WO2019243781A1

WO2019243781A1 - Robot gripper comprising a slippage sensor and operating method of a robot gripper

Info

Publication number: WO2019243781A1
Application number: PCT/GB2019/051653
Authority: WO
Inventors: Saber Mahboubi HEYDARABAD; Steven Davis
Original assignee: University Of Salford Enterprises Limited
Priority date: 2018-06-20
Filing date: 2019-06-14
Publication date: 2019-12-26
Also published as: GB201810156D0; GB2574861A

Abstract

A robot gripper (10) comprises two gripping fingers (1, 2) for gripping an object 4. Operation of respective servo motors (11, 12) causes the fingers (1, 2) to move towards each other such that an object (4) can be grasped. A controller 15 is connected to the servo motors (11, 12) and to slippage sensors (21, 22) and is operable to output control signals to the servo motors (11, 12). The control signals control the level of torque applied by the servo motors (11, 12) to the respective gripping fingers (1, 2) in order to hold object (4) in response to a Super Twisting Sliding Mode Control (STSMC) scheme based on grip force and slip feedback.

Description

ROBOT GRIPPER COMPRISING A SLIPPAGE SENSOR AND OPERATING METHOD

OF A ROBOT GRIPPER

Technical Field of the Invention

The present invention relates to improvements in or relating to robot grippers. Background to the Invention In many applications, robot grippers are utilised to pick, hold or manipulate objects. In order to successfully achieve such tasks, the gripper must be able to apply sufficient pressure to the object that it can be reliably held without slipping. Equally, the gripper must not apply enough pressure to damage the object. This is particularly difficult where a gripper is required to pick, hold or manipulate a range of different objects, rather than a small set of objects with known properties.

As gripping an object with sufficient force to prevent slippage, whilst not damaging or deforming permanently the shape of the object, is a challenge various control schemes for grippers have been proposed. Once such scheme operates by monitoring for slippage and changing the applied grip force on the object in response to any detected slippage. This scheme is inspired by the observation that when starting from a stable hold on an object, the shear force on the gripper increases until it reaches the static friction limit, at which point the object starts sliding. Accordingly, since dynamic friction is lower than static friction, to stop slippage the grip force is increased until the object stops sliding. In order to implement such a control scheme which monitors both applied grip force and slippage, a suitable control methodology has to be used. In this context, it is known to apply sliding mode control (SMC) techniques based around monitoring feedback from gripping force and slippage sensors. Whilst this does provide for practical implementation across a range of sensors designs, such techniques are susceptible to chattering. The term“chattering” describes the phenomenon of finite- frequency, finite-amplitude oscillations are caused by the high-frequency switching of a sliding mode controller exciting unmodeled dynamics in the closed loop. Typically such dynamics may be related to the operation of the sensors and actuators neglected in the principal modelling process. These unwanted effects are difficult to eliminate without compromising performance of the SMC model.

It is therefore an object of the present invention to provide a robot gripper and a control method for a robot gripper that at least partially overcomes or alleviates such problems.

Summary of the Invention

According to a first aspect of the present invention there is provided a robot gripper comprising: at least two gripping fingers; actuation means for actuating each said gripping finger so as to apply a gripping force to the object to be held; monitoring means for determining the gripping force applied by each gripping finger and output a signal indicative thereof; a slippage sensor operable to detect any slippage between each sensing element and the object and output a signal indicative thereof; and a controller operable to control the actuation means so as to adjust the desired gripping force in response to the received signals from the monitoring means and slippage sensor according to a super twisting sliding mode control (STSMC) scheme.

According to a second aspect of the present invention there is provided a method of operating a robot gripper of the type comprising at least two gripping fingers and actuation means for actuating each said gripping finger, the method comprising the steps of: determining the gripping force applied by each gripping finger to the object to be held; detecting any slippage between each sensing element and the object to be held; and controlling the actuation means by control signals such that the desired gripping force applied by each gripping finger is adjusted in response to the applied gripping force and detected slippage according to a super twisting sliding mode control (STSMC) scheme.

