specification of such diagram notations and the required transformations.
Deep Learning has celebrated resounding successes in many application areas of relevance to the I... more Deep Learning has celebrated resounding successes in many application areas of relevance to the Internet-of-Things, for example, computer vision and machine listening. To fully harness the power of deep leaning for the IoT, these technologies must ultimately be brought directly to the edge. The obvious challenge is that deep learning techniques can only be implemented on strictly resource-constrained edge devices if the models are radically downsized. This task relies on different model compression techniques, such as network pruning, quantization, and the recent advancement of XNOR-Net. This paper examines the suitability of these techniques for audio classification on microcontrollers. We present an XNOR-Net for end-to-end raw audio classification and a comprehensive empirical study comparing this approach with pruning-and-quantization methods. We show that raw audio classification with XNOR yields comparable performance to regular full precision networks for small numbers of clas...
Significant efforts are being invested to bring the classification and recognition powers of desk... more Significant efforts are being invested to bring the classification and recognition powers of desktop and cloud systemsdirectly to edge devices. The main challenge for deep learning on the edge is to handle extreme resource constraints(memory, CPU speed and lack of GPU support). We present an edge solution for audio classification that achieves close to state-of-the-art performance on ESC-50, the same benchmark used to assess large, non resource-constrained networks. Importantly, we do not specifically engineer thenetwork for edge devices. Rather, we present a universalpipeline that converts a large deep convolutional neuralnetwork (CNN) automatically via compression and quantization into a network suitable for resource-impoverishededge devices. We first introduce a new sound classification architecture, ACDNet, that produces above state-of-the-art accuracy on both ESC-10 and ESC-50 which are 96.75% and 87.05% respectively. We then compress ACDNet using a novel network-independent ap...
Proceedings of the 7th international conference …, 2002
Sorting and clustering methods inspired by the behavior of real ants are among the earliest metho... more Sorting and clustering methods inspired by the behavior of real ants are among the earliest methods in ant-based meta-heuristics. We revisit these methods in the context of a concrete application and introduce some modifications that yield significant improvements in terms of both ...
Social insect colonies are capable of allocating their workforce in a decentralised fashion; addr... more Social insect colonies are capable of allocating their workforce in a decentralised fashion; addressing a variety of tasks and responding effectively to changes in the environment. This process is fundamental to their ecological success, but the mechanisms behind it remain poorly understood. While most models focus on internal and individual factors, empirical evidence highlights the importance of ecology and social interactions. To address this gap we propose a game theoretical model of task allocation. Individuals are characterised by a trait that determines how they split their energy between two prototypical tasks: foraging and regulation. To be viable, a colony needs to learn to adequately allocate its workforce between these two tasks. We study two different processes: individuals can learn relying exclusively on their own experience, or by using the experiences of others via social learning. We find that social organisation can be determined by the ecology alone, irrespective...
Uploads
Collective Behavior by Bernd Meyer
Leaf harvesting followed a typical diel pattern of a 10–12 h foraging bout followed by a period of inactivity, while fruit harvesting occurred continuously, but with a regular pre‐dawn dip in activity that marked a 24 h cycle.
Although the results of the present study are drawn from a single field colony, the difference found is consistent with a mechanism of negative feedback regulation acting in parallel on two resources that differ in their rates of distribution and processing, creating cycles of formation and depletion of material caches.
This hypothesis should provoke further interest from students of ant behaviour and some simple manipulative experiments that would begin to test it are outlined. Any role of resource caches in regulating foraging by Atta colonies may have similarities to the logistics of warehouse inventories in human economic activity.
We revisit response threshold models with new experiments using temperature regulation in bumblebee colonies as the model system. We show that standard response threshold models do not fit our experiments and present an alternative behavioural model. This captures a fine-grained, time resolved picture of task engagement, which enables us to investigate task allocation with a different set of statistical methods. Using these we show that our model fits the experiment well and explains its salient aspects.
We compare the effectiveness of our model behaviour with that of response threshold models and demonstrate that it can lead to more efficient task management when demands fluctuate. Our results have the potential to provide a basis for the design of more efficient task allocation algorithms for dynamic environments and to elucidate important biological questions, such as the functional role of inter-individual variation.
The common macroscopic description of these systems, based on a Fokker-Planck equation (FPE), suffers from major limitations. Most importantly, it fails at low particle densities and it is impossible to incorporate individual-level experimental observations.
A microscopic Langevin-type individual-based description can – in principle – address these issues but is challenging and computationally expensive to the point that hardware limitations severely restrict their applicability to models of realistic size.
We present a graphics-processor accelerated stochastic simulation solver that obtains performance gains of up to two orders of magnitude even on workstations. We provide a versatile web-interface allowing researcher to perform complex experiments and parameter studies on a dedicated GPU cluster.
Leaf harvesting followed a typical diel pattern of a 10–12 h foraging bout followed by a period of inactivity, while fruit harvesting occurred continuously, but with a regular pre‐dawn dip in activity that marked a 24 h cycle.
Although the results of the present study are drawn from a single field colony, the difference found is consistent with a mechanism of negative feedback regulation acting in parallel on two resources that differ in their rates of distribution and processing, creating cycles of formation and depletion of material caches.
This hypothesis should provoke further interest from students of ant behaviour and some simple manipulative experiments that would begin to test it are outlined. Any role of resource caches in regulating foraging by Atta colonies may have similarities to the logistics of warehouse inventories in human economic activity.
We revisit response threshold models with new experiments using temperature regulation in bumblebee colonies as the model system. We show that standard response threshold models do not fit our experiments and present an alternative behavioural model. This captures a fine-grained, time resolved picture of task engagement, which enables us to investigate task allocation with a different set of statistical methods. Using these we show that our model fits the experiment well and explains its salient aspects.
We compare the effectiveness of our model behaviour with that of response threshold models and demonstrate that it can lead to more efficient task management when demands fluctuate. Our results have the potential to provide a basis for the design of more efficient task allocation algorithms for dynamic environments and to elucidate important biological questions, such as the functional role of inter-individual variation.
The common macroscopic description of these systems, based on a Fokker-Planck equation (FPE), suffers from major limitations. Most importantly, it fails at low particle densities and it is impossible to incorporate individual-level experimental observations.
A microscopic Langevin-type individual-based description can – in principle – address these issues but is challenging and computationally expensive to the point that hardware limitations severely restrict their applicability to models of realistic size.
We present a graphics-processor accelerated stochastic simulation solver that obtains performance gains of up to two orders of magnitude even on workstations. We provide a versatile web-interface allowing researcher to perform complex experiments and parameter studies on a dedicated GPU cluster.