DE102020114738A1 - Method, system and computer program for operating one or more robots, a robot system and / or a robot swarm - Google Patents

Abstract

A method, a system and a computer program are proposed for operating one or more robots, a robot system and / or a robot swarm, the method comprising: inducing or stimulating (S1), a first behavior or a first change in a behavior of one or more first robot (10, 11, 12), - determining (S2) one or more first parameters characterizing a first effect which occurred on the induced or stimulated behavior or the change in the behavior of the one or more first robots (10, 11, 12) - Training (S3) at least one neural network (30) to operate at least one second robot (20, 21, 21) as a function of the determined first parameters.

Description

The invention relates to a method, a system and a computer program for operating one or more robots, a robot system and / or a robot swarm. In particular, the invention relates to a method for operating a robot system comprising a multiplicity of robots or a robot swarm.

Digital communication between different devices (with or without the participation of a human) is known from the prior art. With increasing complexity and diverse combinations of input data and influencing variables of a modern or future robot, the number of conceivable expected effects (even within the functionality of a robot alone but especially in relation to a robot system or robot swarm) can also increase. Even a sensible exchange of information between a large number of robots can possibly lead to undesirable or unforeseeable consequences.

It is an object of this invention to show a method, system and computer program for operating one or more robots, a robot system and / or a robot swarm, which lead to an improved functionality and / or to an improved behavior of the one or more robots, a robot swarm or . of a robot system. In particular, it is the object of the invention to improve the behavior of at least one robot even if this is not related to a specific context, specific parameter, specific object, etc. Furthermore, it is in particular also an object of the invention to enable a transfer of specific, learnable skills from a number of the robots to at least one further robot.

The object is achieved by the features of the independent patent claims. Advantageous embodiments are described in the dependent claims. It is pointed out that additional features of a patent claim dependent on an independent patent claim without the features of the independent patent claim or only in combination with a subset of the features of the independent patent claim can form a separate invention independent of the combination of all features of the independent patent claim, which can be made the subject of an independent claim, a divisional application or a subsequent application. This applies equally to the technical teachings described in the description, which can form an invention that is independent of the features of the independent patent claims.

The following statements on the method according to the invention also apply in a corresponding manner to the system according to the invention and the corresponding computer program. Further features and advantages of the invention emerge from the following description of exemplary embodiments of the invention, with reference to the figures which show details according to the invention, e.g. system components or method steps, and from the claims. The individual features can be implemented individually or collectively in any combination in a variant of the invention.

According to a first aspect, the invention comprises a method for operating one or more robots, a robot system and / or a robot swarm.

The method comprises: causing or stimulating a first behavior or a first change in a behavior of one or more first robots; a determination of one or more first parameters characterizing a first effect which occurred on the induced or stimulated behavior or the change in the behavior of the one or more first robots; and, in particular thereupon: training of at least one neural network to operate at least one second robot as a function of the determined first parameters.

The neural network is preferably trained by a (large) multiplicity of the first parameters. Furthermore, the neural network can (alternatively or additionally) be trained depending on various parameters explicitly described in this document, as well as on the parameters of the features and / or processes described in this document. Training also includes learning, in particular various forms of machine learning.

The one or more of the first robots are stimulated in order to determine the parameters for training the neural network. For example, the at least one induced, stimulated first behavior or the first change in the behavior of the one or more first robots can in itself (almost) not be a functional expedient action for the immediate achievement of a goal or in itself an at least partially counterproductive action. The term “neural network” is preferred to be understood as the DNN, the digital neural network. Alternatively or additionally, various variants of an artificial intelligence, a machine, in particular non-rule-based Learning can be subsumed under the term “neural network” in the context of this document.

In a simplified case, a specific first behavior can be induced by means of a specific behavior parameter and / or behavior pattern. A first behavior can in particular be or include a certain behavior, in particular a behavior that is at least partially abstracted from certain parameters, in particular a behavior pattern. This can also be carried out if a current situation, intention, planning of the first robot is not known or only partially known.

A first change in the behavior (change in behavior) of a first robot can preferably be initiated. This can be a certain first change in behavior, in particular a measure of a difference (difference, factor, proportion, etc., e.g. according to a predefined condition, scale or formula). In this case, a first change in the behavior of the first robot corresponding to a specific function, a temporal and / or spatial gradient, a specific tendency can be initiated. This can also be carried out if the current behavior of the first robot is not known, or only partially, to a small extent.

For example, it does not have to be known what the (previous, current and / or planned) behavior of the robot was, e.g. in an unknown environment and / or towards an unknown object (e.g. before the behavior was changed). It does not have to be known whether the (previous, current or planned) behavior of the first robot is correct, appropriate, appropriate or helpful for its goal or mission.

As an alternative or in addition, a first change in behavior that does not correspond or does not have to correspond to specific values, parameters, etc., can be initiated.

The behavior of the (first) robot can be a behavior pattern or can be understood as a behavior pattern and / or changed. In particular, the behavior, in particular a behavior pattern, of the at least one robot is at least partially related behavior with a concrete context, concrete parameters of its environment, concrete object, in particular the concrete object, in particular an object in connection with its mission, goal, task .

For example, the first behavior and / or a first change in behavior can be a specific pattern, in particular a pattern of one or more actions (for example a pattern that results or forms several parameters of one or more actions), one or more behavior parameters, a behavior pattern and / or a change in a parameter of a pattern. For example, an at least partially new, a specific (new) pattern or a parameter of the pattern and / or at least partially randomly selected pattern of the behavior, in particular a parameter of the behavior pattern, is initiated.

In particular, the term “behavior” can also be understood to mean a mode of action, interaction and / or behavior. In particular, the behavior can be a behavior pattern or be characterized by a specific pattern. In particular, the behavior of one or more specific or specific actions and / or interactions of the robot and / or of a specific or specific context, a specific or specific situation and / or goal is independent, not unambiguously dependent, preferably abstracted from a specific or specific one . As an exception, a behavior can be understood as a sequence and / or pattern of actuations of an actuator of the robot, actions (actions of the robot), action parameters and / or interaction parameters of the robot.

The method is also advantageous, for example, because it allows the behavior of the robots to be changed, e.g. improved, even in a situation that is unknown (for them or for the robot operator), e.g. with a specific goal.

The one or more (first and / or second) robots can be set up robots for providing a service, exploring or colonizing an environment, for producing, assembling or building an object or product, for looking after, growing agricultural products. A robot can be an industrial robot, an agricultural robot, a medical robot (e.g. to care for people or animals, to carry out operations), a construction robot, a service robot, an exploration robot, a reconnaissance robot, a colonization robot, a robot for strategic or defensive applications being. Analogously to this, the method and a corresponding robot system and / or robot swarm can be determined for such purposes.

The first and / or second robot can be a robot that is designed or intended for a (specific) action with respect to or interaction with its environment and / or an object. In this sense, a movement of the Robot (essentially as a whole) and / or a (targeted, regulated) movement of one or more individual parts of the robot (e.g. a robot arm, robot leg, manipulator) in relation to the other parts and / or to its environment or to certain objects being. A robot function and / or a task or mission of the robot can include influencing one or more objects to be influenced. Such a robot functionality and / or mission can include, for example, an investigation (for example in the context of research), modification, processing, assembly, destruction of one or more objects to be influenced or processed.

