Transfer of objects held by robots from receiving poses to target poses

Abstract

The invention relates to a method having the steps of: ascertaining (S60) a predicted result for an attempt to transfer an object (41, 42,..., 45) which is held by a robot (10) from a receiving pose to a target pose of the transfer attempt on the basis of gripping data based on a gripping process for gripping the object in the receiving pose by means of the robot and at least one of the two steps of: providing (S50) the gripping data and / or controlling and / or monitoring (S70) the transfer attempt on the basis of the ascertained predicted result. The invention additionally relates to a system or computer program (product).

Description

[0001] Description

[0002] Transfer of robot-held objects from capture poses to target poses

[0003] The present invention relates to a method for controlling and / or monitoring an attempt to transfer an object held by a robot from a receiving pose to a target pose and / or for determining a predicted result of the transfer attempt, as well as a system or computer program or computer program product for carrying out a method described here.

[0004] An object of an embodiment of the present invention is to improve a transfer of an object held by a robot from a receiving pose to a target pose and / or a prediction of a result of such a transfer attempt.

[0005] This object is achieved in particular by a method having the features of claim 1 and 4. Claims 12 and 13 represent a system and

[0006] A computer program or computer program product for carrying out a method described here is protected. The subclaims relate to advantageous developments.

[0007] According to one embodiment of the present invention, an attempt is made to move an object held by a robot from a receiving pose to a target pose using (a corresponding movement) of the robot, which is referred to herein as a transfer (attempt).

[0008] In one embodiment, the robot has at least three, preferably at least six, in a further development at least seven, joints or (movement) axes, in one embodiment a robot arm with at least three, preferably at least six, in a further development at least seven, joints or

[0009] (Movement) axes, wherein the robot (arm) preferably has, preferably electric (motor) drives for adjusting the joints or (movement) axes and / or an end effector for temporarily or non-destructively releasable holding of objects, the present one is referred to as a gripper as is customary in the art without restriction of generality, in one embodiment a mechanically (acting), magnetically (acting), preferably electromagnetically (acting), and / or pneumatically (acting) or suction gripper, with which it grips the object held (in each case) by the robot, in one embodiment mechanically, preferably frictionally and / or positively, magnetically, preferably electromagnetically, and / or pneumatically or by means of negative pressure, or which is set up or used for this purpose.Accordingly, in one embodiment, an object held by a robot is an object gripped in this way by the robot (arm / gripper), and / or gripping of the object by the robot comprises, in one embodiment, a mechanical gripping action, preferably frictional and / or positive-locking, magnetic, preferably electromagnetic, and / or pneumatic, or achieved by means of negative pressure, and can in particular be such a gripping action. Activating the gripper in one embodiment comprises a closing movement of the gripper and / or the effecting of an (electro)magnetic and / or negative pressure-based holding force, and can in particular consist of this.

[0010] For such robot-assisted gripping processes with subsequent transfer, the present invention is particularly advantageous due to the boundary conditions that exist there, in particular objects that are often difficult and / or unreliable to grip (robotically), rapid transfer movements of the robot or the like.

[0011] A pose within the meaning of the present invention comprises, in one embodiment, a one-, two-, or three-dimensional position and / or one-, two-, or three-dimensional orientation. A pose of an object, in particular of an object held by the robot, is, in one embodiment, a pose of the object relative to the robot, in a further development (in) its gripper, or relative to an environment. A pose of a robot is, in one embodiment, determined by the position(s) of its joints or (motion) axes and / or the position and / or orientation of its gripper. In one embodiment, the pickup and / or target pose are, in particular, are, predetermined, in a further development before the transfer attempt is carried out.

[0012] According to one embodiment of the present invention, a predicted or

[0013] (computationally) predicted result of the transfer attempt is determined on the basis of or in dependence on data which are based on a process of, in particular, an actually attempted and, preferably sensor-detected, gripping of the object (gripping process) in the receiving pose by the robot or which depend on this gripping process and are accordingly referred to here as gripping data.

