Method for precisely determining output torque and collaborative robot

Abstract

The invention relates to a method for precisely determining an output-side torque, in particular an output-side torque of an actuator gear mechanism of a joint of a collaborative robot, by means of artificial intelligence designed to output one or more output variables on the basis of an input variable, the artificial intelligence input variable includes a first angle specification corresponding to an input-side angular position and a second angle specification corresponding to an output-side angular position, and additionally, the output variable includes an output torque determined by the artificial intelligence.

Description

Technical Field

[0001] The present invention generally relates to the technical field of robots, autonomous mobile robots (AMRs) and in particular collaborative robots (cobots) as well as safety and precision engineering in workplaces and assembly systems. Background Art

[0002] In recent years, collaborative robotic systems have been used in more and more facilities. This has led to human workers increasingly sharing the same space with collaborative robotic systems. Such robots are designed to be particularly collaborative. For example, if an object or a worker is in the path, the movement should first be stopped to prevent damage and injury. Stress wave gear mechanisms (harmonic drives / harmonic gearboxes) are often used as gear mechanisms. However, the choice of the gear mechanism is mainly a design decision.

[0003] In order to reliably and precisely control collaborative robots and their articulated motors – which makes them particularly suitable for use in delicate environments and in the presence of human workers – it is necessary to measure and / or know the outgoing / output-side torque as accurately as possible.

[0004] Torque sensors are usually used for this purpose. However, torque sensors are usually expensive, often very heavy, interfere with the robot's flexibility during use and increase energy consumption. In addition, the torque sensor only provides one measurement value, which is no longer checked further. This leads to uncontrolled and dangerous situations, especially if the torque sensor experiences technical defects or emits damaged signals.

[0005] Patent specification EP 2 231 369 B1 discloses a robot and a method for monitoring torques on such a robot. A robot is described which has at least two joints and parts which can be moved relative to each other by means of at least one joint, wherein at least one torque detection sensor is arranged on at least one movable part, and wherein two sensors and a redundant evaluation unit for redundant evaluation are provided for redundant detection of the torque, and wherein a device is provided for shutting down the robot or triggering a safety state in the event that the measured values ​​of the same torque recorded by at least two sensors deviate from each other beyond a specified tolerance range, wherein two sensors in the form of a full bridge with strain gauges are arranged on the gear mechanism of the robot so that the two sensors detect the same torque and are connected in a transmitting unit by means of two computer units in the form of integrated circuits of different designs with microcontrollers, in which a first check of the measured torque value is performed. However, the solution described here has the disadvantage that two sensors are required to detect the torque, which is expensive and also has a negative impact on the dynamics and weight of the robot.

[0006] From another prior art, it is known to carry out an indirect determination via the motor current flow. This solution is inexpensive but highly imprecise and therefore no longer meets today's quality standards for high-performance collaborative robot systems.

[0007] Therefore, the object of the present invention is to further improve the collaborative robot system so that it works precisely and safely, especially in a work environment shared with humans. The aim is to ensure high precision and flexibility of the robot and to minimize costs as much as possible without affecting the quality of the product and the results achieved. Summary of the invention

[0008] The invention provides a method according to claim 1. Therefore, a method is provided for accurately determining an output-side torque, in particular an output-side torque of an actuator gear mechanism of a joint of a collaborative robot, by means of an artificial intelligence, which is designed to output one or more output variables based on input variables, wherein the input variables of the artificial intelligence include: a first angle specification corresponding to an input-side angular position; a second angle specification corresponding to an output-side angular position, and wherein, in addition, the output variable includes an output torque determined by the artificial intelligence.

[0009] Artificial intelligence determines the torque output particularly accurately. The accuracy is much higher than using conventional methods. Therefore, this precise measurement enables collaborative robot control and motor control to work and act more precisely.

[0010] The determination does not rely on a direct measurement of the torque, for example by means of a torque sensor which is usually expensive and heavy. If such a torque sensor is still used, the invention has the additional advantage that the redundant determination of the torque further increases the accuracy. The direct output of the torque sensor is an excellent way to train an AI. In addition, the solution according to the invention makes it possible to determine the torque independently of the torque sensor, which is particularly advantageous, for example, in the event of a malfunction of the latter. This avoids dead times and faulty and / or uncontrolled (and therefore often dangerous) behaviors of the collaborative robot.

