Information processing via physical soft body

Informational embodiment: Computational role of information structure in codes and robots

The body morphology plays an important role in the way information is perceived and processed by an agent. We address an information theory (IT) account on how the precision of sensors, the accuracy of motors, their placement, the body geometry, shape the information structure in robots and computational codes. As an original idea, we envision the robot's body as a physical communication channel through which information is conveyed, in and out, despite intrinsic noise and material limitations. Following this, entropy, a measure of information and uncertainty, can be used to maximize the efficiency of robot design and of algorithmic codes per se. This is known as the principle of Entropy Maximization (PEM) introduced in biology by Barlow in 1969. The Shannon's source coding theorem provides then a framework to compare different types of bodies in terms of sensorimotor information. In line with the PEM, we introduce a special class of efficient codes used in IT that reached the Shannon limits in terms of information capacity for error correction and robustness against noise, and parsimony. These efficient codes, which exploit insightfully quantization and randomness, permit to deal with uncertainty, redundancy and compacity. These features can be used for perception and control in intelligent systems. In various examples and closing discussions, we reflect on the broader implications of our framework that we called Informational Embodiment to motor theory and bio-inspired robotics, touching upon concepts like motor synergies, reservoir computing, and morphological computation. These insights can contribute to a deeper understanding of how information theory intersects with the embodiment of intelligence in both natural and artificial systems.

The Drosophila Connectome as a Computational Reservoir for Time-Series Prediction

In this work, we explore the possibility of using the topology and weight distribution of the connectome of a Drosophila, or fruit fly, as a reservoir for multivariate chaotic time-series prediction. Based on the information taken from the recently released full connectome, we create the connectivity matrix of an Echo State Network. Then, we use only the most connected neurons and implement two possible selection criteria, either preserving or breaking the relative proportion of different neuron classes which are also included in the documented connectome, to obtain a computationally convenient reservoir. We then investigate the performance of such architectures and compare them to state-of-the-art reservoirs. The results show that the connectome-based architecture is significantly more resilient to overfitting compared to the standard implementation, particularly in cases already prone to overfitting. To further isolate the role of topology and synaptic weights, hybrid reservoirs with the connectome topology but random synaptic weights and the connectome weights but random topologies are included in the study, demonstrating that both factors play a role in the increased overfitting resilience. Finally, we perform an experiment where the entire connectome is used as a reservoir. Despite the much higher number of trained parameters, the reservoir remains resilient to overfitting and has a lower normalized error, under 2%, at lower regularisation, compared to all other reservoirs trained with higher regularisation.

Reservoir controllers design though robot-reservoir timescale alignment

Natural behavior emerging in nonlinear dynamical systems enables reservoir computers to control underactuated robots by approximating their inverse dynamics. Unlike other model-free approaches, the reservoir controllers are sample-efficient, meaning a weighted average of the reservoir output can be trained with a limited amount of pre-recorded data. However, developing and testing the reservoir controller relies on repetitive experiments that require researchers' proficiency in both robot and reservoir design. In this paper, we propose a design method for reliable reservoir controllers by synchronizing the timescales of the reservoir dynamics with those observed in the robot. The results demonstrate that our timescale alignment test filters out 99% of ineffective reservoirs. We further applied the selected reservoirs to computational tasks including short-term memory and parity checks, along with control tasks involving robot trajectory tracking. Our findings reveal that a higher computational capability reduces the control failure rate, though it concurrently increases the trajectory-tracking error.

Echo state property and memory capacity of artificial spin ice

Physical reservoir computing by using artificial spin ice (ASI) has been proposed on the basis of both numerical and experimental analyses. ASI is a many-body system consisting of ferromagnets with various interactions. Recently, fabricating magnetic tunnel junctions (MTJs) as ferromagnets in an ASI was achieved in the experiment, which enables an electrical detection of magnetic state of each MTJ independently. However, performing a recognition task of time-dependent signal by such an MTJ-based ASI has not been reported yet. In this work, we examine numerical simulation of a recognition task of time-dependent input and evaluate short-term memory and parity-check capacities. These capacities change significantly when the magnitude of the input magnetic field is comparable to a value around which the magnetization alignment is greatly affected by the dipole interaction. It implies that the presence of the dipole interaction results in a loss of echo state property. This point was clarified by evaluating Lyapunov exponent and confirming that the drastic change of the memory capacities appears near the boundary between negative and zero exponents, which corresponds to the edge of echo state property.

Harnessing spatiotemporal transformation in magnetic domains for nonvolatile physical reservoir computing

Combining physics with computational models is increasingly recognized for enhancing the performance and energy efficiency in neural networks. Physical reservoir computing uses material dynamics of physical substrates for temporal data processing. Despite the ease of training, building an efficient reservoir remains challenging. Here, we explore beyond the conventional delay-based reservoirs by exploiting the spatiotemporal transformation in all-electric spintronic devices. Our nonvolatile spintronic reservoir effectively transforms the history dependence of reservoir states to the path dependence of domains. We configure devices triggered by different pulse widths as neurons, creating a reservoir featured by strong nonlinearity and rich interconnections. Using a small reservoir of merely 14 physical nodes, we achieved a high recognition rate of 0.903 in written digit recognition and a low error rate of 0.076 in Mackey-Glass time series prediction on a proof-of-concept printed circuit board. This work presents a promising route of nonvolatile physical reservoir computing, which is adaptable to the larger memristor family and broader physical neural networks.