The one or more of the first robots and / or the second robots can be part of a robot system and / or a robot swarm. The first and the second robot can belong to the same robot system or robot swarm or to different robot systems or robot swarms, e.g., to robot systems or robot swarms that can be operated in parallel with one another. The one or more first robots and one or more second robots can be robots of different types, e.g. types and / or with different tasks, missions, areas of application. As an exception, the one or more first robots and / or the second robots, the robot system can or can be understood as (different) parts of a production plant and / or a manufacturing facility.

The robot swarm according to the invention is to be understood in particular as a (fixed or variable) number of robots that each act at least partially, in particular predominantly independently and / or are designed for communication, cooperation or mutual influencing or influencing. A distinction can be made between the mission, tasks, interests, security, possibly a technically realized ego of a robot and those of the swarm. A robot swarm can be a robot swarm for the exploration or colonization of certain areas of application, terrain, unknown places, e.g. planets. A swarm of robots can nevertheless be a production plant, or a task for the production, construction of a building, collecting, cultivation, cultivation of an object, e.g. an industrial product, a consumer product, a commodity, an agricultural product.

A robot system and / or robot swarm can be characterized by an at least partially decentralized and / or at least partially centralized method, algorithm and / or intelligence. A first portion of the behavior of the robot can be determined with the means of the robot (essentially means that can be operated or operated separately from the other robots) and / or a second portion of the behavior of the robot can be determined with at least partially common, in particular central means. The behavior of a robot is to be understood as meaning, for example, an action by the robot, a reaction to an external influence and / or an interaction of the robot with its environment or with an object. For example, the behavior of the robot can include selecting or executing an action, determining the parameters of an action, adapting an action, executing an action and / or reacting to an external influence.

The initiation of the at least one first behavior and / or a first change in a behavior of the (first) robot can be initiated or stimulated by means of a first stimulus signal. In the case of the second robot, an optionally present second stimulus signal can be provided for such a purpose. The stimulus control signal can identify a specific first behavior of the one or more first robots, which is predetermined or stimulated.

This can be a first behavior of the one or more first robots that does not contribute to the immediate mission, task of the first robot. This can be a non-productive and / or counterproductive behavior from the point of view of the concrete, individual mission or action of the first robot or from the point of view of the interests of the first robot. This can include a non-productive and / or counterproductive action, combination of actions or interactions. The (first) behavior can each comprise a combination of several actions or interactions, parameters of actions or interactions. These can be characterized by a certain behavior or be shaped by the behavior.

For example, a behavior can be characteristic of a movement, e.g. one or more two-dimensional or three-dimensional movements of a part of the robot and / or the exertion of certain forces on a part of the environment or on an object. The movement can be characterized by a certain behavior.

The information to be provided or provided can be determined as a function of, preferably on the basis of, the evaluation. The determination of the information to be provided can essentially be determined during or after the execution of the activation. The provision can, for example, in the next time interval, e.g. after a few seconds, during or before the next use of a functionality of the robot, during the next terminal cycle or when an operating mode is changed (des Robot or a functionality of the robot).

In one example of the method, depending on the information provided, a behavior of at least one second robot (in particular in a certain way) can be influenced, simplified: changed. Furthermore, the information to be provided or provided can be used to increase the safety, reliability, precision and quality of a functionality and / or an action of the robot. Desired and / or undesired trends (according to respective predetermined criteria) can be identified automatically and at an early stage on the basis of the information provided. For example, depending on the information provided, e.g. when a certain (desired or undesired) trend is recognized, one or more measures, such as a change in the behavior of one or more (certain) robots, can be initiated.

Preferably, the information based on the one or more first parameters characterizing the determined first effect of the actuator of a plurality of the first robots is used directly, indirectly, in particular in certain (determinable) quantities for training at least one neural network. The neural network can be configured as a network for operating at least one (or more) second robots. The one or more of the second robots can be operated by means of the neural network, in particular on the basis of a result or intermediate result of the training. The one or more of the second robots can be executed by means of the trained neural network (in particular at least partially the same hardware) and / or by means of a checkpont, weight information, bias terms of the trained neural network (in particular at least partially on different hardware). Furthermore, the checkponts, weight information and bias terms of the trained neural network can be used at least in part to operate second robots that are operated independently of the first robot, e.g. in a different mission, with a different goal.

As an exception, a second robot can coincide with one (of several, in particular a plurality) of the first robots or a number of the second robots can partly coincide with a number of the first robots. As an exception, a second robot can match a first robot from the past, e.g. from previous use, before a software change (e.g. before a software update), before a change to a mission and / or area of application, etc.

The training of the at least one neural network for operating at least one second robot can (also) be understood as the training as a function of at least partially evaluated, in particular compared or processed first parameters with one another or with further data. For example, the evaluation can be a comparison of the information on the basis of the one or more first parameters with at least one first predetermined condition. The evaluation can include, for example, determining a measure of a correlation, similarity, between the first parameters of the at least two different types and / or the first parameters of the same type of different first robots, time intervals, operations, operating phases, etc. Said evaluation can take place on the basis of one or more comparison values, patterns and / or mathematical relationships. These can be applied to the first parameters during the evaluation (thus the first data are also based on the first parameters or parameter values). When evaluating, it can be checked, for example, whether or to what extent the first parameters characterizing the action of the actuator match a pattern, in particular one of several patterns, a predetermined mathematical relationship, e.g. formula, table, etc., match, or with a correlate such.

• - zumindest zwei erste Wirkungen die bei unterschiedlichen Randbedingungen, Betriebsmodi, Aufgaben, Einsatzgebiete und/oder zu unterschiedlichen Zeitintervallen aufgetreten sind, und/oder
• - zumindest zwei erste Wirkungen in Reaktion unterschiedliches erstes Verhalten bzw. erste Änderungen des Verhaltens, und/oder
• - zumindest zwei erste Wirkungen von zwei unterschiedlichen ersten Roboter.

The training of the at least one neural network can be carried out in a dependent manner, in particular on the basis of two or more of the first parameters, the at least two of the first parameters being indicative of:

• - At least two first effects that occurred under different boundary conditions, operating modes, tasks, areas of application and / or at different time intervals, and / or
• - At least two first effects in response to different first behavior or first changes in behavior, and / or
• - At least two first effects from two different first robots.

In a further example, at least one further parameter can (also) be determined by a first robot and used to train the neural network. Such a parameter can be used together with a first parameter. For example, the at least one first parameter can be used, for example, as a comparison value or a guide value, a calibration value. For example, the first parameters, in particular corresponding comparison values or reference values, calibration values, can be determined by means of corresponding stimulation signals.

This can be done depending on the situation or dynamically.

In a further example, the at least one first parameter can be compared, for example, with the results obtained from other (identical or similar) robots in the case of such or similar stimulus control signals. For example, the comparison data can also be selected, e.g. depending on which functionalities, in particular an occurrence of a certain behavior of a robot, is at least likely to be imminent. A data retrieval and comparison can thus relate to relevant data and can be carried out without a delay or a disadvantage that is noticeable to the user.