[0011] The invention also provides a robot, in particular a collaborative robot according to claim 4 (or a structural unit for a collaborative robot). Therefore, a collaborative robot or a collaborative robot component is provided, which comprises: a robot device, in particular a robot arm and / or an assembly system, wherein the robot device has one or more joints that can be moved by a motor, in particular a servomotor; at least one gear mechanism, in particular a reduction gear mechanism; one or more input-side angular position sensors, in particular an input-side encoder, for determining the input-side angular position on the input side relative to the gear mechanism; one or more output-side angular position sensors, in particular an output-side encoder, for determining the output-side angular position on the output side relative to the gear mechanism; an electronic control system, which is designed to take into account the value of the gear mechanism-side output torque during operation of the collaborative robot, which value is provided by means of artificial intelligence based on data including the input-side angular position of the input-side angular position sensor and the output-side angular position of the corresponding output-side angular position sensor. It should be emphasized that the concepts disclosed herein are not limited to collaborative robots, but can be used for any type of robot.

[0012] The calculations can be performed locally and / or non-locally, in particular also distributedly, for example in the context of cloud computing. The collaborative robot according to the invention - usually local - benefits from the invention by improved and more precise behavior and improved working safety.

[0013] In addition, the invention provides an artificial intelligence system and a suitable training method. Therefore, according to the invention, an artificial intelligence for accurately determining an output-side torque, in particular an output-side torque of an actuator gear mechanism of a joint of a collaborative robot is provided, the artificial intelligence being designed to output one or more output variables based on input variables, wherein the input variables of the artificial intelligence include: a first angle specification corresponding to an input-side angular position; a second angle specification corresponding to an output-side angular position, and wherein, in addition, the output variable includes an output torque determined by the artificial intelligence.

[0014] The AI ​​can be trained continuously during operation, which ensures a very high level of training and also takes into account changes that occur over time (e.g., changes in environmental conditions or wear and tear).

[0015] Preferably, the training takes into account direct torque measurements as target values, such as measurements from a torque sensor.

[0016] Unsupervised learning has proven to be particularly suitable. In particular, good results have been achieved using multilayer perceptrons, in particular recursive multilayer perceptrons, as neural networks. The recursive nature ensures sufficient feedback.

[0017] The AI ​​can be arranged in some kind of control loop and / or feedback loop. Thus, unsupervised learning can be effectively supplemented by torque measurement as a target value.

[0018] The invention greatly increases the dynamic availability of collaborative robots under changing conditions. For example, high precision after replacing a spare part on a collaborative robot (or a refurbished collaborative robot, for example) is ensured by the high flexibility of the AI.

[0019] For example, the AI ​​may consider angular velocity, speed, acceleration, and temperature as additional features (input variables) to properly account for these effects. However, the present invention is by no means limited thereto.

[0020] With the help of AI, the present invention enables effects such as friction, wear, material fatigue, reversal allowance, reaction, other nonlinear effects, manufacturing tolerances and / or individual differences in manufacturing to be efficiently evaluated and taken into account in collaborative robot control. These effects are often difficult to simulate and are rarely even available for direct and adequate mathematical description. Artificial intelligence will help overcome these problems.

[0021] The invention is by no means limited to the effects mentioned here (in the sense of an exhaustive list). On the contrary, a particular advantage of using artificial intelligence in this case is that individual effects do not have to be modeled manually, but all effects that occur - and all effects with appropriate magnitudes - are reproduced simultaneously and directly by the AI.

[0022] The present invention makes a lasting contribution to the advancement of the technology in the field of collaborative robotics through accurate and, if necessary, redundant determination of output torque. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The preferred embodiments of the present disclosure are described below with reference to the following drawings:

[0024] Figure 1 : shows a joint of a robot having an input side angular position encoder and an output side angular position encoder.

[0025] Figure 2 : shows a joint of a robot having an output-side torque sensor and an output-side angular position encoder.

