Frontiers in Robotics and AI

RSS Feed for Frontiers in Robotics and AI | New and Recent Articles
Subscribe to Frontiers in Robotics and AI feed

Self-organization offers a promising approach for designing adaptive systems. Given the inherent complexity of most cyber-physical systems, adaptivity is desired, as predictability is limited. Here I summarize different concepts and approaches that can facilitate self-organization in cyber-physical systems, and thus be exploited for design. Then I mention real-world examples of systems where self-organization has managed to provide solutions that outperform classical approaches, in particular related to urban mobility. Finally, I identify when a centralized, distributed, or self-organizing control is more appropriate.

We consider the problem of autonomous acquisition of manipulation skills where problem-solving strategies are initially available only for a narrow range of situations. We propose to extend the range of solvable situations by autonomous play with the object. By applying previously-trained skills and behaviors, the robot learns how to prepare situations for which a successful strategy is already known. The information gathered during autonomous play is additionally used to train an environment model. This model is exploited for active learning and the generation of novel preparatory behaviors compositions. We apply our approach to a wide range of different manipulation tasks, e.g., book grasping, grasping of objects of different sizes by selecting different grasping strategies, placement on shelves, and tower disassembly. We show that the composite behavior generation mechanism enables the robot to solve previously-unsolvable tasks, e.g., tower disassembly. We use success statistics gained during real-world experiments to simulate the convergence behavior of our system. Simulation experiments show that the learning speed can be improved by around 30% by using active learning.

In swarm robotics multiple robots collectively solve problems by forming advantageous structures and behaviors similar to the ones observed in natural systems, such as swarms of bees, birds, or fish. However, the step to industrial applications has not yet been made successfully. Literature is light on real-world swarm applications that apply actual swarm algorithms. Typically, only parts of swarm algorithms are used which we refer to as basic swarm behaviors. In this paper we collect and categorize these behaviors into spatial organization, navigation, decision making, and miscellaneous. This taxonomy is then applied to categorize a number of existing swarm robotic applications from research and industrial domains. Along with the classification, we give a comprehensive overview of research platforms that can be used for testing and evaluating swarm behavior, systems that are already on the market, and projects that target a specific market. Results from this survey show that swarm robotic applications are still rare today. Many industrial projects still rely on centralized control, and even though a solution with multiple robots is employed, the principal idea of swarm robotics of distributed decision making is neglected. We identified mainly following reasons: First of all, swarm behavior emerging from local interactions is hard to predict and a proof of its eligibility for applications in an industrial context is difficult to provide. Second, current communication architectures often do not match requirements for swarm communication, which often leads to a system with a centralized communication infrastructure. Finally, testing swarms for real industrial applications is an issue, since deployment in a productive environment is typically too risky and simulations of a target system may not be sufficiently accurate. In contrast, the research platforms present a means for transforming swarm robotics solutions from theory to prototype industrial systems.

Modeling of complex adaptive systems has revealed a still poorly understood benefit of unsupervised learning: when neural networks are enabled to form an associative memory of a large set of their own attractor configurations, they begin to reorganize their connectivity in a direction that minimizes the coordination constraints posed by the initial network architecture. This self-optimization process has been replicated in various neural network formalisms, but it is still unclear whether it can be applied to biologically more realistic network topologies and scaled up to larger networks. Here we continue our efforts to respond to these challenges by demonstrating the process on the connectome of the widely studied nematode worm C. elegans. We extend our previous work by considering the contributions made by hierarchical partitions of the connectome that form functional clusters, and we explore possible beneficial effects of inter-cluster inhibitory connections. We conclude that the self-optimization process can be applied to neural network topologies characterized by greater biological realism, and that long-range inhibitory connections can facilitate the generalization capacity of the process.

