Hervé Luga | Institut de Recherche en Informatique de Toulouse (original) (raw)
Papers by Hervé Luga
arXiv (Cornell University), Jul 16, 2018
Over the past twenty years, artificial Gene Regulatory Networks (GRNs) have shown their capacity ... more Over the past twenty years, artificial Gene Regulatory Networks (GRNs) have shown their capacity to solve real-world problems in various domains such as agent control, signal processing and artificial life experiments. They have also benefited from new evolutionary approaches and improvements to dynamic which have increased their optimization efficiency. In this paper, we present an additional step toward their usability in machine learning applications. We detail an GPUbased implementation of differentiable GRNs, allowing for local optimization of GRN architectures with stochastic gradient descent (SGD). Using a standard machine learning dataset, we evaluate the ways in which evolution and SGD can be combined to further GRN optimization. We compare these approaches with neural network models trained by SGD and with support vector machines.
Differentiable Architecture Search (DARTS) is a recent neural architecture search (NAS) method ba... more Differentiable Architecture Search (DARTS) is a recent neural architecture search (NAS) method based on a differentiable relaxation. Due to its success, numerous variants analyzing and improving parts of the DARTS framework have recently been proposed. By considering the problem as a constrained bilevel optimization, we present and analyze DARTS-PRIME, a variant including improvements to architectural weight update scheduling and regularization towards discretization. We propose a dynamic schedule based on per-minibatch network information to make architecture updates more informed, as well as proximity regularization to promote well-separated discretization. Our results in multiple domains show that DARTSPRIME improves both performance and reliability, comparable to state-of-the-art in differentiable NAS. 1
Journal of Digital Imaging
Proceedings of the 11th ACM International Conference on Distributed and Event-based Systems, 2017
Recent advances in Web 3D technology have opened a wide area for Collaborative Virtual Environmen... more Recent advances in Web 3D technology have opened a wide area for Collaborative Virtual Environments (CVE). While CVE are often viewed in a concurrency context, they need to provide a satisfying experience in terms of consistency, latency and recovery. Because (i) Event-Driven architectures (EDA) are well-suited for distributed application and (ii) traditional communication architecture (clientserver) can be limited in such situations, this paper presents a loosely-coupled approach combining event sourcing with a hybrid communication architecture. This model aims to ensure a strong versioning system and resource availability for collaborative 3D object manipulation in a web browser. To evaluate acceptance of our system, we conducted a user study on groups of users working simultaneously on 3D cooperative assembly tasks. The results detail the users' involvement evolution, qualitative appreciations of the system's usability and the collaborative features. CCS CONCEPTS •Human-centered computing →Collaborative content creation; •Software and its engineering →Peer-to-peer architectures; •Networks →World Wide Web (network structure);
ABSTRACT In this chapter a virtual ecosystem environment with basic physical law and energy conce... more ABSTRACT In this chapter a virtual ecosystem environment with basic physical law and energy concept has been proposed, this ecosystem is populated with 3D virtual creatures that are living in this environment in order to forage food. Artificial behaviors are developed in order to control artificial creatures. Initially, we study the behavior of herbivore’s creatures, which feed resources available in their environment. A genetic algorithm with an artificial neural network were implemented together to guarantee some of these behaviors like searching food. Foods are presented in different locations in the virtual ecosystem. The evolutionary process uses the physical properties of the virtual creatures and an external fitness function with several objectives that will conduct to the expected behaviors. The experiment evolving locomoting virtual creatures shows that these virtual creatures try to obtain at least one of the food sources presented in their trajectories. Our best-evolved creatures are able to reach multiple food sources during the simulation time.
