Development of tactile sensor with functions of contact force and thermal sensing for attachment to intelligent robot finger tip Jong-Ho Kim, Woo-Chang Choi, Hyun-Joon Kwon, Dae-Im Kang Division of Physical Metrology Korea Research Institute of Standards and Science (KRISS) Daejeon, South Korea
[email protected] Abstract— This paper describes the design and fabrication of a flexible tactile sensor using polymer micromachining technologies, which can measure the three-component force and temperature caused by thermal conductivity. The tactile sensor is comprised of sixteen micro force sensors with 4 x 4 arrays and its size being 13 mm x 18 mm. Each micro force sensor has a square membrane, and its force capacity is 0.6 N in the three axis directions. The normal and shear forces are obtained from four thin-film metal strain gauges embedded in a polyimide thin-film membrane. The optimal positions of the strain gauges are determined by the strain distribution obtained form finite element analysis. In order to acquire force and temperature signals simultaneously from the fabricated tactile sensor, we fabricated a PCB based on multiplexer, operational amplifier and microprocessor with CAN network function. The fabricated flexible tactile sensor was evaluated from the evaluation apparatus. Also we confirmed that the sensor has the function of temperature sensing throughout feasibility test such as contacting with some objects.
I. INTRODUCTION Generally a robot must have sensory input similar to or superior to the human senses in order to replace or serve as extensions of humans in dangerous, delicate, or remote applications. In case of especially dexterous manipulation tasks, robot requires tactile perception of objects through mechanical contact. Thus measurement and processing of contact force information has also been shown to be of great importance in robotic dexterous manipulation. When human’s fingertip comes into contact with an object, a contact force profile has three components: one component oriented normal to the surface and two components oriented tangentially to the surface. Sensing and processing these triaxial force profiles provides humans with a rich source of information about their physical environments. Especially, the shear force component is very of important from a
viewpoint of slip detection between fingertip and object. Shear force information is useful in dexterous robotic manipulation. For example, shear information is required for full grasp force and torque determination, especially when grasp forces are in the range of 0.01 – 0.1 N, where fingertip force-torque sensors are ineffective. In addition, measuring shear force enables prediction and determination of object slip, as well as estimation of static coefficient of friction[1]. On the other hand, recently, some researchers have developed MEMS-based tactile sensors. Especially work has mainly focused on silicon based sensors that use piezoresisitive[2] or capacitive sensing[3], and polymerbased approaches that use piezoelectric polymer films[4] for sensing. Actually, the silicon substrate used for micomachined tactile sensors presents fundamental limitations. The silicon material is mechanically brittle, and hence not capable of sustaining large deformation and sudden impact. Additionally, the silicon substrate is rigid. It is thus difficult to form continuous and conformal tactile sensor that can be used to cover contoured surfaces. Meanwhile, flexible MEMS-based tactile sensors have been fabricated using polyimide film[5]. However, the sensor was mainly capable of measuring only normal force component. On the other hand, in case of fingertip of robot hand, sensors should satisfy requirements as: linearity, low hysteresis, durability, mechanical flexibility, and robustness. Polymer materials can be satisfied with the points mentioned above. In order to apply for robot fingertip with size like human’s fingertip, miniature polymer-based three-axial tactile sensors are able to be a solution. Therefore the primary goal of the paper is to design a three-component force sensor array capable of measuring full triaxial force component. In addition, we present the flexible tactile sensor which can also measure temperature caused by thermal conductivity simultaneously like human’s fingertip. Commercial polyimide film is used as the substrate of sensor
to give the flexibility which makes it possible to attach to the round surface such as column. Four strain gauges are positioned independently in order to measure the three components of applied forces through a resistance change. Finally, the fabricated tactile sensor is evaluated using the equipment with three-component load cell. II.
DESIGN OF UNIT CELL AND TACTILE SENSOR
The sensor consists of a flexible sensing structure with four strain gauges. An external force applied to the sensor can be resolved into its three components x, y, z as shown in Fig 1. In order to obtain maximum sensitivity, regions of high stress have been identified and the dimensions and position of strain gauges have been determined accordingly. We designed a unit of the three-axial tactile sensor as shown in Fig 1(a). The size of the unit has 2500 m x 2500 m and the thickness of membrane is 50 m. The thickness of membrane has been decided that the force capacity of sensor is equivalent to the maximum load for tactile sensors to be applied in artificial fingertips. The size of membrane is 1500 m x 1500 m. On the other hand, the substrate of sensor use DuPont Kapton HN200 polyimide sheet with the thickness of 125 m. The schematic diagram of the designed three-axial tactile sensor with four resistors (S1 ~ S4) is shown in Fig 1(b). In case of Fx or Fy loading, the loading block can improve the sensitivity due to the effect of moment. The loading block is column type with the diameter of 600 m and the length of 600 m and above. Meanwhile the loading block and the side block were assumed as a rigid body due to thick thickness compared with thin membrane of sensing element.