The present work is a collaborative research aimed at testing the effectiveness of the robot-assisted intervention administered in real clinical settings by real educators. Social robots dedicated to assisting persons with autism spectrum disorder (ASD) are rarely used in clinics. In a collaborative effort to bridge the gap between innovation in research and clinical practice, a team of engineers, clinicians and researchers working in the field of psychology developed and tested a robot-assisted educational intervention for children with low-functioning ASD (N = 20) A total of 14 lessons targeting requesting and turn-taking were elaborated, based on the Pivotal Training Method and principles of Applied Analysis of Behavior. Results showed that sensory rewards provided by the robot elicited more positive reactions than verbal praises from humans. The robot was of greatest benefit to children with a low level of disability. The educators were quite enthusiastic about children's progress in learning basic psychosocial skills from interactions with the robot. The robot nonetheless failed to act as a social mediator, as more prosocial behaviors were observed in the control condition, where instead of interacting with the robot children played with a ball. We discuss how to program robots to the distinct needs of individuals with ASD, how to harness robots' likability in order to enhance social skill learning, and how to arrive at a consensus about the standards of excellence that need to be met in interdisciplinary co-creation research. Our intuition is that robotic assistance, obviously judged as to be positive by educators, may contribute to the dissemination of innovative evidence-based practice for individuals with ASD.

Brain signals represent a communication modality that can allow users of assistive robots to specify high-level goals, such as the object to fetch and deliver. In this paper, we consider a screen-free Brain-Computer Interface (BCI), where the robot highlights candidate objects in the environment using a laser pointer, and the user goal is decoded from the evoked responses in the electroencephalogram (EEG). Having the robot present stimuli in the environment allows for more direct commands than traditional BCIs that require the use of graphical user interfaces. Yet bypassing a screen entails less control over stimulus appearances. In realistic environments, this leads to heterogeneous brain responses for dissimilar objects—posing a challenge for reliable EEG classification. We model object instances as subclasses to train specialized classifiers in the Riemannian tangent space, each of which is regularized by incorporating data from other objects. In multiple experiments with a total of 19 healthy participants, we show that our approach not only increases classification performance but is also robust to both heterogeneous and homogeneous objects. While especially useful in the case of a screen-free BCI, our approach can naturally be applied to other experimental paradigms with potential subclass structure.

We present a reinforcement learning-based (RL) control scheme for trajectory tracking of fully-actuated surface vessels. The proposed method learns online both a model-based feedforward controller, as well an optimizing feedback policy in order to follow a desired trajectory under the influence of environmental forces. The method's efficiency is evaluated via simulations and sea trials, with the unmanned surface vehicle (USV) ReVolt performing three different tracking tasks: The four corner DP test, straight-path tracking and curved-path tracking. The results demonstrate the method's ability to accomplish the control objectives and a good agreement between the performance achieved in the Revolt Digital Twin and the sea trials. Finally, we include an section with considerations about assurance for RL-based methods and where our approach stands in terms of the main challenges.

Motor skill learning of dental implantation surgery is difficult for novices because it involves fine manipulation of different dental tools to fulfill a strictly pre-defined procedure. Haptics-enabled virtual reality training systems provide a promising tool for surgical skill learning. In this paper, we introduce a haptic rendering algorithm for simulating diverse tool-tissue contact constraints during dental implantation. Motion forms of an implant tool can be summarized as the high degree of freedom (H-DoF) motion and the low degree of freedom (L-DoF) motion. During the H-DoF state, the tool can move freely on bone surface and in free space with 6 DoF. While during the L-DoF state, the motion degrees are restrained due to the constraints imposed by the implant bed. We propose a state switching framework to simplify the simulation workload by rendering the H-DoF motion state and the L-DoF motion state separately, and seamless switch between the two states by defining an implant criteria as the switching judgment. We also propose the virtual constraint method to render the L-DoF motion, which are different from ordinary drilling procedures as the tools should obey different axial constraint forms including sliding, drilling, screwing and perforating. The virtual constraint method shows efficiency and accuracy in adapting to different kinds of constraint forms, and consists of three core steps, including defining the movement axis, projecting the configuration difference, and deriving the movement control ratio. The H-DoF motion on bone surface and in free space is simulated through the previously proposed virtual coupling method. Experimental results illustrated that the proposed method could simulate the 16 different phases of the complete implant procedures of the Straumann® Bone Level(BL) Implants Φ4.8–L12 mm. According to the output force curve, different contact constraints could be rendered with steady and continuous output force during the operation procedures.

