No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Empowered neural cellular automata
... Empowerment-a domain independent, information-theoretic metrichas previously been shown to assist in the evolutionary search for neural cellular automata (NCA) capable of homeostasis when employed as a fitness function [11,17]. In our previous study, we successfully extended empowerment, defined as maximum timelagged mutual information between agents' actions and future sensations, to a distributed sensorimotor system embodied as an NCA. ...
... A brief overview of the NCA and experimental framework are provided. For more detail please refer to [11]. ...
Empowerment -- a domain independent, information-theoretic metric -- has previously been shown to assist in the evolutionary search for neural cellular automata (NCA) capable of homeostasis when employed as a fitness function. In our previous study, we successfully extended empowerment, defined as maximum time-lagged mutual information between agents' actions and future sensations, to a distributed sensorimotor system embodied as an NCA. However, the time-delay between actions and their corresponding sensations was arbitrarily chosen. Here, we expand upon previous work by exploring how the time scale at which empowerment operates impacts its efficacy as an auxiliary objective to accelerate the discovery of homeostatic NCAs. We show that shorter time delays result in marked improvements over empowerment with longer delays, when compared to evolutionary selection only for homeostasis. Moreover, we evaluate stability and adaptability of evolved NCAs, both hallmarks of living systems that are of interest to replicate in artificial ones. We find that short-term empowered NCA are more stable and are capable of generalizing better to unseen homeostatic challenges. Taken together, these findings motivate the use of empowerment during the evolution of other artifacts, and suggest how it should be incorporated to accelerate evolution of desired behaviors for them. Source code for the experiments in this paper can be found at: https://github.com/caitlingrasso/empowered-nca-II.
... In summary, an empowerment-driven agent takes actions that maximise its ability to influence the external world in ways that are perceivable by its own sensors. In multi-agent settings, it has been shown that empowerment-maximisation for individual agents leads to spontaneous coordination among the collective [40,70,71]. This coordination arises because shared information enhances an individual's empowerment, or informally, its ability to make an influence. ...
Collective behaviours are frequently observed to self-organise to criticality. Existing proposals to explain these phenomena, such as Self-organised Criticality (SOC), are fragmented across disciplines and only partially answer the question. This paper investigates the underlying, intrinsic, utilities that may explain self-organisation of collective behaviours near criticality. We focus on information-driven approaches such as predictive information, empowerment, and active inference, as well as thermodynamic efficiency, which incorporates both information-theoretic and thermodynamic quantities. By interpreting the Ising model as a perception-action loop, we compare how different intrinsic utilities shape collective behaviour and analyse the distinct characteristics that arise when each is optimised. In particular, we highlight that at the critical regime thermodynamic efficiency balances the predictability gained by the system and its energy costs. Finally, we propose the Principle of Super-efficiency, suggesting that collective behaviours self-organise to the critical regime where optimal efficiency is achieved with respect to the entropy reduction relative to the thermodynamic costs.
Morphological regeneration is an important feature that highlights the environmental adaptive capacity of biological systems. Lack of this regenerative capacity significantly limits the resilience of machines and the environments they can operate in. To aid in addressing this gap, we develop an approach for simulated soft robots to regrow parts of their morphology when being damaged. Although numerical simulations using soft robots have played an important role in their design, evolving soft robots with regenerative capabilities have so far received comparable little attention. Here we propose a model for soft robots that regenerate through a neural cellular automata. Importantly, this approach only relies on local cell information to regrow damaged components, opening interesting possibilities for physical regenerable soft robots in the future. Our approach allows simulated soft robots that are damaged to partially regenerate their original morphology through local cell interactions alone and regain some of their ability to locomote. These results take a step towards equipping artificial systems with regenerative capacities and could potentially allow for more robust operations in a variety of situations and environments. The code for the experiments in this paper is available at: http://github.com/KazuyaHoribe/RegeneratingSoftRobots.
A method for evolving p rograms that construct multicellular structures (organisms) is described. The paper concentrates on the difficult problem of evolv- ing a cell program that constructs a fixed size French flag. We obtain and analyze an organism that shows a remarkable a bility to repair itself when subjected to severe damage. Its behaviour resembles the regenerative power of some living organisms.
Measures of complexity are of immediate interest for the field of autonomous robots both as a means to classify the behav- ior and as an objective function for the autonomous devel- opment of robot behavior. In the present paper we consider predictive information in sensor space as a measure for the behavioral complexity of a chain of two-wheel robots which are passively coupled and controlled by a closed-loop reac- tive controller for each of the individual robots. The predic- tive information, the mutual information between the past and the future of a time series, is approximated by restricting the time horizons to a single time step. This is exact for Marko- vian systems but seems to work well also for our robotic sys- tem which is strongly non-Markovian.When in a maze with many obstacles, the approximated predictive information of the sensor values of an individual robot is found to have a clear maximum for a controller which realizes the sponta- neous cooperation of the robots in the chain so that large areas of the maze can be visited.