Use of an STSMC scheme for a robot gripper improves the robustness of control and helps to avoid chattering phenomena.

In one embodiment, the gripper may comprise two gripping fingers. The skilled man will of course appreciate that grippers that the present invention may be applied to comprising more than two gripping fingers.

Preferably, slippage relating to each gripping finger is measured independently. This may be achieved by providing each gripping finger with an independent slippage sensor. The slippage sensor may comprise an optical flow sensor. Preferably, each gripping finger is independently actuated. This may be achieved by providing each gripping finger with an independent actuation means. The actuation means may comprise a motor. In one embodiment, the actuation means may comprise a servo motor.

The applied gripping force may be measured directly. To achieve such operation, the monitoring means may comprise a force sensor integrated into the gripping finger. In a preferred embodiment, the applied gripping force may be determined by monitoring the operation of the actuation means. In such embodiments, the monitoring means may be operable to monitor the operation of the actuation means to determine the applied gripping force. In one example, where the actuation means is a motor, the monitoring means may be operable to read the torque output signal of the motor. The torque output signal may result from a pre-set determination based on motor inputs or may result from one or more sensors integrated into the motor. In some embodiments, the motor torque may be estimated by measuring eth current supplied to the motor.

The controller may control the actuation means by outputting control signals. The control signals may be operable to cause the actuation means to generate a desired gripping force through the fingers. In embodiments where the actuation means are motors, the control signals may contain an indication of a desired torque output or a desired rotational speed output.

The controller is preferably operable to minimise the applied gripping force. This minimises the risk of object deformation whilst being held.

The controller may be operable to select an initial value of the desired gripping force that is small and positive. In some embodiments, the initial value of the desired gripping force is based on the lightest weight object expected to be handled. This allows adjustment of desired gripping force to an optimal value to take place relatively efficiently and with minimum risk of crushing lighter objects.

The robot gripper may be incorporated into a prosthesis. The prosthesis may comprise a prosthetic hand or arm. In such instances, the gripper may comprise one or more biosensors operable to detect input from a user fitted with the gripper as a prosthesis. The gripper may additionally or alternatively comprise feedback means operable to impart feedback to a user fitted with the gripper as a prosthesis. Detailed Description of the Invention

In order that the invention may be more clearly understood one or more embodiments thereof will now be described, by way of example only, with reference to the accompanying drawings, of which: Figure 1 is a schematic model of a robot gripper according to the present invention;

Figure 2 is a schematic block diagram of a control system for a robot gripper according to the present invention;

Figure 3 is a schematic model of the gripped object and gripping fingers of the present invention;

Figure 4 is a schematic block diagram of an STSMC scheme for a robot gripper according to the present invention;

Figure 5 illustrates an acceleration profile for a pick-and-place task carried out by a gripper according to the present invention; Figure 6 shows the variation over time of (a) slippage, (b) applied gripping force and (c) sliding variable s for a lettuce subjected to the acceleration profile of figure 5;

Figure 7 shows the variation over time of (a) slippage, (b) applied gripping force and (c) sliding variable s for a partially filled bottle subjected to the acceleration profile of figure 5; Figure 8 shows the variation over time of (a) slippage, (b) applied gripping force and (c) sliding variable s for an aluminium block subjected to the acceleration profile of figure 5;

Figure 9 shows the variation over time of (a) slippage, (b) applied gripping force and (c) sliding variable s for an egg subjected to the acceleration profile of figure 5;

Figure 10 is a schematic block diagram of an SMC scheme for a robot gripper according to the prior art;

Figure 11 shows the grasp response of a gripper controlled according to the SMC scheme of figure 10 for (a) a foam block, (b) a piece of wood and (c) an RSJ steel beam when using a chain of step functions as desired force input; and

Figure 12 shows a comparison between the grasp responses of a gripper controlled according to the SMC scheme of figure 10 and a gripper controlled using the STSMC scheme of figure 4 for (a) a foam block, (b) a piece of wood and (c) an RSJ steel beam when using a common test desired input force trajectory.

