EP3328595A2

EP3328595A2 - Method and system for controlling a robot

Info

Publication number: EP3328595A2
Application number: EP16744303.5A
Authority: EP
Inventors: Ricardo PÉREZ SILVA
Original assignee: KUKA Roboter GmbH
Current assignee: KUKA Deutschland GmbH
Priority date: 2015-07-30
Filing date: 2016-07-11
Publication date: 2018-06-06
Also published as: DE102015009892A1; WO2017016641A2

Description

description

Method and system for controlling a robot

The present invention relates to a method and a system for controlling a robot as well as a robot arrangement with the system and a

Computer program product for carrying out the method.

Especially in medical applications such as the

robot-assisted sonography, it may be desirable to detect a contact of the robot or robotic medical device and then provide a force feedback to a haptic input device of the robot.

Robotics, May 9, 2014, propose two different methods for collision detection of robots: a method determines a fuzzy inference system ("Fuzzy inference system", FIS) from a position error, a commanded speed and the measured at joints

Total torques joint external joint moments, as the determination of external forces from a dynamic model of the robot is problematic. According to another method, an autoregressive model ("AutoRegressive" AR), again without a dynamic model of the robot, detects joint torques, where a collision is detected based on a deviation of the joint predicted from the AR from the measured joint torques.

The object of the present invention is to control a robot

improve.

This object is achieved by a method having the features of claim 1. Claims 7 to 9 provide a system or computer program product

Implementation of a method described here or a robot arrangement with a system described here under protection. The subclaims relate to advantageous developments. According to an embodiment of the present invention, a method for controlling a robot, in particular a medical application, comprises the steps:

Determining a collision force on the basis of a model-based determined external force resulting from an external load of the robot, a

Positional error of the robot and a pose and / or speed of the robot; and

Detecting a collision of the robot based on this collision force.

A system for controlling a robot, in particular a medical

Application, according to an embodiment of the present invention, in particular hardware and / or software, in particular program technology, set up for carrying out a method described herein and / or has:

- Means for determining a collision force based on a model-based

detected external force resulting from an external load of the robot, a position error of the robot and a pose and / or

Speed of the robot; and

Means for detecting a collision of the robot on the basis of this collision force,

As a result, a collision of a robot can advantageously be detected in one embodiment. In particular, by taking into account a model-based determined external force resulting from an external load of the robot, instead of a measured G (esamtg) elenkkraft detection can be improved, in particular by a fuzzy system, which then no longer the functionality of a model for determining the external force must comply with.

In one embodiment, the robot has at least six, in particular at least seven, actuatable or actuated joints. In an embodiment in which a medical application of the robot is controlled, it carries a medical instrument, in particular an ultrasound head or the like.

For a more compact representation, an anti-parallel force pair or torque is generally referred to herein as a force. A force, in particular the determined collision force, the model-based determined external force and / or a feedback force explained below, can in particular be a Cartesian force with components in one or more Cartesian directions, in particular one Inertial or robot base fixed or robot end-member fixed coordinate system include, in particular.

In one embodiment, a position error depends on a deviation between an actual, in particular measured, pose of the robot and a desired, in particular commanded, pose of the robot, in particular of one

Deviation between an actual position and a desired position and / or orientation of a robot-fixed reference, in particular of the TCP, can in particular indicate or be the deviation.

In one embodiment, a pose of the robot depends on the position of its joints, in particular drives, and / or the position and / or orientation of a robot-fixed reference, in particular of the TCP, it can indicate or be particular. Generalizing is thus present for a more compact representation

in particular also understood a position and / or orientation of a robot-fixed reference, in particular of the TCPs, as a pose of the robot. Accordingly, in one embodiment, a speed of the robot depends on the speed of its joints, in particular drives, and / or the

translational and / or rotational speed of a robot-fixed reference, in particular of the TCP, from, they can specify or be this particular.

In general, for the purpose of a more compact representation, a speed of a robot-fixed reference is also understood here to be the speed of the robot.

In one embodiment, the method, in particular its step, comprises one

Determining a collision force, the steps:

- Determine Qeweils) one or more, in particular at least two,

in particular at least three degrees of membership of one or more

Components of the model-based determined external force, (respectively) one or more, in particular at least two, in particular at least three, degrees of membership of one or more components of the position error of the robot and (respectively) one or more, in particular at least two, in particular at least three, degrees of affiliation one or more Components of the pose and / or velocity of the robot based on a fuzzy set or membership functions of a fuzzy system; and

Determining one or more components of the collision force on the basis of the fuzzy system as a function of these degrees of membership. Accordingly, the system, in particular its means for determining a collision force, in one embodiment:

- means for determining (in each case) one or more, in particular at least

two, in particular at least three, degrees of membership of one or more components of the model-based determined external force, (in each case) one or more, in particular at least two, in particular at least three,

Degrees of membership of one or more components of the positional error of the robot and (respectively) one or more, in particular at least two, in particular at least three, degrees of membership of one or more components of the pose and / or speed of the robot on the basis of membership functions of a fuzzy system; and

- Means for determining one or more components of the collision force based on the fuzzy system as a function of these degrees of membership.