Physical reservoir computing on a soft bio-inspired swimmer

Bio-inspired Autonomous Underwater Vehicles with soft bodies provide significant performance benefits over conventional propeller-driven vehicles; however, it is difficult to control these vehicles due to their soft underactuated bodies. This study investigates the application of Physical Reservoir Computing (PRC) in the swimmer's flexible body to perform state estimation. This PRC informed state estimation has potential to be used in vehicle control. PRC is a type of recurrent neural network that leverages the nonlinear dynamics of a physical system to predict a nonlinear spatiotemporal input-output relationship. By embodying the neural network into the physical structure, PRC can process the response to an environment input with high computational efficiency. This study uses a soft bio-inspired propulsor embodied as a physical reservoir. We evaluate its ability to predict different state estimation tasks including hydrodynamic forces and benchmark computational tasks in response to the forcing applied to the artificial muscles during actuation. The propulsor's nonlinear fluid-structural dynamics act as the physical reservoir and the kinematic feedback serves as the reservoir readouts. We show that the bio-inspired underwater propulsor can predict the hydrodynamic thrust and benchmark tasks with high accuracy under specific input frequencies. By analyzing the frequency spectrum of the input, readouts, and target signals, we demonstrate that the system's dynamic response determines the frequency contents relevant to the task being predicted. The propulsor's ability to process information stems from its nonlinearity, as it is responsible to transform the input signal into a broader spectrum of frequency content at the readouts. This broad band of frequency content is necessary to recreate the target signal within the PRC algorithm, thereby improving the prediction performance. The spectral analysis provides a unique perspective to analyze the nonlinear dynamics of a physical reservoir and serves as a valuable tool for examining other types of vibratory systems for PRC. This work serves as a first step towards embodying computation into soft bio-inspired swimmers.

Iono-Magnonic Reservoir Computing With Chaotic Spin Wave Interference Manipulated by Ion-Gating

Physical reservoirs are a promising approach for realizing high-performance artificial intelligence devices utilizing physical devices. Although nonlinear interfered spin-wave multi-detection exhibits high nonlinearity and the ability to map in high dimensional feature space, it does not have sufficient performance to process time-series data precisely. Herein, development of an iono-magnonic reservoir by combining such interfered spin wave multi-detection and ion-gating involving protonation-induced redox reaction triggered by the application of voltage is reported. This study is the first to report the manipulation of the propagating spin wave property by ion-gating and the application of the same to physical reservoir computing. The subject iono-magnonic reservoir can generate various reservoir states in a single homogenous medium by utilizing a spin wave property modulated by ion-gating. Utilizing the strong nonlinearity resulting from chaos, the reservoir shows good computational performance in completing the Mackey-Glass chaotic time-series prediction task, and the performance is comparable to that exhibited by simulated neural networks.

An Entropy Dynamics Approach to Inferring Fractal-Order Complexity in the Electromagnetics of Solids

A fractal-order entropy dynamics model is developed to create a modified form of Maxwell's time-dependent electromagnetic equations. The approach uses an information-theoretic method by combining Shannon's entropy with fractional moment constraints in time and space. Optimization of the cost function leads to a time-dependent Bayesian posterior density that is used to homogenize the electromagnetic fields. Self-consistency between maximizing entropy, inference of Bayesian posterior densities, and a fractal-order version of Maxwell's equations are developed. We first give a set of relationships for fractal derivative definitions and their relationship to divergence, curl, and Laplacian operators. The fractal-order entropy dynamic framework is then introduced to infer the Bayesian posterior and its application to modeling homogenized electromagnetic fields in solids. The results provide a methodology to help understand complexity from limited electromagnetic data using maximum entropy by formulating a fractal form of Maxwell's electromagnetic equations.

Toward mechanical proprioception in autonomously reconfigurable kirigami-inspired mechanical systems

Mechanical metamaterials have recently been exploited as an interesting platform for information storing, retrieval and processing, analogous to electronic devices. In this work, we describe the design and fabrication a two-dimensional (2D) multistable metamaterial consisting of building blocks that can be switched between two distinct stable phases, and which are capable of storing binary information analogous to digital bits. By changing the spatial distribution of the phases, we can achieve a variety of different configurations and tunable mechanical properties (both static and dynamic). Moreover, we demonstrate the ability to determine the phase distribution via simple probing of the dynamic properties, to which we refer as mechanical proprioception. Finally, as a simple demonstration of feasibility, we illustrate a strategy for building autonomous kirigami systems that can receive inputs from their environment. This work could bring new insights for the design of mechanical metamaterials with information processing and computing functionalities. This article is part of the theme issue 'Origami/Kirigami-inspired structures: from fundamentals to applications'.

Embedding Bifurcations into Pneumatic Artificial Muscle

Harnessing complex body dynamics has long been a challenge in robotics, particularly when dealing with soft dynamics, which exhibit high complexity in interacting with the environment. Recent studies indicate that these dynamics can be used as a computational resource, exemplified by the McKibben pneumatic artificial muscle, a common soft actuator. This study demonstrates that bifurcations, including periodic and chaotic dynamics, can be embedded into the pneumatic artificial muscle, with the entire bifurcation structure using the framework of physical reservoir computing. These results suggest that dynamics not present in training data can be embedded through bifurcation embedment, implying the capability to incorporate various qualitatively different patterns into pneumatic artificial muscle without the need to design and learn all required patterns explicitly. Thus, this study introduces a novel approach to simplify robotic devices and control training by reducing reliance on external pattern generators and the amount and types of training data needed for control.