[0014] This is based on the realization that during a (gripping) process of grasping the object, various parameters, for example forces, images or the like, have values ​​that allow a more precise and / or reliable prognosis or prediction of an outcome, in particular success, of the transfer, and the idea based on this to record these parameter values ​​and use them for prognosis or prediction.

[0015] Such a predicted result can be used particularly advantageously, in particular for controlling and / or monitoring the transfer attempt.

[0016] Accordingly, according to one embodiment of the present invention, the transfer attempt (for which the result is predicted or the predicted result is determined) is controlled and / or monitored on the basis of the determined predicted result.

[0017] Thus, according to one embodiment of the present invention, a method according to the invention comprises the step:

[0018] - Determining a predicted result of an attempt to transfer an object held by a robot from a receiving pose to a target pose based on grasping data based on a (grasping) operation of grasping the object in the receiving pose by the robot; and the step:

[0019] - Providing this gripping data (based on a gripping operation of the object in the pickup pose by the robot); and / or the step:

[0020] - Controlling and / or monitoring the transfer attempt based on the determined predicted result. Providing within the meaning of the present invention can include, in particular, determining, in particular capturing, storing, transmitting and / or receiving, retrieving or loading (from a memory), and / or processing the gripping data, and in particular can consist of these.

[0021] In one embodiment, controlling and / or monitoring the transfer attempt comprises terminating this attempt, in a further development, releasing the object (held by the robot) before reaching the target pose, in particular instead of further traversing a transfer trajectory of the robot specified for the transfer, in one embodiment

[0022] - dropping of the object by the robot or

[0023] - a setting down of the object (by the robot) in a setting down pose, which is determined in one execution on the basis of the determined predicted result.

[0024] By aborting the transfer (attempt) in this way, unnecessary work of the robot and thus in particular energy and / or (process or cycle) time can be saved.

[0025] By deliberately setting down the object in a determined setting down pose, the chance of success of a further transfer attempt can be improved in one embodiment, in particular by setting down the object in an advantageous further development in a setting down pose in which it can subsequently be picked up well or better by the robot.

[0026] In one embodiment, a transfer trajectory of the robot comprises a path of the robot, preferably its gripper, and in a further development, also a speed when traveling along this path. A transfer attempt within the meaning of the present invention comprises, in one embodiment, an attempt to travel along a transfer trajectory or path with the robot.

[0027] In addition or alternatively to aborting the transfer (attempt), in one embodiment, controlling and / or monitoring the transfer attempt comprises initiating, in particular carrying out, a one- or multi-stage countermeasure, in a further development - changing the holding of the object by the robot, preferably during the transfer attempt; and / or

[0028] - changing one or the specified transfer trajectory of the robot for the transfer, preferably during the transfer attempt; in a further development

[0029] - changing, preferably at least temporarily increasing, a holding force with which the robot holds the object; and / or

[0030] - changing a pose of the object, which is preferably still held, relative to the robot, in particular its gripper.

[0031] Additionally or alternatively, in one embodiment, controlling and / or monitoring the transfer attempt

[0032] - a change in the holding of an object, in particular the same object, by the robot during a further, in particular repeated, transfer attempt; and / or

[0033] - changing one or the specified transfer trajectory of the robot during or for a further, in particular repeated, transfer attempt; in a further training

[0034] - changing, preferably increasing, a holding force with which the robot holds the object during the further, in particular renewed, transfer attempt; and / or

[0035] - changing a pose of the object relative to the robot, in particular its gripper, during the further, in particular repeated, transfer attempt.

[0036] In one embodiment, this makes it possible to advantageously react to a predicted problem, in particular a predicted failure, in the current transfer attempt, in particular a predicted error or failed attempt, and / or to improve the prospects of success of the current and / or subsequent transfer attempt, for example by the robot re-engaging and / or gripping more firmly or the like in the current or subsequent transfer attempt, and / or by the robot following a planned path more slowly or replanning or the like in the case of a predicted problem.

[0037] Additionally or alternatively, in one embodiment, controlling and / or monitoring the transfer attempt includes controlling a process dependent on the transfer, in particular the transfer performed, in particular further handling of the object after the transfer. In one embodiment, this process dependent on the transfer can thereby be improved.