A degree of affiliation can in a conventional manner based on a

Membership function of the fuzzy system can be determined. In other words, the component (s) of the model-based determined external force, the

Position error, the pose and / or speed fuzzifiziert and from these fuzzified sizes or degrees of membership in particular rule-based by the fuzzy system, in particular again defuzzified, component (s) of

Collision force determined. For this purpose, reference is made in particular to the above-mentioned publication F. Dimeas et al., "Human-robot collision detection and

Identification based on fuzzy and time series modeling, "Robotica, 9 May 2014, and incorporated by reference in its entirety into the present disclosure

included. For example, a first membership function of a component of the model-based determined external force / position error /

Pose / velocity (in each case) assign a degree of affiliation to a first fuzzy set, eg "small" and a further degree of affiliation to a second fuzzy set, eg "large", optionally also one or more further degrees of membership to further fuzzy sets, eg "medium" or similar In one embodiment, the fuzzy system is an FIS ("fuzzy inference system"), in particular Sugeno or Mamdani. In one embodiment, it includes the membership functions that map or fuzzify the input values to degrees of membership, rules, or a rule base, respectively In one embodiment, the FIS evaluates its output by weighting the rule base with the degrees of membership, and then weighting it into a different fuzzy set

Defuzzification and this results in the outcome of the FIS (Mamdani). In another embodiment (Sugeno) the defuzzification is carried out by means of the weighting of the rule base and by means of the evaluation of a linear function of the input variables.

It has surprisingly been found that a fuzzy system with the input variables mentioned external force, position error and pose or speed is particularly advantageous for detecting collisions, since in particular non-collision-related variations in the external force can be advantageously filtered.

In one embodiment, a component of the collision force in at least a first direction (respectively) is dependent on one or more

Belonging degrees of a component of the model-based determined external force in this first direction, one or more degrees of membership of a

Component of the position error of the robot in this first direction and / or one or more degrees of membership of a position of a robot-fixed

Reference determined in this first direction. Accordingly, in one embodiment, the system or its means for

Determining at least one component of the collision force means for determining a component of the collision force in at least a first direction (respectively) depending on one or more degrees of membership of a component of the model-based determined external force in this first direction, one or more degrees of membership of a component the position error of the robot in this first direction and / or one or more degrees of membership of a position of a robot-fixed reference in this first direction. A first direction may in particular be a robot (member) or space-fixed and / or Cartesian direction, in particular of the working space of the robot. In one embodiment, a first direction is the vertical direction, a horizontal direction, or a direction inclined to the vertical and horizontal. In a

A first direction is a collision direction, in particular an anticipated and / or main collision direction, and / or perpendicular to a potential, in particular predetermined, collision plane. For example, in an ultrasound examination of a patient lying horizontally, the vertical is the expected main collision direction. It has been found that for a favorable determination of a component of the collision force in certain directions, in particular in one

(Main) collision direction, advantageous in addition to the position error and the component of the external force, the position of a robot-fixed reference in this direction is particularly suitable. Additionally or alternatively, in one embodiment, a component of the collision force in at least one second direction (respectively) is one or more depending on one or more degrees of membership of a component of the model-based determined external force in that second direction

Degrees of membership of a component of the positional error of the robot in this second direction and / or one or more degrees of membership of a robot

Speed of a robot-fixed reference in this second direction, determined.

Accordingly, in one embodiment, the system or its means for determining at least one component of the collision force additionally or alternatively comprises means for determining a component of the collision force in at least one second direction (respectively) in dependence on one or more

Belonging degrees of a component of the model-based determined external force in this second direction, one or more degrees of membership of a

Component of the position error of the robot in this second direction and / or one or more degrees of membership of a speed

robot-fixed reference in this second direction. It has been found that for a favorable determination of a component of the collision force in certain directions, in particular obliquely to a

(Main) collision direction, advantageously in addition to the position error and the component of the external force, the speed of a robot-fixed reference in this, in particular to the (main) direction of collision inclined direction is particularly suitable.

