CA2135463A1 - Control of robotic gripping by detection of acoustic emissions - Google Patents

Abstract

A robotic gripper carries piezoelectric sensors on its gripping surfaces to detect acoustic noise for control of the gripping force applied to a grasped object. The applied force passes through the piezoelectric sensors and the sensors are closely acoustically coupled to the gripping interface to detect acoustic noise. The gripping force is increased when detected acoustic emissions indicate that slip is imminent.

Description

213~463 TITLE: CONTROL OF ROBOTIC GRIPPING BY DBTBCTION OF
ACOUBTIC ENIBBIONB

Field of the Invention This invention relates to sensors in the field of robotics. More particularly it relates to robotic gripping mechanisms that control their grip through the detection of acoustic emissions that are a precursor or incipient indicator of slippage.

Background to the Invention Acoustic emissions are vibratory stress waves that occur during the deformation of solids. Such waves arise when solids fracture, deform or undergo phase transitions, and particularly occur at the interface between objects in contact when slippage develops between such objects.
The detection of slippage, or its anticipation, is important in the field of robotics because it relates to the gripping function. When a robotic manipulator grasps an object control must be exercised over the applied grasping force. Too great an applied force can crush an object. Too little an applied force will result in the premature release of the object.
In the prior art, the emission of acoustic energy in conjunction with the onset of slippage between two objects that are frictionally engaged has been the ~35~63 subject of several studies. References relevant to this field include the following:
(1) D. Dornfeld and C. Handy, "Slip Detection Using Acoustic Emission Analysis," Raleigh, North Carolina Proc 1987, IEEEE International Conference Robotics Automation, pp. 1868-75;

(2) S. Rangwala, F. Forouhar, and D. Dornfeld, "Application of Acoustic Emission Sensing to Slip Detection in Robotic Grippers". Int J. Mach Tools Manufacturing Vol 28 No. 3 pp 207-215, 1988, Pergamon Press.

(3) Robert D. Howe; Nicholas Popp; Prasad Akella; Imin Kao and Mark R. Cutkosky, "Grasping, Manipulations, and Control with Tactile Sensing". Proc 1990, IEEE
Conf. Cincinnati, Ohio, May 13 - 18, 1990 at pp 1258 - 1263.

(4) R.D. Howe and M. R. Cutkosky, "Sensing Skin Acceleration for Slip and Texture Perception", Proc. 1989 IEEE International Conference of Robotics and Automation, Scottsdale Arizona, May 1989, pp 145-150.

(5) J. L. Cuttino; C. O. Huey; T. D. Taylor "Tactile Sensing of Incipient Slip", Proceedings of the USA-Japan Symposium on Flexible Automation, 1988 pp.
547-555.

