sensors Article

Flexible Piezoelectric Tactile Sensor Array for Dynamic Three-Axis Force Measurement Ping Yu, Weiting Liu *, Chunxin Gu, Xiaoying Cheng and Xin Fu The State Key Laboratory of Fluid Power and Mechatronic Systems, Department of Mechanical Engineering, Zhejiang University, Hangzhou 310027, China; [email protected] (P.Y.); [email protected] (C.G.); [email protected] (X.C.); [email protected] (X.F.) * Correspondence: [email protected]; Tel./Fax: +86-571-8795-2226 Academic Editor: Vittorio Passaro Received: 11 April 2016; Accepted: 26 May 2016; Published: 3 June 2016

Abstract: A new flexible piezoelectric tactile sensor array based on polyvinylidene fluoride (PVDF) film is proposed for measuring three-axis dynamic contact force distribution. The array consists of six tactile units arranged as a 3 ˆ 2 matrix with spacing 8 mm between neighbor units. In each unit, a PVDF film is sandwiched between four square-shaped upper electrodes and one square-shaped lower electrode, forming four piezoelectric capacitors. A truncated pyramid bump is located above the four piezoelectric capacitors to improve force transmission. A three-axis contact force transmitted from the top of the bump will lead to the four piezoelectric capacitors underneath undergoing different charge changes, from which the normal and shear components of the force can be calculated. A series of dynamic tests have been carried out by exerting sinusoidal forces with amplitudes ranging from 0 to 0.5 N in the x-axis, 0 to 0.5 N in the y-axis, and 0 to 1.5 N in the z-axis, separately. The tactile units show good sensitivities with 14.93, 14.92, and 6.62 pC/N in the x-, y-, and z-axes, respectively. They can work with good linearity, relatively low coupling effect, high repeatability, and acceptable frequency response in the range of 5–400 Hz to both normal and shear load. In addition, dynamic three-axis force measurement has been conducted for all of the tactile units. The average errors between the applied and calculated forces are 10.68% ˘ 6.84%. Furthermore, the sensor array can be easily integrated onto a curved surface, such as robotic and prosthetic hands, due to its excellent flexibility. Keywords: piezoelectric tactile sensor; three–axis dynamic force; PVDF; flexible; truncated pyramid bump

1. Introduction Tactile sensing plays an important role in human dexterous manipulation and object perception. This sensing ability depends on two kinds of tactile corpuscles: fast-adapting and slowly-adapting corpuscles in human skin. The fast-adapting corpuscle exhibits sensitivity to dynamic skin deformation of relatively high frequency, and can respond to dynamic events such as making and breaking contact, slippage, and texture perception, while the slowly-adapting corpuscle exhibits sensitivity to lower-frequency skin deformation and can be utilized to detect sustained contact forces [1]. To emulate human perception for sophisticated manipulation, different kinds of tactile sensors have been developed for robotic and prosthetic hands [2,3]. These sensors can be classified according to the sensing mechanism employed, including piezoresistive [4,5], piezoelectric [6–8], capacitive [9,10], and optical [11] sensing, etc. Among these sensing mechanisms, piezoelectric sensing features excellent sensitivity to dynamic mechanical stimulation over a wide frequency range and, thus, are widely utilized in developing tactile sensors for mimicking the fast-adapting tactile corpuscle [12,13].

Sensors 2016, 16, 819; doi:10.3390/s16060819

www.mdpi.com/journal/sensors

Sensors 2016, 16, 819

2 of 15

For piezoelectric tactile sensors aiming to mimic the fast-adapting tactile corpuscle and applied in limited spaces and curved surfaces, such as robotic and prosthetic hands, the following design requirements should be satisfied: (i) the tactile sensor should be able to measure a dynamic three-axis force in the designed frequency range. In actual cases, when the hand makes contact with an object, forces between the tactile sensor and the object always occur in arbitrary directions, generating normal and shear force components [9]. In addition, slippage and touching perception of surface features are inseparable from tangential motion and accompany with small vibrations [14]. To extract valuable information in these events, tactile sensors should have good dynamic response capability to follow oscillations in both normal and shear force; (ii) sensing elements, microstructures, and package material should be flexible and conform to the three-dimensional surface in order to be deployed on the curved surface of fingertips [9]; and (iii) the fabrication process of the tactile sensor should be compatible with well-developed techniques, such as the surface micromachining process which enables the possibility of miniaturization, good spatial resolution, and mass production [15]. During the last few decades, several piezoelectric tactile sensors have been developed. A piezoelectric tactile sensor array with 8 ˆ 8 sensing units was realized by coupling a polymer polyvinylidene fluoride (PVDF) film to a monolithic silicon integrated circuit. The response of a sensing unit is linear for normal loads spanning 0.008–1.35 N [16]. To improve the scalability and flexibility, a scalable technique based on inkjet printing was provided to pattern electrodes on a piezoelectric film. With this technique, a flexible piezoelectric transducer array sensitive to normal forces was developed [17]. To make a sensor with a better signal-to-noise ratio, faster response, wider bandwidth, and better force sensitivity, a P(VDF-TrFE) (polyvinylidene fluoride-trifluoroethylene) film was directly spin-coated onto the gate area of a metal oxide semiconductor(MOS)transistor [6,7]. The above-mentioned sensor prototypes focused only on dynamic normal force detection. In order to enable shear force sensing, other research groups tried to integrate PVDF sensing elements vertically into an elastomeric structure [18,19]. Furthermore, to simultaneously measure both dynamic normal and shear force components, four piezoelectric sensing elements are symmetrically arranged on the surface of a vertically-placed cylindrical shell. When an external three-axis force is applied at the tip of the shell, the normal force component causes the same electric voltage on all four piezoelectric sensing elements, while the shear force component causes a reversal of voltage on two opposite piezoelectric elements. From signals from the four elements, the direction angles and magnitude of the three-axis force can be calculated [20]. One problem of these shear force and three-axis force sensors is that the piezoelectric sensing elements were vertically placed, thus, the suggested fabrication process could not be compatible with surface micromachining technology and limited further reduction of the sensor size, which is extremely important in space-limited applications, such as robotic and prosthetic hands. Recently, a structure consisting of a bump layer and a horizontally-arranged sensing array layer has been successfully introduced in capacitive and piezoresistive tactile sensor to measure static three–axis forces [9,21]. Inspired by this, the authors here propose a piezoelectric tactile sensor array meeting the entire three design requirements mentioned above with a designed multilayer flexible structure. Each tactile unit in the sensor array consists of one compliant truncated pyramid bump, which serves as contact force translator; underneath the bump, four separate piezoelectric capacitors based on PVDF are horizontally distributed, which enables three-axis force measurement and makes the fabrication process compatible with the surface micromachining process. Detail design and fabrication of the sensor array are presented, followed by the experimental testing. 2. Structure and Operating Principle 2.1. Sensor Structure Figure 1 illustrates the conceptual diagram of the proposed three–axis tactile sensor array. Figure 1a is the overview of the sensor array, which contains six three-axis tactile units arranged as a 3 ˆ 2 matrix. The center distance between every two adjacent units is 8 mm. Figure 1b,c illustrate