Reservoir-computing based associative memory and itinerancy for complex dynamical attractors

Traditional neural network models of associative memories were used to store and retrieve static patterns. We develop reservoir-computing based memories for complex dynamical attractors, under two common recalling scenarios in neuropsychology: location-addressable with an index channel and content-addressable without such a channel. We demonstrate that, for location-addressable retrieval, a single reservoir computing machine can memorize a large number of periodic and chaotic attractors, each retrievable with a specific index value. We articulate control strategies to achieve successful switching among the attractors, unveil the mechanism behind failed switching, and uncover various scaling behaviors between the number of stored attractors and the reservoir network size. For content-addressable retrieval, we exploit multistability with cue signals, where the stored attractors coexist in the high-dimensional phase space of the reservoir network. As the length of the cue signal increases through a critical value, a high success rate can be achieved. The work provides foundational insights into developing long-term memories and itinerancy for complex dynamical patterns.

Self-learning physical reservoir computer

A self-learning physical reservoir computer is demonstrated using an adaptive oscillator. Whereas physical reservoir computing repurposes the dynamics of a physical system for computation through machine learning, adaptive oscillators can innately learn and store information in plastic dynamic states. The adaptive state(s) can be used directly as physical node(s), but these plastic states can also be used to self-learn the optimal reservoir parameters for more complex tasks requiring virtual nodes from the base oscillator. Both this self-learning property for reconfigurable computing and the morphable logic gate property of the adaptive oscillator make this an ideal candidate for a multipurpose neuromorphic processor.

Highly-integrable analogue reservoir circuits based on a simple cycle architecture

Physical reservoir computing is a promising solution for accelerating artificial intelligence (AI) computations. Various physical systems that exhibit nonlinear and fading-memory properties have been proposed as physical reservoirs. Highly-integrable physical reservoirs, particularly for edge AI computing, has a strong demand. However, realizing a practical physical reservoir with high performance and integrability remains challenging. Herein, we present an analogue circuit reservoir with a simple cycle architecture suitable for complementary metal-oxide-semiconductor (CMOS) chip integration. In several benchmarks and demonstrations using synthetic and real-world data, our developed hardware prototype and its simulator exhibit a high prediction performance and sufficient memory capacity for practical applications, showing promise for future applications in highly integrated AI accelerators.

Pressure-Based Soft Agents

Biological agents have bodies that are composed mostly of soft tissue. Researchers have resorted to soft bodies to investigate Artificial Life (ALife)-related questions; similarly, a new era of soft-bodied robots has just begun. Nevertheless, because of their infinite degrees of freedom, soft bodies pose unique challenges in terms of simulation, control, and optimization. Herein I propose a novel soft-bodied agents formalism, namely, pressure-based soft agents (PSAs): spring-mass membranes containing a pressurized medium. Pressure endows the agents with structure, while springs and masses simulate softness and allow the agents to assume a large gamut of shapes. PSAs actuate both locally, by changing the resting lengths of springs, and globally, by modulating global pressure. I optimize the controller of PSAs for a locomotion task on hilly terrain, an escape task from a cage, and an object manipulation task. The results suggest that PSAs are indeed effective at the tasks, especially those requiring a shape change. I envision PSAs as playing a role in modeling soft-bodied agents, such as soft robots and biological cells.

Magnetization Vector Rotation Reservoir Computing Operated by Redox Mechanism

Physical reservoir computing is a promising way to develop efficient artificial intelligence using physical devices exhibiting nonlinear dynamics. Although magnetic materials have advantages in miniaturization, the need for a magnetic field and large electric current results in high electric power consumption and a complex device structure. To resolve these issues, we propose a redox-based physical reservoir utilizing the planar Hall effect and anisotropic magnetoresistance, which are phenomena described by different nonlinear functions of the magnetization vector that do not need a magnetic field to be applied. The expressive power of this reservoir based on a compact all-solid-state redox transistor is higher than the previous physical reservoir. The normalized mean square error of the reservoir on a second-order nonlinear equation task was 1.69 × 10-3, which is lower than that of a memristor array (3.13 × 10-3) even though the number of reservoir nodes was fewer than half that of the memristor array.

Few- and single-molecule reservoir computing experimentally demonstrated with surface-enhanced Raman scattering and ion gating

Molecule-based reservoir computing (RC) is promising for achieving low power consumption neuromorphic computing, although the information-processing capability of small numbers of molecules is not clear. Here, we report a few- and single-molecule RC that uses the molecular vibration dynamics in the para-mercaptobenzoic acid (pMBA) detected by surface-enhanced Raman scattering (SERS) with tungsten oxide nanorod/silver nanoparticles. The Raman signals of the pMBA molecules, adsorbed at the SERS active site of the nanorod, were reversibly perturbated by the application of voltage-induced local pH changes near the molecules, and then used to perform time-series analysis tasks. Despite the small number of molecules used, our system achieved good performance, including >95% accuracy in various nonlinear waveform transformations, 94.3% accuracy in solving a second-order nonlinear dynamic system, and a prediction error of 25.0 milligrams per deciliter in a 15-minute-ahead blood glucose level prediction. Our work provides a concept of few-molecular computing with practical computation capabilities.