ArXiv, 2018
In the brain, learning signals change over time and synaptic location, and are applied based on t... more In the brain, learning signals change over time and synaptic location, and are applied based on the learning history at the synapse, in the complex process of neuromodulation. Learning in artificial neural networks, on the other hand, is shaped by hyper-parameters set before learning starts, which remain static throughout learning, and which are uniform for the entire network. In this work, we propose a method of deep artificial neuromodulation which applies the concepts of biological neuromodulation to stochastic gradient descent. Evolved neuromodulatory dynamics modify learning parameters at each layer in a deep neural network over the course of the network's training. We show that the same neuromodulatory dynamics can be applied to different models and can scale to new problems not encountered during evolution. Finally, we examine the evolved neuromodulation, showing that evolution found dynamic, location-specific learning strategies.
Lecture Notes in Computer Science, 2004
In this article we present our chess engine Tempo. One of the major difficulties for this type of... more In this article we present our chess engine Tempo. One of the major difficulties for this type of program lies in the function for evaluating game positions. This function is composed of a large number of parameters which have to be determined and then adjusted. We propose an alternative which consists in replacing this function by an artificial neuron network (ANN). Without topological knowledge of this complex network, we use the evolutionist methods for its inception, thus enabling us to obtain, among other things, a modular network. Finally, ...
Proceedings of the 36th Annual ACM Symposium on Applied Computing, 2021
The class distribution of data is one of the factors that regulates the performance of machine le... more The class distribution of data is one of the factors that regulates the performance of machine learning models. However, investigations on the impact of different distributions available in the literature are very few, sometimes absent for domain-specific tasks. In this paper, we analyze the impact of natural and balanced distributions of the training set in deep learning (DL) models applied on histological images, also known as whole slide images (WSIs). WSIs are considered as the gold standard for cancer diagnosis. In recent years, researchers have turned their attention to DL models to automate and accelerate the diagnosis process. In the training of such DL models, filtering out the non-regions-of-interest from the WSIs and adopting an artificial distribution-usually a balanced distribution-is a common trend. In our analysis, we show that keeping the WSIs data in their usual distribution-which we call natural distribution-for DL training produces fewer false positives (FPs) with comparable false negatives (FNs) than the artificially-obtained balanced distribution. We conduct an empirical comparative study with 10 random folds for each distribution, comparing the resulting average performance levels in terms of five different evaluation metrics. Experimental results show the effectiveness of the natural distribution over the balanced one across all the evaluation metrics. CCS CONCEPTS • Computing methodologies → Supervised learning; Image processing; Image segmentation; • Applied computing → Health informatics.
Genome Biology and Evolution, 2021
The large spectrum of hearing sensitivity observed in primates results from the impact of environ... more The large spectrum of hearing sensitivity observed in primates results from the impact of environmental and behavioral pressures to optimize sound perception and localization. Although evidence of positive selection in auditory genes has been detected in mammals including in Hominoids, selection has never been investigated in other primates. We analyzed 123 genes highly expressed in the inner ear of 27 primate species and tested to what extent positive selection may have shaped these genes in the order Primates tree. We combined both site and branch-site tests to obtain a comprehensive picture of the positively selected genes (PSGs) involved in hearing sensitivity, and drew a detailed description of the most affected branches in the tree. We chose a conservative approach, and thus focused on confounding factors potentially affecting PSG signals (alignment, GC-biased gene conversion, duplications, heterogeneous sequencing qualities). Using site tests, we showed that around 12% of the...
In order to help the user to accomplish a task, teleoperation systems have to integrate different... more In order to help the user to accomplish a task, teleoperation systems have to integrate different tools such as visualization, divers interaction devices, planning tools, etc.... The interface must be able to give complete information of the real world and the user can, using a distributed platform, be helped by others users as well as by autonomous robots. The objective of our project is to combine teleoperation, virtual reality and adaptive systems to improve the control of teleoperation missions. Keywords: teleoperation, adaptive systems, virtual reality, cooperative work. 1.