~13~63 -

(6) M.R. Tremblay; W. J. Packhard; and M.R. Cutkosky, "Utilizing Sensed Incipient Slip Signals for Grasp Force Control", Japan/USA Symposium on Flexible Automation, Vol. 2 ASME 1992 pp 1237-1243.
The detection of acoustic emissions has been used in the past to control the grasping force applied to an object. The emissions that have arisen upon the occurrence of minute amounts of slippage have been detected and used to increase the grasping force to arrest the occurrence of further slippage: vis Ref's (1) and (2). See also U.S. patent 4,605,354 to Daly where a slip-grip mechanism is not based on detection of acoustic emissions described.
Piezoelectric sensors have been employed to detect acoustic emissions for use in control of the grip of a robotic manipulator. In the above cited Reference (2) an acoustic emission sensor was attached to an aluminium block pair which were subjected to increasing differential applied forces to induce slippage. At slip initiations, a distinct spike of acoustic emissions was detected. In this report, the acoustic emission detector is depicted as being placed in contact with the side of a grasping end effector, outside the path of the applied force.
The micro-AE signals that arise when slippage is incipient must be detected against a background of 213~463 -noise. The placement of the AE sensor therefore becomes of importance if this phenomena is to be exploited.
It is known to incorporate a piezoelectric sensor beneath the rubber skin on a robotic fingertip located at the point of contact to detect local contact stresses vis, Ref (3). Such a sensor is thus employed to measure applied forces. This same reference describes the use of an accelerometer positioned behind the skin layer of the finger, but not in the path of the gripping force, to measure vibrations (p 1259):
"generated by a number of manipulation events, including the making or breaking of contact, incipient slip and collisions between the grasped object and other objects in the environment".
Such an arrangement is also described in a further Ref.
(4).
The problem of detection of true incipient slip Per se is addressed in Reference (5). This paper reports that:
"Acoustic signals were detected during the transition from static friction to slipping, but, the conditions that yielded signals were difficult to reproduce and no signals at all were obtained if the normal force was high enough to ensure firm, intimate contact over the entire friction surface."
To create useable signals, the authors introduced hard, abrasive particles carried on a rubbery object that is being held in contact with the other, hard, surface at the friction interface, to serve as secondary emitters. In all cases the pick-up point for acoustic emissions is reported as being located collaterally to the force path being applied to the articles being subjected to slip testing.
The use of an incipient slip signal to control a grasping force is described in Reference (6). In this reference an accelerometer mounted behind a rubber nib-covered "skin" is used as a sensor for AE. This detector is reported as having a dynamic range of 1 Hz to 25 kHz.
A further accelerometer was mounted in the object. This paper expressly recommends mounting the finger-tip accelerometer so as to isolate it from the surrounding foam substrate that lies in the force path backing-up the finger-tip skin.
The described experiment in Reference (6) allowed the grasping force to decay until an incipient slip signal was detected, whereupon the grasping force was increased in order to repeat the cycle. The incipient slip signal that was detected arose from the release of elastically deformed nibs formed on the ~135463 _ surface of the contacting skin and not from acoustic emissions generated from generally featureless contacting surfaces.
It is also known to provide robotic manipulators with an array of sensors that provide pressure information over a spatial field. A typical grid of 10 x 10 elements with 2mm spacing is described in Ref [6]. The graphic data shows the presence of pressure sensor signals that commence co-incidently with the initiation of displacements on the order of 1/2 millimetre.
Notwithstanding all of the foregoing disclosures, the inventors of the invention hereafter described have developed new arrangements based on the critical placement of piezoelectric sensors to detect acoustic emissions at the moment when slippage of a grasped object is about to occur. Such signals can usefully be applied to control the gripping force to be applied to a grasped object by a robotic manipulator.
The invention in its general form will first be described, and then its implementation in terms of specific embodiments will be detailed with reference to the drawings following hereafter. These embodiments are intended to demonstrate the principle of the invention, and the manner of its implementation. The invention in its broadest and more specific forms will then be further described, and defined, in each of the individual claims which conclude this Specification.

SummarY of the Invention According to the invention in one of its broader aspects a robotic gripper having gripping surfaces has located beneath at least a portion of such surfaces a piezoelectric acoustic emission sensor material, which sensor material is so positioned as to be subject to stress arising from the gripping of a grasped object by the robotic gripper.
By a further feature of the invention the piezoelectric sensors are closely coupled acoustically to the object being grasped by the presence of an interface layer having an interface surface that is in contact with the grasped object and overlies the piezoelectric material, being efficiently acoustically coupled to such piezoelectric material.
By a further feature of the invention all of the grasping force applied to the grasped object is transmitted through piezoelectric material that is acting as a sensor to detect acoustic emissions. Such force may be directed through multiple piezoelectric sensors that are separately used to detect acoustic emissions.
According to one feature of the invention, the gripping force of the robotic gripper is controlled by a ~13~63 controller which responds to acoustic emissions detected by the sensor and, in particular, cyclically loosens and re-asserts the gripping force based on the detection of acoustic emissions characterized by a trigger signal S level that is above the level of the background noise which occur when slippage between the grasped object and the gripper is imminent. This trigger signal may be set at twice the level of the background noise. More preferably, the trigger signal level may be between 5 and 15. Even more preferably it may be set at substantially ten times the level of the background noise.
The foregoing summarizes the principal features of the invention and some of its optional aspects. The invention may be further understood by the description of the preferred embodiments, in conjunction with the drawings, which now follow.

SummarY of the Figures Figure 1 is a graphic display of the output of the piezoelectric ceramic sensor as a function of time over the period from before to after the commencement of slippage.
Figure 2 is a pictorial depiction of a robotic gripper of the invention supporting an object.
Figure 3 is a cross-section of the gripping interface of the gripper fingers of Figure 2.

~13S~63 _ Figure 4 is a schematic of the control system for Figure 2.