Photonic Physical Reservoir Computing with Tunable Relaxation Time Constant

Recent years have witnessed a rising demand for edge computing, and there is a need for methods to decrease the computational cost while maintaining a high learning performance when processing information at arbitrary edges. Reservoir computing using physical dynamics has attracted significant attention. However, currently, the timescale of the input signals that can be processed by physical reservoirs is limited by the transient characteristics inherent to the selected physical system. This study used an Sn-doped In2 O3 /Nb-doped SrTiO3 junction to fabricate a memristor that could respond to both electrical and optical stimuli. The results show that the timescale of the transient current response of the device could be controlled over several orders of magnitude simply by applying a small voltage. The computational performance of the device as a physical reservoir is evaluated in an image classification task, demonstrating that the learning accuracy could be optimized by tuning the device to exhibit appropriate transient characteristics according to the timescale of the input signals. These results are expected to provide deeper insights into the photoconductive properties of strontium titanate, as well as support the physical implementation of computing systems.

Computation harvesting from nature dynamics for predicting wind speed and direction

Natural phenomena generate complex dynamics because of nonlinear interactions among their components. The dynamics can be exploited as a kind of computational resource. For example, in the framework of natural computation, various natural phenomena such as quantum mechanics and cellular dynamics are used to realize general purpose calculations or logical operations. In recent years, simple collection of such nature dynamics has become possible in a sensor-rich society. For example, images of plant movement that have been captured indirectly by a surveillance camera can be regarded as sensor outputs reflecting the state of the wind striking the plant. Herein, based on ideas of physical reservoir computing, we present a methodology for wind speed and direction estimation from naturally occurring sensors in movies. Then we demonstrate its effectiveness through experimentation. Specifically using the proposed methodology, we investigate the computational capability of the nature dynamics, revealing its high robustness and generalization performance for computation.

Reservoir computing with higher-order interactive coupled pendulums

The reservoir computing approach utilizes a time series of measurements as input to a high-dimensional dynamical system known as a reservoir. However, the approach relies on sampling a random matrix to define its underlying reservoir layer, which leads to numerous hyperparameters that need to be optimized. Here, we propose a nonlocally coupled pendulum model with higher-order interactions as a novel reservoir, which requires no random underlying matrices and fewer hyperparameters. We use Bayesian optimization to explore the hyperparameter space within a minimal number of iterations and train the coupled pendulums model to reproduce the chaotic attractors, which simplifies complicated hyperparameter optimization. We illustrate the effectiveness of our technique with the Lorenz system and the Hindmarsh-Rose neuronal model, and we calculate the Pearson correlation coefficients between time series and the Hausdorff metrics in the phase space. We demonstrate the contribution of higher-order interactions by analyzing the interaction between different reservoir configurations and prediction performance, as well as computations of the largest Lyapunov exponents. The chimera state is found as the most effective dynamical regime for prediction. The findings, where we present a new reservoir structure, offer potential applications in the design of high-performance modeling of dynamics in physical systems.

Redox-based ion-gating reservoir consisting of (104) oriented LiCoO2 film, assisted by physical masking

Reservoir computing (RC) is a machine learning framework suitable for processing time series data, and is a computationally inexpensive and fast learning model. A physical reservoir is a hardware implementation of RC using a physical system, which is expected to become the social infrastructure of a data society that needs to process vast amounts of information. Ion-gating reservoirs (IGR) are compact and suitable for integration with various physical reservoirs, but the prediction accuracy and operating speed of redox-IGRs using WO3 as the channel are not sufficient due to irreversible Li+ trapping in the WO3 matrix during operation. Here, in order to enhance the computation performance of redox-IGRs, we developed a redox-based IGR using a (104) oriented LiCoO2 thin film with high electronic and ionic conductivity as a trap-free channel material. The subject IGR utilizes resistance change that is due to a redox reaction (LiCoO2 ⟺ Li1-xCoO2 + xLi+ + xe-) with the insertion and desertion of Li+. The prediction error in the subject IGR was reduced by 72% and the operation speed was increased by 4 times compared to the previously reported WO3, which changes are due to the nonlinear and reversible electrical response of LiCoO2 and the high dimensionality enhanced by a newly developed physical masking technique. This study has demonstrated the possibility of developing high-performance IGRs by utilizing materials with stronger nonlinearity and by increasing output dimensionality.

The morphological paradigm in robotics

In the paper, we are going to show how robotics is undergoing a shift in a bionic direction after a period of emphasis on artificial intelligence and increasing computational efficiency, which included isolation and extreme specialization. We assemble these new developments under the label of the morphological paradigm. The change in its paradigms and the development of alternatives to the principles that dominated robotics for a long time contains a more general epistemological significance. The role of body, material, environment, interaction and the paradigmatic status of biological and evolutionary systems for the principles of control are crucial here. Our focus will be on the introduction of the morphological paradigm in a new type of robotics and to contrast the interests behind this development with the interests shaping former models. The article aims to give a clear account of the changes in principles of orientation and control as well as concluding general observation in terms of historical epistemology, suggesting further political-epistemological analysis.

Effect of temporal resolution on the reproduction of chaotic dynamics via reservoir computing

Reservoir computing is a machine learning paradigm that uses a structure called a reservoir, which has nonlinearities and short-term memory. In recent years, reservoir computing has expanded to new functions such as the autonomous generation of chaotic time series, as well as time series prediction and classification. Furthermore, novel possibilities have been demonstrated, such as inferring the existence of previously unseen attractors. Sampling, in contrast, has a strong influence on such functions. Sampling is indispensable in a physical reservoir computer that uses an existing physical system as a reservoir because the use of an external digital system for the data input is usually inevitable. This study analyzes the effect of sampling on the ability of reservoir computing to autonomously regenerate chaotic time series. We found, as expected, that excessively coarse sampling degrades the system performance, but also that excessively dense sampling is unsuitable. Based on quantitative indicators that capture the local and global characteristics of attractors, we identify a suitable window of the sampling frequency and discuss its underlying mechanisms.

