Method and system for manipulating a load cell with a robotic gripper

Abstract

The method for handling a load unit (21) with a robot (10) according to the present application comprises the following steps: - activating a lifting state of a gripper (12) of the robot for load lifting; - determining, during the movement of the lifted load unit, a parameter of a time curve of a force quantity dependent on the load unit by means of at least one sensor of the robot; - classifying the load unit lifted by the gripper based on the determined parameter by means of a machine learning model, in particular during the movement of the lifted load unit and / or above a receiving area in which the load unit to be lifted has been located, in particular above a receiving area of a receiving station and / or above or in a receiving container (20); and at least one of the following steps: - if the load unit is classified into a first category, performing a first process (S141) with the robot; and / or - if the load unit is classified into a second category, performing a second process (S142) with the robot.

Description

Technical Field

[0001] The present invention relates to a method for manipulating a payload unit using a robot, and a system and computer program product for performing the method. Background Technology

[0002] A robot gripper with multiple suction elements is known from patent document DE102016220643A1.

[0003] When using such a gripper, it is particularly likely that too many objects or loads will be lifted together and then, for example, picked up together into a storage container. Summary of the Invention

[0004] One embodiment of the present invention aims to better utilize the robot (gripper) to manipulate the load unit, advantageously eliminating the problem of unintentionally taking away too much and / or the wrong load or object.

[0005] The object of the present invention is achieved by a method having the features described in claim 1. Claims 9 and 10 claim protection for a system or computer program product for performing the method described herein. Preferred extensions are given in the dependent claims.

[0006] According to one embodiment of the present invention, a method for manipulating a payload unit using a robot is proposed. In one embodiment, the method includes the following steps performed sequentially or continuously:

[0007] - Activate the lifting state of the robot's gripper to increase the load;

[0008] - During the movement of the lifted load unit (by means of or with the aid of one or more robots), one-dimensional or multi-dimensional parameters of the time curve of the force parameters dependent on the load unit are determined by means of one or more sensors of the robot; and

[0009] - Based on the detected parameters, the load units lifted by the gripper are classified by means of, in one embodiment, a machine learning model; in another embodiment, the classification is performed during the movement of the lifted load units (by means of (one or more) robots); and / or the classification is performed (especially during movement) above a set or defined storage area where load units that need to be lifted, especially those that have already been lifted, are found, in one embodiment above the storage area of ​​the storage station and / or above or in the storage container, that is, especially during the movement of the lifted load units (by means of (one or more) robots) above the storage area.

[0010] And includes at least one of the following steps:

[0011] - If the payload unit has been (already) classified into or assigned to the first category, then the robot performs the first process; and / or

[0012] -If the load unit has (already) been classified into the second category or assigned to the second category,

[0013] Then a robot is used to perform the second process.

[0014] In one implementation, by classifying the load units lifted by the gripper according to the parameters of the time curves of the force parameters dependent on the load units, using a machine learning model, it is advantageous to be able to react accordingly even during the movement of the lifted load units (by or by means of (multiple) robots) and / or to avoid or reduce stopping to measure the weight of the load units after (vibration) dynamic decay, thereby shortening processing time. Additionally or alternatively, in one implementation, simpler and therefore particularly more reliable and / or cheaper sensors can be used. This is especially true because different load units differ in terms of the time curves of the force parameters dependent on the load units, making it possible to classify them particularly well using a machine learning model.

[0015] In one implementation, the method includes the step of: if the payload unit has been classified into or assigned to a third category, then performing a third process using a robot; in an extended implementation, the method includes the step of: if the payload unit has been classified into or assigned to a fourth category, then performing a fourth process using a robot. In one implementation, one or more other categories may also be provided, wherein if the payload unit is assigned to an other category (one of these categories), then performing corresponding, category-specific other processes using a robot.

[0016] Accordingly, in one implementation, it is advantageous to make (more) specialized responses for different load units.

[0017] Therefore, if a load cell is classified as a load cell with (only) one (unique) object, then the object being lifted can be picked normally (first process); if a load cell is classified as a load cell with (exactly) two objects, then the two objects can be picked sequentially in two stages (second process); and if a load cell is classified as a load cell with (exactly) three objects, then the three objects can be picked sequentially in three stages (third process).