La simulation en synthese d'images a d'abord utilise des modeles cognitifs ou proceduraux... more La simulation en synthese d'images a d'abord utilise des modeles cognitifs ou proceduraux. Elle se tourne maintenant vers la recherche de comportements realistes passant par une plus grande independance des acteurs. Notre contribution se situe dans le domaine de l'etude de phenomenes emergents et prolonge nos travaux sur le simulateur invitram. Elle porte sur la generation automatique de personnages dans des univers de simulation par le biais de deux vecteurs : la generation automatique de formes et celle de comportements. Nous utilisons pour cela des paradigmes fournis par la vie artificielle et notamment les systemes evolutionnistes. Ces techniques vont permettre de creer des acteurs evoluant a des niveaux de complexite elevee et notamment des systemes chaotiques ou a la frontiere du chaos. Nous montrons donc dans un premier temps des applications statiques de nos systemes pour la generation de formes. L'objectif est ici la definition d'objets par generation au...
We evolve a Genetic Regulatory Network (GRN) in a three dimensional morphogen gradient environmen... more We evolve a Genetic Regulatory Network (GRN) in a three dimensional morphogen gradient environment to determine the topology of the neurons in a Spiking Neural Network (SNN). A genetic algorithm is used to optimize the GRN, selecting individuals based on the performance of the SNN grown by the GRN. Performance is measured on two tasks: visual discrimination and robotic foraging. Early results show potential for this method as both an indirect encoding and on-line regulator of neural networks.
This paper presents a reconstruction of the News Divine Discoteque, one of the most well-known an... more This paper presents a reconstruction of the News Divine Discoteque, one of the most well-known and welldocumented evacuation catastrophes in Mexico in recent years. A simulated evacuation is compared to the official reports using mean squared error and an error percentage.
Our proposed research project is to enable 3D distributed visualization and manipulation involvin... more Our proposed research project is to enable 3D distributed visualization and manipulation involving collaborative effort through the use of web-based technologies. Our project resulted from a wide collaborative application research fields: Computer Aided Design (CAD), Building Information Modeling (BIM) or Product Life Cycle Management (PLM) where design tasks are often performed in teams and need a fluent communication system. The system allows distributed remote assembling in 3D scenes with real-time updates for the users. This paper covers this feature using hybrid networking solution: a client-server architecture (REST) for 3D rendering (WebGL) and data persistence (NoSQL) associated to an automatically built peer-to-peer (P2P) mesh for real-time communication between the clients (WebRTC). The approach is demonstrated through the development of a web-platform prototype focusing on the easy manipulation, fine rendering and light update messages for all participating users. We prov...
Following Karl Sims seminal works, many approaches from the literature aims at generating artific... more Following Karl Sims seminal works, many approaches from the literature aims at generating artificial creatures using body-brain co-evolution. However, in simulation, creatures are not very realistic, they cannot be tested in physical robots. In this paper, we propose a system that can generate realistic walking artificial creatures. We co-evolve the morphology and the controller of virtual modular robots using GA. The morphology is generated by Graphtals while the global behavior of a creature is done by the cooperation of robot's modules moves. Each module has its own local controller, here based on an ANN. We integrate our system in Gazebo, a popular realistic robotic simulator. Experimental results show the capacity of our approach to generate realistic morphologies and behaviors with simulator parameters set up with realistic values. We expect the virtual robots generated with our system and trained in a realistic robotic simulator to better bridge the gap to reality.
In this paper we propose a new computational model of cell cycle to study the dynamics of cells p... more In this paper we propose a new computational model of cell cycle to study the dynamics of cells population in 2D monolayer culture. Whereas most of the models are phase-orientated our model deals with a checkpoint orientated paradigm and uses the phase orientation as an output to provide the biologists with a relevant view of the simulation result. Through this paper we will present the genericity of our model, able to reproduce the exponential growth phase of different cellular processes.