Computational capability of ecological dynamics

Ecological dynamics is driven by complex ecological networks. Computational capabilities of artificial networks have been exploited for machine learning purposes, yet whether an ecological network possesses a computational capability and whether/how we can use it remain unclear. Here, we developed two new computational/empirical frameworks based on reservoir computing and show that ecological dynamics can be used as a computational resource. In silico ecological reservoir computing (ERC) reconstructs ecological dynamics from empirical time series and uses simulated system responses for information processing, which can predict near future of chaotic dynamics and emulate nonlinear dynamics. The real-time ERC uses real population dynamics of a unicellular organism, Tetrahymena thermophila. The temperature of the medium is an input signal and population dynamics is used as a computational resource. Intriguingly, the real-time ecological reservoir has necessary conditions for computing (e.g. synchronized dynamics in response to the same input sequences) and can make near-future predictions of empirical time series, showing the first empirical evidence that population-level phenomenon is capable of real-time computations. Our finding that ecological dynamics possess computational capability poses new research questions for computational science and ecology: how can we efficiently use it and how is it actually used, evolved and maintained in an ecosystem?

Tapered whisker reservoir computing for real-time terrain identification-based navigation

This paper proposes a new method for real-time terrain recognition-based navigation for mobile robots. Mobile robots performing tasks in unstructured environments need to adapt their trajectories in real-time to achieve safe and efficient navigation in complex terrains. However, current methods largely depend on visual and IMU (inertial measurement units) that demand high computational resources for real-time applications. In this paper, a real-time terrain identification-based navigation method is proposed using an on-board tapered whisker-based reservoir computing system. The nonlinear dynamic response of the tapered whisker was investigated in various analytical and Finite Element Analysis frameworks to demonstrate its reservoir computing capabilities. Numerical simulations and experiments were cross-checked with each other to verify that whisker sensors can separate different frequency signals directly in the time domain and demonstrate the computational superiority of the proposed system, and that different whisker axis locations and motion velocities provide variable dynamical response information. Terrain surface-following experiments demonstrated that our system could accurately identify changes in the terrain in real-time and adjust its trajectory to stay on specific terrain.

Robotic automation and unsupervised cluster assisted modeling for solving the forward and reverse design problem of paper airplanes

Although often regarded a childhood toy, the design of paper airplanes is subtly complex. The design space and mapping from geometry to distance flown is highly nonlinear and probabilistic where a single airplane design exhibits a multitude of trajectory forms and flight distances. This makes optimization and understanding of their behavior challenging for humans. By understanding the behavior of paper airplanes and predicting flight behavior, there is a potential to improve the design of aerial vehicles that operate at low Reynolds numbers. By developing a robotic system that can fabricate, test, analyze, and model the flight behavior in an unsupervised fashion, a wide design space can be reliably characterized. We find there are discrete behavioral groups that result in different trajectories: nose dive, glide, and recovery glide. Informed by this characterization we propose a method of using Gaussian mixture models to extract the clusters of the design space that map to these different behaviors. This allows us to solve both the forward and reverse design problem for paper airplanes, and also to perform efficient optimization of the geometry for a given target flight distance.

A Perovskite Memristor with Large Dynamic Space for Analog-Encoded Image Recognition

Reservoir computing (RC) is a computational architecture capable of efficiently processing temporal information, which allows low-cost hardware implementation. However, the previously reported memristor-based RC mostly utilized binarized data sets to reduce the difficulty of signal processing of the memristor, which inevitably induces data distortion to a certain extent, leading to poor network computing performance. Here, we report on a RC system in a fully memristive architecture based on solution-processed perovskite memristors. The perovskite memristor exhibits 10000 conductance states with a modulation range of more than 4 orders of magnitude. The obtained tens of thousands of finely spaced conductance states with a near-ideal analog property provide a sufficiently large dynamic range and enough intermediate states, which were further applied as a reservoir to map the feature information on different sequential inputs in an analog way. The computing capability of the image classification task of a Fashion-MNIST data set with a high recognition accuracy of up to 90.1% shows that the excellent analog and short-term properties of our perovskite memristor allow the hardware implementation of neuromorphic computing with a reduced training cost.

Physical deep learning with biologically inspired training method: gradient-free approach for physical hardware

Ever-growing demand for artificial intelligence has motivated research on unconventional computation based on physical devices. While such computation devices mimic brain-inspired analog information processing, the learning procedures still rely on methods optimized for digital processing such as backpropagation, which is not suitable for physical implementation. Here, we present physical deep learning by extending a biologically inspired training algorithm called direct feedback alignment. Unlike the original algorithm, the proposed method is based on random projection with alternative nonlinear activation. Thus, we can train a physical neural network without knowledge about the physical system and its gradient. In addition, we can emulate the computation for this training on scalable physical hardware. We demonstrate the proof-of-concept using an optoelectronic recurrent neural network called deep reservoir computer. We confirmed the potential for accelerated computation with competitive performance on benchmarks. Our results provide practical solutions for the training and acceleration of neuromorphic computation.