[0018] Alternatively, if a load cell is classified as a load cell with a number of incorrect objects, instead of the two-stage and / or three-stage sequential picking, for example, the lifted load cell may be delivered (Abgabe) (second or third process), particularly above the receiving area, without picking it thereon or accordingly.

[0019] Similarly, in one implementation, if a load unit is classified as a load unit with the correct type of object, the lifted object can be picked (first process), and the lifted load unit can be delivered above the storage area (second process); if a load unit is classified as a load unit with the wrong object type, it is not picked at this location or accordingly. Alternatively, if a load unit is classified as a load unit with a second type of object, the object can also be picked at a second picking station, for example (second process).

[0020] In one embodiment, the robot has a robotic arm with at least three, particularly at least six, and in one embodiment at least seven joints, particularly rotary joints, and a gripper is arranged on the (end) flange of the robotic arm, which is detachably fixed in one embodiment.

[0021] This invention is particularly advantageous in this regard due to the flexible application possibilities of such robots.

[0022] In one embodiment, activating the lifting state includes switching to an operating mode in which the gripper can lift a load or object, or apply a corresponding clamping force to one or more loads or objects. In one embodiment, if the gripper is a pneumatic gripper with one or more suction units, activating its lifting state includes activating said one or more, preferably all, suction units, particularly generating negative pressure. Additionally or alternatively, in one embodiment, if the gripper is a mechanical gripper with one or more movable, particularly pivotable, clamps, activating its lifting state includes closing said one or more, preferably all, clamps. Accordingly, in one embodiment, deactivation includes deactivating the suction units or opening the clamps.

[0023] By classifying the load units while they are still above the storage area, particularly above or in the storage container, in one implementation, it is advantageous to make it (more) easier to put back or put down (fallengelassen) objects or loads that have been mistakenly or excessively lifted, and then lift them again.

[0024] As described above, embodiments of the present invention are particularly suitable for eliminating unintentionally lifting too little and / or, more advantageously, too much load or object. Accordingly, in one embodiment, the first and second categories, as well as possible third, fourth, and / or other categories, of the same type in one embodiment, of loads or objects lifted by the gripper, differ in quantity.

[0025] In one implementation, the first category is a category that has only one object or only one load (“Einfachgriff”).

[0026] Additionally or alternatively, in one implementation, the second category and / or the third category and / or the fourth category (respectively) are categories with multiple objects or loads (“multi-grip”), and / or one of the second category and the possible third and possible fourth categories is a category without objects or loads (“error grip”). Here, in an extended embodiment, the second category is a category with more than one object, or the second category is a category with exactly two objects, while the third category is a category with more than two objects. In an extended embodiment, when the gripper is empty after activating the lifting state, the load unit lifted by the gripper is classified as a category without objects or loads, or, for the sake of conciseness, the case where the gripper is empty after activating the lifting state is also generally described as the load unit lifted by the gripper having no objects or being of no quantity.

[0027] Therefore, in one implementation, it is advantageous to make (more) specific responses to different load cells.

[0028] Therefore, in one embodiment, the present invention can be used to lift load cells that have different numbers of objects or loads of the same type in one embodiment, but is not limited thereto.

[0029] Similarly, in one implementation, in addition to the number of objects or loads being lifted, a first process, a second process, or possibly a third and / or a fourth process may be performed depending on the classification type of the load being lifted by the gripper.

[0030] Therefore, specifically, the first category can be a category of objects or loads of a first type, and the second category can be a category of another (type) of objects, which in one embodiment can be a category of objects or loads of a second type. In an extended embodiment, the first process can include transferring the lifted load unit to a first picking station by means of a robot, and the second process can include, differently, picking the lifted load unit or transferring the lifted load unit to a second picking station by means of a robot, or it can also include deactivating the lifting state of the gripper above the storage area and making a new lifting attempt (for the correct object or load type).

[0031] In one embodiment, additionally or in addition to the number and / or type of the object or load being lifted, the first process, the second process, or possibly the third and / or fourth process may be performed based on the distribution of the load lifted by the gripper, or in one embodiment, based on the distribution or (multiple) positions relative to the gripper, particularly on the gripper.