Cartesian Genetic Programming (CGP) has many modifications across a variety of implementations, s... more Cartesian Genetic Programming (CGP) has many modifications across a variety of implementations, such as recursive connections and node weights. Alternative genetic operators have also been proposed for CGP, but have not been fully studied. In this work, we present a new form of genetic programming based on a floating point representation. In this new form of CGP, called Positional CGP, node positions are evolved. This allows for the evaluation of many different genetic operators while allowing for previous CGP improvements like recurrency. Using nine benchmark problems from three different classes, we evaluate the optimal parameters for CGP and PCGP, including novel genetic operators.
Artificial embryogeny aims at developing a complete organism starting from a unique cell. Nowaday... more Artificial embryogeny aims at developing a complete organism starting from a unique cell. Nowadays many algorithms exist to synthesize artificial creature shapes or behaviours. With the purpose of shape and high-level behaviour joint evolution, one of the key aspects is the synthesis of positional information. Such pieces of information, called morphogens, are in many developmental models embedded in the environment and interactions are made through simple protein receptors. In this paper, we propose a new and original approach to solve the morphogen-positioning problem. We use a hydrodynamic model to replace the classical spreading algorithm. Mechanical constraints (the cell shape) and a dynamic activity are integrated. Thanks to this improvement, the cell behaviour can affect the spreading algorithm: cells can apply forces on the hydrodynamic environment to create substrate flows. Through experiments, this paper shows the way to develop complex shapes using this kind of simulator ...
Resource allocation and deployment is a ubiquitous logistics optimization problem, e.g. in delive... more Resource allocation and deployment is a ubiquitous logistics optimization problem, e.g. in delivery services, industrial production, and business management. We describe a bioinspired optimization system for allocating resources, based on Gene Regulatory Networks (GRNs)(see (Disset et al.(2017)) for details). Specifically, our system allocates intervention teams to be deployed during catastrophic cli- matic events to sites with failures in an electrical distribution grid covering 17000 km 2 . Our work is concerned with the minimization of the breakdown time for the users of the power grid. Here, we describe our system, including the allocation problem, the choice of GRN as control system and its optimization with an Evolutionary Algorithm (EA), and a test scenario of the deployment of teams to repair failures during a heavy storm.
arXiv (Cornell University), Jul 16, 2018
Over the past twenty years, artificial Gene Regulatory Networks (GRNs) have shown their capacity ... more Over the past twenty years, artificial Gene Regulatory Networks (GRNs) have shown their capacity to solve real-world problems in various domains such as agent control, signal processing and artificial life experiments. They have also benefited from new evolutionary approaches and improvements to dynamic which have increased their optimization efficiency. In this paper, we present an additional step toward their usability in machine learning applications. We detail an GPUbased implementation of differentiable GRNs, allowing for local optimization of GRN architectures with stochastic gradient descent (SGD). Using a standard machine learning dataset, we evaluate the ways in which evolution and SGD can be combined to further GRN optimization. We compare these approaches with neural network models trained by SGD and with support vector machines.