A first parameter of the first effect can be a qualitative and / or quantitative measure of a result of the first behavior or the first change in behavior. The first effect may not or not necessarily be a benefit or a functionally meaningful, positive or creative achievement. This can be a concurrent, insignificant, small especially infinitesimal effect, a side effect or a counterproductive effect (e.g. with regard to the goal of the first robot). In a simplified case, the one or more first parameters (in each case) can be indicative of a physical effect or characterize a pattern of parameter values of a physical reaction that occurs in response to a specific stimulus control signal in the environment of the first robot and / or from the at least one (first) object was recorded. A pattern recognition, an assignment to one or more predetermined patterns, can be carried out. The one or more parameters characterizing a first effect can be determined with a means (e.g. sensor, measuring device) of the robot itself, another robot and / or with a means in the environment. For example, such a means can be a camera, a microphone, etc. with or without a correspondingly configured signal processing device.

For example, influencing, in particular adapting a functionality and / or behavior of the second robot, including an automated and / or remote-controlled actuation of an actuator of the one or more second robots, in particular an actuation to carry out a mission, task, action, by means of the trained neural Network and / or a checkpoint, weight information, bias terms of the trained neural network are executed.

For example, by means of the trained neural network and / or depending on a checkpoint, weight information, bias terms of the trained neural network, certain (critical and / or desired) trends can be recognized before their effect on the (normal, typical for the mission) operating parameters can become critical.

The at least one stimulus control signal for causing or stimulating the first behavior and / or a first change in behavior can include, for example, a predefined sequence of different control signals or control variables (of the same or different types) that are diverse, in particular dynamic, for example dependent on a specific need. The one or more stimulus control signals can be selected, for example, in such a way that a resulting first effect (e.g. desired and / or undesired effects, an effect that can be perceived by a person, e.g. user, etc.) is limited or within a target value (with the target value a target value range is also meant) are.

The stimulus control signal preferably differs from a control signal for functional actuation and / or remote control of the (first) robot. This can, for example, only specify a certain behavior or a change in behavior without controlling a certain action (completely). Rather, decisions can be made about (possible) actions, interactions with the environment with the means of the first robot, e.g. in a manner that is adapted to the situation.

The stimulus control signal can be identified separately (e.g. by a corresponding identifier or header) or contain values, e.g. control variables or combinations of control variables, which differ significantly from those in a typical control signal. A non-functional actuation of the robot is preferably initiated with the stimulus control signal. As an alternative or in addition, a stimulus control signal can (also) cause a functional effect and / or a certain change in the effect, e.g. a certain change in a functionality or action of the robot. It can be a control of the actuator which (in particular additionally) changes a certain functionality and / or action, e.g. a parameter of the robot function, at least partially or temporarily. The stimulus control signal can preferably be used to initiate an actuation that differs from an actual (e.g. functional, function-based) intended use of the actuator and / or from a (qualitative and functional) functional effect of the actuator (resulting from the purpose of the actuator) / or quantitative).

For example, causing or stimulating a first behavior or a change in the behavior of one or more first robots can be carried out in such a way that no or no significant effect on the environment or perceptibility from the environment results. For example, a brief actuation of the one or more actuators, a very brief movement and / or movement with a low amplitude of at least part of the robot can be brought about (controlled) with the stimulus control signal.

The actuator can be an actuator for a movement, in particular for executing a robot functionality, e.g. for influencing one or more objects to be influenced. The one or more actuators can be actuators for operating a manipulator of the robot, e.g. one or more actuators in a robot arm. The actuator can be a hydraulic and / or pneumatic actuator or a part of a hydraulic or pneumatic system of the robot.

The corresponding first parameters of the first effect can then be received and evaluated (e.g. by means of the robot and / or backends that are set up for an operative connection with the robot). These can, for example, be compared with values and patterns on the basis of corresponding statistical data and / or a result of a simulation.

Particularly preferably, a first parameter of the first effect can characterize a behavior or a change in the behavior of an object with which the first robot is acting or interacting.

For example, training of at least one neural network to operate at least one second robot can (also) be based on a plurality of parameters characterizing first effects that are carried out as a function of a plurality of specific stimulus control signals. The training is preferably also carried out as a function of the stimulus control signals which have led to certain first effects.

In a further example, the trained neural network and / or a checkpont, weight information, bias terms of the trained neural network can be used to identify a state, a trend (a specific tendency and / or measure of a tendency) and / or a trend for a State of at least a part of the first robot and / or the at least one second robot are executed.

A prediction of a state, a trend (a certain tendency and / or measure of a tendency) of a second robot can preferably be carried out by means of the trained neural network and / or a checkpont, weight information, bias terms of the trained neural network.

For example, a certain state, trend or a measure of a state or trend, e.g. a state or trend outside of a norm, in particular a norm value, can be recognized or predicted.

The training of at least one neural network to operate at least one second robot can be carried out at least partially inside the robot and / or outside the robot, preferably in a backend arranged at a distance from the first robot and the second robot, e.g. based on the data transmitted for this purpose the first parameter.

The initiation or stimulation of a first behavior or a change in a behavior of one or more first robots can be carried out in the method from or from a backend. Upon receipt of a corresponding stimulus control signal, in particular with a stimulus control signal with predetermined parameters, the first behavior or a first change in a behavior of one or more first robots (e.g. with means of the first robot in question, in particular in a manner adapted to the situation and manner). A computer unit that is set up, for example, to manage, operate, configure, control the robot, the robot system or the robot swarm, etc., can be regarded as the backend. The result is a very high degree of flexibility, which allows, for example, robots that are already in use (possibly imperceptible to the user or staged for a certain perceptibility by the user) to stimulate information, e.g. the first parameters, the information to be provided, to generate or provide (to deliver).

For example, depending on various criteria, e.g. depending on an evaluation of various data on the robot and / or the user of the robot, a need to make sense of certain information to be provided for one or more second robots and / or a certain actuation of the actuator can be made in the backend first robot to be determined. Advantageously, training of at least one neural network to operate at least one second robot can also include diverse, specific and / or non-specific data, data in connection with the robot system or robot swarm, for example from User data, other data, such as "big data", etc. include.

For example, by means of the trained neural network and / or a checkpoint, weight information, bias terms of the trained neural network can be determined which information to be provided (such as which type of information, in relation to which system of the robot, etc.) is desired for a particular one Purpose, e.g. useful or required to operate a second robot. The one or more first behavior or first changes in the behavior of the one or more first robots can then be actuated (in order to provide the required information). In other words, the information to be provided can be determined “on demand”, so to speak.

The backend is to be understood as a computing unit, e.g. server, mainframe, etc., which is arranged at a distance (in particular statically) from the at least one first robot or second robot. The backend is set up, in particular, to aggregate the data for executing the method, in particular data from larger periods of time and / or from a large number of robots. An action carried out from the backend, e.g. initiating a stimulus control signal in one or more first robots, can also take place directly or indirectly in response to an action by a person, e.g. a dispatcher.

The method can also aggregate the data identified in this document, the similar or additional data, e.g. further robot data, mission data and / or user data of the user (s) of the one or more robots, from a large number of time intervals and / or from different ones Robots or models of the robots include.

The at least one stimulus control signal can characterize or include a sequence of gradually increasing control variables for controlling the actuator or a request for a gradually increasing first effect. For example, such a sequence is identified in the stimulus control signal, e.g.

In a simplified case, causing or stimulating a first behavior or a first change in a behavior of one or more first robots can be carried out in a sequence of effects or a sequence of certain effects until a certain measure of the first effect, for example a certain one Intensity of the reaction of an environment of the robot and / or of an object in the vicinity of the robot is achieved.