[0038] In one embodiment, the predicted result is determined by processing the provided gripping data using data processing based at least partially on machine learning, in a further development using at least one machine-learned model or machine learning algorithm, which preferably links the gripping data and predicted results or maps gripping data to predicted results, in one embodiment using at least one artificial neural network. In one embodiment, the artificial neural network has 2D convolutional layers, which is particularly advantageous for (data) processing gripping data that includes image data. Additionally or alternatively, in one embodiment, the artificial neural network has 1D convolutional layers, which is particularly advantageous for (data) processing gripping data that includes time series.

[0039] By means of (such) at least partially machine-learned or machine-learning-based data processing, results of robot-assisted transfer attempts based on gripping data based on a gripping process of gripping an object by the robot can be predicted particularly well, in particular more precisely, reliably and / or for different and / or complex processes.

[0040] In one embodiment, the method comprises the steps:

[0041] - Provide

[0042] - from learning data based on grasping operations from the grasping of objects by a robot, preferably the robot or a similar robot, and

[0043] - of result data, in particular labels or evaluations based on attempts to transfer these objects by this robot, in particular on a preferably automated labelling or evaluation of these attempts; and

[0044] - Training, in particular machine-based, preferably supervised, learning or training, of data processing based on these learning data and result data. This allows the results of robot-assisted transfer attempts to be predicted particularly well, in particular more precisely, reliably, and / or for different and / or complex processes, based on gripping data derived from a gripping process of an object by the robot.

[0045] In one embodiment, a gripping operation comprises gripping an object in the sense of the present invention

[0046] - approaching the object with the robot and / or

[0047] - contact of the object with the robot, in particular its gripper, and / or

[0048] - activating one or the gripper of the robot (for gripping, in particular mechanically, preferably frictionally and / or positively, magnetically, preferably electromagnetically, and / or pneumatically or by means of negative pressure) and / or

[0049] - a retraction movement of the robot after activation of the gripper or when the gripper is activated, in particular with the object held, in a further development, an approach to a start pose of a or the transfer trajectory with the gripper activated.

[0050] In particular, such grasping processes can provide parameter values ​​that allow a more precise and / or reliable prognosis or prediction of an outcome, in particular the success, of the transfer.

[0051] In one embodiment, in a further development, the gripping data and / or the learning data and / or the result data (each) are determined using one or more robot-side or robot-fixed sensors, in a further development one or more drive sensors and / or one or more joint sensors and / or one or more gripper sensors.

[0052] Such sensors are often advantageously present anyway for control and / or monitoring. Additionally or alternatively, such sensors can be used in one embodiment to determine or provide advantageous data even when the robot is relocated and / or in confined spaces. Additionally or alternatively, in a further development, the gripping data and / or the learning data and / or the result data are (each) determined using one or more (robot-)external, in particular environmental, sensors, preferably using one or more 2D cameras and / or one or more 3D cameras.

[0053] Such sensors can often determine or provide particularly meaningful data that has not previously been used in robot-assisted grasping and transfer processes and / or allow (even) more precise and / or reliable prognoses or predictions of an outcome, in particular the success, of the transfer.

[0054] Additionally or alternatively, in a further development, the gripping data and / or the learning data and / or the result data are determined in one embodiment (each) using at least one data model of the robot and / or at least one data model of the object, in a further development of a CAD data model or the like.

[0055] Such data models can, for example, enable or improve the determination of a pose of the object, in particular relative to the robot (gripper), in particular increase precision and / or reduce computational effort.

[0056] Additionally or alternatively, in one embodiment, the gripping data and / or the learning data and / or the result data (each) depend on forces, in a further development on drive forces and / or contact forces, and / or holding forces, and can in particular specify or describe these. For a more compact representation, antiparallel force pairs or torques are generally referred to as forces.

[0057] Additionally or alternatively, in one embodiment, the gripping data and / or the learning data and / or the result data (each) depend on poses of the robot and / or object, and can specify or describe these and / or temporal derivatives thereof, in particular speeds and / or accelerations of the robot or object. In a further development, the gripping data and / or the learning data and / or the result data (each) comprise images or image data of the robot and / or the object.