The commercial finite element program, ANSYS ver. 5.7, is used in order to design sensing element. The used material is polyimide film that has elastic modulus of 3.4 GPa, ultimate stress of 190 MPa, and Poisson’s ratio of 3.3. We use the finite element model that consists of the shell element with four nodes. The supporting block of the sensing element except for membrane with thickness of 50 µm is constrained at bottom. Normal load, Fz, is applied uniformly to the upper surface of bump. Each shear load, Fx, Fy is applied uniformly to the lateral face of bump respectively. The applying maximum force is 0.6 N considering the grasping force of robot hand. Fig 2 shows the strain distribution in the direction x at the location along centerline of membrane when each maximum load is applied in three directions respectively. In case of Fz loading, the strain distribution along x-axis of membrane shows the symmetric distribution with respect to y-axis. Meanwhile, when the force, Fx, or Fy is applied to the lateral face of bump, the strain distribution along the centerline indicates the asymmetric inclination with respect to origin point. In case of Fy loading, the strain distribution along the x-axis has almost zero value. Using these strain distribution values, the shape of strain gauge and its size was determined in order to maximize the sensitivity of the sensor. When a strain gauge is located around the periphery of membrane, the average strain value of strain gauge has 4500 µm in case of Fz loading. Meanwhile the average strain value is 1000 µm when Fx loading is applied. Thus the ratio of the sensitivity of Fz loading has 4.5 times over Fx loading. On the other hand, when the sensing element is subjected to three loadings, Fx, Fy and Fz simultaneously, the strain distributions is very important in order to extract each force component signals without interference from signals of four strain. Using the strain distributions at locations of strain gauges, the solution of the decoupling method for obtaining three force components signals can be represented as follows. 0.06
FZ Loading FX Loading
0.04
(a) Strain
0.02 0.00 -0.02 -0.04 -0.06 -1200
-800
-400
0
400
800
1200
Distance (µm)
(b) Figure 1. Schematic diagram of the dimension for the designed force sensor; (b) Schematic perspective view of a single tactile sensor with four strain gauges S1 ~ S4.
Figure 2. Strain distribution obtained from FEA analysis at the location along the centerline of membrane when each loading (Fz, Fx) of maximum load 0.6 N is applied respectively.
In order to measure three-component force and temperature due to thermal conductivity simultaneously, we suggest the shape of resistors shown in Fig 1. with S1, S3 in
symmetry with respect to y-axis, and S2, S4 in symmetry with respect to x-axis, respectively. In case of Fz loading, the S2 and S4 resistors increase due to the tensile strain, while the S1 and S3 resistors decrease due to the shape of resistors[6]. Finally the triaxial force and temperature components are as follows: ∆VF z = ∆VS 1 − ∆VS 2 + ∆VS 3 − ∆VS 4 ,
(1)
for Fx loading, using strain gauge signal S1 and S3, ∆VF x = ∆VS 1 − ∆VS 3 ,
(2)
for Fy loading, in a similar way, ∆VF y = ∆VS 2 − ∆VS 4 ,
(3)
for temperature component, (4)
∆VTemp = ∆VS 1 + ∆VS 2 + ∆VS 3 + ∆VS 4
Here ∆V means the change of voltage due to the change of resistor. The number of line for data collection is needed to cut down in order to minimize the integration circuit. Fig 3(a) presents the schematic diagram of a flexible tactile sensor which is composed of 16 (4 x 4) three-component force sensors. In case of tactile sensor shown in Fig 3(a), if constant current source or voltage source is used, the number of input lines to extract each force component without interference is 8 and that of output lines is also 8. That is to say, an input line is connected to five force sensors simultaneously. An output line is connected to 16 five force sensors. For example, let us imagine a tactile sensor with three-component force sensors of n x n matrix. Then the number of input lines is 2n and that of output lines being also 2n [7].
(a)
(b)
Figure 3. Schematic diagram of a tactile sensor having triaxial force sensors(4 x 4) and one thermal sensor, and its fabricated sensor using microfabrication technology.