[0032] Therefore, specifically, the first category can be a category including a first load drawn together by a subset of a set of a plurality of suction devices of the gripper and a second load drawn by one or more other suction devices of the gripper, and the second category can be a category including the first and second loads drawn by suction devices in other states. Accordingly, in an extended embodiment, the first process may include deactivating one or more additional suction devices, and the second process may include a different partial deactivation, in which at least one suction device is also deactivated, while one or more suction devices identified as carrying the first load remain active. In other words, in one implementation, a machine learning model identifies the distribution or(s) of the lifted load relative to the gripper, particularly on the gripper, wherein, subsequently in the first, second, and possibly third and fourth processes, different lifting devices of the gripper, particularly suction devices, are selectively deactivated to provide one or more such loads, while at least one additional lifting device identified as carrying the load remains active.

[0033] In one implementation, these categories differ accordingly in terms of the quantity, distribution, and / or type of load lifted by the gripper.

[0034] As described above, in one embodiment, the gripper has two or more lifting devices, and in an extended version, it has two or more, preferably at least three suction devices. This invention is particularly advantageous because such suction grippers are more likely to unintentionally lift excessive loads or objects.

[0035] In one embodiment, at least one sensor is arranged on the gripper, between the gripper and the robot arm to which the gripper is fixed, or on the robot arm. In one embodiment, the sensor is a force sensor and / or a torque sensor, and / or is arranged on the robot arm flange, or between the gripper and the robot arm, or on one of the joints of the robot (arm). In one embodiment, this allows for particularly advantageous, and especially reliably and / or (more) easily, the detection of time curves of force parameters dependent on the load cell.

[0036] In one embodiment, the one-dimensional or multi-dimensional force parameters depend on at least one force component dependent on the load element, in one embodiment being a vertical force component, and / or depend on at least one torque component dependent on the load element, particularly having, and particularly being (multiple) such force or torque components. These are particularly advantageous for classifying the lifted load elements.

[0037] In one embodiment, the one-dimensional or multi-dimensional time curve of the force parameters dependent on the load element is detected by one or more sensors, and / or in one embodiment lasts for more than 0.5 seconds, particularly at least 1 second, and in another embodiment lasts for more than 2 seconds. In one embodiment, the time curve is sampled at a sampling rate of at least 0.01 kHz.

[0038] In one implementation, the one-dimensional or multi-dimensional parameters of the time curve have the time curve itself, and in particular, may be the time curve. Additionally or alternatively, they may also have, in particular, one or more parameters or values ​​derived from the time curve, which in one implementation are parameters or values ​​derived in signal analysis techniques and / or numerical form, particularly one or more parameters of frequency or Fourier analysis, integrals and / or derivatives of the time curve, averages and / or extrema, etc.

[0039] In one implementation, the load cells lifted by the gripper are classified based on the parameters of the time curves of the force parameters dependent on the load cells, and other determined actual values, which in one implementation are actual values ​​of kinematics determined by at least one sensor of the robot, such as axis position, axis velocity and / or axis acceleration, etc., and in another implementation are their time curves, using a machine learning model.

[0040] Thus, in one implementation, load cells can be classified, particularly accurately, quickly and / or reliably, by means of two or more of the aforementioned features, respectively, through a machine learning model.

[0041] In one implementation, the first process includes: transferring the lifted load unit by a robot, in one implementation transferring it from a storage area or storage container and / or transferring it to a (first) picking station, in an extended implementation delivering the lifted load unit to the (first) picking station, and in particular, it may be a combination of the above.

[0042] Additionally or alternatively, in one embodiment, the second process includes: completely or partially deactivating the lifting state of the gripper, particularly deactivating one or more of its lifting devices. In one embodiment, this deactivation does not occur at the (first) picking station; in another embodiment, this deactivation does not occur at any picking station.

[0043] Additionally or alternatively, in one embodiment, the third process includes: completely or partially deactivating the lifting state of the gripper, particularly partially deactivating it differently from the second process, particularly deactivating another set of its lifting devices. In one embodiment, this deactivation does not occur at the (first) picking station; in another embodiment, this deactivation does not occur at any picking station.

[0044] In one extended embodiment, the second and / or third processes include deactivating the lifting state (completely or partially) above the storage area. Thus, in one implementation, overlifted or incorrectly lifted objects or loads can be quickly and / or reliably (re-lifted) in subsequent cycles.