[0026] Figure 3 : shows a schematic diagram of a drive train for a robot joint with an output-side torque sensor and angular position encoders on both sides according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] Figure 1 A robotic joint 100 is shown having an input-side angular position encoder and an output-side angular position encoder.

[0028] In the present case, the input-side angle encoder or angle position encoder 1 is provided by an absolute encoder 1 .

[0029] Furthermore, the output-side angle encoder or angle position encoder 2 is provided by an incremental rotary encoder 2 .

[0030] Due to the dual assembly with angular position encoders, the collaborative robot can be used in the present invention.

[0031] Figure 2 A robotic joint 100 is shown with an output-side torque sensor 210 and an output-side angular position encoder 202 .

[0032] A stress wave gear mechanism, a wave gear mechanism or a sliding wedge gear mechanism 203 is used to reduce the transmission train of the robot joint.

[0033] Figure 3 A schematic diagram of a drive train for a robotic joint with an output-side torque sensor and angular position encoders on both sides according to an embodiment of the present invention is shown.

[0034] The electric motor 330 drives the shaft 320 into rotation. A voltage converter 340 provides the required energy, while a motor electronic control system 350 controls the motor 330. The shaft 320 terminates on the output side in a reducer 360. For example, a gear mechanism provides a gear reduction.

[0035] A rotary encoder or an angular position sensor 301 (input side angular position sensor 301) is arranged on the input side. Another rotary encoder or an angular position sensor 302 (output side angular position sensor 302) is arranged on the output side. According to the present invention, the torque output on the output side of the joint is determined based on the measured values ​​of the angular position sensors 301, 302 (and / or the difference between the measured values) by means of artificial intelligence.

[0036] It is also possible to determine the torque directly via a torque sensor on the output side. In particular, a redundant torque can be determined. This is particularly precise. Failure of the torque sensor is also prevented. In addition, the torque measured directly by the torque sensor can be used to train the AI ​​in a prominent way.

[0037] For example, in continuous operation, torque sensors are used to further train the AI.

[0038] Later, for example, the torque sensor can be removed, for example temporarily. But the torque sensor can also remain there. If the torque sensor fails at a certain point, the training stops and the excellent and extensively trained AI alone takes over the torque determination (based on the output values ​​of the two encoders).

[0039] The invention has the further advantage that the AI, in particular the AI ​​through continuous training, dynamically takes into account environmental effects (eg temperature) as well as signs of wear, material fatigue, manufacturing tolerances of components etc.

[0040] Thus, the present invention facilitates improved control of a collaborative robot as the collaborative robot may be operated with a more accurately determined torque output.

[0041] The technical effect of the invention thus occurs when the torque determination is performed redundantly (by rotary encoder + AI and directly by torque sensor) and when the torque sensor fails. In the latter case, downtimes and faulty behavior of the collaborative robot are particularly avoided.

[0042] This brings economic benefits.

[0043] The embodiments schematically described here can be further developed by numerous details of the invention, in particular by details already described above.

[0044] Although some aspects have been described in the context of an apparatus, it is apparent that these aspects also constitute a description of a corresponding method, wherein a block or apparatus corresponds to a method step or a function of a method step. Similarly, aspects described in the context of a method step also represent a description of a characteristic of a corresponding block or element or a corresponding apparatus.