Turning now to figure 1, a schematic model of a robot gripper 10 is shown. The gripper 10 comprises two gripping fingers 1, 2 each connected to the shaft of a corresponding servo motor 11, 12 (not shown in figure 1) through joints Jl and J2 respectively. Operation of the respective servo motors 11, 12 causes the fingers 1, 2 to move towards each other such that an object 4 can be grasped. Turning now to figure 2, a schematic block diagram of a control system for the gripper 10 is shown. The control system comprises a controller 15 which is connected to the servo motors 11, 12 and to slippage sensors 21, 22 associated with corresponding gripping fingers 1, 2. The slippage sensors 21, 22 are typically optical flow sensors integrated into the respective gripping fingers 1, 2 in the region where the gripping fingers 1, 2 contact the object 4. The slippage sensors 21, 22 are each operable to detect any slippage between the respective gripping element 1, 2 and the object 4 and output a signal indicative thereof

The servo motors 11, 12 may each comprise a monitoring means 31, 32 respectively. Each monitoring means 31, 32 is operable to generate an output signal indicative of the torque output by the servo motor 11, 12. This output signal may be generated by determining the expected torque output from an input signal applied to the motor 11, 12 or may be generated by a suitable sensing arrangement within the motor 11, 12. The controller 15 is operable to output control signals to the servo motors 11,

12. The control signals control the level of torque applied by the servo motors 11, 12 to the respective gripping fingers 1, 2 in order to hold object 4. The control signals are generated in response to the output torque signals and the output slippage signals. The control scheme by which such control signals are generated is described further below. In order to simplify modelling, the object 4 can be represented as a sphere and two coordinate frames are used, as shown in figure 1. The frames comprise a reference coordinate frame OXYZ at the palm 5 of the gripper 10, and an object coordinate frame O X Y_OZ_O which is parallel to the reference coordinate frame, in a way that 0₀ is in the geometric centre of the object 4. According to the chosen reference frames, the gripping force F_n always acts along the X axis of both gripper 10 and object 4.

Each servo motor 11, 12 independently drives one finger 1, 2 of the gripper 10. In the example shown, there is no mechanical linkage between the fingers 1, 2. Nevertheless, the gripper 10 is preferably operated by the controller 15 such that the gripping action is performed symmetrically with respect to the Y axis.

In the example shown, the gripping force applied to the object 4 is linked to the motor torque by the relationship: n(t) = T(t) r cos y(t) (1) where y is the angle between the gripping link and the link connected to the motor around the Z axis. As highlighted in (1), F_n is a function of both T and y: the closer y to 0 the better the gripper performs when tuning the gripping action (smaller variations of the torque T are involved).

Turning now to figure 3, a simplified dynamic model of the gripper 10 is described through a parallel spring-damper characterized by the stiffness coefficient K, the damping coefficient B and mass m. The force counteracting weight and external disturbances is provided by the contact friction pF_n, where F_n and m are the normal force along X and the coefficient of dynamic friction respectively. From the system dynamics a second order equation can be derived:

where y indicates the displacement caused by the slippage of the object with respect to the reference frame in the direction of the Y axis, y its acceleration, m the weight of the object, and g the gravity constant. The variable d_t represents the time-variant disturbance acting upon the object in the form of an externally applied load. D is the disturbance term which is a function of d_t and m.