DescriPtion of Preferred Embodiments A piezoelectric ceramic disk 49.5 mm in diameter and 2.5 mm thick was used as an acoustic sensor.
This sensor is of a type that may be acquired from any commercial source that meets United States Military Standard 2376. It was bonded directly to the base of aluminum plate for the test set-up. The material used as the "gripper" interface surface, namely respective layers of alumina, steel and rubber in separate tests, was then bonded to the piezoelectric ceramic disk to provide good m~ch~n;cal and acoustically coupled contact. The test object, consecutively having differing materials at the slip interface, was placed with its bottom surface on the gripper surface with a pivoting beam resting on its top surface. Weights could then be placed on the upper surface of the beam to control the contact force between the test object and the gripper surface. A constant tangential force was applied to the test object using a known weight hanging from a string attached to the test object. When the force was applied, the test object would slide along the gripper surface.
A sensitive linear variable differential transformer (I.VDT) was used to measure the displacement -Z135~63 , of the test object as the constant tangential force was applied and sliding began. The acoustic emission from a slipping object was correlated with its displacement.
The piezoceramic sensor output was amplified by a multi-stage band-pass amplifier wherein it was filtered to limit the measured signal to within a frequency band from about 100 to 2800 kHz. The amplified output of both the piezoceramic sensor and the LVDT were digitized by a high-speed 16-bit analog-to-digital converter installed in a personal computer. The digitized signals were plotted on the computer screen for interpretation. An example of the appearance of the screen after a test run is shown in Figure 1.
Two values were calculated during each test run:
(a) warning time; and (b) displacement at a preselected trigger level of detected acoustic emissions that is an indicator of incipient slip.
The warning time is the delay (in milliseconds) between when the level of signals from the piezoceramic sensor reaches a trigger threshold value and the moment where the test object has moved a total of 0.1 mm from its initial location. The larger this value, the more time is available to apply corrective action to prevent significant slippage from occurring. A preferred trigger ~13S463 level is ten times the level of background noise although values as low as two times or between 5 and lS times may be used. Tests were done at the ten times level.
The displacement at trigger level is the 5distance that the test object has moved from its position at the beginning of the test at the moment when the trigger threshold signal occurs in the output of the piezoceramic sensor. In other words, this is the distance that the test object has moved ~t the point when 10the sensor has determined that slip is occurring or is about to occur.
Both the net displacement of the test object at the trigger point and the warning time are displayed in the upper right corner of the display shown in Figure 1.
15It can be seen from this example that a slip signal could be detected before any noticeable motion of the test object had occurred. That is, the acoustic emissions due to the existence of an incipient slip condition were detected.
20Before the data found in the tables below was collected, the noise characteristics of the test set-up were investigated. The noise level in the output of the piezoceramic was identified to be less than approximately + 10 mV or 10% of the voltage trigger level used in all 25of the experiments. The noise in the displacement signal 2135~63 -from the LVDT corresponded to an error of approximately +3 um.
Experiments were performed using this test set-up to investigate the effects of the following factors on the generation of acoustic emissions prior to and during slip:
(a) the combination of materials used for the gripper surface and the test object;
(b) the degree of surface roughness of the gripper surface and test object; and (c) the normal force applied to the test object.
For each set of test conditions, 20 test runs were performed. The statistics for each set of test conditions are based on the results of each of the 20 tests.
In Table 1, the results for five different test object surfaces sliding on alumina (Al203) as the interface surface are presented.

213546~

Table 1: Various Test Objects on an Alumina Gripper Surface Slip Warning (ms) Displacement at Mean (Std.Dev.) Trigger (um) Test Obiect Mean (Std.Dev.) Alumina 78 (20) o (2) Steel 70 (15) 1 (2) Aluminum (rough) 51 (13) 8 (8) Aluminum (smooth) 60 (17) 7 (7) Rubber 29 (10) 14 (11) In Table 2, the warning time results for all combinations of test object and gripper interface surface using alumina, steel and rubber are presented.

Table 2: Warning Times for all Combinations of Alumina, Steel and Rubber Alumina Steel Rubber Test Ob~ect Interface Interface Interface Alumina 78 (20) 66 (14) 36 (14) Steel 70 (15) 63 (13) 25 (11) Rubber 29 (10) 29 (9) 26 (8) Table 3 presents the corresponding values of the net displacement at the trigger point.

213~463 .

Table 3: Net Displacement at Trigger for all Combinations of Alumina. Steel and Rubber Displacement Of Test Object At Trigger Threshold (um), Mean (Std.Dev.) Alumina Steel Rubber Test Obiect Interface Interface Interface Alumina o (2) 3 (2) 27 (20) Steel 1 (2) 1 (2) 32 (24) Rubber 14 (11) 8 (10) 7 (6) In Table 4, the warning times for two different normal forces applied to the test object are presented for steel, alumina and rubber sliding on an alumina gripper.