Differentiable Architecture Search (DARTS) is a recent neural architecture search (NAS) method ba... more Differentiable Architecture Search (DARTS) is a recent neural architecture search (NAS) method based on a differentiable relaxation. Due to its success, numerous variants analyzing and improving parts of the DARTS framework have recently been proposed. By considering the problem as a constrained bilevel optimization, we present and analyze DARTS-PRIME, a variant including improvements to architectural weight update scheduling and regularization towards discretization. We propose a dynamic schedule based on per-minibatch network information to make architecture updates more informed, as well as proximity regularization to promote well-separated discretization. Our results in multiple domains show that DARTSPRIME improves both performance and reliability, comparable to state-of-the-art in differentiable NAS. 1
Journal of Digital Imaging
Proceedings of the 11th ACM International Conference on Distributed and Event-based Systems, 2017
Recent advances in Web 3D technology have opened a wide area for Collaborative Virtual Environmen... more Recent advances in Web 3D technology have opened a wide area for Collaborative Virtual Environments (CVE). While CVE are often viewed in a concurrency context, they need to provide a satisfying experience in terms of consistency, latency and recovery. Because (i) Event-Driven architectures (EDA) are well-suited for distributed application and (ii) traditional communication architecture (clientserver) can be limited in such situations, this paper presents a loosely-coupled approach combining event sourcing with a hybrid communication architecture. This model aims to ensure a strong versioning system and resource availability for collaborative 3D object manipulation in a web browser. To evaluate acceptance of our system, we conducted a user study on groups of users working simultaneously on 3D cooperative assembly tasks. The results detail the users' involvement evolution, qualitative appreciations of the system's usability and the collaborative features. CCS CONCEPTS •Human-centered computing →Collaborative content creation; •Software and its engineering →Peer-to-peer architectures; •Networks →World Wide Web (network structure);
ABSTRACT In this chapter a virtual ecosystem environment with basic physical law and energy conce... more ABSTRACT In this chapter a virtual ecosystem environment with basic physical law and energy concept has been proposed, this ecosystem is populated with 3D virtual creatures that are living in this environment in order to forage food. Artificial behaviors are developed in order to control artificial creatures. Initially, we study the behavior of herbivore’s creatures, which feed resources available in their environment. A genetic algorithm with an artificial neural network were implemented together to guarantee some of these behaviors like searching food. Foods are presented in different locations in the virtual ecosystem. The evolutionary process uses the physical properties of the virtual creatures and an external fitness function with several objectives that will conduct to the expected behaviors. The experiment evolving locomoting virtual creatures shows that these virtual creatures try to obtain at least one of the food sources presented in their trajectories. Our best-evolved creatures are able to reach multiple food sources during the simulation time.
ArXiv, 2018
In the brain, learning signals change over time and synaptic location, and are applied based on t... more In the brain, learning signals change over time and synaptic location, and are applied based on the learning history at the synapse, in the complex process of neuromodulation. Learning in artificial neural networks, on the other hand, is shaped by hyper-parameters set before learning starts, which remain static throughout learning, and which are uniform for the entire network. In this work, we propose a method of deep artificial neuromodulation which applies the concepts of biological neuromodulation to stochastic gradient descent. Evolved neuromodulatory dynamics modify learning parameters at each layer in a deep neural network over the course of the network's training. We show that the same neuromodulatory dynamics can be applied to different models and can scale to new problems not encountered during evolution. Finally, we examine the evolved neuromodulation, showing that evolution found dynamic, location-specific learning strategies.
Lecture Notes in Computer Science, 2004
In this article we present our chess engine Tempo. One of the major difficulties for this type of... more In this article we present our chess engine Tempo. One of the major difficulties for this type of program lies in the function for evaluating game positions. This function is composed of a large number of parameters which have to be determined and then adjusted. We propose an alternative which consists in replacing this function by an artificial neuron network (ANN). Without topological knowledge of this complex network, we use the evolutionist methods for its inception, thus enabling us to obtain, among other things, a modular network. Finally, ...
Proceedings of the 36th Annual ACM Symposium on Applied Computing, 2021
The class distribution of data is one of the factors that regulates the performance of machine le... more The class distribution of data is one of the factors that regulates the performance of machine learning models. However, investigations on the impact of different distributions available in the literature are very few, sometimes absent for domain-specific tasks. In this paper, we analyze the impact of natural and balanced distributions of the training set in deep learning (DL) models applied on histological images, also known as whole slide images (WSIs). WSIs are considered as the gold standard for cancer diagnosis. In recent years, researchers have turned their attention to DL models to automate and accelerate the diagnosis process. In the training of such DL models, filtering out the non-regions-of-interest from the WSIs and adopting an artificial distribution-usually a balanced distribution-is a common trend. In our analysis, we show that keeping the WSIs data in their usual distribution-which we call natural distribution-for DL training produces fewer false positives (FPs) with comparable false negatives (FNs) than the artificially-obtained balanced distribution. We conduct an empirical comparative study with 10 random folds for each distribution, comparing the resulting average performance levels in terms of five different evaluation metrics. Experimental results show the effectiveness of the natural distribution over the balanced one across all the evaluation metrics. CCS CONCEPTS • Computing methodologies → Supervised learning; Image processing; Image segmentation; • Applied computing → Health informatics.