[0058] In particular, such force, pose or image data can allow a more precise and / or reliable prognosis or prediction of an outcome, in particular success, of the transfer.

[0059] Additionally or alternatively, in one embodiment, the gripping data and / or the learning data depend on values ​​that have been recorded during a gripping process of gripping an object by a robot, preferably with the aid of the sensor or sensors, in one embodiment, in a particularly preferred development, on values ​​that - preferably (in each case) with the aid of the sensor or sensors,

[0060] - during the approach of the object or the recording pose with the robot and / or

[0061] - during contact of the object with the robot, in particular its gripper, and / or

[0062] - during the activation of a gripper of the robot (for gripping, in particular mechanical, preferably frictional and / or positive, magnetic, preferably electromagnetic, and / or pneumatic or by means of negative pressure) and / or

[0063] - with a gripper activated (to hold the object) and / or

[0064] - during a or the retraction movement of the robot after activation of the gripper or with activated gripper, in particular with (possibly) held object, in a further development during a or the approach to a start pose of a or the transfer trajectory, have been recorded, in one embodiment.

[0065] Accordingly, in one embodiment, a method according to the invention comprises the step of: providing gripping data based on a gripping process of gripping the object in the pickup pose by the robot, which gripping data depend on values ​​detected during the gripping process, preferably with the aid of the sensor(s), in a further development also the step of: detecting, preferably with the aid of the sensor(s), values ​​during the gripping process and providing the gripping data based on these detected values, wherein these values ​​are particularly preferably

[0066] - during the approach of the object or the recording pose with the robot and / or

[0067] - during contact of the object with the robot, in particular its gripper, and / or

[0068] - during the activation of a gripper of the robot (for gripping, in particular mechanical, preferably frictional and / or positive, magnetic, preferably electromagnetic, and / or pneumatic or by means of negative pressure) and / or

[0069] - with a gripper activated (to hold the object) and / or

[0070] - during a or the retraction movement of the robot after activation of the gripper or with activated gripper, in particular with (possibly) held object, in a further development during a or the approach to a start pose of a or the transfer trajectory, have been recorded, in one embodiment.

[0071] As explained, this embodiment is based on the knowledge that during a gripping process parameters, for example forces, scenes or the like, have values, for example force values, image data or the like, which allow a more precise and / or reliable prognosis or prediction of a result, in particular the success of the transfer, and the idea based on this is that these parameter values ​​are recorded and used for prognosis or prediction. This becomes particularly clear from forces recorded when contacting the object and / or activating the gripper and / or activated gripper, in particular drive and / or contact forces, which accordingly form particularly advantageous values ​​or gripping data, without the invention being restricted to this. Likewise, image orPose data and the like, which occur particularly when contacting the object and / or activating the gripper and / or when the gripper is activated, enable particularly good predictions.

[0072] Additionally or alternatively, in one embodiment, the predicted outcome is determined before or during the trans-test. This allows for a particularly advantageous response to the forecast.

[0073] In one embodiment, the grasping data and / or learning data (respectively) comprise image data and / or time series.

[0074] In particular, such data may allow a more precise and / or reliable prognosis or prediction of an outcome, in particular the success, of the transfer.

[0075] In one embodiment, the grasping data and / or learning data (respectively) comprise no image data and / or no time series.

[0076] In one embodiment, this can improve machine learning or training (of data processing), in particular by accelerating and / or simplifying it.

[0077] In one embodiment, the determined predicted result and / or the result data depend on a loss of the object during the transfer attempt. In particular, a loss can lead to a prognosis or evaluation as a worse or poor result. In general, a determined predicted result within the meaning of the present invention can include, in particular, a success of the transfer attempt.

[0078] Additionally or alternatively, the determined predicted result and / or the result data in one embodiment depend on a pose of the object at the end of the transfer attempt and / or a predefined value range. In particular, a better or good result can be predicted or labeled if a deviation between the predicted or actual and a desired or target pose lies within a predefined tolerance range, or the determined predicted result can indicate the deviation or the degree of deviation, in one embodiment in continuous form or discretized in two or more stages. Thus, in an advantageous, particularly simple, embodiment when training the data processing, a T(raining) transfer attempt can be evaluated or labeled as failed or bad if the object is no longer held by the robot after the T(raining) transfer journey, and evaluated or labeled as successful or good.be labeled if the object is (still) held by the robot after the training transfer.

