Fig 5 - uploaded by Kazuya Horibe
Content may be subject to copyright.
Recovery of locomotion. The regrown morphologies are shown semi-translucent. From left to right: Biped, Tripod, and Multipod.
Source publication
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...
Context in source publication
Context 1
... comparison, we then measured the locomotion of the original, damaged, and regrown morphology with an evaluation time of 0.5s for 10 cycles in VoxCad. The ratio of regrowth and travel distance to the original morphology are shown in Table 1 and its locomotion in Fig. 5. The damaged Biped maintained 67% of its original locomotion ability; it replicated a similar locomotion pattern to the one observed in the L-Walker. As the Tripod lost one of its three legs, it was incapable of successful locomotion. Furthermore, the Multiped lost all locomotion -the robot simply collapsed at the starting ...
Similar publications
Recent efforts in stimuli‐responsive soft materials have enabled wirelessly controlled actuation with increasing degrees of freedom, yielding miniature robots capable of various locomotion in open environments such as on a plane or inside fluids. However, grand challenges remain in harnessing photomechanical deformation to induce locomotion and con...
Recently, owing to the significant advancement in the field of soft robots, various studies have been conducted on soft actuators. Among the several soft actuators, the double helix twisted and coiled actuator (DTCA), which is fabricated by coiling spandex and nylon under twisted conditions, exhibits considerable performance in terms of large contr...
Wearable ultraviolet (UV) detectors have attracted considerable interest in the military and civilian realms. However, semiconductor-based UV detectors are easily interfered by elongation due to the elastic modulus incompatibility between rigid semiconductors and polymer matrix. Polymer detectors containing UV responsive moieties seriously suffer f...
Currently soft robots primarily rely on pneumatics and geometrical asymmetry to achieve locomotion, which limits their working range, versatility, and other untethered functionalities. In this paper, we introduce a novel approach to achieve locomotion for soft robots through dynamically tunable friction to address these challenges, which is achieve...
Citations
... Originally introduced as mathematical models to simulate the intricacies of natural systems, including the one discussed in this article, CA operates on the fundamental concept of state transformations among neighboring cells (Alexan et al., 2022). The system's evolution is contingent on transition rules, which can be either deterministic or probabilistic (Horibe et al., 2021). ...
The Cerrado biome in Brazil plays a vital role in preserving biodiversity, providing essential ecosystem services, and supporting agriculture, making it a crucial and valuable natural resource. Sete Cidades National Park stands out for its rock formations, 10,000-year-old cave paintings and its Cerrado vegetation. The Cerrado is known for being a pyrobiome, so its patrolling becomes essential. In this context, this paper introduces a novel approach to swarm robotics patrolling in the unique ecosystem of the Sete Cidades National Park, located within the Cerrado biome. The study presents three distinct cellular automata models designed for the task, aiming to enhance the efficiency and coverage of patrolling efforts. The key difference between our model presented in this research and the previous one is our focus on a map that encompasses diverse vegetation types, specifically designed to represent the Sete Cidades National Park, with a primary goal of monitoring forest fires. The results demonstrated that the best-performing model was the Forest Tabu Inverted Ant Cellular Automata, which achieved an average of 21.77 complete patrol cycles with 95% confidence. This outcome was obtained using three robots, a tabu queue of |Q| = 80, and a maximum pheromone per cell equals to ρ = 103. These parameters highlight the efficacy of this model in optimizing patrol cycles and the efficient use of resources for environmental surveillance in the Sete Cidades National Park, particularly in the context of fire prevention.
... An agent is defined as a system or entity that interacts directly with its surrounding environment through its physical embodiment to perform tasks and achieve objectives. For self-evolving repair [21], soft robots might be the key to achieving this advanced skill, which can significantly enhance the durability and adaptability of robots in terms of embodied intelligence. For mimicking soft biological sensing mechanisms [22], organisms perceive the external environment through organs like skin and tentacles, integrating pressure sensors and temperature sensors to achieve multimodal sensing for a more comprehensive perception and response to environmental changes. ...
Soft robotics is closely related to embodied intelligence in the joint exploration of the means to achieve more natural and effective robotic behaviors via physical forms and intelligent interactions. Embodied intelligence emphasizes that intelligence is affected by the synergy of the brain, body, and environment, focusing on the interaction between agents and the environment. Under this framework, the design and control strategies of soft robotics depend on their physical forms and material properties, as well as algorithms and data processing, which enable them to interact with the environment in a natural and adaptable manner. At present, embodied intelligence has comprehensively integrated related research results on the evolution, learning, perception, decision making in the field of intelligent algorithms, as well as on the behaviors and controls in the field of robotics. From this perspective, the relevant branches of the embodied intelligence in the context of soft robotics were studied, covering the computation of embodied morphology; the evolution of embodied AI; and the perception, control, and decision making of soft robotics. Moreover, on this basis, important research progress was summarized, and related scientific problems were discussed. This study can provide a reference for the research of embodied intelligence in the context of soft robotics.