Using xi and x₂ as the state values we can write xi = y (3)

*i = ^x2 (4) x₂ = u (5) u = g— T(pi, m)R_h + D(d_t, m ) (6) where u is the input of the system and is characterised by a controllable value F_n (the

gripping force), a constant value g and a disturbance D. G(th,m) =— can be

considered time-invariant if the surface of the grasped object is homogeneous and represents a model uncertainty if no previous knowledge of the object is assumed. The disturbance D is associated to a wide range of causes coming from the surrounding environment and as such must be dealt with as an uncertainty too. F_n can be controlled through the motors 11, 12. The kinematic model of the fingers 1, 2 can be written as follows:

where V_c is the velocity of the contact point along the X_c axis and co/ the angular velocity of the fingers around their joints Jl and J2, which, in turn, is a function of the rotational velocity of the motor u>_m, as shown in (7). As friction plays a significant role in determining the grip force, having a rough estimation of the friction coefficient of the contact surface of the object, along with its weight and stiffness, can significantly reduce the chances of slippage and deformation when the object is held. In the prior art, determination of the desired gripping force has be carried out using feedback from force sensors and slippage sensors using Sliding Mode Control (SMC) schemes. Figure 10 shows an example of a proportional integral sliding mode control scheme for a gripper 10. In this scheme, position, velocity and force feedback control with an anti-windup layout is used to control the fingers’ position, velocity and grip force. In particular, Gp is the proportional gain for the position control, Kp the proportional gain for the velocity control, Ki the integral gain for the velocity control and K_aw anti-windup gain, respectively. The feedback gain K_b is a conversion constant of the shaft’s position and angular velocity.

These SMC schemes however commonly suffer from chattering phenomena. Whilst SMC schemes exhibiting chattering may still be reasonably robust in certain circumstances, chattering increases power consumption in actuation means and reduces their lifetime. This is illustrated in figure 11 which shows the results of a series of tests using a chain of step functions with increasing amplitudes (5, 10, 15, 20, 25, 30, 35, 40 N) as the desired force input function of the gripper 10. The motors 11, 12 were actuated in response to these reference force inputs and with an object 4 placed between the fingers 1, 2. The objects 4 used in this test were (a) a foam block, (b) a piece of wood and (c) a RSJ steel beam. In the figure, the dashed lines indicate the desired input force applied to each object whereas the solid lines are the measured gripping force applied to the respective objects 4. The figure clearly illustrates failure to reach the desired force in many instances, including some steady state errors. Additionally, the figure also illustrates considerable overshoots and chatter in other instances, particularly for higher desired forces.

One of the key issues leading to this poor performance can be explained by the fact that the sliding mode control model does not take into account the stiffness and damping coefficients of the object to be grasped. Turning back to the model of figure 3, before the fingers 1, 2 touch the object 4, the stiffness of the object 4 has no effect on the grasp model and hence is negligible. However, as soon as the fingers start touching the object, the stiffness of the system and the stiffness of the object does impact on the grasp model.

Unfortunately, in reality, it is impractical to calculate the stiffness of the unknown objects to be grasped. This makes the grasp model inaccurate. To solve the uncertainty problem explained above, the present invention uses a hybrid sliding mode PID controller as shown in Figure 4. The sliding mode controller will increase the robustness of the proportional integral control to the uncertainty of the object’s mechanical impedance (KOBJ and BOBJ) and hence reduce the undesired steady-state error and response overshoot. The present invention proposes to use a Super Twisting Sliding Mode Control (STSMC) scheme based on grip force and slip feedback. STSMC is a particular type of second order sliding mode control, with the advantage compared to other such controllers that only real-time measurements of the sliding variable s can be used to guarantee the condition s = s = 0. This allows STSMC to be used for systems of relative degree 1 instead of standard SMC. In order to improve performance, the STSMC model used in the present invention takes into account the mechanical impedance (stiffness and damping) of the object 4. For simplicity, it is assumed that the object 4 has isotropic and homogeneous density, stiffness and damping properties. It is also assumed that the deformation of the object 4, when grasped, is within the elastic range of the material. This assumption is justified since the intention is for the gripper 10 not to cause any permanent deformation or even break the object. Under these assumptions stiffness coefficient K is constant, and from Fig. 3, the relationship between the grip force and the velocity of the contact point is given by: F_n = K J V_cdt + BV_C (8)

Turning now to figure 4, a block diagram of STSMC according to the present invention is shown. In figure 4, it is assumed that the control signals for driving motors 11, 12 to apply a desired gripping force are based on a desired value of the motor shaft speed amid calculated by the control law explained below.