Sensors 2016, 16, 819

3 of 15

the schematic view and exploded view of a tactile unit. It consists of five layers: a PDMS bump layer, Sensors 2016, 16, 819(Al) electrode layer, a PVDF film layer, a lower Al electrode layer and a 3 of 15 an upper aluminum PDMS substrate layer. The upper and lower Al electrode layers with thickness of 200 nm are developed upper aluminum (Al) electrode layer, a PVDF film layer, a lower Al electrode layer and a PDMS through wet etching of commercial two-side metalized PVDF film. The PVDF film is sandwiched substrate layer. The upper and lower Al electrode layers with thickness of 200 nm are developed between four square-shaped upper electrodes and one square-shaped lower electrode, thus forming through wet etching of commercial two-side metalized PVDF film. The PVDF film is sandwiched 2 A PDMS four piezoelectric Each piezoelectric capacitor has the area of 1.5 ˆ 1.5 mm between fourcapacitors. square-shaped upper electrodes and one square-shaped lower electrode, thus .forming truncated bump with dimensions of 4 mm ˆ 4 mm mm of is 1.5 fixed at mm the 2center of the fourpyramid piezoelectric capacitors. Each piezoelectric capacitor has ˆ the2 area × 1.5 . A PDMS unit with its four bottom corners to of the4 mm corner of each upper square shown truncated pyramid bump with aligned dimensions × 4 mm × 2 mm is fixed at theelectrodes, center of theas unit with in Figure its 1d.four bottom corners aligned to the corner of each upper square electrodes, as shown in Figure 1d.

Figure 1. Conceptual diagram of the proposed three–axis tactile sensor array. (a) Top view of the array;

Figure 1. Conceptual diagram of the proposed three–axis tactile sensor array. (a) Top view of the array; (b) schematic view of a tactile unit; (c) exploded view of a tactile unit; and (d) cross-sectional view of (b) schematic view of a tactile unit; (c) exploded view of a tactile unit; and (d) cross-sectional view of a tactile unit. a tactile unit.

2.2. Conceptual Operation Principle

2.2. Conceptual Operation Principle

PVDF is a piezoelectric plastic material that generates charges when it is mechanically deformed [22]. The relationship between electric displacement charge density), electric field, PVDF is a piezoelectric plastic material (output that generates charges when it and is mechanical mechanically stress is expressed as follows [23]: deformed [22]. The relationship between electric displacement (output charge density), electric

field, and mechanical stress is expressed as6 follows [23]: 3

Di   diqTq    ikT Ek ,i  1, 2, 3

(1)

6 q 1 3k 1 ÿ ÿ Di “ diq Tq ` ε Tik Ek , i “ 1, 2, 3

(1)

where Di is electric displacement (output charge density) in the i direction, Tq is mechanical stress q “1 k “1 in the q direction, Ek is electric field in the k direction, diq is the piezoelectric constant, and εTik is constant under constant stress. where dielectric Di is electric displacement (output charge density) in the i direction, Tq is mechanical stress in T is In the event of the PVDF being used in the thickness mode, the relationconstant, 1 can be written follows: the q direction, E is electric field in the k direction, d is the piezoelectric and εas dielectric k

iq

ik

constant under constant stress. T (2) D3 thickness  d33T3  mode, E 33 3 the relation 1 can be written as follows: In the event of the PVDF being used in the In our proposed sensor, the PVDF film D3 “ d33mainly T3 ` ε T33works E3 on external electric field. So, the relation 2 can be written as follows:

33 (thickness)

mode and with no (2)

Sensors 2016, 16, 819

4 of 15

In our2016, proposed external Sensors 16, 819 sensor, the PVDF film mainly works on d33 (thickness) mode and with no 4 of 15 electric field. So, the relation 2 can be written as follows:

D d T

33T3 D3 3“ d33 3

(3)

(3)

Figure 2 shows the principle of operation for sensing a three-axis force in a tactile unit. The Figure shows theforce principle operation sensing a into three-axis forceforce in a tactile unit. (F The applied applied2three-axis on the of bump can be for decomposed one normal component z ) and three-axis force on the bump can be decomposed into one normal force component (F ) and z two shear force components (Fx and Fy ) perpendicular to each other, respectively. The normal and two

shearshear forcedirections components (Fx and Fy ) perpendicular each respectively. The normal and shear are defined as shown in Figure 2a, in to which theother, normal direction is the through-thickness directions are defined asnormal shownforce in Figure 2a, in which theisnormal direction is the through-thickness direction. Under the component, the bump compressed and the four piezoelectric capacitors are the subjected to force the same compressive (P11 P12 , P21 and P22 ). By thefour piezoelectric direction. Under normal component, thestress bump is, compressed and the piezoelectric and Q (P)11with and piezoelectric polarity effect are of the PVDF film, ( Q11 , Q12 , Q21 stress capacitors subjected to thecharges same compressive , P12the , P21same and quantity P22 ). By the 22 effect(negative) of the PVDF film, charges (Qupper , Q21 and with the same quantity and polarity (negative) are generated on the electrodes ofQ the piezoelectric capacitors, while equivalent 22 )four 11 , Q12 positive charges generated on the of lower Under thecapacitors, shear force components, the bump are generated on theare upper electrodes the electrodes. four piezoelectric while equivalent positive deforms and generates a torque at the fixed end. The two piezoelectric capacitors on the left side are charges are generated on the lower electrodes. Under the shear force components, the bump deforms subjected to tensile stress, whereas other two on the right side are subjected to compressive stress, as and generates a torque at the fixed end. The two piezoelectric capacitors on the left side are subjected shown in Figure 2c,d. As a result, the developed charges on the left electrodes and right electrodes to tensile stress, whereas other two on the right side are subjected to compressive stress, as shown are of opposite polarity. Thus, Fz can be calculated by taking the average values of in Figure 2c,d. As a result, the developed charges on the left electrodes and right electrodes are of Q11 , Q12 , Q21 and Q , while Fx and Fy can be calculated from the differences among opposite polarity. Thus,22 Fz can be calculated by taking the average values of Q11 , Q12 , Q21 and Q22 , Q , Q12 , Q21 and Q . 22 while F11x and Fy can be calculated from the differences among Q11 , Q12 , Q21 and Q22 .

Figure 2. Principle of operation sensingnormal normal and thethe tactile unit. (a) Reference Figure 2. Principle of operation forforsensing and shear shearforces forcesinin tactile unit. (a) Reference coordinates; (b) response to normal force; (c) response to shear force in the x-axis direction; and (d) and coordinates; (b) response to normal force; (c) response to shear force in the x-axis direction; response to shear force in the y-axis direction. (d) response to shear force in the y-axis direction.

3. Fabrication

3. Fabrication

The fabrication process of the flexible piezoelectric three-axis tactile sensor array is shown as

The fabrication thecompatible flexible piezoelectric three-axis tactile sensor follows (Figure 3),process which isof fully with the surface micromachining process.array is shown as follows (Figure 3), which is fully compatible with the surface micromachining process.

Sensors 2016, 16, 819 Sensors 2016, 16, 819

5 of 15 5 of 15

Figure 3. Fabrication process of the proposed three–axis tactile sensor array. (a) Electrode layers; (b)

Figure 3. Fabrication process of the proposed three–axis tactile sensor array. (a) Electrode layers; PDMS protecting layer; (c) PDMS bump layer; and (d) position and bonding. (b) PDMS protecting layer; (c) PDMS bump layer; and (d) position and bonding.