Genome Biology and Evolution, 2021
The large spectrum of hearing sensitivity observed in primates results from the impact of environ... more The large spectrum of hearing sensitivity observed in primates results from the impact of environmental and behavioral pressures to optimize sound perception and localization. Although evidence of positive selection in auditory genes has been detected in mammals including in Hominoids, selection has never been investigated in other primates. We analyzed 123 genes highly expressed in the inner ear of 27 primate species and tested to what extent positive selection may have shaped these genes in the order Primates tree. We combined both site and branch-site tests to obtain a comprehensive picture of the positively selected genes (PSGs) involved in hearing sensitivity, and drew a detailed description of the most affected branches in the tree. We chose a conservative approach, and thus focused on confounding factors potentially affecting PSG signals (alignment, GC-biased gene conversion, duplications, heterogeneous sequencing qualities). Using site tests, we showed that around 12% of the...
In order to help the user to accomplish a task, teleoperation systems have to integrate different... more In order to help the user to accomplish a task, teleoperation systems have to integrate different tools such as visualization, divers interaction devices, planning tools, etc.... The interface must be able to give complete information of the real world and the user can, using a distributed platform, be helped by others users as well as by autonomous robots. The objective of our project is to combine teleoperation, virtual reality and adaptive systems to improve the control of teleoperation missions. Keywords: teleoperation, adaptive systems, virtual reality, cooperative work. 1.
La simulation en synthese d'images a d'abord utilise des modeles cognitifs ou proceduraux... more La simulation en synthese d'images a d'abord utilise des modeles cognitifs ou proceduraux. Elle se tourne maintenant vers la recherche de comportements realistes passant par une plus grande independance des acteurs. Notre contribution se situe dans le domaine de l'etude de phenomenes emergents et prolonge nos travaux sur le simulateur invitram. Elle porte sur la generation automatique de personnages dans des univers de simulation par le biais de deux vecteurs : la generation automatique de formes et celle de comportements. Nous utilisons pour cela des paradigmes fournis par la vie artificielle et notamment les systemes evolutionnistes. Ces techniques vont permettre de creer des acteurs evoluant a des niveaux de complexite elevee et notamment des systemes chaotiques ou a la frontiere du chaos. Nous montrons donc dans un premier temps des applications statiques de nos systemes pour la generation de formes. L'objectif est ici la definition d'objets par generation au...
We evolve a Genetic Regulatory Network (GRN) in a three dimensional morphogen gradient environmen... more We evolve a Genetic Regulatory Network (GRN) in a three dimensional morphogen gradient environment to determine the topology of the neurons in a Spiking Neural Network (SNN). A genetic algorithm is used to optimize the GRN, selecting individuals based on the performance of the SNN grown by the GRN. Performance is measured on two tasks: visual discrimination and robotic foraging. Early results show potential for this method as both an indirect encoding and on-line regulator of neural networks.
This paper presents a reconstruction of the News Divine Discoteque, one of the most well-known an... more This paper presents a reconstruction of the News Divine Discoteque, one of the most well-known and welldocumented evacuation catastrophes in Mexico in recent years. A simulated evacuation is compared to the official reports using mean squared error and an error percentage.
Our proposed research project is to enable 3D distributed visualization and manipulation involvin... more Our proposed research project is to enable 3D distributed visualization and manipulation involving collaborative effort through the use of web-based technologies. Our project resulted from a wide collaborative application research fields: Computer Aided Design (CAD), Building Information Modeling (BIM) or Product Life Cycle Management (PLM) where design tasks are often performed in teams and need a fluent communication system. The system allows distributed remote assembling in 3D scenes with real-time updates for the users. This paper covers this feature using hybrid networking solution: a client-server architecture (REST) for 3D rendering (WebGL) and data persistence (NoSQL) associated to an automatically built peer-to-peer (P2P) mesh for real-time communication between the clients (WebRTC). The approach is demonstrated through the development of a web-platform prototype focusing on the easy manipulation, fine rendering and light update messages for all participating users. We prov...
