Tipo de investigación

Artificial tactile sensitive skin development is an active research field, with numerous pre- vious interesting designs. One of the most widely used sensors in the (humanoid) robotics community to cover large areas of robots was presented by Cannata et al. [CMMS08]. By cleverly cutting large flex-print-PCBs into separated triangle shaped units, each equipped with 12 capacitive tactile cells, surfaces with moderate curvatures can be covered. The authors demonstrated this by skinning large parts of the child-sized humanoid robot iCub

[SMM+11]. The same technology, although in a slightly miniaturized form was used by

the authors to equip the iCub robot fingertips with a sense of touch. Their sensor can accomplish up to 5kPa sensitivity, but compared to resistive tactile sensor technologies it is more sensitive to noise and further miniaturization is strongly limited due to crosstalk between the tactile cells.

Mittendorfer & Cheng presented hexagonal sensor modules [MC11] that could also cover large robot surfaces. Instead of direct force or pressure measurement sensing, they used optical proximity sensors to detect close-by surfaces. The disadvantage of this technology is that it cannot directly sense force or pressure. As one of the very few artificial skin proposals for technical systems, their hexagonal modules are also equipped with temperature sensing, allowing easier differentiation of human contact from non-human contact.

Technical systems with considerable mass and power, which are typically the features required to be able to perform tasks such as everyday object lifting and manipulation, might pose a health risk if they unintentionally come into contact with other objects or humans. If such systems are to operate in unstructured environments with unpredictable humans, unwanted contacts are unavoidable. Therefore such mobile technical systems must be able to sense collisions and be equipped with logic for taking countermeasures to avoid harm or damage. As a moving mass cannot be instantly stopped, it is important that the outer layer of technical systems is soft and compliant. Elkmann, Fritzsche and Schulenburg et al. [FE09, EFS11] presented such a soft tactile sensor for safe human-robot interaction. The authors developed a clever sleeve for the Kuka LBR robot arm using flexible sensor mats. Unfortunately their design suffered from significant false positives when flexed and stretched around a cylindrical robotic arm [Figure 2.8].

In a recent work, Strohmayr [Str12] presented a flexible tactile sensor skin with an ad- mirable 1.25mm spatial resolution. A grid of intersecting triangle shaped extruded minia-

ture conductive polymer wires formed the sensing element. By applying a load to an

intersecting location of two triangle shaped wires, the sharp edges of these wires are pushed into each other, effectively lowering the electrical resistance between the wires. The sensor is compliant and can achieve up to 8kPa sensitivity. Long time performance has not yet been demonstrated, which is especially important as natural abrasion can have a strong

(a) Kuka LBR robot arm sleeve with flex- ible sensor mats.

(b) Sensor shown to be unfortunately exhibiting false posi- tives on areas with no contact.

Figure 2.8: Live demonstration of tactile sensors [FE09, EFS11] at Fraunhofer Institutes booth in Hannover Fair 2009. (Photographs taken during experiment with permission of booth personnel.)

negative effect on the performance of this sensor design, which depends on the sharp edges of the triangle shaped wires.

Fritzsche et al. [FSE12] recently demonstrated a 3D shaped tactile sensor using ther- moforming. A flat piezoresistive sensor foil was heated and formed into the desired shape by pressing it into a mold. The technical specifications were optimized towards collision detection (10N detection threshold and large tactile cell areas) and as such, the sensor is not usable for fine force or pressure measurements.

Sato et al. [SPH12] presented Touch´e, a clever sensor design that uses swept frequency

capacitive sensing to augment arbitrary conductive objects with sensing capabilities to distinguish between different touches. Although not directly usable for force or pressure measurement, the sensor enables classification of persons and grasp types. The sensor’s working principle is based on the fact that the measured capacitance of the object being touched changes at different sampling frequencies depending on the person touching the object and also on the used grasp type. A frequency sweep over the whole spectrum of the sensor reveals patterns unique to the specific contact situations, allowing recognition of previously learned contact situations.

Large area touch mats have received broad acceptance in a variety of consumer electronic devices. Sensor grid arrays are used to capture the control signals from users using custom digitizer pens or fingers alone. Current state-of-the-art trackpads and touchscreens that

use this technology can detect multiple contact points simultaneously [VIR+09, App]. Fig-

ure 2.9 displays the capacitive sensor array of the Apple Inc. Magic Mouse that can detect finger gestures. For capacitive sensors to detect a change, the dielectric near the electrode has to change from the ambient electric field. Thus, such sensors are not able to detect arbitrary materials (for example wearing a glove typically renders fingers undetectable). The capacitive sensor grids, mostly used in such products, are used as presence detectors,

Figure 2.9: The foil based electrode sheet of capacitive touch sensors of the ges-

ture recognition capable Magic Mouse by Apple Inc. Notice the cutouts

necessary to fold the flex-print PCB foil with electrodes into the con-

cave 3D-surface. (Image included with permission from iFixit, source:

http://d3nevzfk7ii3be.cloudfront.net/igi/UyyKmUFBEB5WBCOW)

but they cannot directly measure the interaction forces or pressures.