It is widely thought that sensorimotor synchronization, underpinning cultural domains such as music and dance, played a critical role in the evolution of human sociality. Here, we present virtual legged robots controlled by central pattern generators (CPGs) that evolve to synchronize motion to rhythmic sensory input in real time. Multi-stage, multi-objective evolutionary algorithms were used to maximize flexibility of the CPGs with respect to control parameters, and then to optimize a neural input layer for wide-ranging susceptibility to rhythmic inputs. The evolved CPGs self-organize to accommodate the input sequence over a range of frequencies and patterns while keeping the agents upright. We show how this behaviour can be scaled up to multiple interacting agents, including with differing morphologies, to produce novel behaviours. We then outline how spike timing dependent plasticity can be used for the acquisition of new motor patterns. Finally, taking inspiration from biocultural evolution and cognitive neuroscience, we suggest ways in which real-time social adaptation can play a key role in the evolution of complex social behaviours in robots.

In a high-stakes cooperative survival game, self-interested behaviours reward the individual in the short-term but may have a detrimental impact on the collective in the long-term. Such situations can be solved by introducing social contracts between players that reduce the set of possible actions. In the absence of an empowered authority capable of enforcement, however, a player will only uphold such a contract so long as they believe that the other players will do the same. We term this buy-in. In this context, we envision a cooperative survival game that extends the scope of the ‘conventional’ Mexican standoff (a three-player Hawk-Dove game) to n -players, from which we design and implement a self-organising multiagent system. We devise a set of experiments across varying degrees of initial buy-in and examine its impact on social contracts and the voluntary restriction of self-interest. In particular, we show that there is a cyclical, non-transitive dependency between the three that is both ring-reinforcing and critical for systemic stability.

Insect inspired navigation strategies have the potential to unlock robotic navigation in power-constrained scenarios as they can function effectively with limited computational resources. One such strategy, familiarity-based navigation, has successfully navigated routes of up to 60m using a single layer neural network trained with an Infomax learning rule in online robotic applications. Here we challenge Infomax to navigate longer routes, investigating the relationship between performance, view size, view acquisition rate and network size. By doing so, we determine the parameters at which Infomax operates effectively and explore the profile with which it fails. We show that effective memorisation of familiar views is possible for longer routes than previously attempted, but that this length decreases for reduced input view dimensions. In the selection of an ideal view acquisition rate, we also show that this must be increased with route length for consistent performance. In investigating the applicability to small, lower-power robots, we demonstrate that computational and memory savings may be made with equivalent performance by reducing the network size. Finally, we investigate the profile with which failure occurs, demonstrating increased confusion occurring across the route as it extends in length. These findings are being used to inform theories of insect navigation and improve practical deployment of view based navigation for long routes.

It is often postulated that robots will eventually face conditions, whether on extraterrestrial bodies or deep underwater, that could not have been predicted by their designers. In such conditions, truly autonomous robots should be able to describe and talk about their environments in order to collectively find appropriate solutions. We designed an emergent naming systems for such purposes. This paper focuses on a shortest-path discovery scenario in an unstructured environment, where landmarks are collectively named, by a swarm of robots, as they are discovered. The robots use those landmarks as beacons for navigation and score them according to their relevance to the task at hand. Meanwhile the naming system enables the swarm to update these scores asynchronously, using very little bandwidth. We compare our naming-based navigation performances with swarms that do not communicate and swarms with prior knowledge of the environment, and find that our approach performs similarly to the latter. This has significant implications on the link between space conceptualisation and language, as this proto-language enables the robots to find a topological path without individually mapping the environment.

We propose asynchronous cellular automata fashioned model of true slime mold Physarum polycephalum plasmodium equipped with a dynamic feedback mechanism based on Bayesian and inverse Bayesian inference. These are implemented as feedback from dynamical protoplasmic flow into local tubular structures in slime mold. Because inverse Bayesian inference replaces conditional probabilities with empirical ones and relaxes the probability space, the model can behave robustly and adaptively. We describe a brief overview of our model in this paper.

Various models have been developed to shed light on neuronal mechanisms of homeostatic plasticity (HP). We focus on one such model implemented on continuous-time-recurrent neural networks. Though this HP mechanism encourages oscillatory dynamics by preventing node saturation, it was curiously detrimental to behavioral fitness when compared to non-plastic networks on several tasks (Williams, 2004, 2005). When we set out to explain this result, we discovered a type of oscillation that depends on HP’s continued regulation of circuit parameters. If HP is turned off, oscillation stops. This suggests that HP can play an enabling role in central pattern generation which has not been explored in modelling or experimental contexts. We first situate this phenomenon within the space of possibilities for HP’s involvement in oscillation. Then, we show that these “HP-enabled” oscillations are extraordinarily common in random circuits of various sizes. Finally, we describe how the degree of timescale separation between HP and neural dynamics affects HP-enabled oscillation. This analysis suggests promising avenues for dialogue between modeling and experiment.

