Search

Robotic chefs are a promising technology that can bring sizeable health and economic benefits when deployed ubiquitously. This deployment is hindered by the costly process of programming the robots to cook specific dishes while humans learn from observation or freely available videos. In this paper, we propose an algorithm that incrementally adds recipes to the robot’s cookbook based on the visual observation of a human chef, enabling the easier and cheaper deployment of robotic chefs. A new recipe is added only if the current observation is substantially different than all recipes in the cookbook, which is decided by computing the similarity between the vectorizations of these two. The algorithm correctly recognizes known recipes in 93% of the demonstrations and successfully learned new recipes when shown, using off-the-shelf neural networks for computer vision. We show that videos and demonstrations are viable sources of data for robotic chef programming when extended to massive publicly available data sources like YouTub

Soft continuum robots are highly flexible and adapt- able, making them ideal for unstructured environments such as the human body and agriculture. However, their high compliance and manoeuvrability make them difficult to model, sense, and control. Current control strategies focus on Cartesian space control of the end-effector, but few works have explored full-body control. This study presents a novel image-based deep learning approach for closed-loop kinematic shape control of soft continuum robots. The method combines a local inverse kinematics formulation in the image-space with deep convolutional neural networks for accurate shape control that is robust to feedback noise and mechanical changes in the continuum arm. The shape controller is fast and straightforward to implement; it takes only a few hours to generate training data, train the network, and deploy, requiring only a web camera for feedback. This method offers an intuitive and user-friendly way to control the robot’s 3D shape and configuration through teleoperation using only 2D hand-drawn images of the desired target state without the need for further user instruction or consideration of the robot’s kinematics.

Embodiment is a crucial concept for the autonomy and adaptivity of systems working in the physical world with high degrees of uncertainty and complexity. The physical bodies of autonomous adaptive systems heavily influence the information flow from the environment to the central processing (and vice versa), requiring us to consider the full triad of brain, body, and environment to investigate intelligent behavior. This article provides a structured review of embodied intelligence with a special emphasis on the concept of timescales and their role in self-organization and the emergence of complex behavior. We classify embodied interactions into three types—cross-timescale matching, separation, and nontemporal sequences—and discuss how these interactions were studied in the past as well as how they can contribute to the systematic investigation of complex autonomous and adaptive systems in both biological and artificial entities.

Soft robotic sensors have been limited in their applications due to their highly nonlinear time variant behavior. Current studies are either looking into techniques to improve the mechano-electrical properties of these sensors or into modelling algorithms that account for the history of each sensor. Here, we present a method for combining multi-material soft strain sensors to obtain equivalent higher quality sensors; better than each of the individual strain sensors. The core idea behind this work is to use a combination of redundant and disjoint strain sensors to compensate for the time-variant hidden states of a soft-bodied system, to finally obtain the true strain state in a static manner using a learning-based approach. We provide methods to develop these variable sensors and metrics to estimate their dissimilarity and efficacy of each sensor combinations, which can double down as a benchmarking tool for soft robotic sensors. The proposed approach is experimentally validated on a pneumatic actuator with embedded soft strain sensors. Our results show that static data from a combination of nonlinear time variant strain sensors is sufficient to accurately estimate the strain state of a system. © 2016 IEEE.

The mechanical properties of a sensor strongly affect its tactile sensing capabilities. The role of morphology and stiffness on the quality of the tactile data has already been the subject of several studies, which focus mainly on static sensor designs and design methodologies. However, static designs always come with trade-offs: considering stiffness, soft compliant sensors ensure a better contact, but at the price of mechanically filtering and altering the detected signal. Conversely, online adaptable filters can tune their characteristics, becoming softer or stiffer when needed. We propose a magneto-active elastomer filter which, when placed on top of the tactile unit, allows the sensor to change its stiffness on demand. We showcase the advantages provided by online stiffening adaptation in terms of information gained and data structure. Moreover, we illustrate how adaptive stiffening influences classification, using 9 standard machine learning algorithms, and how adaptive stiffening can increase the classification accuracy up to 34% with respect to static stiffness control.