[0079] In general, in one execution (binary): “(expected) successful” or “(expected) unsuccessful” can be predicted or determined as the result.

[0080] In another, also particularly advantageous embodiment, for example, when training the data processing, a T(rainingst)transfer attempt can be evaluated or labeled according to how large a deviation is between a target or (the) target pose and the achieved pose of the object after the T(rainingst)transfer run, in one further development the smaller the deviation the better, in another further development as failed or bad as soon as the deviation lies outside a specified tolerance range.

[0081] Accordingly, for example, here too (binary) “(probably) successful” or “(probably) unsuccessful” or the (probably) deviation between the desired or target pose and the (probably) achieved pose of the object can be predicted as a result or determined as a predicted result, in particular as a deviation discretized in steps or in continuous, possibly scaled, form.

[0082] According to one embodiment of the present invention, a system, in particular hardware and / or software, in particular program-technical, is set up to carry out a method described here and / or has one or preferably several of the following means:

[0083] - Means for determining a predicted result of an attempt to transfer an object held by a robot from a receiving pose to a target pose (transfer attempt) based on gripping data based on a gripping process of gripping the object in the receiving pose by the robot; - Means for controlling and / or monitoring the transfer attempt based on the determined predicted result, in particular

[0084] - means for terminating the attempt, in one embodiment dropping off the object before reaching the target pose, in a further development dropping off the object in a determined dropping pose;

[0085] - means for initiating, in particular carrying out, a countermeasure, in particular for changing a holding of the object by the robot and / or a predetermined transfer trajectory of the robot for the transfer; and / or

[0086] - means for changing a holding of an object by the robot and / or a predetermined transfer trajectory of the robot in a further transfer attempt; and / or

[0087] - means for controlling a process dependent on the transfer, in particular further handling of the object after the transfer;

[0088] - means for providing the gripping data;

[0089] - a data processing means based at least partially on machine learning, in particular at least one artificial neural network, for determining the predicted result by processing the provided gripping data;

[0090] - means for providing learning data based on grasping operations of objects by a robot and result data based on attempts to transfer these objects by the robot;

[0091] - means for training data processing based on these learning data and result data;

[0092] - at least one robot-side sensor for determining the gripping data, learning data and / or result data;

[0093] - at least one external sensor for determining the gripping data, learning data and / or result data;

[0094] - at least one data model of the robot and / or object for determining the gripping data, learning data and / or result data;

[0095] - Means for providing values ​​acquired during a gripping process of gripping an object by a robot for determining the gripping data, learning data, and / or result data, in particular means for acquiring these values. A system and / or means within the meaning of the present invention can be designed in hardware and / or software, in particular at least one, in particular digital, processing unit, in particular a microprocessor unit (CPU), graphics card (GPU), or the like, preferably data- or signal-connected to a memory and / or bus system, and / or one or more programs or program modules. The processing unit can be designed to execute commands implemented as a program stored in a memory system, to acquire input signals from a data bus, and / or to output output signals to a data bus.A storage system can have one or more, in particular different, storage media, in particular optical, magnetic, solid-state and / or other non-volatile media. The program can be such that it embodies or is capable of carrying out the methods described here, so that the processing unit can carry out the steps of such methods and thus control and / or monitor an attempt to transfer an object held by a robot from a receiving pose to a target pose or determine a predicted result of the attempt. In one embodiment, a computer program product can have, in particular be, a storage medium, in particular a computer-readable and / or non-volatile one, for storing a program or instructions or with a program or instructions stored thereon. In one embodiment, executing this program orof these instructions by a system or a controller, in particular a computer or an arrangement of several computers, the system or the controller, in particular the computer(s), to carry out a method described here or one or more of its steps, or the program or the instructions are set up for this purpose.

[0096] In one embodiment, one or more, in particular all, steps of the method are fully or partially computer-implemented or one or more, in particular all, steps of the method are fully or partially automated, in particular by the system or its means.