Behavior has an understated role in the genesis of complex ecologies. Discussion of ecological regulation describes the phenomenon in terms of coupled feedbacks which have been connected by Harvey (2004) to rein control as introduced by Clynes (1969). These descriptions have motivated the question of how communities that instantiate such feedbacks can evolve in the first place, especially with respect to global regulatory effects such as those supposed in Lovelock and Margulis’ Gaia theory (1974). While Gaian regulation is not incompatible with evolution, it appears there are intermediate steps that are necessary for its establishment, and likely the establishment of coupled ecological regulation at any scale. Here we present a series of dynamical models that show how simple dormancy behavior can help account for that differential survival across a variety of seasonal conditions. Furthermore, the combination of that behavior and a traditional rein control mechanism lead to a significant increase in survivable conditions, providing a hypothesis for how ecological regulation may be scaffolded. Further discussion suggests that effective behavior of pioneer species is a requirement for the establishment of robust ecosystems.

Through the combination of artificial components and living organisms, we can develop a novel methodology for aquatic monitoring. By observing the responses of organisms to changes in their environment, a broad-spectrum sensor was created. One of the organisms broadly used as a biosensor is Daphnia . Its broad distribution and well-studied biology make it a promising element for incorporating into a biohybrid. This Daphnia -based sensor was calibrated against increasing salinity as a preliminary experiment. The swimming behaviour (spinning and movement inhibition) was observed for different salinities. The results showcase significant and observable differences. This and other calibration experiments will be used here as bases for the behavioural results interpretation.

Exploration is a key challenge in Reinforcement Learning, especially in long-horizon, deceptive and sparse-reward environments. For such applications, population-based approaches have proven effective. Methods such as Quality-Diversity deals with this by encouraging novel solutions and producing a diversity of behaviours. However, these methods are driven by either undirected sampling (i.e. mutations) or use approximated gradients (i.e. Evolution Strategies) in the parameter space, which makes them highly sample-inefficient. In this paper, we propose Dynamics-Aware QD-Ext (DA-QD-ext) and Gradient and Dynamics Aware QD (GDA-QD), two model-based Quality-Diversity approaches. They extend existing QD methods to use gradients for efficient exploitation and leverage perturbations in imagination for efficient exploration. Our approach takes advantage of the effectiveness of QD algorithms as good data generators to train deep models and use these models to learn diverse and high-performing populations. We demonstrate that they outperform baseline RL approaches on tasks with deceptive rewards, and maintain the divergent search capabilities of QD approaches while exceeding their performance by ∼ 1.5 times and reaching the same results in 5 times less samples.

Domestication syndrome in cereal grains is commonly thought to be the product of domestication through a combination of direct artificial selection and indirect natural selection by humans. We propose an agent-based model of grain domestication. We simulate cereal grains with four genes that impact their reproductive cycle undergoing harvesting and selective planting by simulated humans. When direct artificial selection is applied to one gene domestication syndrome emerges in the other genes as a result of indirect natural selection. In the absence of direct artificial selection no domestication syndrome emerged, consistent with periods of predomestication cultivation in human history. Domesticated variants are strongest when humans select for traits inconsistent with the wild type traits, and weakest when humans select for traits consistent with the wild type.

Lexicase selection is an effective many-objective evolutionary algorithm across many problem domains. Lexicase can be computationally expensive, especially in areas like evolutionary robotics where individual objectives might require their own physics simulation. Improving the efficiency of Lexicase selection can reduce the total number of evaluations thereby lowering computational overhead. Here, we introduce a fitness agnostic adaptive objective sampling algorithm using the filtering efficacy of objectives to adjust their frequency of occurrence as a selector. In a set of binary genome maximization tasks modeled to emulate evolutionary robotics situations, we show that performance can be maintained while computational efficiency increases as compared to ϵ -Lexicase.

Cellular Automata (CAs) have potential as powerful parallel computational systems, which has lead to the use of CAs as reservoirs in reservoir computing. However, why certain Cellular Automaton (CA) rules, sizes and input encodings are better or worse at a given task is not well understood. We present a method that enables identification and visualization of the specific information content, flow and transformations within the space-time diagram of CA. We interpret each spatio-temporal location in CA’s space-time diagram as a function of its input and call this novel notion the CA’s Canonical Computations (CCs). This allows us to analyze the available information from the space-time diagrams as partitions of the input set. The method also reveals how input-encoder-rule interactions transform the information flow by changing features like spatial and temporal location stability as well as the specific information produced. This general approach for analysing CA is discussed for the engineering of reservoir computing systems.

Camouflage in nature seems to arise from competition between predator and prey. To survive, predators must find prey, and prey must avoid being found. This work simulates an abstract model of that adversarial relationship. It looks at crypsis through evolving prey camouflage patterns (as color textures) in competition with evolving predator vision. During their “lifetime” predators learn to better locate camouflaged prey. The environment for this 2D simulation is provided by a set of photographs, typically of natural scenes. This model is based on two evolving populations, one of prey and another of predators. Mutual conflict between these populations can produce both effective prey camouflage and predators skilled at “breaking” camouflage. The result is an open source artificial life model to help study camouflage in nature, and the perceptual phenomenon of camouflage more generally.