Step 1: Upper and lower electrode layers patterning, as shown in Figure 3a. A 25 μm-thick, metalized (Cr/Al), andlower poled electrode PVDF filmlayers (Piezotech S.A.S., France) is prepared. The 3a. piezoelectricity Step 1: Upper and patterning, as shown in Figure A 25 µm-thick, constant d33 of the is 15PVDF ± 3 pC/N. the film is taped on aisclean four-inch wafer metalized (Cr/Al), andfilm poled film Firstly, (Piezotech S.A.S., France) prepared. Thesilicon piezoelectricity with care such that the film is stretched uniformly over the wafer. The tape should cover all sides of with constant d33 of the film is 15 ˘ 3 pC/N. Firstly, the film is taped on a clean four-inch silicon wafer the PVDF film, overlapping by 2 mm [24]. Secondly, a 2.5 μm-thick RZJ-304 positive photoresist (PPR) care such that the film is stretched uniformly over the wafer. The tape should cover all sides of the is spin-coated onto one side of the PVDF film and then soft baked at 60 °C for 60 min. The reason why PVDF film, overlapping by 2 mm [24]. Secondly, a 2.5 µm-thick RZJ-304 positive photoresist (PPR) soft bake was chosen is that the curie temperature of the PVDF film is only about 135 °C and the is spin-coated onto one side of the PVDF film and then soft baked at 60 ˝ C for 60 min. The reason polarized electrets are thermodynamically stable up to about 90 °C [25]. For security, soft bake at 60 °C for ˝ C and why 60 soft bake was chosen is that thecooling curie down temperature of the PVDFthe film is only 135 min is carried out. Thirdly, after to room temperature, PVDF film isabout torn off from the polarized are thermodynamically stabletoup about 90 ˝ as C [25]. For security, soft bake at the siliconelectrets wafer, turned over and then taped again theto silicon wafer described above. Fourthly, ˝ 60 Cwith for 60 carried PPR out. isThirdly, afteronto cooling downside to room the PVDF film is torn themin sameisprocess, spin-coated the other of the temperature, PVDF film. Fifthly, the two-side PPR-coated PVDF film is inserted between two aligned masksto which containwafer patterns the upper off from the silicon wafer, turned over and then taped again the silicon as for described above. and lower electrode layers. Using an ultraviolet source, PPR on both sides are exposed. Sixthly, the Fourthly, with the same process, PPR is spin-coated onto the other side of the PVDF film. Fifthly, the exposed film is developed, rinsed, dried under nitrogen, and baked at 60which °C forcontain 60 min.patterns Seventhly, two-side PPR-coated PVDF film is inserted between two aligned masks for the the PVDF film is immersed into an aluminum etchant, a mixture containing phosphoric acid, acetic upper and lower electrode layers. Using an ultraviolet source, PPR on both sides are exposed. Sixthly, acid, nitric acid, and water with weight ratio of 16:4:1:1, for about 10 min to pattern the upper and the exposed film is developed, rinsed, dried under nitrogen, and baked at 60 ˝ C for 60 min. Seventhly, lower electrodes. At last, the residual PPR is removed by acetone. the PVDF film is immersed into an aluminum etchant, a mixture containing phosphoric acid, acetic acid, nitric acid, and water with weight ratio of 16:4:1:1, for about 10 min to pattern the upper and lower electrodes. At last, the residual PPR is removed by acetone.

Sensors 2016, 16, 819 Sensors 2016, 16, 819

6 of 15 6 of 15

Step 2: PDMS film coating, as shown in Figure 3b. The liquid-state PDMS (Sylgard 184, Step 2: PDMS film coating, as shown in Figure 3b. The liquid-state PDMS (Sylgard 184, A:B = 10:1 A:B = 10:1 in weight) is spin-coated onto the lower electrode layer at 2000 rpm for 30 s to form a 50 in weight) is spin-coated onto the lower electrode layer at 2000 rpm for 30 s to form a 50 µm-thick μm-thick PDMS substrate layer and then cured at room temperature for 24 h. The same process is PDMS substrate layer and then cured at room temperature for 24 h. The same process is used to form used to form a 50 μm-thick PDMS film on the upper electrode layer for bonding. a 50 µm-thick PDMS film on the upper electrode layer for bonding. Step 3: Bump layer fabrication (Figure 3c). The liquid-state PDMS (10:1) is poured into a stainless Step 3: Bump layer fabrication (Figure 3c). The liquid-state PDMS (10:1) is poured into a stainless steel mold which is manufactured by a milling machine, and then cured at 80 ˝°C for 3 h to replicate steel mold which is manufactured by a milling machine, and then cured at 80 C for 3 h to replicate the shape of the bump. the shape of the bump. Step 4: Sensor array assembling. After being activated by oxygen plasma, the PDMS bump layer Step 4: Sensor array assembling. After being activated by oxygen plasma, the PDMS bump layer fabricated in Step 3, the PVDF layer with patterned electrodes and PDMS coats fabricated in Step 2 fabricated in Step 3, the PVDF layer with patterned electrodes and PDMS coats fabricated in Step 2 are are aligned and bonded using a three-axis manual stage (Newport 460 P), as shown in Figure 3d. aligned and bonded using a three-axis manual stage (Newport 460 P), as shown in Figure 3d. Figure 4a shows the fabricated PVDF film with patterned upper and lower electrodes. The Figure 4a shows the fabricated PVDF film with patterned upper and lower electrodes. electrodes are well fabricated and have high uniformity. Figure 4b is the magnified view close to The electrodes are well fabricated and have high uniformity. Figure 4b is the magnified view close electrodes of a tactile unit, where four square-shaped upper electrodes aligned well with one squareto electrodes of a tactile unit, where four square-shaped upper electrodes aligned well with one shaped lower electrode. Figure 4c shows the fabricated piezoelectric three-axis tactile sensor array, square-shaped lower electrode. Figure 4c shows the fabricated piezoelectric three-axis tactile sensor which has 6 tactile units with high flexibility, as demonstrated in Figure 4d. array, which has 6 tactile units with high flexibility, as demonstrated in Figure 4d.

Figure 4. (a) (a)Photograph Photographofofthe thefabricated fabricated PVDF film with patterned upper lower electrodes; Figure 4. PVDF film with patterned upper andand lower electrodes; (b) (b) magnified view close to electrodes of a tactile unit; (c) photograph of the fabricated three-axis tactile magnified view close to electrodes of a tactile unit; (c) photograph of the fabricated three-axis tactile sensor sensor array; array; and and (d) (d) photograph photograph of of the the fabricated fabricated sensor sensor array array mounted mounted on on aa curved curved surface. surface.