[0045] In one extended embodiment, the second process includes: picking different from the first process, in one implementation being sequential picking, particularly delivering a first portion of the lifted load unit at a (first) picking station, and delivering another portion of the lifted load unit at another picking station or at the same (first) picking station after reprocessing the delivered first portion, particularly after it is transported away.

[0046] Additionally or alternatively, in an extended embodiment, the third process includes: picking different from the first and / or second processes, in one embodiment being sequential picking, particularly delivering a first portion of the lifted load unit at a (first) picking station, and after reprocessing the delivered first portion, particularly after it has been transported away, delivering a second portion of the lifted load unit at another picking station or at the same (first) picking station, and after reprocessing the delivered second portion, particularly after it has been transported away, delivering another portion of the lifted load unit at another picking station or at the same (first) picking station. Similarly, in one embodiment, the second process may also include: picking different from the first process, particularly sequential picking; and the third process includes: deactivating the lifting state of the gripper above the receiving area, thereby in one embodiment (only) sequentially picking two lifted objects or loads, and more than two lifted objects or loads being delivered to the receiving area.

[0047] In one implementation, the second and / or third processes (respectively) include a placement movement (Ablagebewegung) of the lifted load unit using a robot, whereby the robot guides or continues to carry the load unit (“guided placement movement of the lifted load unit”), particularly in the storage area and / or before deactivation in the lifted state. Thus, the lifted object or load can be protected during deactivation in the lifted state. Similarly, the object or load can be lowered by deactivation in the lifted state, thereby saving process time in particular.

[0048] In one embodiment, deactivating the lifting state of the gripper in the second and / or third processes (respectively) includes deactivating at least one lifting device while at least one additional lifting device identified as bearing the load remains active. As previously described, in one embodiment, the categories differ in terms of the load lifted by the gripper relative to the gripper or its distribution on the gripper; therefore, in one embodiment, all lifting devices of the gripper are selectively deactivated except for the lifting device identified as bearing the load or except for a group of lifting devices identified as sharing the same load.

[0049] In particular, in (classified) multiple gripping, it is advantageous to selectively deactivate specific lifting devices, especially suction devices, so that the robot can continue to manipulate only the desired number or type of objects or loads.

[0050] In one implementation, the load units lifted by the gripper are classified based on the motion of the lifted load units detected by the robot, using a machine learning model. It has been shown that additionally considering the detected motion of the lifted load units when classifying using a machine learning model can be particularly advantageous, especially in improving the accuracy, reliability, and / or speed of classification.

[0051] In one implementation, the model has at least one artificial neural network, which in an extended embodiment is a network with 1D convolutions. This machine learning model is particularly advantageous for the classification described in this application, especially because of its learning characteristics and / or its accuracy, reliability, and / or speed.

[0052] In one implementation, specifically, the model performs machine learning by using the robot's gripper to lift one or more load cells, which are generally referred to herein as (multiple) learning load cells. In an extended embodiment, the category of at least one learning load cell is input manually or automatically, in one implementation by means of automatic recognition, for example, through image processing or image analysis, or pre-set by only partially activating the gripper, in another implementation by activating only one of its lifting mechanisms. Additionally or alternatively, in one implementation, the model can perform machine learning by using the robot's gripper to lift multiple different learning load cells.

[0053] Thus, in one implementation, machine learning can be greatly improved, individually or in particular, in the form of combinations of two or more of the aforementioned features.

[0054] According to one embodiment of the invention, a system for manipulating a payload unit using a robot is proposed, the system being specifically designed, in particular, by hardware and / or software techniques, especially programming techniques, to perform the methods described herein, and / or having:

[0055] - A device for activating the lifting state of the robot's gripper to lift the load;

[0056] - A device for determining, by means of at least one sensor of a robot, parameters of a time curve that depend on the force parameters of the load unit during the movement of the lifted load unit.

[0057] - A device for classifying load cells lifted by a gripper based on determined parameters using a machine learning model, particularly during the movement of the lifted load cells.

[0058] Or sorting may be performed above the storage area where the lifted load unit is already located, particularly above the storage area of ​​the storage station and / or above or within the storage container; and

[0059] - A means for performing a first process using a robot when the load unit is classified into a first category; and / or performing a second process using a robot when the load unit is classified into a second category.