[0045] Exemplary embodiments of the present invention can be implemented in a computer system. The computer system can be a local computer device with one or more processors and one or more storage devices (e.g., an embedded control device in a personal computer, a laptop computer, a tablet computer, a cellular phone or a robot), or can be a distributed computer system (e.g., a cloud computing system with one or more processors or one or more storage devices distributed in various locations, such as distributed in a local client and / or one or more remote server farms and / or a data center). The computer system may include any circuit or combination of circuits. In an exemplary embodiment, the computer system may include one or more processors, and the one or more processors may be of any type. As used herein, a processor may mean any type of computing circuit, such as but not limited to a microprocessor, a microcontroller, a complex instruction set computer (CISC), a reduced instruction set computer (RISC), a very long instruction word (VLIW) microprocessor, a graphics processor, a digital signal processor (DSP), a multi-core processor, a field programmable gate array (FPGA), or any other type of processor or processing circuit. Other types of circuitry that a computer system may include may be custom circuits, application specific integrated circuits (ASICs), and the like, such as one or more circuits (e.g., communications circuits) for use with wireless devices such as cellular phones, tablet computers, laptop computers, two-way radios, and similar electronic systems. A computer system may include one or more storage devices, which may include one or more storage elements suitable for a particular application, such as a main memory in the form of a random access memory (RAM), one or more hard disks, and / or one or more drives for handling removable media such as CDs, flash memory cards, DVDs, and the like. A computer system may also include a display device, one or more speakers, and a keyboard and / or controller, which may include a mouse, trackball, touch screen, voice recognition device, or any other device that allows a system user to input information into and receive information from a computer system.

[0046] Some or all of the method steps may be performed by (or using) hardware devices such as processors, microprocessors, programmable computers or electronic circuits. In some exemplary embodiments, one or more of the key method steps may be performed by such devices.

[0047] Depending on the specific implementation requirements, exemplary embodiments of the present invention may be implemented using hardware or software. The implementation may be implemented using a non-volatile storage medium, such as a digital storage medium, such as a floppy disk, DVD, Blu-ray disc, CD, ROM, PROM and EPROM, EEPROM or flash memory, on which electronically readable control signals are stored, which interact (or can interact) with a programmable computer system to perform the corresponding method. Therefore, the digital storage medium may be computer readable.

[0048] Some exemplary embodiments according to the invention comprise a data carrier having electronically readable control signals, which are capable of interacting with a programmable computer system, such that the system performs one of the methods described herein.

[0049] Generally, the exemplary embodiments of the present invention can be implemented as a computer program product with a program code, wherein when the computer program product is running on a computer, the program code effectively performs one of the methods described.For example, the program code can be stored on a machine-readable medium.

[0050] Further exemplary embodiments comprise the computer program for performing one of the methods described herein, stored on a machine-readable medium.

[0051] In other words, exemplary embodiments of the present invention are, therefore, a computer program having a program code for performing one of the methods described herein, when the computer program runs on a computer.

[0052] A further exemplary embodiment of the present invention is therefore a storage medium (or a data carrier or a computer-readable medium), which storage medium comprises: a computer program stored thereon for performing one of the methods described herein when the computer program is executed by a processor. The data carrier, the digital storage medium or the recorded medium is typically tangible and / or non-transitory. A further exemplary embodiment of the present invention is an apparatus as described herein, comprising a processor and a storage medium.

[0053] Therefore, a further exemplary embodiment of the present invention is a data stream or a sequence of signals representing the computer program for performing one of the methods described herein.For example, the data stream or the sequence of signals may be configured to be transmitted via a data communication connection, for example via the Internet.

[0054] Another exemplary embodiment comprises a processing means, for example a computer or a programmable logic device, configured to or adapted to perform any of the methods described herein.

[0055] A further exemplary embodiment comprises a computer on which the computer program for performing one of the methods described herein is installed.

[0056] Another exemplary embodiment according to the invention comprises an apparatus or system configured to send (e.g., electronically or optically) a computer program for performing one of the methods described herein to a receiver. The receiver may be, for example, a computer, a mobile device, a storage device, etc. The apparatus or system may, for example, comprise a file server for sending the computer program to the receiver.

[0057] In some exemplary embodiments, a programmable logic device (e.g., a field programmable gate array, FPGA) may be used to perform some or all of the functions of the methods described herein. In some exemplary embodiments, a field programmable gate array may collaborate with a microprocessor to perform any of the methods described herein. Typically, the method is preferably performed by each hardware device.