Mind the matter: Active matter, soft robotics, and the making of bio-inspired artificial intelligence

Philosophical and theoretical debates on the multiple realisability of the cognitive have historically influenced discussions of the possible systems capable of instantiating complex functions like memory, learning, goal-directedness, and decision-making. These debates have had the corollary of undermining, if not altogether neglecting, the materiality and corporeality of cognition-treating material, living processes as "hardware" problems that can be abstracted out and, in principle, implemented in a variety of materials-in particular on digital computers and in the form of state-of-the-art neural networks. In sum, the matter in se has been taken not to matter for cognition. However, in this paper, we argue that the materiality of cognition-and the living, self-organizing processes that it enables-requires a more detailed assessment when understanding the nature of cognition and recreating it in the field of embodied robotics. Or, in slogan form, that the matter matters for cognitive form and function. We pull from the fields of Active Matter Physics, Soft Robotics, and Basal Cognition literature to suggest that the imbrication between material and cognitive processes is closer than standard accounts of multiple realisability suggest. In light of this, we propose upgrading the notion of multiple realisability from the standard version-what we call 1.0-to a more nuanced conception 2.0 to better reflect the recent empirical advancements, while at the same time averting many of the problems that have been raised for it. These fields are actively reshaping the terrain in which we understand materiality and how it enables, mediates, and constrains cognition. We propose that taking the materiality of our embodied, precarious nature seriously furnishes an important research avenue for the development of embodied robots that autonomously value, engage, and interact with the environment in a goal-directed manner, in response to existential needs of survival, persistence, and, ultimately, reproduction. Thus, we argue that by placing further emphasis on the soft, active, and plastic nature of the materials that constitute cognitive embodiment, we can move further in the direction of autonomous embodied robots and Artificial Intelligence.

Edge-of-chaos learning achieved by ion-electron-coupled dynamics in an ion-gating reservoir

Physical reservoir computing has recently been attracting attention for its ability to substantially reduce the computational resources required to process time series data. However, the physical reservoirs that have been reported to date have had insufficient computational capacity, and most of them have a large volume, which makes their practical application difficult. Here, we describe the development of a Li⁺ electrolyte-based ion-gating reservoir (IGR), with ion-electron-coupled dynamics, for use in high-performance physical reservoir computing. A variety of synaptic responses were obtained in response to past experience, which were stored as transient charge density patterns in an electric double layer, at the Li⁺ electrolyte/diamond interface. Performance for a second-order nonlinear dynamical equation task is one order of magnitude higher than memristor-based reservoirs. The edge-of-chaos state of the IGR enabled the best computational capacity. The IGR described here opens the way for high-performance and integrated neural network devices.

Leveraging plant physiological dynamics using physical reservoir computing

Plants are complex organisms subject to variable environmental conditions, which influence their physiology and phenotype dynamically. We propose to interpret plants as reservoirs in physical reservoir computing. The physical reservoir computing paradigm originates from computer science; instead of relying on Boolean circuits to perform computations, any substrate that exhibits complex non-linear and temporal dynamics can serve as a computing element. Here, we present the first application of physical reservoir computing with plants. In addition to investigating classical benchmark tasks, we show that Fragaria × ananassa (strawberry) plants can solve environmental and eco-physiological tasks using only eight leaf thickness sensors. Although the results indicate that plants are not suitable for general-purpose computation but are well-suited for eco-physiological tasks such as photosynthetic rate and transpiration rate. Having the means to investigate the information processing by plants improves quantification and understanding of integrative plant responses to dynamic changes in their environment. This first demonstration of physical reservoir computing with plants is key for transitioning towards a holistic view of phenotyping and early stress detection in precision agriculture applications since physical reservoir computing enables us to analyse plant responses in a general way: environmental changes are processed by plants to optimise their phenotype.

Counting with Cilia: The Role of Morphological Computation in Basal Cognition Research

"Morphological computation" is an increasingly important concept in robotics, artificial intelligence, and philosophy of the mind. It is used to understand how the body contributes to cognition and control of behavior. Its understanding in terms of "offloading" computation from the brain to the body has been criticized as misleading, and it has been suggested that the use of the concept conflates three classes of distinct processes. In fact, these criticisms implicitly hang on accepting a semantic definition of what constitutes computation. Here, I argue that an alternative, mechanistic view on computation offers a significantly different understanding of what morphological computation is. These theoretical considerations are then used to analyze the existing research program in developmental biology, which understands morphogenesis, the process of development of shape in biological systems, as a computational process. This important line of research shows that cognition and intelligence can be found across all scales of life, as the proponents of the basal cognition research program propose. Hence, clarifying the connection between morphological computation and morphogenesis allows for strengthening the role of the former concept in this emerging research field.

Mechanical neural networks: Architected materials that learn behaviors

Aside from some living tissues, few materials can autonomously learn to exhibit desired behaviors as a consequence of prolonged exposure to unanticipated ambient loading scenarios. Still fewer materials can continue to exhibit previously learned behaviors in the midst of changing conditions (e.g., rising levels of internal damage, varying fixturing scenarios, and fluctuating external loads) while also acquiring new behaviors best suited for the situation at hand. Here, we describe a class of architected materials, called mechanical neural networks (MNNs), that achieve such learning capabilities by tuning the stiffness of their constituent beams similar to how artificial neural networks (ANNs) tune their weights. An example lattice was fabricated to demonstrate its ability to learn multiple mechanical behaviors simultaneously, and a study was conducted to determine the effect of lattice size, packing configuration, algorithm type, behavior number, and linear-versus-nonlinear stiffness tunability on MNN learning as proposed. Thus, this work lays the foundation for artificial-intelligent (AI) materials that can learn behaviors and properties.