[0097] In one embodiment, the system comprises the robot. Further advantages and features emerge from the subclaims and the exemplary embodiments. The following shows, partially schematically:

[0098] Fig. 1 : a system according to an embodiment of the present invention; and

[0099] Fig. 2: a method according to an embodiment of the present invention.

[0100] Fig. 1 shows a system according to an embodiment of the present invention with a robot 10, a data processing device 20 and cameras (sensors) 31, 32.

[0101] The robot 10 uses its gripper 11 to successively grasp one of the objects 41, 42, ... arranged in a container 40 and transports it to a storage area 400 or attempts to do so.

[0102] During a training phase, measured values ​​are initially recorded during each gripping process of gripping one of the objects by the robot 10 using the camera 31 and / or one or more robot-side sensors, of which a contact or contact force sensor 12 on the gripper 11 is indicated as a particularly advantageous example, and, if necessary after further processing such as filtering or the like, are stored as learning data (Fig. 2: step S10). Additionally or alternatively, drive forces can also be recorded and, if necessary after further processing such as filtering or the like, stored as learning data.Preferably, the learning data are based on (measurement) values, at least some of which have been or are being recorded during an approach and / or contact of the object with the gripper, during an activation of the gripper and / or with the gripper activated, in particular during a retraction movement of the robot from the pick-up pose to a start pose of a transfer trajectory or path.

[0103] After this gripping process, a transfer (attempt) is carried out in each case (Fig. 2: step S20) and its result is evaluated and stored as result data (Fig. 2: step S30), for example by means of the camera 32 and image processing a deviation of the achieved from a desired target or target pose of the respective object at the end of the transfer is quantified and / or by means of the sensor 12 a possible loss of the object during the transfer (attempt) is detected and the corresponding transfer (attempt) is accordingly evaluated as successful or unsuccessful or as success or failure.

[0104] With the learning and result data thus obtained, a data processing system based at least partially on machine learning, in the exemplary embodiment at least one artificial neural network 21, is trained in a step S40.

[0105] Subsequently, in an operative phase, analogous to the learning data, gripping data are acquired during a gripping process of gripping one of the objects 41, 42, ... by the robot 10 (Fig. 2: step S50).

[0106] Then, the data processing device 20 uses the trained artificial neural network 21 to determine a predicted result of a transfer attempt following this grasping process (Fig. 2: step S60).

[0107] Based on this predicted result, the data processing device 20 controls and / or monitors the transfer attempt (Fig. 2: step S70).

[0108] For example, it can drop an object grasped in step S50 into the center of the container 40 or place it in a pose (particularly) more suitable for grasping and then pick up one of the objects again, in one embodiment specifically this dropped object, and attempt transfer again if a failure is predicted based on the gripping data obtained during the gripping process of grasping the object in the pick-up pose by the robot. Likewise, in step S70, the data processing device 20 can also increase a holding force for holding the object during the transfer attempt and / or change a grip. Additionally or alternatively, in step S70, a process dependent on the transfer, for example, transporting the object away after the transfer, can also be controlled based on the determined predicted result.

[0109] By way of example, Fig. 1 shows recording poses of the objects 41, 42 and 43, a target pose of the object 45 and a transfer trajectory or path of the object 44 held by the robot 10 or its gripper 11. Although exemplary embodiments have been explained in the preceding description, it should be noted that a large number of modifications are possible.

[0110] For a more compact illustration, the training of the data processing or the artificial neural network 21 and the subsequent control or monitoring was explained using the same container 40 with objects 41, 42, .... Preferably, a separate training phase can be carried out beforehand with other objects and / or another robot, which are preferably similar to the objects whose transfer is subsequently to be controlled or monitored in an operative phase, or are similar to the robot gripper 11 used therein, wherein the data processing or the artificial neural network 21 can advantageously also be further trained during this operative phase.

[0111] Furthermore, it should be noted that the exemplary embodiments are merely examples and are not intended to limit the scope of protection, applications, or structure in any way. Rather, the preceding description provides the skilled person with a guide for implementing at least one exemplary embodiment. Various modifications, particularly with regard to the function and arrangement of the described components, may be made without departing from the scope of protection as defined by the claims and equivalent combinations of features.