... CA can simulate complex systems with realism, replacing numerical simulations from differential equations. By using a transition rule, either deterministic or probabilistic, the state transformation of CA cells can simulate the dynamic action of different systems [16]. CA has been applied in various contexts, including robotics [11,12], disease modeling [14,17,18], cryptography [19][20][21], and even modeling fires [15] or pedestrian evacuation in panic [22][23][24]. ...
Wildfires pose a significant threat to both biodiversity and human communities, and understanding their behavior and the rate at which they burn through different vegetation types is crucial for effective management and conservation. In this research, we present a comprehensive analysis of wildfire behavior and vegetation burning rates in the unique ecosystem of Sete Cidades National Park. To achieve this, we adopt a qualiquantitative approach that combines both qualitative and quantitative methodologies, considering the multifaceted variables at play, including wind conditions, various vegetation types, and the dynamics of fire progression. We conducted an extensive dataset comprising $ 100 $ simulations for each of three distinct scenarios, ensuring robustness in our data for statistical analysis. By incorporating qualitative data obtained through field observations and expert opinions, we gain a deeper understanding of the contextual nuances specific to Sete Cidades National Park. This approach enriches the interpretation of our quantitative results, providing valuable context and real-world relevance. Our materials include a cellular automaton lattice with $ 200 \times 200 $ cells, representing the diverse landscape of the study area. We used MATLAB to visualize this landscape, generating distinct representations of the scenarios. Our findings reveal the distribution of different vegetation types across these scenarios, emphasizing the resilience of Rupestrian Cerrado, the diversity of Typical Cerrado, and the importance of Riparian Forest in preserving aquatic ecosystems. This research contributes to the broader understanding of wildfire management, considering the interdisciplinary aspects of environmental science, forestry, and meteorology. By integrating knowledge from diverse fields, we provide a holistic analysis that can inform effective conservation strategies and wildfire management practices.
... The introduction of neural networks (NNs) in cellular automata allowed for complex update rules that can be learned. This type of CA has started to emerge and be introduced in different areas such as texture generation and regeneration [11], [12], 2D or 3D models growth and regeneration [13], [14], [15], [16], areas that were impossible to tackle with hand-written rules. ...
Neural cellular automata have been proven effective in simulating morphogenetic processes. Developing such automata has been applied in 2D and 3D processes related to creating and regenerating complex structures and enabling their behaviors. However, neural cellular automata are inherently uncontrollable after the training process. Starting from a neural cellular automaton trained to generate a given shape from one living cell, this paper aims to gain insight into the behavior of the automaton, and to analyze the influence of the different image characteristics on the training and stabilization process and its shortcomings in different scenarios. For each considered shape, the automaton is trained on one RGB image of size 72 × 72 pixels containing the shape on an uniform white background, in which each pixel represents a cell. The evolution of the automaton starts from one living cell, employing a shallow neural network for the update rule, followed by backpropagation after a variable number of evolutionary steps. We studied the behavior of the automaton and the way in which different components like symmetry, orientation and colours of the shape influence its growth and alteration after a number of epochs and discussed this thoroughly in the experimental section of the paper. We further discuss a pooling strategy, used to stabilize the model and illustrate the influence of this pooling on the training process. The benefits of this strategy are compared to the original model and the behavior of the automaton during its evolution is studied in detail. Finally, we compare the results of models using different filters in the first stage of feature selection. The main results of our study are the insights gained into how the neural cellular automaton works, what it is actually learning, and what influence this learning, as there are observable result differences depending on the characteristics of the input images and the filters used in the model.
... VSRs, the modular soft robots we employ in this work, naturally fit this scheme, and hence constitute a relevant test bed for assessing the body-brain interplay in a controlled simulated environment. Moreover, they are particularly well-suited for studying how real-world phenomena can be ported in the artificial domain, as, e.g., development [13,14] or morphological regeneration [15]. ...
... The desired robots are supposed to reconfigure their morphology in response to any damage in their components. Such devices can be best described as synthetic living machines [106,229]. ...
Although the currently available pharmacological assays can cure most pathological disorders, they have limited therapeutic value in relieving certain disorders like myocardial infarct, peripheral vascular disease, amputated limbs, or organ failure (e.g. renal failure). Pilot studies to overcome such problems using regenerative medicine (RM) delivered promising data. Comprehensive investigations of RM in zebrafish or reptilians are necessary for better understanding. However, the precise mechanisms remain poorly understood despite the tremendous amount of data obtained using the zebrafish model investigating the exact mechanisms behind their regenerative capability. Indeed, understanding such mechanisms and their application to humans can save millions of lives from dying due to potentially life-threatening events. Recent studies have launched a revolution in replacing damaged human organs via different approaches in the last few decades. The newly established branch of medicine (known as Regenerative Medicine aims to enhance natural repair mechanisms. This can be done through the application of several advanced broad-spectrum technologies such as organ transplantation, tissue engineering, and application of Scaffolds technology (support vascularization using an extracellular matrix), stem cell therapy, miRNA treatment, development of 3D mini-organs (organoids), and the construction of artificial tissues using nanomedicine and 3D bio-printers. Moreover, in the next few decades, revolutionary approaches in regenerative medicine will be applied based on artificial intelligence and wireless data exchange, soft intelligence biomaterials, nanorobotics, and even living robotics capable of self-repair. The present work presents a comprehensive overview that summarizes the new and future advances in the field of RM.