Reservoir computing based on an external-cavity semiconductor laser with optical feedback modulation

We numerically and experimentally investigate reservoir computing based on a single semiconductor laser with optical feedback modulation. In this scheme, an input signal is injected into a semiconductor laser via intensity or phase modulation of the optical feedback signal. We perform a chaotic time-series prediction task using the reservoir and compare the performances of intensity and phase modulation schemes. Our results indicate that the feedback signal of the phase modulation scheme outperforms that of the intensity modulation scheme. Further, we investigate the performance dependence of reservoir computing on parameter values and observe that the prediction error improves for large injection currents, unlike the results in a semiconductor laser with an optical injection input. The physical origin of the superior performance of the phase modulation scheme is analyzed using external cavity modes obtained from steady-state analysis in the phase space. The analysis indicates that high-dimensional mapping can be achieved from the input signal to the trajectory of the response laser output by using phase modulation of the feedback signal.

Reservoir computing on a silicon platform with a ferroelectric field-effect transistor

Reservoir computing offers efficient processing of time-series data with exceptionally low training cost for real-time computing in edge devices where energy and hardware resources are limited. Here, we report reservoir computing hardware based on a ferroelectric field-effect transistor (FeFET) consisting of silicon and ferroelectric hafnium zirconium oxide. The rich dynamics originating from the ferroelectric polarization dynamics and polarization-charge coupling are the keys leading to the essential properties for reservoir computing: the short-term memory and high-dimensional nonlinear transform function. We demonstrate that an FeFET-based reservoir computing system can successfully solve computational tasks on time-series data processing including nonlinear time series prediction after training with simple regression. Due to the FeFET’s high feasibility of implementation on the silicon platform, the systems have flexibility in both device- and circuit-level designs, and have a high potential for on-chip integration with existing computing technologies towards the realization of advanced intelligent systems.

Kasidit Toprasertpong and colleagues describe reservoir computing hardware with potential for on-chip integration with existing computing technologies. The approach is based on a ferroelectric field-effect transistor, and can solve computational tasks on time series data including nonlinear time series prediction after training with simple regression.

Lessons for Robotics From the Control Architecture of the Octopus

Biological and artificial agents are faced with many of the same computational and mechanical problems, thus strategies evolved in the biological realm can serve as inspiration for robotic development. The octopus in particular represents an attractive model for biologically-inspired robotic design, as has been recognized for the emerging field of soft robotics. Conventional global planning-based approaches to controlling the large number of degrees of freedom in an octopus arm would be computationally intractable. Instead, the octopus appears to exploit a distributed control architecture that enables effective and computationally efficient arm control. Here we will describe the neuroanatomical organization of the octopus peripheral nervous system and discuss how this distributed neural network is specialized for effectively mediating decisions made by the central brain and the continuous actuation of limbs possessing an extremely large number of degrees of freedom. We propose top-down and bottom-up control strategies that we hypothesize the octopus employs in the control of its soft body. We suggest that these strategies can serve as useful elements in the design and development of soft-bodied robotics.

Spintronic reservoir computing without driving current or magnetic field

Recent studies have shown that nonlinear magnetization dynamics excited in nanostructured ferromagnets are applicable to brain-inspired computing such as physical reservoir computing. The previous works have utilized the magnetization dynamics driven by electric current and/or magnetic field. This work proposes a method to apply the magnetization dynamics driven by voltage control of magnetic anisotropy to physical reservoir computing, which will be preferable from the viewpoint of low-power consumption. The computational capabilities of benchmark tasks in single MTJ are evaluated by numerical simulation of the magnetization dynamics and found to be comparable to those of echo-state networks with more than 10 nodes.

Information Processing Capacity of a Single-Node Reservoir Computer: An Experimental Evaluation

Physical dynamical systems are able to process information in a nontrivial manner. The machine learning paradigm of reservoir computing (RC) provides a suitable framework for information processing in (analog) dynamical systems. The potential of dynamical systems for RC can be quantitatively characterized by the information processing capacity (IPC) measure. Here, we evaluate the IPC measure of a reservoir computer based on a single-analog nonlinear node coupled with delay. We link the extracted IPC measures to the dynamical regime of the reservoir, reporting an experimentally measured nonlinear memory of up to seventh order. In addition, we find a nonhomogeneous distribution of the linear and nonlinear contributions to the IPC as a function of the system operating conditions. Finally, we unveil the role of noise in the IPC of the analog implementation by performing ad hoc numerical simulations. In this manner, we identify the so-called edge of stability as being the most promising operating condition of the experimental implementation for RC purposes in terms of computational power and noise robustness. Similarly, a strong input drive is shown to have beneficial properties, albeit with a reduced memory depth.