For example, an induced or stimulated behavior or change in behavior is increased, in particular gradually, until a certain first effect is recognized. The parameters requested by the stimulus control signal can (also) be compared with the predetermined condition until a certain effect is achieved. For example, the first robot can perform a change or action in relation to a first object, e.g. a known object or test object. One or more second robots can then, depending on the information provided, carry out or change a change, action or behavior in relation to a second object. In a further embodiment, a parameter of a first boundary condition is specified in which the actuation of the actuator of the (first and / or second) robot is to be carried out with the stimulus control signal.

In this case, the at least one first parameter of the boundary condition can characterize a predefined time interval, position information and / or a parameter of an environmental condition. The boundary condition for controlling an actuator of the robot with a certain stimulus control signal can also relate to a certain boundary condition within the robot and / or a boundary condition in the vicinity of the robot. Such a boundary condition can be, for example, a range of values of a measured variable in the environment and / or within the robot, e.g. an outside temperature, air humidity, a temperature of a part of the robot, the finding of a certain object in the environment, etc.

For example, the at least one (desired) parameter of a boundary condition can be determined in a backend and / or in the robot itself, a device in the vicinity of the robot, etc. For example, parameters of a boundary condition can be specified for a number of the first robots in which (when the corresponding parameter values or parameter value ranges are reached) execution of a specific control of the one or more actuators, e.g. a specific action or action of a specific type, in particular in a specific area of application (e.g. a specific environment). Such a predetermined parameter of the boundary condition can, for example, be a measure of a temperature, for example at certain points or parts inside or outside the robot, in a certain range of values.

A request (or a corresponding specification) for controlling (also called actuation) an actuator of the one or more first robots can be determined by a first robot at a first time interval, transmitted to the first robot, and / or received by the first robot and then carried out in the robot at a (respectively suitable or suitable, later) second time interval at which the at least one boundary condition is met and / or a predefined criterion, for example relating to sufficient security, is met.

In one embodiment, first data characterizing the at least one first effect in the at least one stimulus control signal with second data characterizing a first effect of the actuator in the past, in particular in the case of the same or a comparable stimulus control signal and / or a stimulus control signal with a specific deviation can be compared. A first parameter can characterize a difference between the first effects. A lack of a first effect can also be regarded as the first effect.

For example, the evaluation can include that features are determined which have a comparatively longer-term change, for example a change taking place on the order of a few months or years, in particular a measure of the aging or wear and tear of a part of one or more first robots, in particular in the form of a qualitative and / or show quantitative identifiers. For example, one or more (e.g. correspondingly corresponding) statistical measures can be determined on the basis of the parameters from a large number of the first robots. Furthermore, one or more such features, e.g. as second characteristic quantities of the same type, of the at least one second robot can be determined. Furthermore, a comparison of a second feature, in particular a second identifier of a second robot with a measure of one or more first features, in particular parameters, can be carried out on the basis of a number of the first robots (from the plurality of first robots). The operation of the second robot can then be influenced as a function of the result of the comparison. For example, one or more first robots and one or more second robots can have the same and / or different mission, area of application, etc.

For example, one or more second identifiers for long-term changes in a second robot can in turn be compared with an aggregated measure on the basis of a large number of the first identifiers, for example in relation to a specific period of time. A measure can then be carried out in the at least one second robot if the comparison is outside a specified standard value or if a deviation results that exceeds a certain amount.

For example, depending on the comparison of the data, an influencing of a functionality and / or a certain (functional) action or an action of a certain type of the robot, including at least one release, restriction, change and / or blocking of at least one performance feature of the functionality or one Feature of the action, in particular with regard to certain operating states, missions, tasks, actions carried out.

According to an optional embodiment, the actuator of the robot is activated under two different boundary conditions, in particular the area of use, mission of the robot and / or operating parameters of the robot or operating parameters of the actuator. A comparison of the parameters (parameter data) characterizing the effects of the actuator in the case of the at least two activations of the actuator can then take place.

For example, two or more different specifications for controlling the at least one actuator can be generated. These can relate to different parameter values of the boundary conditions, different operating modes, different time intervals, which can be identified in the specifications. When the respective boundary conditions corresponding to the at least two different specifications are recognized, the respective controls of the actuator are carried out in the at least one robot and / or a number of robots.

For example, the initiation or stimulation of a first behavior or a first change in a behavior of the one or more first robots is carried out with the stimulus control signal in such a way that a certain effect or an effect of a certain type in relation to the local environment of the robot and / or is evoked on a certain object or an object of a certain kind. The determination of the one or more parameters can characterize the corresponding effect of the actuator on the local environment of the robot, on a specific object or an object of a specific type, and / or on (on inducing or stimulating a first behavior or a first change in behavior) following reaction of the environment of the robot or the object. Thereupon can (but does not have to) from the effect of the actuator on the local environment of the robot and / or from a reaction of the environment of the robot or the object to the effect of the actuator, also information characterizing a property of the environment, a Property of the object and / or information on how to deal with the environment or with the object can be determined. This can be provided as part of the said information to be provided or as further information provided. It can also be information that does not allow any or no explicit conclusions to be drawn about the environment or object.

According to an optional embodiment, the evaluation relates to a large number of the same or different first behavior or a first change in a behavior of one or more first robots. According to an optional embodiment, the evaluation relates to a large number of different first robots. A plurality of the actuations with different first stimulus control signals is preferably carried out in the one or more first robots. Corresponding parameters of the effect can be aggregated in the procedure, e.g. within a longer period of time. A large number of the different first behaviors or the different first changes in a behavior are preferably caused or stimulated under different boundary conditions, different areas of use, different operating modes and / or at different time intervals. For example, during the evaluation, the one or more first parameters can be compared which characterize an effect in response to the same, the same, comparable or different stimulus control signals. For example, one or more first parameters can be compared which characterize an effect of the actuator of the first robot on the same stimulus control signal. A mathematical, in particular a statistical, ratio relating to the first parameters, which for example in one or more first robots to a stimulus control signal and / or in different areas of application with different boundary conditions, can then be determined.

For example, a parameter of the (next) inducement or stimulation of a first behavior or a first change in a behavior of one or more first robots can be determined as a function of the result of the previous evaluation.

For example, the actuator of one or more first robots is controlled with two different boundary conditions and / or with two different operating parameters of the actuator, and the first parameters characterizing the effects of the actuator of one or more robots are taken into account for the two or more boundary conditions and / or operating parameters.

In a further example, the method comprises controlling at least one actuator of the first robot with the stimulus control signal, the actuator of the first robot producing a specific effect in relation to the local environment of the robot, and identifying the determination of the one or more parameters the effect of the actuator on the local environment of the first robot and / or on a reaction of the environment of the first robot to the effect of the actuator takes place.

In particular, a performance feature of the functionality for a certain second robot, for a certain group or number of second robots or for all second robots that are, for example, in the vicinity or in a certain environment, released, restricted, changed, for example adapted and / or be blocked. Preferably, the performance feature or a number of the performance features (for example in a certain way) can be influenced, which may make a possible at least partially automated and / or remote-controlled actuation of the actuator of the robot (at least when corresponding criteria apply) necessary. For example, depending on a result of the evaluation of the at least one parameter, the effect of the actuator of the one or more first robots, a release, restriction, change and / or blocking of at least one performance feature of the corresponding functionality and / or a specific action or action can be used of a certain type of a second robot.