... Embodied intelligence can be viewed as the enterprise of determining how to change not only one's behavior to adapt to changing circumstances but one's body as well. Soft materials incorporated into autonomous machines provides a diverse set of options for morphological change, including change in shape [16], material properties [18,17,19], numerosity [14,12], and mass [13,6,7]. However, such plasticity provides avenues for incorporating physical change into machines. ...
The histories of the Artificial Intelligence and robotics communities have been primarily those of bioinspiration: observing, distilling, and instantiating a wealth of biological structures and mechanisms into machines. However, this transferal from biology to technology has been uneven: the machine learning community has focused on non-physical mechanisms, like synaptic plasticity, while the robotics community has focused on physical structures, like the quadrupedal body plan. Organisms however offer a wealth of physical mechanisms, like bones that strengthen in response to load or tissues that regenerate in response to amputation, that have yet to be instantiated in machines. Biological robots—machines built entirely from biological tissue–exhibit many of these morphological mechanisms ‘for free’, suggesting that biobots could be designed in future to exaggerate these adaptive mechanisms, or provide hints for how they can be incorporated into morphologically self-editing robots.
... Minimizing the need for communication among modules would permit to disassemble robots and to reassemble them differently, to cope with different tasks. Early controllers for VSRs relied on a single centralized neural network [65] or fixed, rigid intermodule communication between voxels [28,44], effectively making their design closer to a single, large neural controller with shared weights. This rigid Message-Passing (MP) mechanism, while practical, makes the modules not perfectly interchangeable, thus limiting their flexibility. ...
... Inspired by multi-cellular systems, Joachimczak et al. [31] evolved soft-bodied animats in both aquatic and terrestrial environments, showcasing the concept of metamorphosis in simulation. In another soft robot simulation work, partially damaged robots regenerated their original morphology through local cell interactions in a neural CA system [28]. Moreover, Sudhakaran et al. [63] grew complex functional entities in Minecraft through a neural CA-based morphogenetic process. ...
... While the mechanisms behind natural morphological regeneration are still not fully understood, artificial methods such as neural cellular automata have proven successful tools for their modelling. For instance, the work presented in this paper is an extension of our previous conference paper [9], where we evolve neural cellular automata to grow locomoting soft robots capable of some morphological regeneration. Neural cellular automata are based on the simpler cellular automata. ...
... An overview of the approach is shown in Figure 1. As before [9], we first use an evolutionary algorithm to discover neural cellular automata that can grow the control and morphology of a diversity of soft voxel-based robots. The NCA for a particular robot is then efficiently trained through differentiable programming (i.e. ...
... The morphologies of each of these three robots are shown in Fig. 5a and the locomotion patterns in Fig. 4d. We use two different approaches to train regeneration NCAs, the first using an evolutionary approach as previously reported in [9]. The second approach is based on gradient-based optimization, i.e. differentiable programming. ...
Biological systems are very robust to morphological damage, but artificial systems (robots) are currently not. In this paper we present a system based on neural cellular automata, in which locomoting robots are evolved and then given the ability to regenerate their morphology from damage through gradient-based training. Our approach thus combines the benefits of evolution to discover a wide range of different robot morphologies, with the efficiency of supervised training for robustness through differentiable update rules. The resulting neural cellular automata are able to grow virtual robots capable of regaining more than 80% of their functionality, even after severe types of morphological damage.
... shape of the cellular automaton and a pre-specified target shape [8,18,28,31,37,40]. ...
Information-theoretic fitness functions are becoming increasingly popular to produce generally useful, task-independent behaviors. One such universal function, dubbed empowerment, measures the amount of control an agent exerts on its environment via its sensorimotor system. Specifically, empowerment attempts to maximize the mutual information between an agent's actions and its received sensor states at a later point in time. Traditionally, empowerment has been applied to a conventional sensorimotor apparatus, such as a robot. Here, we expand the approach to a distributed, multi-agent sensorimotor system embodied by a neural cellular automaton (NCA). We show that the addition of empowerment as a secondary objective in the evolution of NCA to perform the task of morphogenesis, growing and maintaining a pre-specified shape, results in higher fitness compared to evolving for morphogenesis alone. Results suggest there may be a synergistic relationship between morphogenesis and empowerment. That is, indirectly selecting for coordination between neighboring cells over the duration of development is beneficial to the developmental process itself. Such a finding may have applications in developmental biology by providing potential mechanisms of communication between cells during growth from a single cell to a multicellular, target morphology. Source code for the experiments in this paper can be found at: \url{https://github.com/caitlingrasso/empowered-nca}.