Wave-packet spreading in disordered soft architected structures

We study the dynamical and chaotic behavior of a disordered one-dimensional elastic mechanical lattice, which supports translational and rotational waves. The model used in this work is motivated by the recent experimental results of Deng et al. [Nat. Commun. 9, 1 (2018)]. This lattice is characterized by strong geometrical nonlinearities and the coupling of two degrees-of-freedom (DoFs) per site. Although the linear limit of the structure consists of a linear Fermi-Pasta-Ulam-Tsingou lattice and a linear Klein-Gordon (KG) lattice whose DoFs are uncoupled, by using single site initial excitations on the rotational DoF, we evoke the nonlinear coupling between the system's translational and rotational DoFs. Our results reveal that such coupling induces rich wave-packet spreading behavior in the presence of strong disorder. In the weakly nonlinear regime, we observe energy spreading only due to the coupling of the two DoFs (per site), which is in contrast to what is known for KG lattices with a single DoF per lattice site, where the spreading occurs due to chaoticity. Additionally, for strong nonlinearities, we show that initially localized wave-packets attain near ballistic behavior in contrast to other known models. We also reveal persistent chaos during energy spreading, although its strength decreases in time as quantified by the evolution of the system's finite-time maximum Lyapunov exponent. Our results show that flexible, disordered, and strongly nonlinear lattices are a viable platform to study energy transport in combination with multiple DoFs (per site), also present an alternative way to control energy spreading in heterogeneous media.

Durable Pneumatic Artificial Muscles with Electric Conductivity for Reliable Physical Reservoir Computing

A McKibben-type pneumatic artificial muscle (PAM) is a soft actuator that is widely used in soft robotics, and it generally exhibits complex material dynamics with nonlinearity and hysteresis. In this letter, we propose an extremely durable PAM containing carbon black aggregates and show that its dynamics can be used as a computational resource based on the framework of physical reservoir computing (PRC). By monitoring the information processing capacity of our PAM, we verified that its computational performance will not degrade even if it is randomly actuated more than one million times, which indicates extreme durability. Furthermore, we demonstrate that the sensing function can be outsourced to the soft material dynamics itself without external sensors based on the framework of PRC. Our study paves the way toward reliable information processing powered by soft material dynamics.

Local Discrimination Based on Piezoelectric Sensing in Robots Composed of Soft Matter with Different Physical Properties

The coronavirus epidemic has attracted significant attention to the applications of pet robots which can be used to treat and entertain people in their homes. However, pet robots are fabricated using hard materials and it is difficult for them to communicate with people through contact. Soft robots are expected to realize communication through contact similar to that of actual pets. Soft robots provide people with a sense of healing and security owing to their softness and can extract rich information through external stimuli by applying a machine learning framework called physical-reservoir computing. It is crucial to determine the differences between the physical properties of soft materials that affect the information extracted from a soft body to develop an intelligent soft robot. In this study, two owl-shaped soft robots with different softnesses were developed to analyze the characteristics of the signal data obtained via piezoelectric film sensors embedded in models with different physical properties. An accuracy of 94.2% and 95.9% was obtained for touched part classification using 1D CNN and logistic regression models, respectively. Additionally, the relationship between the softness of material and classification performance was investigated by comparing the distribution of part classification accuracy for different hyper-parameters of two owl models.

Dynamic effects on reservoir computing with a Hopf oscillator

Limit cycle oscillators have the potential to be resourced as reservoir computers due to their rich dynamics. Here, a Hopf oscillator is used as a physical reservoir computer by discarding the delay line and time-multiplexing procedure. A parametric study is used to uncover computational limits imposed by the dynamics of the oscillator using parity and chaotic time-series prediction benchmark tasks. Resonance, frequency ratios from the Farey sequence, and Arnold tongues were found to strongly affect the computation ability of the reservoir. These results provide insights into fabricating physical reservoir computers from limit cycle systems.

Abstract Reservoir Computing

Noise of any kind can be an issue when translating results from simulations to the real world. We suddenly have to deal with building tolerances, faulty sensors, or just noisy sensor readings. This is especially evident in systems with many free parameters, such as the ones used in physical reservoir computing. By abstracting away these kinds of noise sources using intervals, we derive a regularized training regime for reservoir computing using sets of possible reservoir states. Numerical simulations are used to show the effectiveness of our approach against different sources of errors that can appear in real-world scenarios and compare them with standard approaches. Our results support the application of interval arithmetics to improve the robustness of mass-spring networks trained in simulations.

Photonic reinforcement learning based on optoelectronic reservoir computing

Reinforcement learning has been intensively investigated and developed in artificial intelligence in the absence of training data, such as autonomous driving vehicles, robot control, internet advertising, and elastic optical networks. However, the computational cost of reinforcement learning with deep neural networks is extremely high and reducing the learning cost is a challenging issue. We propose a photonic on-line implementation of reinforcement learning using optoelectronic delay-based reservoir computing, both experimentally and numerically. In the proposed scheme, we accelerate reinforcement learning at a rate of several megahertz because there is no required learning process for the internal connection weights in reservoir computing. We perform two benchmark tasks, CartPole-v0 and MountanCar-v0 tasks, to evaluate the proposed scheme. Our results represent the first hardware implementation of reinforcement learning based on photonic reservoir computing and pave the way for fast and efficient reinforcement learning as a novel photonic accelerator.

Towards enduring autonomous robots via embodied energy

Autonomous robots comprise actuation, energy, sensory and control systems built from materials and structures that are not necessarily designed and integrated for multifunctionality. Yet, animals and other organisms that robots strive to emulate contain highly sophisticated and interconnected systems at all organizational levels, which allow multiple functions to be performed simultaneously. Herein, we examine how system integration and multifunctionality in nature inspires a new paradigm for autonomous robots that we call Embodied Energy. Whereas most untethered robots use batteries to store energy and power their operation, recent advancements in energy-storage techniques enable chemical or electrical energy sources to be embodied directly within the structures and materials used to create robots, rather than requiring separate battery packs. This perspective highlights emerging examples of Embodied Energy in the context of developing autonomous robots.