For example, a specific specification or regulation for influencing a functionality of the one or more second robots can be determined for several boundary conditions and (immediately or again later) when the respective wheel conditions (later) apply. On this basis, for example, a route clearance or a map for releasing, restricting, changing and / or blocking at least one performance feature of the corresponding functionality can be determined and taken into account. In particular, it can be a functionality of the robot, including an automated and / or remote-controlled movement of the robot. A graded distinction can be made, which can be made in fine steps. In other words, the entire functionality of the second robot does not have to be deactivated in the event of a slight increase in the risk or in the event of the smallest suspected case. Rather, it should be adapted adaptively. Preferably, depending on the determined parameters for the operational readiness and / or failure probability of the actuator, for example on the basis of a stored rule, the most favorable or best possible combination of properly executable performance features can be used in each case to be chosen. This has the advantage that the best possible functionality (for example, based on the user's point of view) can be offered in each case, which would be executable under the given boundary conditions.

For example, by means of the trained neural network and / or a checkpoint, weight information, bias terms of the trained neural network, an adjustment of a part of an automatic (independent, "self-determined") action, part of a remotely controlled and / or externally determined, e.g. by a person (user or dispatcher) controlled action for the operation of at least part of a functionality and / or action of a second robot.

In particular, such a portion can be adapted in relation to a certain functionality of the robot and / or an action (to be carried out by the second robot) which may make a possible actuation of the robot (at least when corresponding criteria apply) necessary. Using the trained neural network and / or a checkpoint, weight information, bias terms of the trained neural network, a prediction or prediction about the consequences, risks, and precision of the action can be made or carried out in the second robot.

In particular, the object of a certain type can be an object with a certain quality. In particular, the object is characterized by one or more specific features and / or at least two features that are related to one another in a specific scientific, e.g. mathematical, relationship. In the latter case, for example, an absolute parameter of a feature can be comparatively unimportant if a certain proportion is met with another feature.

Furthermore, the object can be part of the first robot (the same first robot itself). For example, the control of the one or more actuators of one or more first robots with a first stimulus control signal can specify a self-related action of the respective first robot. In a simplified example, one or more robot arms can be relocated to interact with one another. In other words, the stimulus control signal can be used to "pat yourself on the shoulder" or "click your suspenders" from a first robot. At least one neural network can then be trained to operate at least one second robot.

For example, a measure can be carried out with respect to one or more second robots by means of the trained neural network and / or a checkpont, weight information, bias terms of the trained neural network, which in particular with regard to an expected boundary condition and / or to a specific area of application , in particular, limits the predicted or expected level of risk to an acceptable level. For example, depending on the trained neural network and / or a checkpoint, weight information, bias terms of the trained neural network, an adaptive change or blocking of at least one performance feature or a specific selection of the performance features of a functionality of a second robot can be carried out. A graded distinction can be made, which is carried out in useful fine steps.

For example, a functionality of the at least one second robot can be adapted in such a way that a predefined level of the risk (for example that can be expected in operation in general or in an upcoming operation) is not exceeded. For example, a target value for a measure of one or more specific risks is specified for the neural network. Various parameters of the at least one second robot can then be controlled by the trained neural network in such a way that the level of risk is maintained.

To carry out the method, a database can be created, for example in a backend, which includes comparison data on the effects of the at least one actuator or one type of actuator in one or more types of the robot on certain stimulus control signals. These can include, for example, comparative values for normal, deviating, conspicuous values related to specific use, wear and tear or age of the components or robots. In this context, a “value” is preferably to be understood as a pattern. The sections of the comparison data can preferably be linked to the respectively valid boundary conditions. Depending on the boundary conditions of a robot in the field, suitable comparison data for evaluating one or more requirements for training at least one neural network for operating at least one second robot can then be determined. Information for the development, operation, maintenance and / or for influencing the behavior of the robots already in use or on a mission (different or the same number of robots) can be carried out in the near future, for example days, weeks, months will.

The invention can be applied, for example, in relation to a control of the robot with a Mobile user device, e.g. smartphone, Smart-Close or with a robot remote control, in which the user is outside the robot in the vicinity of the robot. However, the method also relates to such an automated and / or remote-controlled operation of the robot (with or without the presence, readiness for action and / or physical presence of a person) from a great distance, e.g. by a person (user, dispatcher) and / or a algorithm located outside the robot, is controlled remotely. Automated and / or remote-controlled actuation of the robot can thereby be initiated, released or provided. The robot can be operated more or less immediately, ie with a corresponding control command that is interpreted in the robot as a command for operating the actuator, or only indirectly, for example with a control command that, in combination with a further condition, results in a Actuation of the at least one actuator leads.

The method can be carried out more or less regularly and at suitable times, preferably under certain boundary conditions and / or in certain areas of application, with regard to controlling at least one actuator of one or more first robots with a stimulus control signal.

According to a further embodiment, the one or more first robots and one or more second robots are operated for at least partially the same or similar missions and / or for the same or similar goals, in particular at least one common mission and / or a common goal. These missions or goals can be separated by longer time intervals. In other words, an “experience” of the one or more first robots can be passed on, in particular inherited, to the one or more of the second robots (e.g. by means of the trained neural network and / or a check pontoon, weight information, bias terms of the trained neural network).

The at least one first robot and the at least one second robot particularly preferably form a robot system and / or a robot swarm. The robot system or the robot swarm can have at least one common mission and / or at least one goal. The mission and / or goal can be at least partially predetermined and / or at least partially adaptable, in particular dynamically (for example: in the course of the mission or depending on the achievement or attainability of the goal). In particular, it can be a mission and / or goal that can be at least partially (e.g., by an operator) specified and / or at least partially determined or (in particular dynamically) adapted by the robot system or by the robot swarm.

In particular, the at least one first robot and the at least one second robot (in particular at the same time) can have at least partially different roles, tasks and / or short-term goals. The at least one mission and / or the at least one long-term goal can be at least partially, in particular essentially independently, independently determinable, or independently adaptable, and / or at least partially abstracted from the roles, tasks and / or a long-term goal.

The formation of a robot system or the robot swarm can take place at least temporarily, in particular depending on one or more upcoming goals. For example, one or more robot systems and / or swarms of robots can be formed for a specific mission or each with at least one specific goal, for example from a number of robots that can be operated with the neural network.

For example, one or more robot systems or one or more robot swarms can be formed as a function of goals located or prioritized in a list of goals. The at least one robot system or the robot swarm can be formed by means of a neural network, in particular by means of said neural network.

Such a configuration of the said robot system or the robot swarm can be strategically advantageous, in particular in the case of missions, goals, situations, solutions and / or tasks that are not known a priori.

Similar behavior, similar behavioral changes, similar contexts, similar situations can in particular be at least partially or predominantly different behavior, behavioral changes, contexts, situations. Similar behavior, similar changes in behavior, similar contexts, similar situations are characterized in particular by the fact that they are essentially assigned to the same class (or qualitative type), their parameters have the same value range, exceed at least a certain degree of similarity, a certain degree of Fall below dissimilarity, and / or correspond in each case to the same or similar patterns.