Body representations, which have multimodal receptive fields in the peripersonal space where individuals interact with the environment within their reach, show plasticity through tool use and are necessary for adaptive and skillful use of external tools. In this study, we propose a neural network model that develops a multimodal and body-centered peripersonal space representation of the plastic body representation through tool use, whereas previous developmental models can only explain the plastic body representation as a non-body-centered one. Our proposed model reconstructs visual and tactile sensations corresponding to proprioceptive sensations after integrating visual and tactile sensations through a Transformer based on a self-attention mechanism. By learning through camera vision and arm touch of a simulated robot and proprioception of camera and arm postures, a body representation was developed that localizes tactile sensations on a simultaneously developed peripersonal space representation. In particular, learning during tool use causes the body representation to have plasticity due to tool use, and the peripersonal space representation is shared by sharing part of the visual and tactile decoding modules. As a result, the model obtains the plastic body representation on the body-centered multimodal peripersonal space representation.

Identifying conditions that promote egalitarian major transitions, where unlike replicating units unite to form a higher-level unit, is an open problem with far-reaching implications. We propose that egalitarian major transitions can only begin in ecological communities that are conducive to them. To formalize this idea, we introduce the concept of “transition-ability”, which describes the extent to which a community is poised to undergo an egalitarian major transition. We hypothesize that transitionability is a property of ecological interaction networks, which represent the set of pairwise interactions among members of a community. Using a digital artificial ecology that simulates interactions between species based on a static interaction network, we test the transition-ability of interaction networks created by a range of graph-generation techniques, as well as some real-world ecological networks. To measure the extent to which a community is moving towards a major transition, we quantify the increase in community-level fitness relative to individual-level fitness across five different fitness proxies. We find that some network generation protocols produce more transitionable networks than others. In particular, interaction strengths (i.e. edge weights) have a substantial impact on transitionability, despite receiving low attention in the literature.

Pairing a neuro-symbolic model with library learning to facilitate program induction seems a promising way of fostering open-ended innovation, by leveraging the robustness, expressivity, and extrapolative capabilities of programs. This paper investigates how Open-Ended Dreamer (OED), an unsupervised diversity-oriented neuro-symbolic learner built upon DreamCoder (Ellis et al., 2021), may support open-ended program discovery. By alternating between phases of generation, selection, and abstraction, OED aims to expand a hierarchical library of diversity-enabling building blocks (in the form of programs), which are subsequently reused and composed in later iterations. As a first test-bed, we apply OED to a tower building domain and investigate the impact of library learning, neural guidance, innate priors, and language or environmental pressures on the formation of symbolic knowledge. Our experiments suggest that promoting greater exploration and stochasticity is crucial to offset the bias introduced by the growing language, and foster more creative divergence.

Scientists have long tried to predict evolutionary outcomes in order to design vaccines for next year’s diseases, stabilize endangered ecosystems, or make better choices in designing evolutionary algorithms. To predict, however, we must first be able to retroactively identify the key steps that determined the evolved state. Researchers have long examined the role of historical contingency in evolution; when do small, seemingly insignificant mutations substantially shift the probabilities of what traits or behaviors ultimately evolve? Practitioners of experimental evolution have recently begun to investigate this question using a new technique: analytic replay experiments. We can found many populations with a given genotype in order to measure the probability of a particular trait evolving from that starting point; we call this the “potentiation” of that genotype. Moving along a lineage, we can identify which mutations altered potentiation. Here we used digital organisms to conduct a high-resolution analysis of how individual mutations affected the potentiation of associative learning. We find that the probability of evolving associative learning can increase suddenly – even with a single mutation that appeared innocuous when it occurred. While there was no obvious signal to identify potentiating mutations as they arose, we were able to retrospectively identify mechanisms by which these mutations influenced subsequent evolution. Many of the most interesting and complex evolutionary adaptations that occur in nature are exceptionally rare. Here, we extend techniques for understanding these rare evolutionary events and the patterns and processes that produce them.

Embodied AI, the main direction of AI engaged in the synthetic modeling of natural cognition, is generally based on hardware-software approaches. Here we present the general lines of a proposal referred as the “wetware route” to Embodied AI , construed as the ideation and implementation of chemical models of cognitive processes inspired by the theory of autopoiesis. Our approach proposes to focus on the organizational traits that characterize autopoietic systems as cognitive systems, and accordingly attempts to define a viable organizational Embodied AI research program in the nascent wetware sci-tech arena of Synthetic Biology.

This research paper presents a chemical compiler developed to find optimal configurations of a platform for synthesizing specific branched oligomers in an artificial chemistry, along with exemplary compiler output and benchmarks where the platform configuration suggested by the compiler is compared to other configurations in simulation. The compiler operates as a pipeline with two stages: labelling and optimization. The report explains the structure of the compiler target and its interpretation, followed by a code walk-through of the compiler stages with code snippets and examples. The compiler can be used as a code generator for reactions in a chemical simulator and to derive loading schemes for multilevel droplets. The results obtained in simulations suggest that the container system can efficiently optimize the yield of coupled reaction networks and that multi-level droplets can lead to significant improvements.