4. Measurementsand andDiscussion Discussion 4. Measurements 4.1. Sensor Unit Calibration 4.1. Sensor Unit Calibration To study the sensing performance of the sensor array, calibration experiments have been To study the sensing performance of the sensor array, calibration experiments have been conducted. Figure 5a,b show the schematic and photograph of the experimental setup. The tactile conducted. Figure 5a,b show the schematic and photograph of the experimental setup. The tactile sensor array is located on a shaker (LDS V455, Brüel&Kjær, Darmstadt, Germany) which can apply sensor array is located on a shaker (LDS V455, Brüel&Kjær, Darmstadt, Germany) which can apply random dynamic forces up to 489 N, with a frequency in the range 5 Hz–7.5 kHz. For normal force random dynamic forces up to 489 N, with a frequency in the range 5 Hz–7.5 kHz. For normal force loading, the tactile sensor array is horizontally fixed on the shaker’s platform (Figure 5c). As for loading, the tactile sensor array is horizontally fixed on the shaker’s platform (Figure 5c). As for shear shear force loading, the sensor array is vertically fixed on the shaker’s platform (Figure 5d), with force loading, the sensor array is vertically fixed on the shaker’s platform (Figure 5d), with an upright an upright standing plate. By rotating the sensor array, shear force in any direction can be applied. standing plate. By rotating the sensor array, shear force in any direction can be applied. A round A round loading bar with a diameter of 8 mm is used to press the bump of a tactile unit against loading bar with a diameter of 8 mm is used to press the bump of a tactile unit against the shaker to the shaker to add preloading forces. The loading bar is attached to a six-axis load cell (Nano43, ATI add preloading forces. The loading bar is attached to a six-axis load cell (Nano43, ATI Industrial Industrial Automation, Apex, NC, USA), which has 0.0078 N resolution, ˘36 N measurement range Automation, Apex, NC, USA), which has 0.0078 N resolution, ±36 N measurement range and 0–2800 Hz frequency range. With this strategy, the applied force can be precisely measured by the six–axis

Sensors 2016, 16, 819

7 of 15

and 0–2800 Hz16, frequency range. With this strategy, the applied force can be precisely measured by the Sensors 2016, 819 7 of 15 six–axis load cell. The six-axis load cell is fixed onto a three-axis positioner (M460P, Newport, Miami, load cell. The six-axis load cell is fixed onto a three-axis (M460P, Newport, Miami, FL, USA) FL, USA) to enable the alignment capability betweenpositioner the axis of loading bar and the bump center to enable the alignment capability between the axis of loading bar and the bump center of a tactile of a tactile unit. Charges developed by the tactile sensor array are measured by a charge amplifier unit. Charges developed by the tactile sensor array are measured by a charge amplifier (DH5862, (DH5862, Donghua Testing Technology Co., LTD, Jingjiang, China) which has 16 channels, 10 mv/pC Donghua Testing Technology Co., LTD, Jingjiang, China) which has 16 channels, 10 mv/pC sensitivity, and 0.3–30 KHz frequency range. Output voltages of both the six-axis load cell and the sensitivity, and 0.3–30 KHz frequency range. Output voltages of both the six-axis load cell and the charge amplifier areare collected bybya aNI-DAQ (USB-6343,National National Instruments Corporation, charge amplifier collected NI-DAQ device device (USB-6343, Instruments Corporation, Austen, TX, USA) andand displayed onona acomputer basedononLabview Labview software. Parameters Austen, TX, USA) displayed computer screen screen based software. Parameters of theof the mechanical stimulus (waveform, frequency, areset setbybya signal a signal generator (33500B, Agilent mechanical stimulus (waveform, frequency,amplitude) amplitude) are generator (33500B, Agilent Technologies, PaloPalo Alto, CA, USA) bya apower power amplifier (PA1000L, Brüel&Kjær, Technologies, Alto, CA, USA)and andthen then amplified amplified by amplifier (PA1000L, Brüel&Kjær, Darmstadt, Germany). Darmstadt, Germany).

Figure 5. (a)5.Schematic of of thethe experimental for sensor sensorunit unitcalibration.; calibration.; photograph Figure (a) Schematic experimental setup setup for (b) (b) photograph of theof the experimental setup; close-upview viewof of the the contact contact between loading barbar andand PDMS bump in experimental setup; (c)(c) close-up betweenthe the loading PDMS bump in normal loading; close-up viewofofthe thecontact contact between between the PDMS bump normal forceforce loading; andand (d)(d) close-up view theloading loadingbar barand and PDMS bump in shearloading. force loading. shearinforce

Sinusoidal normal and shear forces with constant frequency (20 Hz) and variable amplitude

Sinusoidal and shear forcesunit. with constantoutputs frequency variable were applied,normal respectively, on a tactile Meanwhile from(20 the Hz) unit and and the six-axisamplitude load werecell applied, respectively, onalla the tactile Meanwhile outputs thebeunit andon thethe six-axis were recorded. Before tests,unit. a preload force (about 2 N)from should applied top of load the tactile unit. Without force, a contactforce loss between the bar the bump maytop of cell were recorded. Before aallpreload the tests, a preload (about 2 N)loading should beand applied on the occur in the negative half cycle of stimulation which leads the applied stimulus to being uncontrolled the tactile unit. Without a preload force, a contact loss between the loading bar and the bump may generating anhalf erratic force pattern impact on the loading bar surface. It should be noted that occurand in the negative cycle of stimulation which leads the applied stimulus to being uncontrolled piezoelectric transducers only respond to the dynamic forces and, hence, the static preload does not that and generating an erratic force pattern impact on the loading bar surface. It should be noted affect the measurements [7]. In addition, all the tests were performed under ambient conditions with piezoelectric transducers only respond to the dynamic forces and, hence, the static preload does not a temperature around 20 °C. The experimental procedure is as follows. (1) Normal load test: the tactile affect the measurements [7]. In addition, all the tests were performed under ambient conditions with array was horizontally located, as in Figure 5c. The shaker was excited to apply a sinusoidal 20 Hz a temperature around 20 ˝ C. The experimental procedure is as follows. (1) Normal load test: the normal force with variable amplitude (Fzm ) in the range of 0–1.5 N on the tactile unit; (2) Shear load tactile array wasthe horizontally located, as invertically Figure 5c. The as shaker excited apply a sinusoidal test (x-axis): tactile sensor array was placed, shownwas in Figure 5d.toThe shaker was 20 Hz normal force with variable amplitude (Fzm ) in the range of 0–1.5 N on the tactile unit; (2) Shear load test (x-axis): the tactile sensor array was vertically placed, as shown in Figure 5d. The shaker