[0058] Exemplary embodiments may be based on artificial intelligence, in particular the use of machine learning models or machine learning algorithms. Machine learning may refer to algorithms and statistical models that a computer system can use to perform specific tasks without explicit instructions, rather than relying on models and reasoning. For example, in machine learning, instead of rule-based data transformations, data transformations that can be derived from the analysis of historical and / or training data can be used. For example, a machine learning model or a machine learning algorithm may be used to analyze the content of an image or other data such as sensor data or measurement data. In order for a machine learning model to analyze the content of an image, a training image may be used as input and training content information may be used as output to train the machine learning model. By training a machine learning model with a large number of training images and / or training sequences (e.g., words or sentences) and associated training content information (e.g., labels or annotations), the machine learning model "learns" to recognize the content of the image, so that the machine learning model can be used to recognize the content of the image not included in the training data. The same principle can also be used for other types of sensor data: by training a machine learning model using training sensor data and a desired output, the machine learning model "learns" the conversion between sensor data and output, which can be used to provide output based on non-training sensor data provided to the machine learning model. The provided data (e.g., sensor data, metadata, and / or image data) may be pre-processed to obtain a feature vector, which is used as input to a machine learning model.

[0059] The machine learning model can be trained using training input data. The above example uses a training method known as supervised learning. In supervised learning, a machine learning model is trained using multiple training sample values, wherein each sample value may include multiple input data values ​​and multiple expected output values, i.e., each training sample value is associated with an expected output value. By specifying both the training sample values ​​and the expected output values, the machine learning model "learns" which output value to provide based on input sample values ​​similar to the sample values ​​provided as part of the training. In addition to supervised learning, semi-supervised learning can also be used. In semi-supervised learning, some of the training sample values ​​lack the expected output value. Supervised learning can be based on supervised learning algorithms (e.g., classification algorithms, regression algorithms, or similarity learning algorithms). Classification algorithms can be used when the output is limited to a limited set of values ​​(classification variables), i.e., when the input is classified as one of a limited set of values. Regression algorithms can be used when the output shows any numerical value (within a certain range). Similarity learning algorithms can be similar to both classification algorithms and regression algorithms, but are based on learning from examples using similarity functions that measure the similarity or correlation between two objects. In addition to supervised learning or semi-supervised learning, unsupervised learning can also be used to train machine learning models. In unsupervised learning, (only) input data can be provided, and an unsupervised learning algorithm can be used to find structure in the input data (e.g., by grouping or clustering the input data to find commonalities in the data). Clustering is the distribution of input data including multiple input values ​​into subsets (clusters) so that input values ​​within the same cluster are similar according to one or more (predefined) similarity criteria, while these input values ​​are dissimilar to input values ​​included in other clusters.

[0060] Reinforcement learning is a third group of machine learning algorithms. In other words, reinforcement learning can be used to train machine learning models. In reinforcement learning, one or more software actors (i.e., "software agents") are trained to perform actions in an environment. Rewards are calculated based on the actions performed. Reinforcement learning is based on training one or more software agents to select actions such that the cumulative reward increases, thereby causing the software agent to become better at the task given to it (as evidenced by increasing rewards).

[0061] In addition, some techniques can be applied to some of the machine learning algorithms in the machine learning algorithm. For example, feature learning can be used. In other words, the machine learning model can be trained at least in part using feature learning, and / or the machine learning algorithm can include a feature learning component. Feature learning algorithms, known as representation learning algorithms, can retain the information in their input but transform the information in a way that makes it useful, usually as a preprocessing stage before performing classification or prediction. Feature learning can be based on principal component analysis or cluster analysis, for example.

[0062] In some examples, anomaly detection (i.e., outlier detection) can be used, which is intended to provide identification of input values ​​that are suspicious because they are significantly different from the majority of input data and training data. In other words, the machine learning model can be trained at least in part using anomaly detection, and / or the machine learning algorithm can include an anomaly detection component.

[0063] In some examples, a machine learning algorithm may use a decision tree as a prediction model. In other words, a machine learning model may be based on a decision tree. In a decision tree, observations about an object (e.g., a set of input values) may be represented by branches of the decision tree, and output values ​​corresponding to the object may be represented by leaves of the decision tree. A decision tree may support both discrete and continuous values ​​as output values. If discrete values ​​are used, the decision tree may be referred to as a classification tree; if continuous values ​​are used, the decision tree may be referred to as a regression tree.