Illusory ownership can be induced in a virtual body by visuo-motor synchrony. Our aim was to test the possibility of a re-association of the right thumb with a virtual left arm and express the illusory body ownership of the re-associated arm through a synchronous or asynchronous movement of the body parts through action and vision. Participants felt that their right thumb was the virtual left arm more strongly in the synchronous condition than in the asynchronous one, and the feeling of ownership of the virtual arm was also stronger in the synchronous condition. We did not find a significant difference in the startle responses to a sudden knife appearance to the virtual arm between the two synchrony conditions, as there was no proprioceptive drift of the thumb. These results suggest that a re-association of the right thumb with the virtual left arm could be induced by visuo-motor synchronization; however, it may be weaker than the natural association.

In this experiment, we aimed to measure the conscious internal representation of one's body appearance and allow the participants to compare this to their ideal body appearance and to their real body appearance. We created a virtual representation of the internal image participants had of their own body shape. We also created a virtual body corresponding to the internal representation they had of their ideal body shape, and we built another virtual body based on their real body measures. Participants saw the three different virtual bodies from an embodied first-person perspective and from a third-person perspective and had to evaluate the appearance of those virtual bodies. We observed that female participants evaluated their real body as more attractive when they saw it from a third-person perspective, and that their level of body dissatisfaction was lower after the experimental procedure. We believe that third-person perspective allowed female participants to perceive their real body shape without applying the negative prior beliefs usually associated to the “self”, and that this resulted in a more positive evaluation of their body shape. We speculate that this method could be applied with patients suffering from eating disorders, by making their body perception more realistic and therefore improve their body satisfaction.

The real world is highly variable and unpredictable, and so fine-tuned robot controllers that successfully result in group-level “emergence” of swarm capabilities indoors may quickly become inadequate outside. One response to unpredictability could be greater robot complexity and cost, but this seems counter to the “swarm philosophy” of deploying (very) large numbers of simple agents. Instead, here I argue that bioinspiration in swarm robotics has considerable untapped potential in relation to the phenomenon of phenotypic plasticity: when a genotype can produce a range of distinctive changes in organismal behavior, physiology and morphology in response to different environments. This commonly arises following a natural history of variable conditions; implying the need for more diverse and hazardous simulated environments in offline, pre-deployment optimization of swarms. This will generate—indicate the need for—plasticity. Biological plasticity is sometimes irreversible; yet this characteristic remains relevant in the context of minimal swarms, where robots may become mass-producible. Plasticity can be introduced through the greater use of adaptive threshold-based behaviors; more fundamentally, it can link to emerging technologies such as smart materials, which can adapt form and function to environmental conditions. Moreover, in social animals, individual heterogeneity is increasingly recognized as functional for the group. Phenotypic plasticity can provide meaningful diversity “for free” based on early, local sensory experience, contributing toward better collective decision-making and resistance against adversarial agents, for example. Nature has already solved the challenge of resilient self-organisation in the physical realm through phenotypic plasticity: swarm engineers can follow this lead.

Predictions and predictive knowledge have seen recent success in improving not only robot control but also other applications ranging from industrial process control to rehabilitation. A property that makes these predictive approaches well-suited for robotics is that they can be learned online and incrementally through interaction with the environment. However, a remaining challenge for many prediction-learning approaches is an appropriate choice of prediction-learning parameters, especially parameters that control the magnitude of a learning machine's updates to its predictions (the learning rates or step sizes). Typically, these parameters are chosen based on an extensive parameter search—an approach that neither scales well nor is well-suited for tasks that require changing step sizes due to non-stationarity. To begin to address this challenge, we examine the use of online step-size adaptation using the Modular Prosthetic Limb: a sensor-rich robotic arm intended for use by persons with amputations. Our method of choice, Temporal-Difference Incremental Delta-Bar-Delta (TIDBD), learns and adapts step sizes on a feature level; importantly, TIDBD allows step-size tuning and representation learning to occur at the same time. As a first contribution, we show that TIDBD is a practical alternative for classic Temporal-Difference (TD) learning via an extensive parameter search. Both approaches perform comparably in terms of predicting future aspects of a robotic data stream, but TD only achieves comparable performance with a carefully hand-tuned learning rate, while TIDBD uses a robust meta-parameter and tunes its own learning rates. Secondly, our results show that for this particular application TIDBD allows the system to automatically detect patterns characteristic of sensor failures common to a number of robotic applications. As a third contribution, we investigate the sensitivity of classic TD and TIDBD with respect to the initial step-size values on our robotic data set, reaffirming the robustness of TIDBD as shown in previous papers. Together, these results promise to improve the ability of robotic devices to learn from interactions with their environments in a robust way, providing key capabilities for autonomous agents and robots.

