F. Germagnoli

Towards the realization of an artificial tactile system: fine-form discrimination by a tensorial tactile sensor array and neural inversion algorithms

This paper describes techniques and methodologies so far developed to investigate object fine-form discrimination by means of artificial tactile sensors. Sensor arrays, selectively sensitive to stress-tensor components and based on piezoelectric polymer technology, have been realized. Sensor output data are used to solve inverse elastic contact problems, by means of neural networks suitably trained to learn regularized inverse maps. Two possible neural network designs are considered: one is based on the multilayer perceptron trained with the standard backpropagation algorithm, and the other is based on the use of radial basis functions. In both cases, reconstruction of object shapes is demonstrated to be effective and robust with both simulated and real data. >

Skin-like tactile sensor arrays for contact stress field extraction

Abstract Continuous scanning of variable contact forces occurring during object exploration and manipulation by means of miniature sensors is becoming increasingly important in medical prosthetics and orthotics, advanced robotics, teleoperation and telepresence. In this paper a system able to provide robots with tactile sensitivity is presented. Skin-like tactile sensor arrays, which are selectively sensitive to stress-tensor components and are based on piezoelectric polymer technology, have been implemented. The basic characteristics of the sensors, together with the mechanical devices and the electronics developed for setting up the artificial tactile system, are described. An investigation of the expected responses of the sensor has been carried out by modelling the mechanics of object-sensor interaction and by finding the stress solutions of a forward elastic contact problem. The system has been tested by comparing the theoretical results with data measured experimentally from the sensor.

A neural network-based system for tactile exploratory tasks

This paper deals with the problem of object identification by touch. In experiments carried out on human subjects, five tactile primitives were identified, controlling the exploratory strategy of the fingertip on the object. These primitives have been used for setting up the algorithms driving a robotic gripper during the exploration of an unknown object. An artificial tactile system has been designed for the recognition of prism shaped rigid objects. It is based on artificial tactile sensors attached on each fingertip of a two finger robot gripper, giving a sampled image of the normal stress distribution on the sensors surface. The stress images are processed by a three-layer MLP which recognizes the tactile primitive and gives its direction on the sensor surface. The basic strategy consists of reaching an edge of a face and following all the edges of the same face. At the end of a face exploration, an ART neural network classifies the face touched. When all the faces have been explored, a second ART network extracts the shape of the touched prism, by combining the trajectories accomplished by the grip with the characteristics of faces.

An artificial tactile sensor for fine-form discrimination

A new artificial tactile sensor has been developed to provide reliable data about the exploration of small geometrical details of an object (fine-form discrimination). The main characteristics of this sensor are: high spatial resolution, small packaging and flexibility of the support which implies the chance of wearing it as a finger of a glove. This sensor will be applied in both robotic and rehabilitation fields.

A System For Testing A Multielement Tactile Sensor Array

This paper describes the hardware of a system built for testing a piezoelectric tactile sensor array, which has been conceived and realized for recognizing the contact stress components of an object loading its surface. The system is capable of generating a controlled contact on the sensor surface of known load and geometry and measuring the electric responses of the array, by which the six independent components of the stress tensor field are computed. By means of this system a great variety of stress shapes can be tested and the generated space-time dicretized stress field on the sensor can be measured, acquired and stored for each contact applied. In the world of engineering, artificial tactile perception, that can be obtained during object exploration and manipulation by means of miniaturized sensors, is becoming increasingly important in robotics, telemanipulation and advanced prosthetics and orthotics. In the field of artificial taction, a new tactile sensor array has been conceived and realized at the University of Pisa (Italy) [l]. One of the purposes of this research program is to face the problem of fine-form discrimination of an object in non conformal contact with the sensor surface. This problem implies to solve, analytically or numerically, the inverse elastic problem in order to reconstruct the geometry of the touching object from the spatially discretized stress tensor field [2]. Thus, a very precise apparatus is needed to experimentally test the sensor, to acquire and to record data for the comparison and the validation of theorethical solutions of the inverse elastic problem. Moreover, since the sensor has been realized as an array of sensing elements, the single element signals have to be recorded by a controlled acquisition chain, which takes into account the gain and the time constant of each detected response. The multielement tactile sensor is based on the piezoelectric polymer technology, which allows to implement arrays selectively sensitive to stress-tensor components. The six independent components of the tensor field in each point are computed from a linear combination of the measured electric responses of six miniaturized elements of the sensor, each of them sensitive to a particular direction of the stress. On the sensor surface 7 small zones have been realized, each one containing the six elements, in order to reconstruct a 7 point

A computerized system for helping blind people to learn braille code

Thls paper describes a syslem capable or helping bllnd people to learn lht Italian braille code The system is b d on Ibc mognlllon of a braille tut by means of an wWfklal tsrlllcsclrsor, whlch senses the outdenlation of the paper. A scl of srlectcd masagcs are reproduced through a loudspeaker, leaching the bllnd about the ouldented kxt be Is louchtn& A backpropagnlion n e d nehvork has been Lmplemeoled far lhe rccoguilioa of the braille alphabet trough the

Tactile fine form discrimination by a backpropagation neural net

A backpropagation neural network is presented which models the process of tactile fine form discrimination starting from a tensorial image of the stress, detected by the skin tactile receptors inside the dermis. This approach can be used for building artificial tactile systems.

Human strategies in exploring objects by touch

The goal of this study is to investigate the strategy of actively exploring an object by touch; in particular it examined the contour following which provides information about the global shape of an object. Two experiments have been carried out in which two classes of subjects were requested to discriminate between pairs of different objects, using only their forefinger tip. By these experiments, five tactile primitives were identified, which seem to drive the control of the exploratory strategy on a low relief picture.

Neural network techniques for fine-form discrimination

The ability of neural network (NN) techniques in processing tensorial tactile images have been investigated. Tactile images were provided by a sensor array of 21/spl times/21 sensitive elements, realized with piezoelectric polymers, able to detect the six components of the stress tensor. Two different types of NN have been used: a multilayer perceptron trained with the backpropagation algorithm, to filter and pre-process the data (reconstruction phase) and an ART network, to recognize different object typology (classification phase).