[0064] Association rules are another technique that can be used in machine learning algorithms. In other words, a machine learning model can be based on one or more association rules. Association rules are created by identifying relationships between variables in a large data set. A machine learning algorithm can identify and / or use one or more ratio rules that represent knowledge derived from the data. These rules can be used, for example, to store, manipulate, or apply knowledge.

[0065] Machine learning algorithms are typically based on machine learning models. In other words, the term "machine learning algorithm" may refer to a set of instructions that can be used to create, train, or use a machine learning model. The term "machine learning model" may refer to a data structure and / or rule set that represents learned knowledge (e.g., based on training performed by a machine learning algorithm). In an exemplary embodiment, using a machine learning algorithm may imply using an underlying machine learning model (or multiple underlying machine learning models). Using a machine learning model may imply that the machine learning model and / or the data structures / rule sets that make up the machine learning model are trained by a machine learning algorithm.

[0066] For example, a machine learning model can be an artificial neural network (ANN). ANN is a system inspired by biological neural networks, such as those found in the retina or brain. ANN includes multiple interconnected nodes and multiple connections between nodes - so-called edges. There are usually three types of nodes: input nodes that receive input values, hidden nodes that are (only) connected to other nodes, and output nodes that provide output values. Each node can represent an artificial neuron. Each edge can send information from one node to another node. The output of a node can be defined as a (non-linear) function of the input (e.g., the sum of its inputs). The input of a node can be used in a function based on the "weight" of the edge or the node that provides the input. As part of the learning process, the weights of the nodes and / or edges can be adjusted. In other words, training an artificial neural network can include adjusting the weights of the nodes and / or edges of the artificial neural network, that is, to achieve the desired output for a specific input.

[0067] Alternatively, the machine learning model can be a support vector machine, a random forest model or a gradient enhancement model. A support vector machine (i.e., a support vector network) is a supervised learning model with an associated learning algorithm that can be used to analyze data (e.g., in classification or regression analysis). A support vector machine can be trained by providing an input with multiple training input values ​​belonging to one of two categories. A support vector machine can be trained so that new input values ​​are assigned to one of two categories. Alternatively, the machine learning model can be a Bayesian network, which is a probabilistic directed acyclic graph model. A Bayesian network can use a directed acyclic graph to represent a set of random variables and their conditional dependencies. Alternatively, a machine learning model or its training can be based on a genetic algorithm, which is a search algorithm and heuristic technique that simulates a natural selection process.

[0068] Reference numerals list

[0069] 1 Angle position encoder / absolute encoder (input side)

[0070] 2 Angular position encoder / incremental rotary encoder (output side)

[0071] 3 Gear mechanism

[0072] 20 Drive shaft

[0073] 100 Robot joints

[0074] 202 Angle position encoder (output side)

[0075] 203 gear mechanism

[0076] 210 torque sensor (output side)

[0077] 301 Angle position encoder (input side)

[0078] 302 Angle position encoder (output side)

[0079] 310 torque sensor (output side)

[0080] 320 drive shaft

[0081] 330 Motor / Electric Motor

[0082] 340 Voltage Converter / Voltage Source

[0083] 350 Electronic control systems / units for motors

[0084] 360 reducer

Claims

1. A method for accurately determining an output-side torque, in particular an output-side torque of an actuator gear mechanism (203) of a joint (100) of a collaborative robot, by means of artificial intelligence, in particular a machine learning model or a machine learning algorithm, The artificial intelligence is designed to output one or more output variables based on input variables, in, The input variables of the artificial intelligence include: - a first angular specification corresponding to the angular position of the input side; - a second angular specification corresponding to the angular position on the output side, And wherein, additionally, the output variable includes an output torque determined by the artificial intelligence.

2. The method according to claim 1, wherein: The first angle description is provided by an angular position sensor (301), in particular an encoder, in particular an absolute encoder, on an input side relative to the gear mechanism, and the second angle description is provided by an angular position sensor (302), in particular an encoder, in particular an incremental rotary encoder, on an output side relative to the gear mechanism.

