Tactile sensing is an essential component in human-robot interaction and object manipulation. Soft sensors allow for safe improved gripping performance. Here we present the TacTip family of sensors: a range soft optical tactile with various morphologies fabricated through dual-material 3D printing. All these are inspired by same biomimetic design principle: transducing deformation surface via movement pins analogous to function intermediate ridges within human fingertip. The performance...

Embodied Choice considers action performance as a proper part of the decision making process rather than merely means to report decision. The central statement embodied choice is existence bidirectional influences between and decisions. This implies that for expressed by an action, dynamics its constraints (e.g. current trajectory kinematics) influence process. Here we use perceptual task compare three types model: serial decision-then-action model, parallel decision-and-action model where...

Slip detection helps to prevent robotic hands from dropping grasped objects and would thus enable complex object manipulation. Here we present a method of detecting slip with biomimetic optical tactile sensor-the TacTip-that operates by measuring the positions internal pins embedded in its compliant skin. We investigate whether local pin movement is strong signal slip. Accurate robust discrimination between static slipping obtained support vector machine (accuracy 99.88%). then demonstrate...

Deep learning has the potential to have impact on robot touch that it had vision. Optical tactile sensors act as a bridge between subjects by allowing techniques from vision be applied touch. In this paper, we apply deep an optical biomimetic sensor, TacTip, which images array of papillae (pins) inside its sensing surface analogous structures within human skin. Our main result is application CNN can give reliable edge perception and thus robust policy for planning contact points move around...

Tactile sensing is used by humans when grasping to prevent us dropping objects. One key facet of tactile slip detection, which allows a gripper know grasp failing and take action an object being dropped. This study demonstrates the detection capabilities recently developed Model O (T-MO) robotic hand using support vector machines detect test multiple scenarios including responding onset in real time with 11 different objects various grasps. In this article, we demonstrate benefits testing...

Deep learning combined with high-resolution tactile sensing could lead to highly capable dexterous robots. However, progress is slow because of the specialist equipment and expertise. The DIGIT sensor offers low-cost entry touch using GelSight-type sensors. Here we customize have a 3D-printed surface based on TacTip family soft biomimetic optical DIGIT-TacTip (DigiTac) enables direct comparison between these distinct types. For this comparison, introduce robot system comprising desktop arm,...

Although superresolution has been studied to huge impact in visual imaging, it is relatively unexplored tactile robotics. Here we demonstrate a novel optical sensor design (the TacTip) capable of achieving 40-fold localization 0.1mm accuracy compared with 4mm resolution between elements. This reached for localizing 40mm diameter hemicylinder finger pad also diameter. Deformations the surface are measured as displacements molded internal pins, pin separation thus defining resolution. Active...

A key unsolved problem in tactile robotics is how to combine perception and control interact robustly intelligently with the surroundings. Here, we focus on a prototypical task of exploration over surface features such as edges or ridges, which principal exploratory procedure humans recognize object shape. Our methods were adapted from an approach for biomimetic active touch that perceives stimulus location identity while controlling aid perception. With minor modification policy, rotate...

In this work, we present an active tactile perception approach for contour following based on a probabilistic framework. Tactile data were collected using biomimetic fingertip sensor. We propose control architecture that implements perception-action cycle the exploratory procedure, which allows to react contact whilst regulating applied force. addition' is actively repositioned optimal position ensure accurate perception. The method trained off-line and then testing performed on-line around...

Motivated by the impact of superresolution methods for imaging, we undertake a detailed and systematic analysis localization acuity biomimetic fingertip flat region tactile skin. We identify three key factors underlying that enable perceptual to surpass sensor resolution: 1) is constructed with multiple overlapping, broad but sensitive receptive fields; 2) perception method interpolates between receptors (taxels) attain subtaxel acuity; 3) active ensures robustness unknown initial contact...

This article illustrates the application of deep learning to robot touch by considering a basic yet fundamental capability: estimating relative pose part an object in contact with tactile sensor. We begin surveying applied robotics, focusing on optical sensors, which help link and for vision. then show how can be used train accurate models 3D surfaces edges that are insensitive nuisance variables, such as motion-dependent shear. involves including representative motions unlabeled...

Developing artificial tactile sensing capabilities that rival human touch is a long-term goal in robotics and prosthetics. Gradually more elaborate biomimetic sensors are being developed applied to grasping manipulation tasks help achieve this goal. Here we present the neuroTac, novel neuromorphic optical sensor. The neuroTac combines hardware design from TacTip sensor which mimicks layered papillae structure of glabrous skin, with an event-based camera (DAVIS240, iniVation) algorithms...

A common approach in the field of tactile robotics is development a new perception algorithm for each application existing hardware solutions. In this letter, we present method dimensionality reduction an optical-based sensor image output using convolutional neural network encoder structure. Instead various complex algorithms, and/or manually choosing task-specific data features, unsupervised feature extraction allows simultaneous online deployment multiple simple algorithms on set black-box...

Bringing tactile sensation to robotic hands will allow for more effective grasping, along with a wide range of benefits human-like touch. Here, we present three-dimensional-printed, three-fingered robot hand comprising an OpenHand ModelO customized house TacTip soft biomimetic sensor in the distal phalanx each finger. We expect that combining grasping capabilities this underactuated sophisticated sensing result platform research-the Tactile Model O (T-MO). The design uses three JeVois...

Advancing robot hand dexterity with optical tactile sensing raises questions about humanoid robotics.

We describe a novel, biomimetic tactile sensing system modeled on the facial whiskers (vibrissae) of animals such as rats and mice. The "BIOTACT Sensor" consists conical array modular, actuated hair-like elements, each instrumented at base to accurately detect deflections shaft by whisker-surface contacts. A notable characteristic this is that, like biological sensory it mimics, are moved back-and-forth ("whisked") so make repeated, brief contacts with surfaces interest. Furthermore, these...

Loss of dopamine from the striatum can cause both profound motor deficits, as in Parkinsons's disease, and disrupt learning. Yet effect on striatal neurons remains a complex controversial topic, is need comprehensive framework. We extend reduced model medium spiny neuron (MSN) to account for dopaminergic modulation its intrinsic ion channels synaptic inputs. tune our D1 D2 receptor MSN models using data recent large-scale compartmental model. The new capture input-output relationships...

In this work, we apply active Bayesian perception to angle and position discrimination extend the method perform actions in a sensorimotor task using biomimetic fingertip. The first part of study tests off-line with large dataset edge orientations positions, Monte Carlo validation ascertain classification accuracy. We observe significant improvement over passive methods that lack loop for actively repositioning sensor. second then applies these findings about an example real-time. Using...

Tactile manipulation will be essential for automating industrial and service tasks currently done by humans. However, the application of tactile feedback to dexterous remains a challenging unsolved problem, with robot capabilities lagging far behind those Here, we present thumb (TacThumb): cheap, robust, 3-D-printed optical sensor integrated on Yale GrabLab model M2 gripper. To test capabilities, cylinder is rolled along TacThumb using opposing nontactile finger. The information permits...