This paper presents the design of an assessment process and its outcomes to investigate the impact of Educational Robotics activities on students' learning. Through data analytics techniques, the authors will explore the activities' output from a pedagogical and quantitative point of view. Sensors are utilized in the context of an Educational Robotics activity to obtain a more effective robot–environment interaction. Pupils work on specific exercises to make their robot smarter and to carry out more complex and inspirational projects: the integration of sensors on a robotic prototype is crucial, and learners have to comprehend how to use them. In the presented study, the potential of Educational Data Mining is used to investigate how a group of primary and secondary school students, using visual programming (Lego Mindstorms EV3 Education software), design programming sequences while they are solving an exercise related to an ultrasonic sensor mounted on their robotic artifact. For this purpose, a tracking system has been designed so that every programming attempt performed by students' teams is registered on a log file and stored in an SD card installed in the Lego Mindstorms EV3 brick. These log files are then analyzed using machine learning techniques (k-means clustering) in order to extract different patterns in the creation of the sequences and extract various problem-solving pathways performed by students. The difference between problem-solving pathways with respect to an indicator of early achievement is studied.

This article reports on two studies that aimed to evaluate the effective impact of educational robotics in learning concepts related to Physics and Geography. The reported studies involved two courses from an upper secondary school and two courses from a lower secondary school. Upper secondary school classes studied topics of motion physics, and lower secondary school classes explored issues related to geography. In each grade, there was an “experimental group” that carried out their study using robotics and cooperative learning and a “control group” that studied the same concepts without robots. Students in both classes were subjected to tests before and after the robotics laboratory, to check their knowledge in the topics covered. Our initial hypothesis was that classes involving educational robotics and cooperative learning are more effective in improving learning and stimulating the interest and motivation of students. As expected, the results showed that students in the experimental groups had a far better understanding of concepts and higher participation to the activities than students in the control groups.

In this article we investigate the role of interactive haptic-enabled tangible robots in supporting the learning of cursive letter writing for children with attention and visuomotor coordination issues. We focus on the two principal aspects of handwriting that are linked to these issues: Visual perception and visuomotor coordination. These aspects, respectively, enhance two features of letter representation in the learner's mind in particular, namely the shape (grapheme) and the dynamics (ductus) of the letter, which constitute the central learning goals in our activity. Building upon an initial design tested with 17 healthy children in a preliminary school, we iteratively ported the activity to an occupational therapy context in 2 different therapy centers, in the context of 3 different summer school camps involving a total of 12 children having writing difficulties. The various iterations allowed us to uncover insights about the design of robot-enhanced writing activities for special education, specifically highlighting the importance of ease of modification of the duration of an activity as well as of adaptable frequency, content, flow and game-play and of providing a range of evaluation test alternatives. Results show that the use of robot-assisted handwriting activities could have a positive impact on the learning of the representation of letters in the context of occupational therapy (V = 1, 449, p < 0.001, r = 0.42). Results also highlight how the design changes made across the iterations affected the outcomes of the handwriting sessions, such as the evaluation of the performances, monitoring of the performances, and the connectedness of the handwriting.