Table 4: Warning times for Two Applied Normal Forces Slip Warning (ms) For Test Object;
Means (Std.Dev.) Test Obiect Normal Force = 1 N Normal Force = 4.5N
Steel 70 (27) 62 (15) Aluminum 60 (17) 53 (9) Rubber 29 (10) 24 (9) From the above data the following observations can be made.
With all combinations of materials tested in the course of these experiments, the time delay between 25the slip signal trigger and the test object moving 0.1 mm ~135~63 was consistently greater than 24 milliseconds as indicated in Table 2. Thus if corrective action, such as increasing the grip force, could be applied in less than 24 ms, this slip could be limited to less than 0.1 mm.
Best results are obtained when neither material is compliant. This can be seen in Table 2 where the test object is rubber or the gripper interface surface is rubber. This is most likely due to the less severe local relaxation of stress prior to gross slippage when one of the materials is compliant.
Of the non-complaint materials, steel and alumina appear to provide more notification of impending slip than does aluminum (see Table 2).
As can also be seen from Table 2, slightly more incipient slip warning signal time seemed to be generated when alumina rather than steel was used as the gripper surface. This may be due to the fine surface roughness of the alumina sample as compared to the smoother surface of the steel sample.
With all combinations of hard materials the net displacement of the test object at the time of trigger on the piezoceramic sensor is negligible. From Table 3, the net displacements at the trigger vary from O um to 3 um with a standard deviation of 2 um, which is not statistically significant. Therefore, it is possible to ~135~63 consistently detect the acoustic emissions resulting from incipient slip, prior to any measurable displacement.
Comparing the results shown in Table l for the aluminum test object with rough and smooth surfaces, it can be seen that there is not a statistically significant difference in the warning times. It, therefore, appears that the surface roughness of the test object is not a principle factor, provided there is some degree of rough gripper surface.
Increasing the normal force as shown in Table 4, and therefore the frictional force applied to the test object, does not have a great effect on the slip warning time. Presumably, this would indicate that increasing the frictional force does not have a major effect on the stress relaxation process prior to gross slip, since incipient slip signals are already detected.
As can be seen in Table 2, all combinations of hard materials provide good warning times (consistently in excess of 50 ms). Also, the displacements at the time of the trigger on the piezoceramic sensor output as shown in Tables 1 and 3 are negligible for hard materials. In other words, it is possible to consistently detect the acoustic emissions resulting from incipient slip, prior to any measurable displacement of the test object if the piezoelectric sensor carrying the applied force is _ closely acoustically coupled to the grasped object through acoustically efficient interface surfaces.
For any material sliding on rubber, the level of acoustic vibrations is lower than for the case of a hard sliding surface. The low level of the acoustic vibrations, particularly in the early stages of slip is probably a combination of two causes: the impedance mismatch at the rubber/piezoceramic interface and the plastic behaviour of the rubber at the interface as the tangential force is first applied.
For rubber sliding on rubber, the transition to sliding is quite sudden, implying a large difference between the static and dynamic coefficients of friction and the build-up of the tangential force prior to any significant slippage. The result of the sudden transition is that the warning time available from the slip trigger to the object have moved 0.1 mm is quite small even though the test object has only moved a few microns when slip threshold is triggered.
For hard materials (steel, alumina) sliding on a rubber gripper surface, the transition to sliding appears to be a more gradual process which implies that the difference between the static and dynamic coefficients of friction is much smaller than for rubber sliding on rubber. The object, therefore, slips further before acoustic emissions are detected resulting in ~135~63 larger values for the displacement when the slip threshold is triggered. Also, since the transition to sliding is slower, the warning time available in this case is actually longer than for rubber sliding on rubber.
It should be noted that the warning times calculated from these experiments are based on a fairly narrow range of test object accelerations. In a typical test run, the acceleration of the test object is loapproximately 50 mm/s2. This value tends to vary from run to run, but usually falls within the range 10 mm/sZ
to 200 mm/s2. This implies that the test object velocity typically falls within the range 0.5 mm/s to 10 mm/s for a measured warning time of 50 milliseconds.
15Based on the results of these experiments, the optimum material for the gripper surface should have the following characteristics:
(a) fine surface roughness for the generation of strong incipient slip signals;
20(b) hard wearing to maintain the surface roughness over the life of the gripper; and (c) good impedance match to piezoceramic to maximize the coupling of the acoustic emissions generated at the slip interface into 25the piezoceramic.