The second direction may in particular be a robot (member) or space-fixed and / or Cartesian direction, in particular of the working space of the robot. In one embodiment, the second direction is a horizontal direction, the vertical direction, or a direction inclined to the vertical and horizontal. In a

Embodiment, the second direction is inclined to the first direction, in particular perpendicular to this. In one embodiment, two second directions are perpendicular to each other and / or two first directions perpendicular to each other.

In one embodiment, the fuzzy system, in particular one or more of its membership functions and / or rules and / or its linear function, is or are parameterized by means of a neural network, in particular based on the method of the smallest error rate and / or the method the reverse propagation descending gradient.

Correspondingly, in one embodiment, the system comprises means for parameterizing the fuzzy system, in particular one or more of its membership functions and / or rules, by means of a neural network, in particular based on the method of the smallest error rate and / or the method of

Backward propagation descending gradient.

As a result, in one embodiment, the method or system can advantageously be adapted to different robots and / or applications, in particular rapidly, autonomously and / or robustly.

In one embodiment, the external force is modeled based on a

dynamic model of the robot. Accordingly includes in one

Execution of the system Means for model-based determination of the external force on the basis of a dynamic model of the robot. In one embodiment, the dynamic model maps a pose, speed and / or acceleration of the robot and forces acting thereon, in particular corresponding to a relation M (q) d ² q / dt ² + h (q, dq / dt) = T _mo t - T _e with the mass matrix M, the joint coordinates or the position and orientation of a robot-fixed reference q or their time derivatives, the force vector h, the driving forces T _mo t and the external forces T _e . It can equally be implemented in Cartesian or joint coordinate space. The driving forces T _mot and / or the vector q and its time derivatives can be detected or detected in particular by means of sensors. As mentioned above, in particular the combination of a model-based determination of external forces on the basis of a dynamic model with the

Determining a contact force based on these external forces on the basis of or by a fuzzy system (s) to improve detection of a collision.

Alternatively or additionally, in particular following the dynamic model, a (time) predictive model, in particular an autoregressive model, can be used

In particular, an AR model, be provided to determine the external forces modeled. As a result, a time delay through the fuzzy system can be compensated for, at least partially.

Accordingly, in one embodiment, the external force is determined on the basis of a predictive model, in particular an autoregressive model. Accordingly, in one embodiment, the system comprises means for model-based determination of the external force on the basis of a predictive model, in particular an autoregressive

Model.

The predictive model, in one embodiment, forms an external force, determined in one embodiment based on the dynamic model, on one

(predicted) external force. In addition to a predictive model, reference is additionally made in particular to the above-cited publication F. Dimeas et al., "Robot Robot Collision Detection and Identification based on fuzzy and time series modeling", Robotica, May 9, 2014, and its contents are fully incorporated into the present disclosure included. In one embodiment, after detection of a collision based on the collision force is switched to another determination to determine a feedback force, this other determination can be carried out in an embodiment even parallel to the determination of the collision force to the To improve switching, in particular to reduce a delay caused thereby.

Accordingly, in one embodiment, a feedback force is determined after detecting a collision of the robot without the fuzzy system and / or the predictive model based on, in particular the same, dynamic model of the robot.

Accordingly, in one embodiment, the system includes means for switching upon detection of a collision based on the collision force to another determination to determine a feedback force, or means for determining a feedback force after detecting a collision of the robot without Fuzzy system and / or the predictive model based on, in particular the same, dynamic model of the robot. In one embodiment, the feedback force is a feedback force of a haptic input device of the robot or a force, which is impressed on an operating element of an input device of the robot, in particular by at least one actuator, in order to force the operator a contact of the robot,

in particular to convey a guided by this tool, in particular medical instrument. Accordingly, in one embodiment, the system or the robot assembly includes a haptic input device and means for applying the feedback force to an operating element of the input device to force feedback to a user a contact of the robot, in particular a guided by this tool, in particular medical instrument mediate, on. By switching after detecting the contact or the determination of

Feedback force without fuzzy system and / or predictive model, in one embodiment advantageously distortions, in particular filtering, the force that can be useful for detecting a collision, reduced, and so a force feedback can be improved. Additionally or alternatively, prior to detecting a collision of the robot, the feedback force may be at least substantially equal to zero based on the determined collision force. Also, this can be a force feedback can be improved.

Alternatively, the feedback force in an embodiment prior to detecting a collision of the robot can also be determined on the basis of the determined collision force, in particular corresponding thereto.