3. The method according to claim 1 or 2, wherein: The gear mechanism (203) comprises a stress wave gear mechanism, a wave gear mechanism and / or a sliding wedge gear mechanism.

4. A robot, in particular a collaborative robot or a collaborative robot component, comprising: - a robotic device, in particular a robot arm and / or an assembly system, The robot device has one or more joints (100) that can be moved by a motor (330), in particular a servo motor, - at least one gear mechanism (203), in particular a reduction gear mechanism, one or more input-side angular position sensors (301), in particular input-side encoders, for determining the input-side angular position on the input side relative to the gear mechanism, one or more output-side angular position sensors (302), in particular output-side encoders, for determining the output-side angular position relative to the output side of the gear mechanism, - an electronic control system designed to take into account during operation of the collaborative robot a value of the gear mechanism-side output torque, said value being provided by means of artificial intelligence, in particular a machine learning model or a machine learning algorithm, based on data comprising: ○The input side angular position of the input side angular position sensor (301); and ○The corresponding output side angular position of the output side angular position sensor (302).

5. The collaborative robot or collaborative robot component according to claim 4, further comprising a torque sensor (310), in particular an output-side torque sensor (310), wherein: The output torque provided by means of artificial intelligence, in particular a machine learning model or a machine learning algorithm, and the torque detected by the torque sensor complement each other, Among them, in particular, Using redundancy in the determination of the value of torque to improve the accuracy of said value; and / or The output torque provided by means of artificial intelligence is used in the event that the torque sensor (310) fails or provides an erroneous and / or damaged signal.

6. The collaborative robot or collaborative robot component according to claim 4 or 5 comprises at least one stress wave gear mechanism, a wave gear mechanism and / or a sliding wedge gear mechanism.

7. A method for operating a collaborative robot according to one of claims 4 to 6, comprising the step of taking into account an outgoing torque determined by artificial intelligence based on a method according to one of claims 1 to 3.

8. A computer, a computer system or a computer network configured to perform locally or non-locally the method according to one of claims 1 to 3 or claim 7.

9. An artificial intelligence, in particular a machine learning model or a machine learning algorithm for accurately determining an output-side torque, in particular an output-side torque of an actuator gear mechanism (203) of a joint (100) of a collaborative robot, The artificial intelligence is designed to output one or more output variables based on input variables, in, The input variables of the artificial intelligence include: - a first angular specification corresponding to the angular position of the input side; - a second angular specification corresponding to the angular position on the output side, And wherein, additionally, the output variable includes an output torque determined by the artificial intelligence.

10. The artificial intelligence according to claim 9, which is designed for unsupervised learning, in particular unsupervised learning using a recursive multilayer perceptron.

11. The artificial intelligence according to claim 9 or 10, wherein: The input variables also include: angular velocity, speed, acceleration and / or temperature.

12. A method for training an artificial intelligence according to one of claims 9 to 11 for accurately determining the output torque using the following data set as input data, in particular by unsupervised learning: - a data set comprising data relating to the input-side angular position and the output-side angular position of the gear mechanism (203).

13. The method according to claim 12, wherein: A measured value of the output-side torque, in particular measured by a torque sensor (310), is also used as a target value, in particular as part of a feedback control loop in which the artificial intelligence is arranged.

14. The method according to claim 13, wherein: The training is performed continuously, but is stopped and / or interrupted if it is determined that the value of the output torque measured by means of the torque sensor (310) is impaired or the corresponding signal is interrupted.

15. A computer program comprising instructions which, when the program is executed by a computer, cause the computer to perform the method according to one of claims 1 to 4, claim 7, or claims 12 to 14.

16. Use of a collaborative robot according to one of claims 4 to 6 and / or a computer program according to claim 14 for determining an output-side torque taking into account and for the purpose of taking into account one or more of the following effects: Friction, wear, material fatigue, reversal allowance, reaction, other non-linear effects, manufacturing tolerances and / or individual differences in manufacturing, and / or for determining the output-side torque for the purpose of modeling the dependency of environmental variables and state variables by means of the artificial intelligence, wherein the environmental variables and state variables are in particular one or more of the following: angular velocity, speed, acceleration, temperature.

Images