III. FABRICATION OF TACTILE SENSOR The fabrication processes are as follows [8]. The substrate of sensor use DuPont Kapton HN200 polyimide sheet with the thickness of 125 m. In order to smooth the surface of the film, polyimide layer with thickness of 3 µm is spun on top surface of the film after dehydration baking. The polyimide layer is cured at 350 ℃ for 2 hours. Next, Ni-Cr alloy (Ni-Cr, 80:20) material is used as strain gauges due to its high resistivity and low coefficient of thermal resistance (TCR). 400 Å of Ni-Cr is defined on 100 Å of Cr adhesion layer using e-beam evaporator and lifted off. The first wiring metal layer of 2000 Å of Au is patterned on 200 Å of Cr using the same process as strain gauge’s step. After the patterning of strain gauges and first wiring layer, the first passivation layer is spun on, patterned and cured using photo-definable polyimide (HD4000), its thickness is 3 µm. The HD4000 series polyimde is chosen due to its unique characteristic that is solvent developed and photo-definable. The second wiring layer, 3000 Å of Au is deposited on 300 Å of Cr adhesion layer using previous metal deposition method and then lifted off. In order to protect metal layer and define contact hole, second passivation layer is defined the same method and material as first passivation layer’s. Finally, the Kapton polyimide film is etched to define membrane. Thick photoresist layer is used as etching mask of the film in KOH solution. The used photoresist does not dissolve in KOH solution and is inexpensive compared with plated Cu. And then the wet etching of the film is carried out in KOH solution. After etching at 70 ℃ for 15 minutes, the membrane of thickness has 50 µm. A photograph of the completed tactile sensor is shown in Fig 3(b). The mold for loading bump was made by using steel plate and the polyurethane rubber was inserted into the mold. The assembly is left in place to cure overnight. Then the loading bump was removed from the steel mold. Fig 4 represents a view of the fabricated PCB with microprocessor (C8051F061, Cygnal Co.) and 16 bit A/D converter, 12 bit D/A converter, Mux chip and CAN network for tactile sensor. The signal was amplified with 100 times compared with original signal. The size of the board is 50 mm x 50 mm. The algorithm of choosing the input/output signal line and calculation of each force and temperature component was embedded into microprocessor. The sensing element of tactile sensor was connected to the contact pad of the simple flexible PCB using ACF film (Sony, Co.) bonding technique. The connection of the FPCB to the fabricated signal processing board was achieved by the signal lines. Additionally, the tactile sensor was attached to fingertip of robot hand which has three fingers using adhesive film (3M, Co.) In reference, we used robot hand manipulator fabricated by KIST (Korea Institute of Science and Technology) research center which performs the Intelligent Robotics Development Program. Finally the microprocessor with
CAN network sends the data of tactile array signal to computer which has data display program developed by LabViewTM software[9].
Figure 5. View of an evaluation system based on three-component loadcell (capacity 10 N) for triaxial force sensor.
When the applied load, Fz, increases from 0 to 0.6 N along the z-direction, Fig 6 represents the output voltage with respect to signal obtained from load cell. The slope of output voltage over force shows 0.52 V/N which was obtained by using least square method. On the other hand, in case of Fx loading, the probe tip of load cell moves along xaxis direction as shown in Fig 3(a). Then the tip contacts with the side face of loading bump as shown in Fig 7. Thus the unit cell shows the slope with 0.23 V/N. On the other hand, Fig 8 shows that the slope of voltage over Fy loading was 0.28 V/N. Thus the sensitivity of Fz loading over Fx, Fy loadings was about 2 times. 0.5 0.4
Experiment 0.52 V/N
0.3 Volt (V)
Figure 4. View of a fabricated PCB with microprocessor(C8051F061, Cygnal Co.) and CAN network for tactile sensor with 16 triaxial force sensors and one thermal sensor.
0.2 0.1 0.0 0.0
0.2
0.4
0.6
0.8
Load cell (N)
Figure 6. View of an evaluation system based on three-component loadcell (capacity 10 N) for triaxial force sensor and its result of evaluation under Fz loading. 0.20 Experiment 0.23 V/N
Volt(V)
0.15
0.10
0.05
0.00 0.0
0.2
0.4
0.6
0.8
Load cell(N)
Figure 7. View of an evaluation system based on three-component loadcell for triaxial force sensor and its result of evaluation under Fx loading. 0.20
Experiment 0.28 V/N
0.15 Volt (V)
IV. EXPERIMENT RESULTS AND DISCUSSION The fabricated tactile sensor was calibrated using the evaluation system for triaxial force sensor. Fig 5 shows the fabricated tester based on three-component load cell (capacity 10 N) to evaluate the three-component force sensor. The tester has a universal stage, which makes threecomponent loading, Fx, Fy, and Fz, apply to the tactile sensor by rotating the three axis of the stage. The tester based on three-component load cell has the rated capacity of +10 N in three-component force. The three-component load cell was connected to signal conditioning amplifier (DMC 9012A; HBM Co.). Additionally, the tester has vision system which needs to align the loading bump with probe tip attached to load cell.