A robot assembly according to an embodiment of the present invention includes a robot and a system for controlling the same described herein

Robot. A means in the sense of the present invention may be designed in terms of hardware and / or software, in particular a data or signal-connected, in particular digital, processing, in particular microprocessor unit (CPU) and / or a memory and / or bus system or multiple programs or program modules. The CPU may be configured to execute instructions implemented as a program stored in a memory system, to capture input signals from a data bus, and / or

Output signals to a data bus. A storage system may comprise one or more, in particular different, storage media, in particular optical, magnetic, solid state and / or other non-volatile media. The program may be such that it is capable of embodying or executing the methods described here, so that the CPU can carry out the steps of such methods and thus, in particular, operate the robot or

can control.

In one embodiment, one or more of the steps described herein are at least partially automated, in particular by the system or means described herein.

The invention is particularly in medical applications in which a robot-guided medical instrument temporarily contacts a patient, particularly in robot-assisted sonography or other imaging

Process, advantageous, so that these represent a preferred use and for which is therefore claimed in particular protection. For here in particular a reliable, early and / or gentle detection of a contact and / or a subsequent precise power feedback is important. However, it can equally be used in other applications and is therefore not limited to medical applications.

Further advantages and features emerge from the subclaims and the exemplary embodiments. This shows, partially schematized ,:

1 shows a robot arrangement with a robot and a system for controlling the robot according to an embodiment of the present invention

Invention; and

2 shows the sequence of a method according to an embodiment of the present invention.

1 shows a robot arrangement with a robot 30, which carries an ultrasound head 31, and a system for controlling the robot 30 or a medical application of the robot in the form of a robot controller 10, which communicates with a haptic input device 20 for controlling the robot 30 respectively.

is signal-connected.

The robot controller 10 executes a method explained below with reference to FIGS. 1, 2 according to an embodiment of the present invention and has a corresponding computer program product for this purpose.

In a first step S10, the controller 10 determines joint angles and

Driving forces, in particular torques, of the robot 30 based on a dynamic model 1 of the robot 30, an external force F _e , resulting from an external load of the robot. To do this, the ones from the dynamic model are based on the joint angles and their time derivatives

determined total joint forces subtracted the driving forces and projected this difference by means of the Jacobian matrix in the Cartesian space in which the force F _{e has} three components in the x-, y- and z-direction of a coordinate system whose z-axis oriented vertically or against the gravitational direction is. Subsequently, in a step S20, this external force F _e determined on the basis of the dynamic model 1 1 is determined by an ARMA model 12

component-wise a probable external force [F ' _e , _x , F' _e , _y , F ' _e , _z ] determined.

Thus, on the basis of the dynamic model 11 and the model 12 model-based predictive determines an external force [F _'ex, F' _E, _Y, F _'ez] resulting from an external stress of the robot 30th

In a next step S30, for the components F ' _ex , F' _e , _y and F ' _e , z of this external force respectively by FIS 13X, 13Y and 13Z forming subsystems of an FIS 13, membership degrees B based on membership functions Bi of the fuzzy system 13 determines, as indicated schematically in FIG. 1.

In an analogous manner, in step S30 in the subsystems 13X, 13Y also for horizontal components Δχ, Ay of a position error and horizontal

Components dx / dt, dy / dt a speed of the TCPs of the robot 30 each based on membership functions Bi of the fuzzy system 13th

Membership degrees B determined.

Analogously, in step S30 in the subsystem 13Z, also for a vertical component Az of the position error and a vertical component z of a position of the TCP of the robot 30 are respectively based on

Membership Functions Bi of the Fuzzy System 13 Membership degrees B determined. From these ascertained degrees of membership B, the subsystems 13X, 13Y, 13Z of the fuzzy system 13 determine a vertical rule in a step S40 in a rule-based manner

Component F _cz or horizontal components F _cx , F _{cy of} a collision force.

The rules and / or membership functions Bi of the fuzzy system 13 are parameterized in advance by a neural network. As long as the thus determined collision force [F _cx , F _cy , F _c , ₂ ] remains below a predetermined threshold, a switching means 14 of the controller 10 in a step S50 to the haptic input device 20, a feedback force F ^ equal to zero , alternative The collision force [F _c , _x , F _{c, y} , F _c , _z ] thus determined can also be output to the haptic input device 20 itself as a feedback force F _ft .

On the other hand, the collision force [F _cx , F _{c, y} , F _cz ] exceeds the given one

Threshold, the switching means 14 to the haptic input device 20 in step S50 a feedback force F «, which determines the controller 10 based on the dynamic model 1 1 without the predictive model 12 or the fuzzy system 13, in a simple Case directly the external force F _e . It can also output a corresponding, for example optical and / or acoustic, signal

If the amount falls below one, in particular the same, threshold value, the switching means 14 switches back again and the controller 10 detects the next collision of the ultrasonic head 31 in the manner described above.