Sensors2016, 2016,16, 16,819 819 Sensors

15 88ofof15

excited to apply sinusoidal 20 Hz shear forces in the x-axis with variable amplitude (Fxm ) in the range of was excited to apply sinusoidal 20 Hz shear forces in the x-axis with variable amplitude (F ) in the 0–0.5 N on the tactile unit; (3) Shear load test (y-axis): the tactile sensor array was rotated by 90xm degrees range of 0–0.5 N on the tactile unit; (3) Shear load test (y-axis): the tactile sensor array was rotated by along the z-axis. The shaker was excited to apply a sinusoidal 20 Hz shear forces in y-axis with 90 degrees along the z-axis. The shaker was excited to apply a sinusoidal 20 Hz shear forces in y-axis variable amplitude (Fym ) in the range of 0–0.5 N on the tactile unit. Each test was repeated three times with variable amplitude (Fym ) in the range of 0–0.5 N on the tactile unit. Each test was repeated three to evaluate repeatability of the sensor’s performance. times to evaluate repeatability of the sensor’s performance. Figure 6 presents measured tactile unit’s charge outputs and applied forces in reference six-axis Figure 6 presents measured tactile unit’s charge outputs and applied forces in reference six-axis load cell. Responses of the four piezoelectric capacitors were able to trace the applied forces well in load cell. Responses of the four piezoelectric capacitors were able to trace the applied forces well in all tests. From results of normal load tests showed in Figure 6a,b, we can observe a nearly uniform all tests. From results of normal load tests showed in Figure 6a,b, we can observe a nearly uniform behavior of output charges from all four piezoelectric capacitors in a tactile unit. Comparing Figure 6b behavior of output charges from all four piezoelectric capacitors in a tactile unit. Comparing Figure 6b with Figure 6a, we can find that the amplitude of output charge increases with the amplitude of the with Figure 6a, we can find that the amplitude of output charge increases with the amplitude of the input force. As for the results of shear load tests showed in Figure 6c–f, the polarity of the output charges input force. As for the results of shear load tests showed in Figure 6c–f, the polarity of the output Q11 and Q12 is opposite to that of Q21 and Q22 for the x-axis force, and the polarity of output charges charges Q11 and Q12 is opposite to that of Q21 and Q22 for the x-axis force, and the polarity of output and Q Q21 and is opposite to that oftoQthat and Q22 , exactly as what we described in Section 2.2. Qcharges 11 12 Q21 is opposite of Q 11 12 and Q22 , exactly as what we described in Section 2.2.

Figure Figure6.6.Measured Measuredcharges chargesofofaatactile tactileunit unitand andapplied appliedforces forcesininreference referencesix-axis six-axisload loadcell cellwhen when sinusoidal forces (f = 20 Hz) in normal and shear directions were applied respectively. (a) Normal sinusoidal forces (f = 20 Hz) in normal and shear directions were applied respectively. (a) Normal force = 0.24 N; normal (b) normal force amplitude = 1.45 shear force in xforce amplitude with with amplitude Fzm =Fzm 0.24 N; (b) force withwith amplitude Fzm F=zm1.45 N; N; (c) (c) shear force in x-axis axis amplitude = 0.2 shear force in x-axis with amplitude = 0.49 shear forcein withwith amplitude FxmF=xm0.2 N; N; (d)(d) shear force in x-axis with amplitude FxmFxm = 0.49 N;N; (e)(e) shear force ym = = 0.21 0.21 N; (f) shear force in y-axis with amplitude F ym = 0.48 N. iny-axis y-axiswith withamplitude amplitudeFFym N; (f) shear force in y-axis with amplitude Fym = 0.48 N.

Sensors 2016, 16, 819

9 of 15

Sensors 2016, 16, 819and shear components (Fx , Fy and Fz ) of the applied three-axis force can be decoupled 9 of 15 Normal

by combining the four output charges in different forms, as described in Section 2.2. Combinational output charges are defined as follows: Normal and shear components (Fx , Fy and Fz ) of the applied three-axis force can be decoupled  in different (Q  Q21 )  (Q12asdescribed Q22 ) by combining the four output charges forms, in Section 2.2. Combinational Qx  11  output charges are defined as follows: 2

  Q12 )  (Q21  Q22 ) $Q  (Qp11  Qy “ Q11 ´Q21 q`pQ12 ´Q22 q ’ 22 & x pQ ´Q12q`p Q21 ´Q22 q 11Q  Q22 Q  Qy “Q11 12 21 ’ %Qz  Q11 `Q12 `Q2 21 `Q22 4  Qz “ 4

(4)

Figure 7. Average amplitudes (100 wave periods) of sinusoidal output charges versus sinusoidal input forces (20 Hz). In which symbols ‘˝’, ‘#’, ‘4’ represent combinational output charges Qx , Qy , and Qz , respectively. Each symbol has three colors indicating responses from three repeated tests. Solid lines represent linear fit of average response of the three repeat tests. (a) z-axis; (b) x-axis; (c) y-axis.

(4)

Sensors 2016, 16, 819

10 of 15

Using data in Figure 6, relationships between combinational output charges (Qx , Qy , and Qz ) and applied forces were plotted in Figure 7. These plots relate average amplitudes (100 wave periods) of sinusoidal combinational output charges and corresponding average amplitudes (100 wave periods) of sinusoidal applied forces. Symbols ‘˝’, ‘#’, ‘4’ represent combinational output charges Qx , Qy , and Qz , respectively. Each symbol has three colors indicating responses from three repeated tests. From Figure 7a we can see that the change in normal force mainly causes change in Qz . From Figure 7b,c, we can see that the change in shear force mainly shows change in the corresponding combinational output charge (for instance, when shear force in the x-axis is increased, Qx changes with it). These mean cross-talk effects among the three axes are relatively small. Meanwhile, little deviations were observed from responses of three repeat tests, and the calculated repeatability in measuring x-, y-, and z-directional forces is 0.82%, 1.99%, and 1.35% F.S. (Full Scale), respectively. To evaluate the linearity, least squares linear curves were used to fit average responses of the three repeat tests. They all show good linearity, and the nonlinearity in measuring x-, y-, and z-directional forces is 2.45%, 2.37%, and 1.74% F.S., respectively. Slopes of fitted lines, which represent sensitivities of the tactile unit, have been extracted as 14.93, 14.92, and 6.62 pC/N for x-, y-, and z-axes, respectively. Using these values, three components of the applied force can be calculated as follows: $ ’ & Fx “ Fy “ ’ % F “ z

1 14.93 Q x “ 0.067Q x 1 14.92 Qy “ 0.067Qy 1 6.62 Qz “ 0.1511Qz

(5)

4.2. Frequency Effect Test To evaluate frequency effect, the normal and shear load tests described in Section 3 have been repeated by altering the frequency in the range of interest of the present applications ( f = 5 Hz, 20 Hz,100 Hz, 200 Hz and 400 Hz). This frequency range is suitable for several sensing applications. In the human hand, for example, the two types of fast–adapting mechanoreceptors FAI and FAII, which respond to dynamic events, such as contact, slippage, and texture encoding, are sensitive to dynamic skin deformation in the frequency ranges of ~5 to 50 Hz and ~40 to 400 Hz, respectively [1]. Figure 8 shows average amplitude (100 wave periods) of sinusoidal combinational output charges versus corresponding average amplitude (100 wave periods) of sinusoidal applied forces at different frequencies. As it can be seen, a linear behavior is achieved over the whole explored range (5–400 Hz). The slope of charge–to–force curve, which represents sensitivity of the tactile unit, has small variation over the entire range of tested frequency in both normal and shear load tests. Overall, in normal direction the minimum sensitivity is 80% of the maximum sensitivity, and in the shear direction, the minimum sensitivity is 78% of the maximum sensitivity. According to the 3 db bandwidth requirement, our developed tactile unit shows acceptable frequency characteristic from 5 to 400 Hz in both normal and shear directional tests. Though the variation of sensitivity in the entire range of tested frequency is relatively small, an interesting variation tendency was observed in the frequency range from 5 to 100 Hz. It can be noticed that the sensitivity in the normal direction increased gradually when the frequency was changed from 5 to 100 Hz, and then remain constant. This may be attributed to the high-pass filtering effect of the charge amplifier. Contrarily, the sensitivity in shear (x-axis) direction decreased gradually when the frequency was changed from 5 to 100 Hz. This may be because the PDMS bump experiences greater deformation in the shear loading test than in the normal loading test, which induces a more significant damping effect.