The mechanical properties of a sensor strongly affect its tactile sensing capabilities. By exploiting tactile filters, mechanical structures between the sensing unit and the environment, it is possible to tune the interaction dynamics with the surrounding environment. But how can we design a good tactile filter? Previously, the role of filters' geometry and stiffness on the quality of the tactile data has been the subject of several studies, both implementing static filters and adaptable filters. State-of-the-art works on online adaptive stiffness highlight a crucial role of the filters' mechanical behavior in the structure of the recorded tactile data. However, the relationship between the filter's and the environment's characteristics is still largely unknown. We want to show the effect of the environment's mechanical properties on the structure of the acquired tactile data and the performance of a classification task while testing a wide range of static tactile filters. Moreover, we fabricated the filters using four materials commonly exploited in soft robotics, to merge the gap between tactile sensing and robotic applications. We collected data from the interaction with a standard set of twelve objects of different materials, shapes, and textures, and we analyzed the effect of the filter's material on the structure of such data and the performance of nine common machine learning classifiers, both considering the overall test set and the three individual subsets made by all objects of the same material. We showed that depending on the material of the test objects, there is a drastic change in the performance of the four tested filters, and that the filter that matches the mechanical properties of the environment always outperforms the others., This work was supported by the SMART project, European Union's Horizon 2020 research and innovation under the Marie Sklodowska-Curie (grant agreement ID 860108).

Autonomous dishwasher loading is a benchmark problem in robotics that highlights the challenges of robotic perception, planning, and manipulation in an unstructured environment. Current approaches resort to a specialized solution, however, these technologies are not viable in a domestic setting. Learning‐based solutions seem promising for a general purpose solutions; however, they require large amounts of catered data to be applied in real‐world scenarios. This article presents a novel learning‐based solution without a training phase using pre‐trained object detection networks. By developing a perception, planning, and manipulation framework around an off‐the‐shelf object detection network, we are able to develop robust pick‐and‐place solutions that are easy to develop and general purpose requiring only a RGB feedback and a pinch gripper. Analysis of a real‐world canteen tray data is first performed and used for developing our in‐lab experimental setup. Our results obtained from real‐world scenarios indicate that such approaches are highly desirable for plug‐and‐play domestic applications with limited calibration. All the associated data and code of this work are shared in a public repository.

Dense and distributed tactile sensors are critical for robots to achieve human-like manipulation skills. Soft robotic sensors are a potential technological solution to obtain the required high dimensional sensory information unobtrusively. However, the design of this new class of sensors is still based on human intuition or derived from traditional flex sensors. This work is a first step towards automated design of soft sensor morphologies based on optimization of information theory metrics and machine learning. Elementary simulation models are used to develop the optimized sensor morphologies that are more accurate and robust with the same number of sensors. Same configurations are replicated experimentally to validate the feasibility of such an approach for practical applications. Furthermore, we present a novel technique for drift compensation in soft strain sensors that allows us to obtain accurate contact localization. This work is an effort towards transferring the paradigm of morphological computation from soft actuator designing to soft sensor designing for high performance, resilient tactile sensory networks. © 2020 IEEE.

Impulsive actuation enables robots to perform agile manoeuvres and surpass difficult terrain, yet its capacity to induce continuous and stable locomotion have not been explored. We claim that strictly convex foot shapes can improve impulse effectiveness (impulse used per travelled distance) and locomotion speed by facilitating periodicity and stability. To test this premise, we introduce a theoretical two-dimensional model based on rigidbody mechanics to prove stability. We then implement a more elaborate model in simulation to study transient behaviour and impulse effectiveness. Finally, we test our findings on a robot platform to prove their physical validity. Our results prove, that continuous and stable locomotion can be achieved in the strictly convex case of a disc with off-centred mass. In keeping with our theory, stable limit cycles of the off-centred disc outperform the theoretical performance of a cube in simulation and experiment, using up to 10 times less impulse per distance to travel at the same locomotion speed.