[0060] In one embodiment, the system or its apparatus includes:

[0061] - A device for performing a third process using a robot when the load unit is classified into a third category; particularly for performing a fourth process using a robot when the load unit is classified into a fourth category; and / or

[0062] - An apparatus for enabling a model to learn by lifting at least one learning payload unit using a gripper of a robot; in particular, an apparatus for manually or automatically inputting the category of the at least one learning payload unit or for presetting the category of the at least one learning payload unit by partially activating only the gripper, in particular only one of its lifting devices; an apparatus for enabling a model to learn by lifting multiple different learning payload units using a gripper of a robot.

[0063] The apparatus of this invention can be implemented using hardware and / or software techniques, and in particular includes: a processing unit, preferably connected to a storage system and / or a bus system via data or signal connections, particularly a digital processing unit, particularly a microprocessor unit (CPU), graphics card (GPU), etc.; and / or one or more programs or program modules. The processing unit can be configured to: execute instructions implemented as a program stored in the storage system; detect input signals from the data bus and / or output output signals to the data bus. The storage system can have one or more, particularly different, storage media, particularly optical, magnetic, solid-state, and / or other non-volatile media. The program can be provided to embody or execute the methods described herein, enabling the processing unit to perform the steps of the method, and thereby, in particular, to operate or control a robot. In one embodiment, the computer program product can have, particularly, a non-volatile storage medium for storing the program, or a storage medium on which the program is stored, wherein execution of the program enables a system or controller, particularly a computer, to perform the methods described herein or one or more steps of the method.

[0064] In one implementation, one or more, particularly all, steps of the method, especially by means of the system or its apparatus, are performed fully or partially automatically.

[0065] In one implementation, the system includes a robot.

[0066] In one implementation, "above the storage area" means, in a manner conventional in the art, a position that is at least partially located inside the (outer boundary) of the storage area horizontally.

[0067] One embodiment of the invention is particularly advantageous for clamping unknown object types whose weight is not known in advance or used for this purpose. Attached Figure Description

[0068] Other advantages and features are given in the dependent claims and embodiments. For this purpose, some are illustrated schematically:

[0069] Figure 1 A system according to one embodiment of the present invention; and

[0070] Figure 2 This is a method according to one embodiment of the present invention. Detailed Implementation

[0071] Figure 1 A system for performing a method according to an embodiment of the present invention is shown.

[0072] The robot 10 has a multi-axis robotic arm 11 and a gripper 12 with three exemplary suction devices 12.1-12.3, and picks up objects or loads 21 from the storage container 20 and individually picks them into packing boxes 31 provided there at the picking station 30.

[0073] Therefore, robot controller 13 performs the following: Figure 2 The method illustrated in the figure:

[0074] In step S110, the robot 10 moves its gripper 12 into the storage container 20 and activates its lifting state or all of its suction devices 12.1-12.3. In this way, at least one object 21 can be grasped with a high probability (pneumatically).

[0075] In the second step S120, the robot 10 lifts the gripper 12 carrying the load unit to be lifted upwards. Here, the force-torque sensor 14 between the gripper 12 and the robot arm 11 continuously or at discrete time intervals detects the vertical force component depending on the lifted load unit. The results of the corresponding measurement signals from the sensor 14 form a time curve (parameters) dependent on the force parameters of the load unit. In a variant, additional force components and / or torque components may be detected, and / or additionally, in addition to the force-torque sensor 14, sensors on the axis or joint of the robot arm and / or one or more strain gauges on the gripper 12 may be used, and / or additionally, in addition to the time curve itself, parameters may be used, for example, frequency analysis or the average value of the time curve.

[0076] During this movement, while the lifted load unit is still within or above the receiving container 20, a machine learning model, in this embodiment a trained artificial neural network 13.1, classifies the load unit lifted by the gripper into four purely exemplary categories based on determined parameters: "lifting one object" (= first category), "lifting two objects" (= second category), "lifting more than two objects" (= third category), or "lifting no object" (= fourth category). In a variant, other and / or more or fewer categories may be specifically provided, for example, only {"lifting one object" (= first category) and " The categories are: {"lifting multiple objects" (= second category)}, or {"lifting one object" (= first category), "lifting two objects" (= second category), and "lifting more than two objects" or "no objects lifted" (= third category)}, or {"lifting one object" (= first category), "lifting multiple objects" (= second category), and "no objects lifted" (= third category)}. In one implementation, in step S120, the axis position of the robot may be additionally detected using sensing technology, wherein the artificial neural network 13.1 then classifies the load units lifted by the gripper based on these actual values.