In a further example, depending on the neural network, it is possible to adapt a proportion of the independent action and / or the proportion of the action determined from outside for the operation a functionality of a second robot, in particular in relation to a (common or predetermined) mission and / or (common or predetermined) goal.

According to a further embodiment, the at least one first parameter of the first behavior or the first change in the behavior of the first robot is determined, changed and / or initiated by the neural network. Determining, changing, for example adapting, and / or initiating can depend on a (in particular common) mission and / or on a goal of at least one second robot (in particular on a common and / or long-term goal of the at least one second robot and at least one first Robot).

For example, a parameter of the (induced or still to be induced) behavior or the induced change in behavior is determined by the neural network (also to be understood: by means of the neural network and / or dependent on the data of the neural network and / or dependent on the data that are determined with a run of the neural network) is determined or adapted. At least part of said neural network can then read in the first parameter of the first effect.

Preferably, the same first robot, a different first robot, a second robot, and / or a system of at least one or more first robots, one or more second robots, and a backend can perform a behavior or change the one or more first robots Induce behavior. In particular, a parameter of the behavior or a parameter of a change in behavior can be determined or adapted depending on the second robot, in particular on the current mission of the second robot. For example, causing a specific behavior or a change in the behavior of the first robot can be initiated and / or influenced by at least one second robot. In other words, a second robot can (so to speak at its discretion and / or depending on its own, so to speak “personal” goal and / or role) stimulate, stimulate or (in response to a certain way).

According to a further embodiment, influencing, in particular controlling, a change in behavior of at least one second robot by means of the trained neural network and / or a checkpoint, weight information, bias terms of the trained neural network is carried out.

The influenced, in particular controlled, behavior of the at least one second robot can be a behavior pattern. The behavior or the behavior pattern of the second robot can relate to a second context, second situation and / or second action that has at least a certain degree of similarity and / or at most a certain degree of dissimilarity with the corresponding first context, first situation and / or have the first action of the at least one first robot. Furthermore, the behavior or the behavior pattern of the second robot can also be changed without knowledge of and / or reference to a specific or specific context, a specific or specific situation and / or a specific or specific action. A certain amount of change in behavior or a certain or selectable behavior pattern can be determined by means of the trained neural network ( 30th ) and / or a checkpoint, weight information, bias terms of the trained neural network can be determined or adapted.

According to a further embodiment, a parameter is in particular a mode of interpreting the environment of the second robot and / or a second object by the second robot by means of the at least one trained neural network and / or on the basis of a check pont, bias term, weight information, of the trained neural network Network.

According to a further embodiment, the first effect is a change in an operating parameter of the first robot, a parameter of a specific action of the first robot, a parameter of the environment of the first robot and / or change in the state of a first object in the environment of the first Robot.

According to a further embodiment, the first effect is one or more entries from the following list: a change in a parameter characterizing a behavior of a first object in the environment of the first robot; and / or a change in a parameter characterizing an interaction between the first robot and its environment; and / or a change in a parameter characterizing an interaction of the first robot with a specific first object in a specific environment of the robot, in particular in a specific context and / or in a specific situation. These can be specific changes or specific dimensions, patterns of changes.

Here can the object is, for example, an object related to a mission of the robot. For example, the object can be an object to be processed or treated by the robot, a disruptive object which, for example, hinders a mission of the robot, and / or an interaction object, for example an object with which the robot (e.g. according to its mission ) Interact should or must act.

In particular, the object can be an at least partially unknown object or an object with an at least partially unknown destination. In particular, the behavior of the object can be an at least partially unknown behavior and / or a behavior of an at least partially unknown object.

According to a further embodiment, the object is an object with natural and / or artificial intelligence. The first effect on the induced first behavior or the induced first change in the behavior of the first robot can be an action in relation to the object and / or a change in the behavior of the object.

A biological intelligence or intelligence associated with a biological body can be understood as a natural intelligence. An object with a natural intelligence can be understood as a living being, e.g. human or animal.

According to a further embodiment, the first behavior or a first change in the behavior of the one or more first robots is selected as a function of a random variable and / or using one or more specific conditions.

In particular, the selected or selectable behavior, change in behavior is limited to one or more specific conditions defined by the condition and / or requires the application of (in each case) a condition defined in the condition for its execution. In a simplified case, such a condition can relate to one or more parameters (also to be understood as parameter values), in particular operating parameters, of the (first) robot. In a further simplified case, the at least one condition can include one or more predetermined limit values or (permissible, tolerable, intolerable, impermissible) value ranges, e.g. for an action, a behavior, etc. The at least one support particularly preferably comprises one or more (permissible, tolerable, impermissible) patterns. The patterns can relate to various parameters of the (first) robot, the environment, the object, the action of the robot, the interaction of the robot with the object. The at least one execution condition can relate to the (resulting) behavior or the (resulting) change in a behavior of the at least one robot. For example, the at least one execution condition can characterize certain limits or frameworks of the behavior or the change in behavior.

In a further example, causing, in particular stimulating, a first behavior or a first change in a behavior of one or more first robots can be carried out at least partially by the first robot and / or at least partially by a further first robot and / or a second robot. The first behavior or the first change in behavior (also independent of a specific situation, context, action) can have almost any different variants, parameter values or relate to different contexts, situations. This can still be kept within certain behavioral limits. For example, a particularly effective determination of the respective first parameters can be carried out characterizing the respective first effects.

According to a further embodiment, a first behavior or a first change in a behavior of one or more first robots is caused or stimulated; and / or the determination, in particular a selection of one or more first parameters to be determined and / or processed, of one or more first effects; and / or the processing of one or more first parameters of the one or more first effects is carried out. The execution as such, a condition for the execution of the respective step and / or a parameter with which the respective step is executed is selected as a function of: a parameter of the environment of the at least one second robot; and / or a parameter of a second object in the vicinity of the second robot; and / or a short term target of the second robot; and / or a long-term goal of the second robot; and / or a solution path of the second robot in relation to a short-term goal and / or a long-term goal of the second robot.

According to a further embodiment, the behavior of the second robot is influenced; and / or a desired behavior and / or undesired behavior of the second robot (as such) recognized.

The desired behavior and / or undesired behavior of the second robot can be an admissible behavior and / or an inadmissible behavior.

According to a further embodiment, causing or stimulating a first behavior or a first change in a behavior of one or more first robots, determining one or more first parameters characterizing one or more first effects, and / or evaluating one or more first parameters referring to characterizing the one or more first effects on a multiplicity of identical or similar first behaviors or first changes in the behavior of the one or more first robots, in connection with several different contexts and / or situations, and / or on a multiplicity of different first behaviors or first changes in the behavior of the one or more first robots, in connection with the same or similar contexts and / or situations.

Similar behavior, similar behavioral changes, similar contexts, similar situations can in particular be at least partially or predominantly different behavior, behavioral changes, contexts, situations. Similar behavior, similar changes in behavior, similar contexts, similar situations are characterized in particular by the fact that they are essentially assigned to the same class (or qualitative type), their parameters have the same value range, exceed at least a certain degree of similarity, a certain degree of Fall below dissimilarity, and / or correspond in each case to the same or similar patterns. The “multitude” is to be understood as meaning, in particular, a number of several 1000, several 10,000 or several 1000000 contexts, situations and / or operating cycles of the method.