Although exemplary embodiments have been explained in the foregoing description, it should be understood that a variety of modifications are possible. It should also be noted that it is the exemplary

The explanations are merely examples of the scope of protection

Applications and the structure should in no way limit. Rather, the expert is by the preceding description a guide for the

Implementation of at least one exemplary embodiment given, wherein various changes, in particular with regard to the function and arrangement of the described components, can be made without departing from the scope, as it is apparent from the claims and these equivalent

Feature combinations results.

LIST OF REFERENCE NUMBERS

10 robot control

11 dynamic model

12 AR

13

13X-13Z FIS ("Fuzzy Inference System")

14 switching means

20 haptic input device for remote control of the robot 30 robot

31 ultrasound head

B membership degree

B, membership function

F _e ,

[F'e, x, F ' _e , _y , F' _e , z] external force

[F _c , x, F _c , _y , F _c , _z ] collision force

Ffb Feedback Force

[Δχ, Ay, Δζ] position error

[dx / dt, dy / dt] horizontal speed

z vertical position

Claims

claims

A method of controlling a robot (30), in particular a medical application, comprising the steps of:

Determining (S10-S40) a collision force ([F _c , _x , F _c , _y , F _c , _z ]) on the basis of a model-based determined external force (F _e , [F ' _e , _x , F' _e , _y> F ' _e , _z ]) resulting from an external load on the robot, a position error ([Δχ, Ay, Az]) of the robot and a pose (z) and / or velocity ([dx / dt, dy / dt]) of the robot; and

Detecting (S50) a collision of the robot based on this collision force. 2. The method of claim 1, comprising the steps of:

Determining (S30) at least one degree of affiliation (B) of at least one component (F ' _e , _x , F' _e , _y , F ' _ez ) of the model-based determined external force, at least one degree of affiliation (B) of at least one component (Ax, Ay, Az) of the positional error of the robot and at least one degree of affiliation (B) of at least one component (z, dx / dt, dy / dt) of the pose and / or

Speed of the robot based on membership functions (B,) of a fuzzy system (13, 13X-13Z); and

Determining (S30) at least one component (F _cx , F _cy , F _c , _z ) of the collision force on the basis of the fuzzy system as a function of these membership degrees. 3. The method according to claim 2, characterized in that

a component (F _cz ) of the collision force in at least a first,

in particular vertical, direction depending on at least one

Membership degree of a component (F ' _ez ) of the model-based determined external force in the first direction, at least one degree of membership of a component (Az) of the positional error of the robot in the first direction and / or at least one degree of membership of a position (z) of one

robot-fixed reference in the first direction is determined; and or

a component (F _c , _x , F _c , _y ) of the collision force in at least one second, in particular horizontal, direction as a function of at least one degree of membership of a component (F ' _{e, x} , F' _{e, y} ) of the model-based determined external force in the second direction, at least one degree of affiliation a component (Δχ, Ay) of the positional error of the robot in the second direction and / or at least one degree of membership of a speed ([dx / dt, dy / dt]) of a robot-fixed reference in the second direction is determined.

Method according to one of the preceding claims 2 to 3, characterized

characterized in that the fuzzy system (13, 13X-13Z) is parameterized by means of a neural network.

Method according to one of the preceding claims, characterized in that the external force is determined on the basis of a dynamic model (11) of the robot and / or a predictive model, in particular an autoregressive model (12).

Method according to one of the preceding claims, characterized in that a feedback force (F _fb ), in particular a haptic input device (20) of the robot, after detecting a collision of the robot without the fuzzy system and / or the predictive model based on a dynamic model (1 1) of the robot is determined and / or detected before detecting a collision of the robot on the basis of the determined collision force ([F _cx , F _{c, y} , F _cz ]) or, at least substantially, equal to zero is.

7. System (10, 20) for controlling a robot (30), in particular a

medical application, which is set up to carry out a method according to one of the preceding claims and / or has:

Means (10) for determining a collision force ([F _cx , F _cy , F _c , _z ]) on the basis of a model-based determined external force (F _e , [F ' _e , _x , F' _e , _y , F ' _e , _z ]) resulting from an external load of the robot, a position error ([Δχ, Ay, Δζ]) of the robot and a pose (z) and / or speed ([x dt, dy / dt]) of the robot ; and

Means (10) for detecting a collision of the robot based on this collision force.

Robot arrangement comprising a robot (30) and a system (10, 20) for

Controlling the robot according to one of the preceding claims.

A computer program product having a program code stored on a computer readable medium for carrying out a method according to any one of the preceding claims.