[0077] If a load cell is assigned to the first category, i.e., it is determined that only one object is to be lifted based on the time curve (parameters) of the force parameters dependent on the load cell (S130: "1"), then the robot performs a transfer motion of the lifted load cell from the storage area to the picking station 30, and delivers the object or load cell to the provided packaging box 31 by deactivating all the suction devices 12.1-12.3 (S141).

[0078] If the load unit is assigned to the second category, i.e., it is determined that two objects are lifted based on the time curve (parameters) of the force parameters dependent on the load unit (S130: "2"), then the controller 13 deactivates all suction devices except one suction device 12.1-12.3 in step S142, while the gripper is still in or above the receiving container 20, so that the over-gripped objects fall back into the receiving container, and then the objects held by the still effective (activated) suction devices are picked into the provided packaging box 31 as in step S141.

[0079] Alternatively, in step S142, all suction devices 12.1-12.3 may be deactivated while the grippers remain in or above the receiving container 20, causing all gripped objects to fall back into the receiving container, and then a new attempt to grip only one object may be performed, similar to step S140.

[0080] Alternatively, in step S142, the provided box 31 may first be approached as in step S141, and only a portion of the suction devices above it may be deactivated, specifically the suction devices identified as carrying the same load, while the other suction devices remain active, so that only one object is picked into the provided box 31. Then, after the box is removed and the next box is provided, the other suction devices are deactivated so that the second object can be picked into the newly provided box.

[0081] If the load cell is classified into the third category, i.e., more than two objects are lifted based on the time curve (parameters) of the force parameters dependent on the load cell (S130: "3"), then it can be performed in a similar manner (step S143). As in the case of the second category, here in particular, the excessively clamped objects can be dropped into or above the storage container 20, or picked up sequentially, or all objects can be dropped and a new clamping attempt can be made.

[0082] If the load unit should be classified into the fourth category "No lifting object" (S130: "0"), then a new clamping attempt is made (S140).

[0083] To train the artificial neural network 13.1 for the above method or for a machine learning model, the corresponding object or load type is lifted in advance in step S10. Here, it is determined that only a single object is lifted, either based on the existence (and lifting) of only a single object, or by operator or image recognition. In this way, machine learning can be improved. Additionally or alternatively, in step S10, multiple different types of objects can also be lifted sequentially, preferably one at a time, using the gripper 12, and thereby or based on this, the artificial neural network 13.1 is trained.

[0084] Although exemplary implementations have been described in the preceding description, it should be noted that many variations may exist.

[0085] In particular, steps S140 and / or S143 can be omitted. In one embodiment, if the lifted load unit has been classified into category "single grip" (S130: "1"), then a single object is picked normally (S141 - first process); otherwise, if the lifted load unit is classified into category "multiple grip" (S130: "2"), then all suction devices 12.1-12.3 are deactivated while still in or above the receiving container 20 (S142 - second process). In an extended embodiment, additionally, in the case of being classified into category "misclassified grip", a new lifting attempt can be started directly (S130: "0" or S140 - third process). Alternatively, empty grippers or misclassified grippers can also be identified or handled in other ways, especially in the case of no corresponding category or classification.

[0086] Furthermore, it should be noted that these exemplary embodiments are merely examples and should not be construed as limiting the scope of protection, application, or construction. Rather, the foregoing description provides guidance for those skilled in the art to make modifications to at least one exemplary embodiment, wherein various changes, particularly concerning the function and arrangement of the components, can be made without departing from the scope of the invention, for example, as can be obtained according to the claims and their equivalent combinations of features.

[0087] Therefore, additionally or alternatively to the aforementioned quantity, in particular, the types of the promoted objects can be classified, and type-specific processes can be performed on this basis.

[0088] List of reference numerals

[0089] 10 robots

[0090] 11 Robotic Arms

[0091] 12 clamps

[0092] 12.1-12.3 Suction device (pneumatic lifting device)

[0093] 13 Controllers

[0094] 13.1 Artificial Neural Networks

[0095] 14 Force Sensors

[0096] 20 Storage Containers

[0097] 21 Objects / Loads

[0098] 30 Picking Stations

[0099] 31 Packaging boxes.