Robots are promising tools for promoting engagement of autistic children in interventions and thereby increasing the amount of learning opportunities. However, designing deliberate robot behavior aimed at engaging autistic children remains challenging. Our current understanding of what interactions with a robot, or facilitated by a robot, are particularly motivating to autistic children is limited to qualitative reports with small sample sizes. Translating insights from these reports to design is difficult due to the large individual differences among autistic children in their needs, interests, and abilities. To address these issues, we conducted a descriptive study and report on an analysis of how 31 autistic children spontaneously interacted with a humanoid robot and an adult within the context of a robot-assisted intervention, as well as which individual characteristics were associated with the observed interactions. For this analysis, we used video recordings of autistic children engaged in a robot-assisted intervention that were recorded as part of the DE-ENIGMA database. The results showed that the autistic children frequently engaged in exploratory and functional interactions with the robot spontaneously, as well as in interactions with the adult that were elicited by the robot. In particular, we observed autistic children frequently initiating interactions aimed at making the robot do a certain action. Autistic children with stronger language ability, social functioning, and fewer autism spectrum-related symptoms, initiated more functional interactions with the robot and more robot-elicited interactions with the adult. We conclude that the children's individual characteristics, in particular the child's language ability, can be indicative of which types of interaction they are more likely to find interesting. Taking these into account for the design of deliberate robot behavior, coupled with providing more autonomy over the robot's behavior to the autistic children, appears promising for promoting engagement and facilitating more learning opportunities.

Many insect species, and even some vertebrates, assemble their bodies to form multi-functional materials that combine sensing, computation, and actuation. The tower-building behavior of red imported fire ants, Solenopsis invicta, presents a key example of this phenomenon of collective construction. While biological studies of collective construction focus on behavioral assays to measure the dynamics of formation and studies of swarm robotics focus on developing hardware that can assemble and interact, algorithms for designing such collective aggregations have been mostly overlooked. We address this gap by formulating an agent-based model for collective tower-building with a set of behavioral rules that incorporate local sensing of neighboring agents. We find that an attractive force makes tower building possible. Next, we explore the trade-offs between attraction and random motion to characterize the dynamics and phase transition of the tower building process. Lastly, we provide an optimization tool that may be used to design towers of specific shapes, mechanical loads, and dynamical properties, such as mechanical stability and mobility of the center of mass.

In this paper we describe the control approaches tested in the improved version of an existing soft robotic neck with two Degrees Of Freedom (DOF), able to achieve flexion, extension, and lateral bending movements similar to those of a human neck. The design is based on a cable-driven mechanism consisting of a spring acting as a cervical spine and three servomotor actuated tendons that let the neck to reach all desired postures. The prototype was manufactured using a 3D printer. Two control approaches are proposed and tested experimentally: a motor position approach using encoder feedback and a tip position approach using Inertial Measurement Unit (IMU) feedback, both applying fractional-order controllers. The platform operation is tested for different load configurations so that the robustness of the system can be checked.

This study aimed to investigate whether using a wearable robot applying interactive rhythmic stimulation on the upper limbs of patients with Parkinson's disease (PD) could affect their gait. The wearable robot presented tactile stimuli on the patients' upper limbs, which was mutually synchronized with the swing of their upper limbs. We conducted an evaluation experiment with PD patients (n = 30, Modified Hoehn-Yahr = 1–3, on-state) to investigate the assistance effect by the robot and the immediate after-effect of intervention. The participants were instructed to walk 30 m under four different conditions: (1) not wearing the robot before the intervention (Pre-condition), (2) wearing the robot without the rhythm assistance (RwoA condition), (3) wearing the robot with rhythm assistance (RwA condition), and (4) not wearing the robot immediately after the intervention (Post-condition). These conditions were conducted in this order over a single day. The third condition was performed three times and the others, once. The arm swing amplitude, stride length, and velocity were increased in the RwA condition compared to the RwoA condition. The coefficient of variance (CV) of the stride duration was decreased in the RwA condition compared to the RwoA condition. These results revealed that the assistance by the robot increased the gait performance of PD patients. In addition, the stride length and velocity were increased and the stride duration CV was decreased in the Post-condition compared to the Pre-condition. These results show that the effect of robot assistance on the patient's gait remained immediately after the intervention. These findings suggest that synchronized rhythmic stimulation on the upper limbs could influence the gait of PD patients and that the robot may assist with gait rehabilitation in these patients.