According to a further embodiment, the trained neural network (e.g. after completion or intermediate result of training) and / or on the basis of a checkpoint, bias term, weight information of the trained neural network is used to release, change and / or restrict at least one performance feature of the functionality, and / or an operating parameter of the at least one second robot carried out.

In particular, depending on the information provided, a combination of the (together, simultaneously) activatable performance features of the second robot can be selected that is activated, made activatable, changed, restricted or blocked.

The at least one release, change and / or restriction of at least one performance feature of the functionality and / or an operating parameter of the at least one second robot can be carried out as a function of the recognized permissible behavior and / or impermissible behavior of the second robot.

This can be carried out in such a way that a probability for a specific behavior of the second robot, in particular towards a target value, is changed.

For example, depending on the information provided, an adjustment of a portion of the independent action and / or portion of the action determined from outside for the operation of a functionality of a second robot ( 20th , 21 , 22nd ) respectively.

According to a further embodiment, the method comprises: causing or stimulating a second behavior or a second change in a behavior of a second robot; determining one or more second parameters characterizing a second effect that has occurred on the first behavior or the second change in the behavior of the one or more second robots; comparing information based on the one or more second parameters with information based on the one or more first parameters characterizing the first effects, and then: training the neural network using information based on the one or more second parameters and information based on the one or more first parameters.

In a further example, the at least one trained neural network ( 30th ) and / or on the basis of a checkpoint, bias term, weight information, the trained neural network ( 30th ) an adjustment of a proportion of the independent action and / or proportion of the action determined from outside for the operation of a functionality of a second robot ( 20th , 21 , 22nd ) respectively.

According to a second aspect, the invention comprises a system having at least a first robot and / or a second robot and a neural network, the first robot and / or the second robot and the neural network being set up to execute an operative connection. The system is designed to: induce or stimulate a first behavior or a first change in a behavior of one or more first robots; it is also designed to: determine one or more first parameters characterizing a first effect that occurred on the induced or stimulated behavior or the change in the behavior of the one or more first robots, and is designed to train the neural network to operate at least a second one Robot dependent of the one or more determined first parameters.

The system can include a device provided in a robot and a back end, the device in the robot and the back end and / or the mobile user device being set up to establish or execute an operative connection. The system can be designed to carry out the method in accordance with one or more features of the method. The system can be a system for operating a first robot and / or a second robot. The system can be the robot system and / or robot swarm.

According to a third aspect, the invention comprises a robot system or robot swarm having one or more first robots and one or more second robots, the robot system or robot swarm being set up a, in particular common, predeterminable by means of the at least one first robot and / or the at least one second robot To carry out a mission and / or to strive for a specifiable, in particular common, goal by means of the first robots and / or the second robots, one or more of the first robots being set up, a neural network for operating the at least one second robot by performing the method to train.

According to a fourth aspect, the invention is characterized by a computer program, the computer program being designed to operate a robot, robot system or robot swarm according to one or more features described in this document and / or the method according to one or more features described in this document Features to perform. In particular, the computer program is a software program which, for example, as an app (= “application”) for one or more types of electronic computing units, can preferably run on said electronic control unit. The computer program is preferably also designed to take into account, aggregate a large number of additional information items for operating the system and / or for executing the method.

According to a further aspect, the invention is characterized by a computer program product comprising executable program code and / or data sections, the program code and / or data sections being designed to enable the electronic unit to carry out the method according to one or more features of the device or the method . The computer program product comprises, in particular, a medium that can be read by the data processing device or storage areas on a medium or in a backend or cloud on which the program code and / or data sections are stored. According to a further aspect, the invention is characterized by a product which comprises a permanent or restricted authorized access right to stored data of the computer program product. The computer program product can be designed as an update of a previous computer program which, for example, includes the parts of the computer program or the corresponding program code as part of a function expansion, for example as part of a so-called “remote software update”. The method can therefore also include unrestricted, restricted access to the program code and / or data sections, for example for a group of robots and / or the corresponding users, for example restricted to certain applications or conditions.

The robot in the context of this document is preferably a robot of a type of robot already described above. This results in advantages which are identified in this document and which are easily comprehensible for a person skilled in the art. Alternatively or additionally, the robot can be a machine, in particular at least partially. For example, the robot can be a logistics robot. The robot system or robot swarm can be a logistics system, in particular a logistics system that can be operated predominantly in an automated manner. In a further example, a robot can be an unmanned (or unmanned operable) flying object, an unmanned (or unmanned operable) ship. For example, a robot system can include a number or multiplicity of such robots. In addition to a (more or less automated) movement of the robot as a whole, the at least one robot can be distinguished by a mission, e.g. relating to a certain interaction with its environment or with certain objects. The robot can be movable as a whole and / or in itself (as a part of the robot in relation to a further part of the robot). The robot can be a machine for making certain changes in its environment. With regard to its shape, actions and / or sensors and information processing, a robot can at least partially distinguish itself from the characteristics of a person or an animal. A robot can include one or more biological agents.

In a particular exception, the robot can be any type of machine. One result of the method can be a (comparatively easy to implement) variant of swarm intelligence with regard to the machines. For example, the Robot is an agricultural machine or a machine for carrying out earthworks and / or a digging machine, a machine for mining ores or precious stones.

A particularly great advantage of the invention arises when it is applied to a number of robots (special robots, robots with a common task, mission, area of application, etc.) selected according to predetermined criteria, e.g. belonging to one or more, e.g. cooperating brands or members of an organization , a group of users, a group of a social network and so on. The method can include or the system can also be designed for an expedient exchange of information on the basis of the data of other robots from the number of robots. Furthermore, the invention also includes all other systems or functionalities of the second robots that can be used in the second robot, the weight information based on the invention, e.g. with the system or depending on the information determined in the process, in particular by means of the trained neural network and / or a check pontoon , Bias terms of the trained neural network are expediently operable.

With the method, a number of the (first and / or second) robots can be provided with a type of “swarm intelligence”. In this case, behavior, changes in behavior, behavioral patterns and / or actuation of their actuators or an action resulting from one or more actuations can be initiated in the first robot (possibly not functionally necessary for the first robots or in themselves relatively “meaningless”) , whereby information that can be used (at least for one or more second robots) is determined and provided, for example by means of the trained neural network and / or a checkpoint, weight information, bias terms of the trained neural network.

• 1 zeigt eine schematische Darstellung der Erfindung. Diese bezieht sich auf mehrere erste Roboter 10,11,12, ..., ein Backend 30 und/oder ein mobiles Anwendergerät 40, 40', und mehrere zweiten Roboter 20, 21, 22, .... Die Roboter umfassen unterschiedliche Aktoren 1a, 1b, 1c, 1d bzw. 2a, 2b, 2c, 2d. Es können unterschiedliche Arten von Aktoren sein.
• 2 zeigt ein Ablaufdiagramm des Verfahrens gemäß einer Ausführung der vorliegenden Erfindung. Dieser ist (symbolisch durch ihre Anordnung) auf die entsprechende Recheneinheiten bezogen, die für die jeweiligen Schritte der Erfindung (lediglich beispielhaft) vorgeschlagen werden.

Exemplary embodiments of the invention are explained below, without restricting the generality, with reference to the accompanying drawings. The features of the examples can be combined with one another and / or with all of the other features described in this document.