The development of robotic manipulators and hands that show dexterity, adaptability, and subtle behavior comparable to human hands is an unsolved research challenge. In this article, we considered the passive dynamics of mechanically complex systems, such as a skeleton hand, as an approach to improving adaptability, dexterity, and richness of behavioral diversity of such robotic manipulators. With the use of state-of-the-art multimaterial three-dimensional printing technologies, it is possible to design and construct complex passive structures, namely, a complex anthropomorphic skeleton hand that shows anisotropic mechanical stiffness. We introduce a concept, termed the “conditional model,” that exploits the anisotropic stiffness of complex soft-rigid hybrid systems. In this approach, the physical configuration, environment conditions, and conditional actuation (applied actuation) resulted in an observable conditional model, allowing joint actuation through passivity-based dynamic interactions. The conditional model approach allowed the physical configuration and actuation to be altered, enabling a single skeleton hand to perform three different phrases of piano music with varying styles and forms and facilitating improved dynamic behaviors and interactions with the piano over those achievable with a rigid end effector. Copyright © 2018 The Authors.

During abdominal palpation diagnosis, a medical practitioner would change the stiffness of their fingers in order to improve the detection of hard nodules or abnormalities in soft tissue to maximize the haptic information gain via tendons. Our recent experiments using a controllable stiffness robotic probe representing a human finger also confirmed that such stiffness control in the finger can enhance the accuracy of detecting hard nodules in soft tissue. However, the limited range of stiffness achieved by the antagonistic springs variable stiffness joint subject to size constraints made it unsuitable for a wide range of physical examination scenarios spanning from breast to abdominal examination. In this letter, we present a new robotic probe based on a variable lever mechanism able to achieve stiffness ranging from $\text{0.64}$ to $\text{1.06}$ N $\cdot$ m/rad that extends the maximum stiffness by around $\text{16}$ times and the stiffness range by $\text{33}$ times. This letter presents the mechanical model of the novel probe, the finite element simulation as well as experimental characterization of the stiffness response for lever actuation.

The ability to physically enlarge one’s own body structures plays an important role in robustness and adaptability of biological systems. It is, however, a significant challenge for robotic systems to autonomously extend their bodies. To address this challenge, this paper presents an approach using Hot Melt Adhesives (HMAs) to assemble and integrate extensions into the robotic body. HMAs are thermoplastics with temperature dependent adhesiveness and bonding strength. We exploit this property of HMAs to connect passive external objects to the robot’s own body structures, and investigate the characteristics of the approach. In a set of elementary configurations, we analyze to which extent a robot can self-reconfigure using the proposed method. We found that the extension limit depends on the mechanical properties of the extension, and the reconfiguration algorithm. A five-axis robot manipulator equipped with specialized HMA handling devices is employed to demonstrate these findings in four experiments. It is shown that the robot can construct and integrate extensions into its own body, which allow it to solve tasks that it could not achieve in its initial configuration.

Summary This is a case report of a patient with thyroid cancer with asymptomatic pulmonary metastases, and without obvious progression over 34 years. The patient, a 47-year-old male, was shown to have miliary shadows on chest radiographs from the age of 13; indeed, he was temporarily treated for pulmonary tuberculosis without success. A tumour appeared in the right neck in December 1988 (at age 47). A diagnosis of lymph node metastasis of papillary carcinoma of the thyroid was made by biopsy; he then underwent total thyroidectomy with radical dissection of the neck (April 1989). However, the bilateral metastatic lymph nodes in the neck had invaded the vasculature, preventing complete dissection. Post-operative whole body 131I scintigraphy revealed diffuse intensive uptake in the bilateral lung fields, demonstrating for the first time that the pulmonary lesions were metastases of the thyroid cancer. He remains under periodic effective treatment with 131I.