During embryogenesis, multicellular animals are shaped via cell proliferation, cell rearrangement, and apoptosis. At the end of development, tissue architecture is then maintained through balanced rates of cell proliferation and loss. Here, we take an in silico approach to look for generic systems features of morphogenesis in multicellular animals that arise as a consequence of the evolution of development. Using artificial evolution, we evolved cellular automata-based digital organisms that have distinct embryonic and homeostatic phases of development. Although these evolved organisms use a variety of strategies to maintain their form over time, organisms of different types were all found to rapidly recover from environmental damage in the form of wounds. This regenerative response was most robust in an organism with a stratified tissue-like architecture. An evolutionary analysis revealed that evolution itself contributed to the ability of this organism to maintain its form in the face of genetic and environmental perturbation, confirming the results of previous studies. In addition, the exceptional robustness of this organism to surface injury was found to result from an upward flux of cells, driven in part by cell divisions with a stable niche at the tissue base. Given the general nature of the model, our results lead us to suggest that many of the robust systems properties observed in real organisms, including scar-free wound-healing in well-protected embryos and the layered tissue architecture of regenerating epithelial tissues, may be by-products of the evolution of morphogenesis, rather than the direct result of selection.
The classical approach to using utility functions suffers from the drawback of having to design and tweak the functions on a case by case basis. Inspired by examples from the animal kingdom, social sciences and games we propose empowerment, a rather universal function, defined as the information-theoretic capacity of an agent's actuation channel. The concept applies to any sensorimotor apparatus. Empowerment as a measure reflects the properties of the apparatus as long as they are observable due to the coupling of sensors and actuators via the environment. Using two simple experiments we also demonstrate how empowerment influences sensor-actuator evolution
We are interested in the synthesis of autonomous agents using evolutionary techniques. Most work in this area utilizes a direct mapping from genotypic space to phenotypic space. In order to address some of the limitations of this approach, we present a simplified yet biologically defensible model of the developmental process. The design issues that arise when formulating this model at the molecular, cellular and organismal level are discussed, and for each of these issues we describe how they were resolved in our implementation. We present and analyze some of the morphologies that can be explored using this model, specifically one that has agent-like properties. In addition, we demonstrate that this developmental model can be evolved. 1. Introduction Our long term goal is the co-evolution of bodies and control systems for complete autonomous agents. The design of autonomous agents is a complex task that typically involves a great deal of time and effort when done by hand...
In this report we introduce an artificial evolutionary system, Artificial Ontogeny (AO), that uses a developmental encoding scheme to translate a given genotype into a complete agent, which then acts in a physically-realistic virtual environment. Evolution is accomplished using a genetic algorithm, in which the genotypes are treated as genetic regulatory networks. The dynamics of the regulatory network direct the growth of the agent, and lead to the construction of both the morphology and neural control of the agent. We demonstrate that such a model can be used to evolve agents to perform non-trivial tasks, such as directed locomotion and block pushing in a noisy environment. It is shown that mutations expressed earlier in development tend to have a more variable morphological and behavioural effect than mutations expressed later in development, which tend to have a less pronounced effect. These results support the hypothesis that ontogeny provides artificial evolution with beneficial mutat...
Intelligent creatures can explore their environments and learn useful skills without supervision. In this paper, we propose DIAYN ("Diversity is All You Need"), a method for learning useful skills without a reward function. Our proposed method learns skills by maximizing an information theoretic objective using a maximum entropy policy. On a variety of simulated robotic tasks, we show that this simple objective results in the unsupervised emergence of diverse skills, such as walking and jumping. In a number of reinforcement learning benchmark environments, our method is able to learn a skill that solves the benchmark task despite never receiving the true task reward. In these environments, some of the learned skills correspond to solving the task, and each skill that solves the task does so in a distinct manner. Our results suggest that unsupervised discovery of skills can serve as an effective pretraining mechanism for overcoming challenges of exploration and data efficiency in reinforcement learning
What determines large-scale anatomy? DNA does not directly specify geometrical arrangements of tissues and organs, and a process of encoding and decoding for morphogenesis is required. Moreover, many species can regenerate and remodel their structure despite drastic injury. The ability to obtain the correct target morphology from a diversity of initial conditions reveals that the morphogenetic code implements a rich system of pattern-homeostatic processes. Here, we describe an important mechanism by which cellular networks implement pattern regulation and plasticity: bioelectricity. All cells, not only nerves and muscles, produce and sense electrical signals; in vivo, these processes form bioelectric circuits that harness individual cell behaviors toward specific anatomical endpoints. We review emerging progress in reading and re-writing anatomical information encoded in bioelectrical states, and discuss the approaches to this problem from the perspectives of information theory, dynamical systems, and computational neuroscience. Cracking the bioelectric code will enable much-improved control over biological patterning, advancing basic evolutionary developmental biology as well as enabling numerous applications in regenerative medicine and synthetic bioengineering.