Sensors 2016, 16, 819

11 of 15

100 Hz. This Sensors 2016, 16,may 819 be because the PDMS bump experiences greater deformation in the shear loading 11 of 15 test than in the normal loading test, which induces a more significant damping effect.

Figure 8. 8. Frequency Frequency response response of of the the tactile tactile unit. unit. (a) (a)Average Average amplitude amplitude (100 (100 wave wave periods) periods) of of Figure sinusoidal output charge Q versus sinusoidal input force F at different frequencies; (b) average z at different frequencies; (b) average sinusoidal output charge zz versus amplitude (100 (100 wave wave periods) periods) of of sinusoidal sinusoidaloutput outputcharge chargeQQx versus versus sinusoidal sinusoidal input inputforce force Fx at at amplitude x different frequencies. frequencies. different

4.3. Dynamic Dynamic Three-Axis Three-Axis Force Force Measurement Measurement 4.3. Inthe thecalibration calibration tests, external was applied onaxes three axes separately develop the In tests, external forceforce was applied on three separately to developtothe data–fitted data–fitted model in(4)Equations (4) and (5). inmanipulation, a dexterous manipulation, force in components model in Equations and (5). Actually, in aActually, dexterous force components three axes in three axes are always applied simultaneously, forming resultant a three-axis resultant force.the Todeveloped verify the are always applied simultaneously, forming a three-axis force. To verify developed data–fitted model in the reconstruction of the applied three-axis a dynamic three-axis data–fitted model in the reconstruction of the applied three-axis force, aforce, dynamic three-axis force force measurement hasconducted. been conducted. The illustration and photograph of experimental is measurement has been The illustration and photograph of experimental setup issetup shown shown in 9a,b. Figure 9a,b. The brass loading bar is utilized a dynamic three-axis each in Figure The brass loading bar is utilized to applytoaapply dynamic three-axis force onforce eachontactile tactile unit. Adhesive tape is adhered to the surface bottom of surface of the bar. the bar with contacts unit. Adhesive tape is adhered to the bottom the brass bar.brass When theWhen bar contacts the withsurface the topofsurface of the the tape sticks to the bump. Then, by automatically the top the bump, thebump, tape sticks to the bump. Then, by automatically moving themoving bar in the bar in the x-, yand z-axes simultaneously, a dynamic three-axis force F, which is composed of force -, - and -axes simultaneously, a dynamic three-axis force F, which is composed of force components Fx ,, Fyy and Fzz,, will to the the six-axis six-axis components and F will be be applied applied on on the the tactile unit. The brass bar is attached to load cell which is mounted on an electric three-axis motion platform (M-VP-25XL-XYZ, Newport, RI, load cell which is mounted on an electric three-axis motion platform (M-VP-25XL-XYZ, Newport, RI, USA), as in Figure 9b. The platform can deliver reliable positioning USA), asshown shown in Figure 9b. motion The motion platform canhighly deliver highly reliable performance positioning at a maximum of 25 mm/s. of Magnitudes applied forceofcomponents cancomponents be measuredcan by the performance atspeed a maximum speedMagnitudes of 25 mm/s. applied force be six–axis load cell, while output charges (Q , Q , Q and Q ) of the tactile unit are measured by the measured by the six–axis load cell, while 11output ( 22 , , and ) of the tactile unit 12 charges 21 charge amplifier. are measured by the charge amplifier. Different dynamic dynamicthree-axis three-axisforces forceswere wereapplied appliedonon each tactile unit U(tactile unit located Different each tactile unit unit located at ij (tactile at the i-th row and j-th column, as shown in Figure 1a, i = 1, 2, 3, j = 1, 2). Figure 9c shows the i-th row and j-th column, as shown in Figure 1a, i = 1, 2, 3, j = 1, 2). Figure 9c shows a typical a typical from response from tests, in which aforce three-axis forceofcomposed of force components response one of the one tests,ofinthe which a three-axis composed force components Fx = 0.4 N, Fx “ 0.4 N, Fy “ 0.36 N, and Fz “ ´1.45 N was applied on tactile unit U12 . The variation of applied

Sensors 2016, 16, 819 Sensors 2016, 16, 819

12 of 15 12 of 15

Fy = 0.36 N,and Fz = -1.45 N was applied on tactile unit U12 . The variation of applied force measured by themeasured load cell was shown in the half responses from four piezoelectric force by the load cell wasupper shown in of theFigure upper 9c, halfand of Figure 9c, and responses from four capacitors were shown in the bottom half of Figure 9c. It can be seen that when the that applied force piezoelectric capacitors were shown in the bottom half of Figure 9c. It can be seen when the changes, output charges Q , Q , Q and Q change correspondingly at the same time. By 12 2111 , Q12 ,22Q21 and Q22 change correspondingly at the same time. applied force changes, output11charges Q substituting Q11Q , Q12 ,QQ21, and Q in Equations 4 and 5, three force components F , F and F can 22 By substituting , Q and Q 11 12 21 22 in Equations 4 and 5, three force components xFx , yFy and zFz can be becalculated. calculated.

Figure force on on each eachunit unitof ofthe thesensor sensorarray; array;(b) (b)photograph photograph Figure9.9.(a) (a)Illustration Illustration of of applying applying three-axis force of 1 2 3 1 2 3 the force loading system ( (○ the the fabricated tactile sensor array,

the○brass loading bar, bar, the six-axis of the force loading system fabricated tactile sensor array, the brass loading ○ the 4 three-axial 4 three-axial load cell,

motion platform); and (c) response a tactileof unit to a three-axis six-axis load cell, ○ motion platform); and (c) of response a tactile unit to aforce. three-axis force.

ˇ Table 1 summarizes ˇ applied and calculated forces from all the tests. Percentage errors, calculated ˇ Table 1 summarizes ˇ applied and calculated forces from all the tests. Percentage errors, calculated as ˇpFcalc ´ Fappl q{Fappl ˇˆ100% (Fcalc is the calculated force component, Fappl is the applied force as |( − )/ | × 100% ( is the calculated force component, is the applied force component) are also presented. It can be seen that the calculated forces basically match the applied component) are also presented. It can be seen that the calculated forces basically match the applied forces. However, several relatively large errors were generated as the applied force components are forces. However, several relatively large errors were generated as the applied force components are small. The maximal error as large as 37.5% was observed when an x-axis force component as small as small. The maximal error as large as 37.5% was observed when an x-axis force component as small 0.08 N was applied. Since Equations (4) and (5) are data-fitted models, errors always exist between as 0.08 N was applied. Since Equations (4) and (5) are data-fitted models, errors always exist between the data points and the fitted curves. Meanwhile, errors brought by noise also always exist. So, when the data points and the fitted curves. Meanwhile, errors brought by noise also always exist. So, when the applied force component is small, it is more likely to generate large percentage errors because the the applied force component is small, it is more likely to generate large percentage errors because the applied force component gets close to the error of the fitted-model and noise. Overall, the average applied force component gets close to the error of the fitted-model and noise. Overall, the average error is 10.68% with a standard deviation of 6.84%. This error is acceptable in applications aiming to error is 10.68% with a standard deviation of 6.84%. This error is acceptable in applications aiming to mimic human tactile sensing. It is reported that the just noticeable difference (JND) estimated from the mimic human tactile sensing. It is reported that the just noticeable difference (JND) estimated from sensation produced by passive compression of human finger skin ranges from 89.29% to 22.92% of the sensation produced by passive compression of human finger skin ranges from 89.29% to 22.92% applied basic weight when the basic weight changes from 20 to 200 g (0.2 to 2 N) [26]. of applied basic weight when the basic weight changes from 20 to 200 g (0.2 to 2 N) [26].