Within the STSMC of figure 4, to overcome slippage from an initial stable grasp with minimum deformation of the object, an error function is defined as: e(F, S_L) = K(F_n - F_d) - S_l (9) where e depends on applied force and slippage, and k is a constant related to the stiffness of the grasped object. With respect to the reference frame, F_n is the actual (measured) gripping force of the fingers 1, 2, F_d is the desired force of the fingers 1, 2 in order to stop slippage and Si the slippage or displacement of the object 4, which is acquired through the slip sensors 21, 22. The role of the STSMC implemented by the controller 15 is to minimize the applied force F_n and consequently minimize the object deformation following a slippage event.

If the value of F_d is known, from knowledge of the surface friction and weight of the object 4, the controller 15 will achieve a stable grasp while lifting the object 4. If F_d is not known, F_d should be set to a small positive initial value. In case the gripper 10 has to handle different objects, the value of F_d should correspond to the lightest object 4 to be handled. In this scenario, the gripper 10 will adjust F_n while lifting the heavier objects 4, preventing lighter objects 4 from getting crushed.

The sliding manifold of the error state is given by:

G d 1^l

sq, e) = [- + h\ (10)

In the present case l is equal to 1. This is because the number of derivatives to be included, l, should be l =n-l, where n is the input output relative degree. In the present case n is equal to 2, so we can rewrite (10) as o e, e) = e + hb (11) where an arbitrary positive constant h guarantees the exponential decay of the error.

In order to steer the sliding variable s to zero in finite time, the control function for STSMC is provided in the form:

w = — Wsgn(o) (13) with the following tuning relationships: g = Vc W = l.l c (14) where C, the sliding gain constant, is a sufficiently large constant. Beneficially, a large value for C increases the robustness of the system in the presence of external disturbances. The value of C can be tuned by trial and error.

The system may optionally include feedback from the manipulator’s control cabinet (K_RC4) and a time derivative slippage to the sliding gain constant, C, in order to guarantee the stability of the system.

Turning now to figures 5 to 9, these illustrate the performance of the gripper 10 of the present invention utilising STSMC. Specifically, these figures show the results of a series of pick-and-place tasks where objects 4 are moved between pick and place points through fast accelerations and decelerations. In view of the significant dynamic accelerations involved objects 4 are at relatively high risk of slip or indeed being dropped completely.

Turing specifically to figure 5, this shows the acceleration profile of the object trajectory used in the pick-and-place tasks. Figures 6 to 9 each show the respective (a) slippage, (b) applied gripping force and (c) sliding variable s for different objects during three separate tests, plotted on the same axes. In this context: in figure 6, the object is a lettuce, in figure 7, the object is a partially filled bottle, in figure 8, the object is a block of aluminium, and in figure 9, the object is an egg. Whilst in each instance there is some random variation in slippage events, the figures do illustrate some common features. Firstly, as an initially low value of F_d is used, the slippage and applied force profiles at the start of the test differ depending upon the weight of the object. Subsequently, profiles are very similar and stable for each object indicating successful grip is achieved despite the different object characteristics.

A further illustration of the performance of gripper 10 using STSMC according to the present invention is shown in figure 12. As in figure 11, the gripper 10 was used to manipulate (a) a foam block, (b) a piece of wood and (c) an RSJ steel beam. A desired input force trajectory for each test is illustrated by the dashed lines in the figure. The lighter solid lines illustrate the output grip force resulting from the prior art SMC model of figure 10. The darker solid lines illustrate the output grip force resulting from the STSMC model of the present invention. As can be seen from this figure, the prior art model suffers from steady-state error and overshoot, whereas the STSMC model of the present invention can robustly remove the steady-state errors and overshoots.