'~13~463 -Figure 2 depicts a robot gripper 1 carrying piezoelectric ceramic sensors 2 within its grippers 3, directly beneath the gripping interface 4, so that each sensor 2 is in the force path, and experiences stress, from the application of a gripping force to an object 5.
When activated, a controller 8 governed by software, drives the gripper motor 9 that activates the grippers 3 through mechanical links 6 (that may have a pliable outer sheath 10 for aesthetics) to close the gripper fingers 3 slowly. During this time, the gripper position sensor 7, which senses the position of the fingers 3, is monitored by the controller 8. When the fingers 3 come into contact with the object 5 to be held, the increased load on the gripper motor 9 allows the controller 8 to sense the resistance and the gripper motor 9, under direction from the controller 8, slows to a stop. The software detects this lack of motion from the position sensor 7 and enables the slip detection routine to be implemented.
If the object 5 begins to slip, the software causes the controller 8 to increase the motor control voltage upon detection of a sufficient threshold amplitude of acoustic slip signal. When the object 5 stops slipping, the acoustic slip signal drops below the threshold level, and the software causes the controller 8 to slowly decrease the motor control voltage towards a minimum force level until slip is detected again. This process can be repeated cyclically, or intermittently.
While this control scheme is easy to implement, it relies upon there being a close relationship between the gripping force and the control voltage. Thus, a large, high-torque motor 9 is preferred to drive the gripper in order to produce a sufficient gripping force, without backlash being present as would occur if high reduction gearing is employed. However other options for force-control to improve the performance of the system and decrease the size may be employed.
The detection of the acoustic emissions at low levels is critically dependant on the efficient coupling acoustically of the piezoelectric sensor to the gripping interface.
In Figure 6 the piezoelectric sensor 2 is shown supported by the mechanical link 6 present within the gripper 3. Bonded to the sensor 2 is the interface surface material 4a. This interface material 4 should be an efficient conductor of acoustic emissions and should be efficiently acoustically coupled to the sensor 2.
In use, it will be seen that all of the applied gripping force is transmitted from the link 6, through the sensor 2 and gripping interface material 4a to the gripping interface 4. This produces an efficient means ~13~63 ., .

for detecting low level acoustic emissions, sufficient to provide an early warning signal to detect incipient slip.

Conclusion The foregoing has constituted a description of specific embodiments showing how the invention may be applied and put into use. These embodiments are only exemplary. The invention in its broadest, and more specific aspects, is further described and defined in the claims which now follow.
These claims, and the language used therein, are to be understood in terms of the variants of the invention which have been described. They are not to be restr~icted to such variants, but are to be read as covering the full scope of the invention as is implicit within the invention and the disclosure that has been provided herein.

Claims (6)

THE EMBODIMENTS OF THE INVENTION IN WHICH AN EXCLUSIVE
PROPERTY IS CLAIMED ARE AS FOLLOWS:

1. A robotic gripper for applying a gripping force to a grasped object, the gripper having a gripping surface with piezoelectric acoustic emission sensor material located beneath at least a portion of such surface, which sensor material is so positioned as to be subject to stress arising from the gripping of the grasped object by the robotic gripper.

2. A robotic gripper as in claim 1 wherein the piezoelectric sensor material is closely coupled acoustically to the object being grasped, such gripper comprising an interface layer disposed between the piezoelectric material and the grasped object wherein the interface layer is a good acoustic conductor and is efficiently acoustically coupled to such piezoelectric sensor material.

3. A robotic gripper as in claims 1 or 2 further comprising a controller which responds to acoustic emissions detected by the sensor and cyclically loosens and re-asserts the gripping force of the robotic gripper based on the detection of acoustic emissions arising when slip is incipient.

4. A robotic gripper as in claim 3 wherein the controller responds to a trigger signal level that is at least twice the level of the background noise which occurs when slippage between the grasped object and the gripper is imminent.

5. A robotic gripper as in claim 4 wherein the trigger signal level is between 5 and 15 times the level of the background noise.

6. A robotic gripper as in claim 3 wherein the piezoelectric sensor material that is acting as a sensor to detect acoustic emissions comprises multiple piezoelectric sensors that are separately used to detect acoustic emissions at different locations over the gripping interface.