Sensors 2016, 16, 819

Sensors 2016, 16, 819

13 of 15

Table 1. Applied and calculated force components on the sensor array.

13 of 15

Fy Fz Fx Tactile Unit Appl. Calc. |Err.| Appl. Calc. |Err.| Appl. Calc. |Err.| Table 1. Applied and calculated force components on the sensor array. (N) (N) (%) (N) (N) (%) (N) (N) (%) U11 0.16 0.15 6.25 0.22 13.64 −0.595 −0.55 7.56 Fx F0.19 Fz y U 11 0.34 0.29 14.71 0.33 0.30 9.09 −0.95 −0.885 6.84 Tactile Unit Appl. Calc. |Err.| Appl. Calc. |Err.| Appl. Calc. |Err.| U12 0.08 0.11 37.50 0.32 0.30 6.25 −0.695 −0.655 5.76 (N) (N) (N) (N) (N) (%) U12 0.40 0.42 (%) 5.00 0.36 (N) 0.39 (%) 8.33 −1.45 −1.35 6.90 U 0.16 0.15 ´0.55 7.56 U1121 0.23 0.19 6.25 17.39 0.22 0.08 0.19 0.05 13.64 37.50 ´0.595 −0.545 −0.495 9.17 U11 0.34 0.29 14.71 0.33 0.30 9.09 ´0.95 ´0.885 6.84 U21 0.12 0.14 16.67 0.45 0.49 8.89 −1.02 −0.92 9.80 U12 0.08 0.11 37.50 0.32 0.30 6.25 ´0.695 ´0.655 5.76 U 22 0.43 0.46 6.98 0.41 0.43 4.88 −1.32 −1.2 9.09 U12 0.40 0.42 5.00 0.36 0.39 8.33 ´1.45 ´1.35 6.90 U2122 0.34 0.30 17.39 11.76 0.08 0.12 0.05 0.14 37.50 16.67 ´0.545 −0.745 ´0.495 −0.63 15.44 U 0.23 0.19 9.17 U 0.12 0.14 0.45 ´0.92 9.80 U2131 0.15 0.16 16.67 6.67 0.19 0.49 0.17 8.89 10.53 ´1.02 −0.475 −0.38 20.00 U 0.43 0.46 0.41 ´1.2 9.09 U2231 0.21 0.22 6.98 4.76 0.44 0.43 0.39 4.88 11.36 ´1.32 −0.99 −1.05 6.06 U22 0.34 0.30 11.76 0.12 0.14 16.67 ´0.745 ´0.63 15.44 U32 0.44 0.45 6.67 2.27 0.16 0.17 0.15 10.53 6.25 ´0.475 −0.96 −0.89 7.29 U31 0.15 0.16 0.19 ´0.38 20.00 U 32 0.20 0.19 5.00 0.29 0.30 3.45 −0.73 −0.665 8.90 U31 0.21 0.22 4.76 0.44 0.39 11.36 ´0.99 ´1.05 6.06 U32 0.44 0.45 2.27 U 0.20 0.19 5.00 32 4.4. Signal-to-Noise Ratio (SNR) Analysis

0.16 0.29

0.15 0.30

6.25 3.45

´0.96 ´0.73

´0.89 ´0.665

7.29 8.90

To evaluate the influence of the noise on measurement accuracy, signal-to-noise ratio analysis 4.4. Signal-to-Noise Ratio (SNR) Analysis was carried out. Figure 10a shows the response of a piezoelectric capacitor in a tactile unit when no forceTo was applied.the Figure 10b shows the piezoelectric capacitor when a 20 Hz sinusoidal evaluate influence of theresponse noise on of measurement accuracy, signal-to-noise ratio analysis force in the z-axis direction applied on the top of the tactile unitinusing the unit experimental was carried out. Figure 10a was shows the response of asurface piezoelectric capacitor a tactile when no setup and applied. process described the sensor unit calibration tests. Thecapacitor signal-to-noise calculated by force was Figure 10binshows response of the piezoelectric when aratio 20 Hz sinusoidal Equation (6)z-axis is 40.35 dB. force in the direction was applied on the top surface of the tactile unit using the experimental setup and process described in the sensor unit calibration tests. The signal-to-noise ratio calculated by P SNR  10 log( s ) Equation (6) is 40.35 dB. (6) PPns SNR “ 10 logp q (6) Pn where is the signal-to-noise ratio of the sensor, is the power of the signal, and is the power where SNR is the signal-to-noise ratio of the sensor, Ps is the power of the signal, and Ps is the power of the noise. of the noise.

Figure Figure 10. 10. (a) (a) The The response response of of aa piezoelectric piezoelectric capacitor capacitor in in aa tactile tactile unit unit when when no no force force was was applied; applied; and (b) the response of a piezoelectric capacitor when a 20 Hz sinusoidal force in z-axis direction and (b) the response of a piezoelectric capacitor when a 20 Hz sinusoidal force in z-axis direction with with amplitude of the the tactile tactile unit. unit. amplitude of of about about 55 N N was was applied applied on on top top surface surface of

Sensors 2016, 16, 819

14 of 15

5. Conclusions In this paper, we have successfully demonstrated a new flexible dynamic three-axis tactile sensor array using PVDF film as the sensing material. The PVDF film in the array is horizontally arranged, which makes the fabrication process compatible with surface micromachining technology, enabling accurate control of sensor sizes and good spatial resolution. A preliminary prototype sensor with 3 ˆ 2 tactile units has been fabricated and tested. The results of calibration experiments demonstrate that the tactile unit has good linearity when sinusoidal stimulations with amplitudes in the range of 0–0.5 N, 0–0.5 N and 0–1.5 N were applied on x-, y- and z-axes separately. The calculated nonlinearities for x–, y– and z–axes are 2.45%, 2.37% and 1.74% F.S., respectively. The estimated sensitivities measured with the current setup are 14.93, 14.92 and 6.62 pC/N for x-, y- and z-axes, respectively. Furthermore, the sensor shows relatively low coupling effect, high repeatability, and acceptable frequency response from 5 to 400 Hz in both normal and shear load tests. By the results of calibration experiments, a date-fitted model has been established to calculate three components of the applied dynamic three-axis force. The calculated force components have a discrepancy of 10.68 ˘ 6.84% from the applied three-axis forces. Obviously, structural dimensions of the sensor will affect the sensing performance of the developed sensor array. For example, a higher bump results in a larger torque when a shear force is applied on the top of the bump, which will increase the sensitivities for the x- and y-axes. Therefore, structural dimensions of the sensor will be optimized in our future work. Meanwhile, slippage and texture encoding in artificial hand applications will also be focused on. In addition, external noise is a source of measurement errors in our sensor, especially when the applied force is small. This problem will be paid attention to in future. One of the effective solutions is to add an electromagnetic shield on the surface of the sensor. Acknowledgments: The activity presented in this paper is partially support by the National Basic Research Program (973 Program) of China (2011CB013303), and the Science Fund for Creative Research Groups of National Natural Science Foundation of China (51521064). Author Contributions: Ping Yu and Weiting Liu conceived and designed the sensor; Ping Yu performed the experiments, analyzed the data and wrote the paper; Chunxin Gu and Xiaoying Cheng offered help in the process of experiments; Weiting Liu and Xin Fu contributed reagents and analysis tools. Conflicts of Interest: The authors declare no conflict of interest.