In use, the gripper 10 may be utilised for any suitable pick and place tasks. In addition, it is possible to utilise a gripper according to the present invention to form a prosthesis such as a prosthetic hand or arm. The above embodiment is described by way of example only. Many variations are possible without departing from the scope of the invention as defined in the appended claims.

Claims

1. A robot gripper comprising: at least two gripping fingers; actuation means for actuating each said gripping finger so as to apply a gripping force to the object to be held; monitoring means for determining the gripping force applied by each gripping finger and output a signal indicative thereof; a slippage sensor operable to detect any slippage between each sensing element and the object and output a signal indicative thereof; and a controller operable to control the actuation means so as to adjust the desired gripping force in response to the received signals from the monitoring means and slippage sensor according to a super twisting sliding mode control (STSMC) scheme.

2. A robot gripper as claimed in claim 1 wherein the gripper comprises two gripping fingers.

3. A robot gripper as claimed in claim 2 wherein each gripping finger is provided with an independent slippage sensor.

4. A robot gripper as claimed in any preceding claim wherein each slippage sensor comprises an optical flow sensor.

5. A robot gripper as claimed in any preceding claim wherein each gripping finger is provided with an independent actuation means.

6. A robot gripper as claimed in any preceding claim wherein the actuation means comprise a servo motor.

7. A robot gripper as claimed in preceding claim wherein the monitoring means comprise a force sensor integrated into the gripping finger.

8 A robot gripper as claimed in any preceding claim wherein the monitoring means are operable to monitor the operation of the actuation means to determine the applied gripping force.

9. A robot gripper as claimed in claim 8 when dependent upon claim 6 wherein the monitoring means are operable to read the torque output signal of the motor.

10 A robot gripper as claimed in any preceding claim wherein the controller controls the actuation means by outputting control signals operable to cause the actuation means to generate a desired gripping force through the fingers.

11 A robot gripper as claimed in claim 10 when dependent upon claim 6 wherein the control signals contain an indication of a desired torque output or a desired rotational speed output.

12 A robot gripper as claimed in any preceding claim wherein the controller is operable to minimise the applied gripping force.

13. A robot gripper as claimed in claim 12 wherein the controller is operable to select an initial value of the desired gripping force that is small and positive.

14. A robot gripper as claimed in any preceding claim wherein the robot gripper is incorporated into a prosthetic hand or arm.

15. A robot gripper as claimed in claim 14 wherein the gripper comprises one or more biosensors operable to detect input from a user fitted with the gripper as a prosthesis.

16. A robot gripper as claimed in claim 14 or claim 15 wherein the gripper comprises feedback means operable to impart feedback to a user fitted with the gripper as a prosthesis

17. A method of operating a robot gripper of the type comprising at least two gripping fingers and actuation means for actuating each said gripping finger, the method comprising the steps of: determining the gripping force applied by each gripping finger to the object to be held; detecting any slippage between each sensing element and the object to be held; and controlling the actuation means by control signals such that the desired gripping force applied by each gripping finger is adjusted in response to the applied gripping force and detected slippage according to a super twisting sliding mode control (STSMC) scheme.

18. A method as claimed in claim 17 wherein the gripper comprises two gripping fingers.

19. A method as claimed in claim 17 or claim 18 wherein slippage relating to each gripping finger is measured independently.

20 A method as claimed in any one of claims 17 to 19 wherein each gripping finger is independently actuated.

21. A method as claimed in claims 17 to 20 wherein the applied gripping force is measured directly.

22. A method as claimed in claims 17 to 21 wherein the applied gripping force is determined by monitoring the operation of the actuation means.

23. A method as claimed in claims 17 to 22 wherein the control signals are operable to cause the actuation means to generate a desired gripping force through the fingers.

24. A method as claimed in claims 17 to 23 wherein the control signals are operable to minimise the applied gripping force.

25. A method as claimed in claim 24 wherein the control signals are operable to select an initial value of the desired gripping force that is small and positive.