Claims

1. A method for manipulating a load cell (21) using a robot (10), comprising the following steps: - Activate the lifting state of the gripper (12) of the robot to lift the load; - During the movement of the lifted load unit, parameters of the time curve dependent on the force parameters of the load unit are determined by means of at least one sensor of the robot. - Based on the determined parameters, the load units lifted by the gripper are classified using a machine learning model; And includes at least one of the following steps: - If the payload unit is classified into a first category, then the robot performs a first process (S141); and / or - If the payload unit is classified into the second category, the robot is used to perform the second process (S142). Machine learning is performed by using the gripper of the robot to lift at least one learning payload unit (S10), and the robot is controlled to perform the first process or the second process based on the distribution of the payload relative to the gripper.

2. The method according to claim 1, characterized in that, The sorting is performed during the movement of the lifted load unit and / or above a storage area in which the load unit to be lifted is already located, wherein the storage area is above the storage area of ​​the storage station and / or above or in the storage container (20).

3. The method according to claim 1, characterized in that, Includes the following steps: If the load unit is classified into the third category, the robot is used to perform the third process (S143).

4. The method according to claim 3, characterized in that, Includes the following steps: If the load unit is classified into the fourth category, the robot is used to perform the fourth process (S140).

5. The method according to claim 1, characterized in that, The categories differ in the quantity, distribution, and / or type of loads lifted by the gripper.

6. The method according to claim 1, characterized in that, The gripper has at least two lifting devices, and / or the sensor is arranged on the gripper, between the gripper and the robot arm on which the gripper is fixed, or arranged on the robot arm, and / or the force parameter depends on at least one force component dependent on the load unit and / or at least one torque component dependent on the load unit.

7. The method according to claim 6, characterized in that, The lifting device is a suction device (12.1-12.3), and / or the force component is a vertical force component.

8. The method according to any one of claims 1-7, characterized in that, - The first process includes: transferring the lifted load unit by means of the robot; and / or - The second process includes: completely or partially deactivating the lifting state of the gripper, and / or completely or partially deactivating it above the storage area; and / or sequentially picking or repeating lifting attempts, different from the first process; and / or - The third process includes: completely or partially deactivating the lifting state of the gripper, and / or completely or partially deactivating it above the storage area; and / or sequential picking or repeated lifting attempts, different from the first and second processes.

9. The method according to claim 8, characterized in that, In the third process, unlike the second process, the lifting state of the gripper is deactivated; and / or the transfer movement exits from the receiving area and / or is transferred to the picking station (30); and / or the lifting state of the gripper is completely or partially deactivated by deactivating one or more of its lifting devices.

10. The method according to any one of claims 1-7, characterized in that, The second and / or third processes include: performing a guided placement movement of the lifted load unit by means of the robot; and / or partially deactivating the lifting state of the gripper includes: deactivating at least one lifting device while at least one additional lifting device identified as carrying the load remains active.

11. The method according to any one of claims 10, characterized in that, The guided placement of the lifted load unit by means of the robot is performed in the storage area and / or before the lifted state is deactivated.

12. The method according to any one of claims 1-7, characterized in that, The model has at least one artificial neural network (13.1).

13. The method according to any one of claims 1-7, characterized in that, The category of the at least one learning payload unit is input manually or automatically, or preset by only partially activating the gripper, and / or the model performs machine learning by leveraging the robot's gripper to lift multiple different learning payload units.

14. The method according to claim 13, characterized in that, The category of the at least one learning payload unit is preset by activating only one of its boosters.

15. A system for manipulating a payload unit using a robot (10), said system being designed to perform the method according to any one of claims 1-14, and / or comprising: - A means for activating the lifting state of the gripper (12) of the robot to lift the load; - A means for determining, by means of at least one sensor of the robot, parameters of a time curve dependent on the force parameters of the load unit during the movement of the lifted load unit. - A means for classifying load units lifted by the gripper based on determined parameters using a machine learning model; and - A means for performing a first process using the robot when the load unit is classified into a first category, and / or for performing a second process using the robot when the load unit is classified into a second category.

16. A computer program product comprising a program stored on a computer-readable medium, wherein, The execution of this program causes the system according to claim 15 to perform the method according to any one of claims 1-14.

Images