References 1. 2. 3. 4. 5.

6. 7. 8.

Johansson, R.S.; Flanagan, J.R. Coding and use of tactile signals from the fingertips in object manipulation tasks. Nat. Rev. Neurosci. 2009, 10, 345–359. [CrossRef] [PubMed] Yousef, H.; Boukallel, M.; Althoefer, K. Tactile sensing for dexterous in-hand manipulation in robotics-A review. Sens. Actuators A-Phys. 2011, 167, 171–187. [CrossRef] Kappassov, Z.; Corrales, J.-A.; Perdereau, V. Tactile sensing in dexterous robot hands-Review. Robot. Auton. Syst. 2015, 74, 195–220. [CrossRef] Oddo, C.M.; Beccai, L.; Felder, M.; Giovacchini, F.; Carrozza, M.C. Artificial roughness encoding with a bio-inspired MEMS-based tactile sensor array. Sensors 2009, 9, 3161–3183. [CrossRef] [PubMed] Koiva, R.; Zenker, M.; Schurmann, C.; Haschke, R.; Ritter, H.J. A highly sensitive 3D-shaped tactile sensor. In Proceedings of the IEEE/ASME International Conference on Advanced Intelligent Mechatronics, Wollongong, Australia, 9–12 July 2013; pp. 1084–1089. Dahiya, R.S.; Metta, G.; Valle, M.; Adami, A.; Lorenzelli, L. Piezoelectric oxide semiconductor field effect transistor touch sensing devices. Appl. Phys. Lett. 2009, 95, 034105. [CrossRef] Adami, A.; Dahiya, R.; Collini, C.; Cattin, D.; Lorenzelli, L. POSFET touch sensor with CMOS integrated signal conditioning electronics. Sens. Actuators A-Phys. 2012, 188, 75–81. [CrossRef] Chuang, C.-H.; Liou, Y.-R.; Chen, C.-W. Detection system of incident slippage and friction coefficient based on a flexible tactile sensor with structural electrodes. Sens. Actuators A-Phys. 2012, 188, 48–55. [CrossRef]

Sensors 2016, 16, 819

9. 10. 11.

12. 13. 14. 15. 16. 17.

18.

19.

20. 21. 22. 23. 24. 25. 26.

15 of 15

Lee, H.-K.; Chung, J.; Chang, S.-I.; Yoon, E. Normal and shear force measurement using a flexible polymer tactile sensor with embedded multiple capacitors. J. Microelectromech. Syst. 2008, 17, 934–942. Liang, G.; Wang, Y.; Mei, D.; Xi, K.; Chen, Z. Flexible capacitive tactile sensor array with truncated pyramids as dielectric layer for Three-Axis force measurement. J. Microelectromech. Syst. 2015, 24, 1510–1519. [CrossRef] Iwasaki, T.; Takeshita, T.; Arinaga, Y.; Uemura, K.; Ando, H.; Takeuchi, S.; Furue, M.; Higurashi, E.; Sawada, R. Shearing force measurement device with a built–in integrated micro displacement sensor. Sens. Actuators A-Phys. 2015, 221, 1–8. [CrossRef] Ramadan, K.S.; Sameoto, D.; Evoy, S. A review of piezoelectric polymers as functional materials for electromechanical transducers. Smart. Mater. Struct. 2014, 23, 033001. [CrossRef] Shirafuji, S.; Hosoda, K. Detection and prevention of slip using sensors with different properties embedded in elastic artificial skin on the basis of previous experience. Robot. Auton. Syst. 2014, 62, 46–52. [CrossRef] Cutkosky, M.R.; Ulmen, J. Dynamic tactile sensing. In The Human Hand as an Inspiration for Robot Hand Development; Springer: Basel, Switzerland, 2014; pp. 389–403. Lee, H.-K.; Chang, S.-I.; Yoon, E. A flexible polymer tactile sensor: Fabrication and modular expandability for large area deployment. J. Microelectromech. Syst. 2006, 15, 1681–1686. [CrossRef] Kolesar, E.S.; Dyson, C.S. Object imaging with a piezoelectric robotic tactile sensor. J. Microelectromech. S. 1995, 4, 87–96. [CrossRef] Seminara, L.; Pinna, L.; Valle, M.; Basiricò, L.; Loi, A.; Cosseddu, P.; Bonfiglio, A.; Ascia, A.; Biso, M.; Ansaldo, A. Piezoelectric polymer transducer arrays for flexible tactile sensors. IEEE Sens. J. 2013, 13, 4022–4029. [CrossRef] Choi, B.; Choi, H.R.; Kang, S. Development of tactile sensor for detecting contact force and slip. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems, Edmonton, AB, Canada, 2–6 August 2005; pp. 2638–2643. Drimus, A.; Petersen, M.B.; Bilberg, A. Object texture recognition by dynamic tactile sensing using active exploration. In Proceedings of the 21st IEEE International Symposium on Robot and Human Interactive Communication, Paris, France, 9–13 September 2012; pp. 277–283. Wakatsuki, N.; Kagawa, Y.; Haba, M. Tri-axial sensors and actuators made of a single piezoelectric cylindrical shell. IEEE Sens. J. 2004, 4, 102–107. [CrossRef] Zhang, T.; Liu, H.; Jiang, L.; Fan, S.; Yang, J. Development of a flexible 3–D tactile sensor system for anthropomorphic artificial hand. IEEE Sens. J. 2013, 13, 510–518. [CrossRef] Rajala, S.; Tuukkanen, S.; Halttunen, J. Characteristics of piezoelectric polymer film sensors with solution-processable graphene-based electrode materials. IEEE Sens. J. 2015, 15, 3102–3109. [CrossRef] Dahiya, R.S.; Valle, M.; Lorenzelli, L. SPICE model for lossy piezoelectric polymers. IEEE Trans. Ultroson. Ferr. 2009, 56, 387–395. [CrossRef] [PubMed] Dargahi, J.; Parameswaran, M.; Payandeh, S. A micromachined piezoelectric tactile sensor for an endoscopic grasper–theory, fabrication and experiments. J. Microelectromech. Syst. 2000, 9, 329–335. [CrossRef] Piezotech, S.A.S. Piezoelectric Films Technical Information. Technical Manual. Available online: http://www.piezotech.fr/ (accessed on 20 March 2016). Raj, D.V.; Ingty, K.; Devanandan, M. Weight appreciation in the hand in normal subjects and in patients with leprous neuropathy. Brain 1985, 108, 95–102. [CrossRef] [PubMed] © 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).

Flexible Piezoelectric Tactile Sensor Array for Dynamic Three-Axis Force Measurement.

A new flexible piezoelectric tactile sensor array based on polyvinylidene fluoride (PVDF) film is proposed for measuring three-axis dynamic contact fo...
5MB Sizes 0 Downloads 14 Views