[0112] List of reference symbols

[0113] 10 robots

[0114] 11 grippers

[0115] 12 Sensor 20 Data processing device

[0116] 21 Artificial Neural Network (Data Processing Device)

[0117] 31 , 32 Camera (sensor)

[0118] 40 containers

[0119] 41 - 45 objects 400 storage space

Claims

Patent claims 1. A method for controlling and / or monitoring an attempt to transfer an object (41, 42, ..., 45) held by a robot (10) from a receiving pose to a target pose, the method comprising the steps of: - determining (S60) a predicted result of the transfer attempt based on gripping data based on a gripping process of gripping the object in the pickup pose by the robot; and - Controlling and / or monitoring (S70) the transfer attempt based on the determined predicted result.

2. Method according to claim 1, characterized in that the control and / or monitoring of the transfer attempt - an abort of the experiment, in particular a release of the object before reaching the target pose, in particular a release of the object in a determined release pose; - a countermeasure, in particular a change in the holding of the object by the robot and / or a predetermined transfer trajectory of the robot for the transfer; and / or - changing the robot's holding of an object and / or a predetermined transfer trajectory of the robot during a further transfer attempt; and / or - controlling a process dependent on the transfer, in particular further handling of the object after the transfer.

3. Method according to one of the preceding claims, characterized in that it comprises providing the gripping data (S50) and / or the predicted result is determined according to one of the following claims.

4. A method for determining a predicted result of an attempt to transfer an object (41, 42, ..., 45) grasped by a robot (10) from a receiving pose to a target pose, the method comprising the steps of: - providing (S50) gripping data based on a gripping operation of gripping the object in the pickup pose by the robot; and Determining (S60) a predicted result of the transfer attempt based on the provided grasping data.

5. Method according to the preceding claim, characterized in that the predicted result is determined by processing the provided gripping data by means of data processing based at least partially on machine learning, in particular by means of at least one artificial neural network (21).

6. Method according to the preceding claim, characterized by the steps: - providing (S10) learning data based on grasping operations for grasping objects by a robot and result data based on attempts to transfer these objects by the robot (S30); and - Training (S40) of data processing based on these learning data and result data.

7. Method according to one of the preceding claims, characterized in that a gripping process of gripping an object comprises approaching and / or contacting the object with the robot and / or activating a gripper of the robot and / or a retraction movement of the robot after activating the gripper, in particular approaching a start pose of a transfer trajectory.

8. Method according to one of the preceding claims, characterized in that the gripping data, learning data and / or result data are - at least one robot-side sensor (12), - at least one external sensor (31, 32), and / or - at least one data model of the robot and / or object is determined and / or - forces, in particular driving, contact and / or holding forces, and / or - Poses of the robot and / or object depend; and / or that the grasping data and / or learning data depend on values ​​acquired during a grasping process of an object by a robot; and / or that the predicted result is determined before or during the trans experiment.

9. Method according to one of the preceding claims, characterized in that the gripping data and / or learning data comprise image data and / or time series.

10. Method according to one of the preceding claims 1 to 8, characterized in that the gripping data and / or learning data do not comprise image data and / or time series.

11. Method according to one of the preceding claims, characterized in that the determined predicted result and / or the result data depend on a loss of the object during the transfer attempt and / or a pose of the object at the end of the transfer attempt and / or a predetermined range of values.

12. System which is arranged to carry out a method according to one of the preceding claims and / or - means for determining a predicted result of an attempt to transfer an object (41, 42, ..., 45) held by a robot (10) from a pickup pose to a target pose on the basis of gripping data based on a gripping operation of gripping the object in the pickup pose by the robot; and - means for providing the gripping data and / or means for controlling and / or monitoring the transfer attempt on the basis of the determined predicted result.

13. A computer program or computer program product, wherein the computer program or computer program product contains instructions, in particular stored on a computer-readable and / or non-volatile storage medium, which, when executed by one or more computers or a system according to claim 12, cause the computer(s) or the system to carry out a method according to one of claims 1 to 11.

Images