Following Karl Sims seminal works, many approaches from the literature aims at generating artific... more Following Karl Sims seminal works, many approaches from the literature aims at generating artificial creatures using body-brain co-evolution. However, in simulation, creatures are not very realistic, they cannot be tested in physical robots. In this paper, we propose a system that can generate realistic walking artificial creatures. We co-evolve the morphology and the controller of virtual modular robots using GA. The morphology is generated by Graphtals while the global behavior of a creature is done by the cooperation of robot's modules moves. Each module has its own local controller, here based on an ANN. We integrate our system in Gazebo, a popular realistic robotic simulator. Experimental results show the capacity of our approach to generate realistic morphologies and behaviors with simulator parameters set up with realistic values. We expect the virtual robots generated with our system and trained in a realistic robotic simulator to better bridge the gap to reality.
In this paper we propose a new computational model of cell cycle to study the dynamics of cells p... more In this paper we propose a new computational model of cell cycle to study the dynamics of cells population in 2D monolayer culture. Whereas most of the models are phase-orientated our model deals with a checkpoint orientated paradigm and uses the phase orientation as an output to provide the biologists with a relevant view of the simulation result. Through this paper we will present the genericity of our model, able to reproduce the exponential growth phase of different cellular processes.
Cartesian Genetic Programming (CGP) has many modifications across a variety of implementations, s... more Cartesian Genetic Programming (CGP) has many modifications across a variety of implementations, such as recursive connections and node weights. Alternative genetic operators have also been proposed for CGP, but have not been fully studied. In this work, we present a new form of genetic programming based on a floating point representation. In this new form of CGP, called Positional CGP, node positions are evolved. This allows for the evaluation of many different genetic operators while allowing for previous CGP improvements like recurrency. Using nine benchmark problems from three different classes, we evaluate the optimal parameters for CGP and PCGP, including novel genetic operators.
Artificial embryogeny aims at developing a complete organism starting from a unique cell. Nowaday... more Artificial embryogeny aims at developing a complete organism starting from a unique cell. Nowadays many algorithms exist to synthesize artificial creature shapes or behaviours. With the purpose of shape and high-level behaviour joint evolution, one of the key aspects is the synthesis of positional information. Such pieces of information, called morphogens, are in many developmental models embedded in the environment and interactions are made through simple protein receptors. In this paper, we propose a new and original approach to solve the morphogen-positioning problem. We use a hydrodynamic model to replace the classical spreading algorithm. Mechanical constraints (the cell shape) and a dynamic activity are integrated. Thanks to this improvement, the cell behaviour can affect the spreading algorithm: cells can apply forces on the hydrodynamic environment to create substrate flows. Through experiments, this paper shows the way to develop complex shapes using this kind of simulator ...
Resource allocation and deployment is a ubiquitous logistics optimization problem, e.g. in delive... more Resource allocation and deployment is a ubiquitous logistics optimization problem, e.g. in delivery services, industrial production, and business management. We describe a bioinspired optimization system for allocating resources, based on Gene Regulatory Networks (GRNs)(see (Disset et al.(2017)) for details). Specifically, our system allocates intervention teams to be deployed during catastrophic cli- matic events to sites with failures in an electrical distribution grid covering 17000 km 2 . Our work is concerned with the minimization of the breakdown time for the users of the power grid. Here, we describe our system, including the allocation problem, the choice of GRN as control system and its optimization with an Evolutionary Algorithm (EA), and a test scenario of the deployment of teams to repair failures during a heavy storm.