Cellular Automata (CA) are a remarkable example of morphogenetic system, where cells grow and self-organise through local interactions. CA have been used as abstractions of biological development and artificial life. Such systems have been able to show properties that are often desirable but difficult to achieve in engineered systems, e.g. morphogenesis and replication of regular patterns without any form of centralized coordination. However, cellular systems are hard to program (i.e. evolve) and control, especially when the number of cell states and neighbourhood increase. In this paper, we propose a new principle of morphogenesis based on Compositional Pattern Producing Networks (CPPNs), an abstraction of development that has been able to produce complex structural motifs without local interactions. CPPNs are used as Cellular Automata genotypes and evolved with a NeuroEvolution of Augmenting Topologies (NEAT) algorithm. This allows complexification of genomes throughout evolution with phenotypes emerging from self-organisation through development based on local interactions. In this paper, the problems of 2D pattern morphogenesis and replication are investigated. Results show that CA-NEAT is an appropriate means of approaching cellular systems engineering, especially for future applications where natural levels of complexity are targeted. We argue that CA-NEAT could provide a valuable mapping for morphogenetic systems, beyond cellular automata systems, where development through local interactions is desired.
A major goal of regenerative medicine and bioengineering is the regeneration of complex organs, such as limbs, and the capability to create artificial constructs (so-called biobots) with defined morphologies and robust self-repair capabilities. Developmental biology presents remarkable examples of systems that self-assemble and regenerate complex structures toward their correct shape despite significant perturbations. A fundamental challenge is to translate progress in molecular genetics into control of large-scale organismal anatomy, and the field is still searching for an appropriate theoretical paradigm for facilitating control of pattern homeostasis. However, computational neuroscience provides many examples in which cell networks - brains - store memories (e.g., of geometric configurations, rules, and patterns) and coordinate their activity towards proximal and distant goals. In this Perspective, we propose that programming large-scale morphogenesis requires exploiting the information processing by which cellular structures work toward specific shapes. In non-neural cells, as in the brain, bioelectric signaling implements information processing, decision-making, and memory in regulating pattern and its remodeling. Thus, approaches used in computational neuroscience to understand goal-seeking neural systems offer a toolbox of techniques to model and control regenerative pattern formation. Here, we review recent data on developmental bioelectricity as a regulator of patterning, and propose that target morphology could be encoded within tissues as a kind of memory, using the same molecular mechanisms and algorithms so successfully exploited by the brain. We highlight the next steps of an unconventional research program, which may allow top-down control of growth and form for numerous applications in regenerative medicine and synthetic bioengineering.
Taking a biologically inspired approach to the design of autonomous, adaptive machines.
We propose a multi-objective method for avoiding premature convergence in evolutionary algorithms, and demonstrate a three-fold performance improvement over comparable methods. Previous research has shown that partitioning an evolving population into age groups can greatly improve the ability to identify global optima and avoid converging to local optima. Here, we propose that treating age as an explicit optimization criterion can increase performance even further, with fewer algorithm implementation parameters. The proposed method evolves a population on the two-dimensional Pareto front comprising (a) how long the genotype has been in the population (age); and (b) its performance (fitness). We compare this approach with previous approaches on the Symbolic Regression problem, sweeping the problem difficulty over a range of solution complexities and number of variables. Our results indicate that the multi-objective approach identifies the exact target solution more often that the age-layered population and standard population methods. The multi-objective method also performs better on higher complexity problems and higher dimensional datasets -- finding global optima with less computational effort.
Natural DNA can encode complexity on an enormous scale. Researchers are attempting to achieve the same representational efficiency in computers by implementing developmental encodings, i.e. encodings that map the genotype to the phenotype through a process of growth from a small starting point to a mature form. A major challenge in in this effort is to find the right level of abstraction of biological development to capture its essential properties without introducing unnecessary inefficiencies. In this paper, a novel abstraction of natural development, called Compositional Pattern Producing Networks (CPPNs), is proposed. Unlike currently accepted abstractions such as iterative rewrite systems and cellular growth simulations, CPPNs map to the phenotype without local interaction, that is, each individual component of the phenotype is determined independently of every other component. Results produced with CPPNs through interactive evolution of two-dimensional images show that such an encoding can nevertheless produce structural motifs often attributed to more conventional developmental abstractions, suggesting that local interaction may not be essential to the desirable properties of natural encoding in the way that is usually assumed.
Most simulations of biological evolution depend on a rather restricted set of properties. In this paper a richer model, based on differential gene expression is introduced to control developmental processes in an artificial evolutionary system. Differential gene expression is used to get different cell types and to modulate cell division and cell death. One of the advantages using developmental processes in evolutionary systems is the reduction of the information needed in the genome to encode e.g. shapes or cell types which results in better scaling behavior of the system. My result showed that the shaping of multicellular organisms in 3d is possible with the proposed system. 1 Introduction In the field of artificial evolution, current research tries to imitate biological concepts of evolution and development to simulate or build artificial organisms. This paper reports on a biologically inspired model that has been used to evolve 3d shapes of simulated, multicellular organisms. The...