0.10
0.05
0.00 0.0
0.2
0.4 Load cell (N)
0.6
0.8
Figure 8. View of an evaluation system based on three-component loadcell for triaxial force sensor and its result of evaluation under Fy loading.
The force component, Fx, Fy, showed a linearity under applied force component, while the component, Fz, presents the second-order fitting curve. Using the slopes shown in Figs 6~8, we calculated the interference error of other two force components in case of each applying load. The interference error was about 30% compared with the applying force component. However, the interference error was significantly large than commercial multi-component load cell. Exact location of bump and its size will be needed to decrease the interference error. Meanwhile the sensitivity according to temperature change was not achieved yet. Instead of that, we performed the feasibility test by contacting surface of sensor with some objects such as human being’s finger, metal, and so on. Thus we confirmed the possibility of temperature sensor caused by thermal conductivity as well as three-component force sensor. In near future, more experiments will be conducted by using temperature chamber. On the other hand, the requirement of tactile sensor in the robot manipulator field needs harsh environment test such as impact, wear and repeated loading since the control for grasping work in robot manipulation field requires the robust tactile sensor. Additionally, the signal processing board must be small size to integrate the sensor system to robot hardware system. V.
CONCLUSIONS AND FUTURE PLAN
The three-axial flexible tactile sensor has been designed and fabricated, which is comprised of micro force sensor (4 x 4). Additionally, the sensor can measure the temperature change. Each strain gauge has a square membrane, and its force capacity is 0.6 N in the three-axis direction. The polyimide microfabrication technique was used to fabricate the sensing element of the tactile sensor. In order to confirm the characterization of the sensor, the three-axial tactile sensor has been evaluated by applying normal and tangential force between 0 and 0.6 N using the evaluation system. The three-axial tactile sensor presented the sensitivity of 0.52
V/N in the normal loading and about 0.25 V/N for tangential loading. Additionally the sensor showed the function of temperature sensing caused by thermal conductivity throughout contacting with human being’s finger or metal, and so on. However, modification of loading bump and more tests such as temperature test, impact, fatigue loading tests are needed to develop functional tactile sensor array for use in dexterous robotic manipulation, medicine and industry field. ACKNOWLEDGMENT This paper was performed for the Intelligent Robotics Development Program, one of the 21st Century Frontier R&D Programs funded by the Ministry of Science and Technology of Korea. REFERENCES [1] [2]
[3]
[4] [5]
[6]
[7]
[8]
[9]
K. H. Lee, and H. R. Nicholls, “Tactile sensing for mechatronics-a state of the art survey,” Mechatronics, Vol. 9, pp.1-31, 1999. B. J. Kane, M. R. Cutkosy, and G. T. A. Kovacs, “CMOS-compatible traction stress sensor for use in high-resolution tactile imaging,” Sensor and Actuators (A), vol. 54, pp. 511-516, 1996. Y M. Leineweber, G. Pelz, M. Schmidt, H. Kappert, G. Zimmer, “New tactile sensor chip with silicon rubber cover,” Sensors and Actuators (A), vol. 84, pp. 236-245, 2000. E.S. Kolesar and C.S. Dyson, “Object imaging with a piezoelectric robotic tactile sensor,” JMEMS, vol. 4, pp. 87-96, 1995. J. Engle, J. Chen, Z. Fan, and C. Liu, “Polymer micromachined multimodal tactile sensors,” Sensors and Actuators A. Vol. 117, pp.50-61, 2005. J.H. Kim, H.J. Kwon, D.I. Kang, J.H. Choi, M.S. Kim, Y.K. Park, "Input method of data using tactile sensor ", Patent pending, No. 123118, 2005. 12. Korea. J.H. Kim, D.I. Kang, Y.K. Park, M.S. Kim, "Input output wire for tactile sensor using three-component force sensors", Patent pending, No. 219568, 2005. 11, USA. J.H. Kim, W.C. Choi, Y.K. Park, J.I. Lee, M.S. Kim, J.H. Choi and D.I. Kang, “Fabrication method of flexible tactile sensor using polymer film”, Patent pending, No. 102261, 2005. 10, Korea. J. H. Kim, W. C. Choi, H. J. Kwon, J. I. Lee, S. M. Lee, and D. I. Kang, "State of the art of tactile sensor and its application", in Proceedings of the international sensors conference, pp.37-39, Nov. 2005.