• 1 shows a schematic representation of the invention. This relates to several first robots 10, 11, 12, ..., a backend 30th and / or a mobile user device 40 , 40 ' , and several second robots 20th , 21 , 22nd , .... The robots have different actuators 1a , 1b , 1c , 1d or. 2a , 2 B , 2c , 2d . There can be different types of actuators.
• 2 shows a flow chart of the method according to an embodiment of the present invention. This is related (symbolically by its arrangement) to the corresponding computing units which are proposed for the respective steps of the invention (only by way of example).

In the following description of the figures, the same elements or functions are provided with the same reference symbols. In the invention features of the 1 with the characteristics of 2 be combined.

The described method for providing S4 information for operating a second robot 20th , 21 , 22nd includes driving S1 at least one actuator 1a , 1b , 1c , 1d one or more first robots 10 , 11th , 12th with a stimulus control signal, a determining S2 one or more parameters characterizing a physical effect of the actuator in response to the stimulus control signal, an evaluation S3 at least one parameter characterizing the determined effect of the actuator 1a , 1b , 1c , 1d in the at least one stimulus control signal having at least one first predetermined condition. The information (possibly a section of the data with the information) is then determined and made available S4 the information for operating a robot 20th , 21 , 22nd , and / or a robot system or robot swarm comprising a robot 20th , 21 , 22nd , ... based on the comparison.

The said information can be sent directly or indirectly to the one or more second robots 20th , 21 , 22nd ..., to be provided. In one example, the information is provided by means of a mobile user device 40 , 40 ' and / or a backend 30th provided. One or more (useful) steps for processing, aggregation, preparation of the information in a user device can be performed 40 , 40 ' and / or a backend 30th are executed.

Only in exceptional cases can the information provided also be used in the same robot 10 , 11th , 12th directly or indirectly provided or to operate the robot 10 , 11th , 12th , especially in the same robot 10 , 11th , 12th , be used. a number of the first robots and a number of the second robots can at least partially or temporarily overlap. A second robot can be a first robot after a software update and / or a robot with a different mission.

In a first example, a number of first robots are used 10 , 11th , 12th a request (for example, as corresponding data packets, data with which the corresponding stimulus control signals are triggered) distributed which a specific Control of certain actuators S1 calls. For example, actuators of the robot arms, means of locomotion for the robots, or an output device can be controlled 1a etc. the robot 10 , 11th , 12th etc. are initiated.

A number of first robots 10 , 11th , 12th for control S1 of the actuator or actuators 1a , 1b , 1c , 1d can be selected according to a feature that is dependent on a geographical criterion, area of application, mission, user group (customer group, customer segment, etc.) and / or certain boundary conditions in which the at least one first robot is operated. For example, a number of the robots 10 , 11th , 12th are defined in predominantly controllable conditions, for example in laboratory conditions or in conditions that allow a replica or simulation of the area of application of the second robot 20th , 21 , 22nd corresponds, and / or are operated in the vicinity of a control center and a different number of robots that are operated in a, for example, remote, partially unknown area of application. In this example, these steps of the method can take place in a specific area delimited (by position coordinates) that is or is not affected by a specific problem or issue in connection with the operation of the robots, for example.

In an example that can be combined with further features of the invention, the activation S1 of the one or more actuators 1a , 1b , 1c , 1d one or more first robots the one or more first robots 10 , 11th , 12th with a first stimulus control signal, carry out a (specific) change or action in relation to a first object, for example a known object or test object.

One or more second robots can then 20th , 21 , 22nd execute or change a change, action or behavior in relation to a second object depending on the information provided (determined in this way). The second object can be a different object than the first object. For example, the second object can be a much larger, distant, unknown object.

• - eine hydraulische Größe, insbesondere eine Sequenz oder ein Muster der Druckveränderung in einem, z.B. pneumatischen oder hydraulischen, System des Roboters. Daraufhin kann ein Abgleich des ermittelten (gemessenen) Druckfolgemusters mit einem oder mehreren vorausbestimmten Mustern erfolgen.
• - einen Parameter kennzeichnend einen Bewegungsparameters, Beschleunigungsparameters und/oder Kraftparameters eines Aktors bzw. eines Roboterarms. Beispielsweise kann eine Sequenz oder ein Muster kennzeichnend eine (resultierende) Bewegung, Beschleunigung und/oder bestimmter dabei auftretender Kräfte ermittelt werden. Daraufhin kann ein Abgleich des ermittelten (gemessenen) Musters mit einem oder mehreren vorausbestimmten Mustern erfolgen.
• - einen Parameter kennzeichnend eine Auswirkung eines kurzen an einem Aktor des Roboters mit dem Stimulus-Steuersignal veranlassten Impulses.
• - einen oder mehrere Zeitparameter, beispielsweise Verzögerungszeiten, Totzeiten, etc. Beispielsweise kann ein Parameter der Wirkung des Aktors bei einem Stimulus-Steuersignal ein Zeitintervall zwischen einem bestimmten Stimulus-Steuersignal oder einer Kombination von Stimulus-Signalen und dem Erreichen einer bestimmten physikalischen Wirkung des Aktors kennzeichnen.
• - ein Maß eines Schalls und/oder Körperschalls, der auf die Ansteuerung mit dem Stimulus-Steuersignal gemessen wird. Daraufhin kann ein Abgleich des als Parameter der Wirkung des ermittelten (gemessenen, als die ersten Parameterdaten erfassten) Vibrationsmusters, Druckfolgemusters, Frequenzmusters, Amplitudenmusters, Phasenmusters, eines zeitlichen Gradient einer der genannten Größen, etc. , mit einem oder mehreren vorausbestimmten Mustern erfolgen. Dabei können die Daten kennzeichnend die Wirkung des Aktors beispielsweise mit einem Mikrofon (z.B. einem Mikrofon des Roboters oder einem Mikrofon in der Umgebung zur Erfassung der Umgebung) und/oder einem Körperschall-Sensor ermittelt werden. Ein Maß des Schalls und/oder Körperschalls kann mit einem Körperschallsensor des Roboters oder einem Mikrofon bzw. Körperschallsensor in der Umgebung des Roboters, z.B. von einem weiteren ersten Roboter erfasst werden.

The one or more first parameters characterizing a first effect in a stimulus control signal can (in a simplified case) be characteristic of:

• a hydraulic variable, in particular a sequence or a pattern of the pressure change in a system of the robot, for example a pneumatic or hydraulic system. The determined (measured) print sequence pattern can then be compared with one or more predetermined patterns.
• a parameter characterizing a movement parameter, acceleration parameter and / or force parameter of an actuator or a robot arm. For example, a sequence or a pattern characterizing a (resulting) movement, acceleration and / or specific forces occurring in the process can be determined. The determined (measured) pattern can then be compared with one or more predetermined patterns.
• - a parameter characterizing an effect of a short pulse caused by the stimulus control signal on an actuator of the robot.
• - One or more time parameters, for example delay times, dead times, etc. For example, a parameter of the effect of the actuator in the case of a stimulus control signal can be a time interval between a certain stimulus control signal or a combination of stimulus signals and the achievement of a certain physical effect of the actuator mark.
• - A measure of a sound and / or structure-borne sound that is measured in response to the activation with the stimulus control signal. The vibration pattern, print sequence pattern, frequency pattern, amplitude pattern, phase pattern, a temporal gradient of one of the variables mentioned, etc., can then be compared with one or more predetermined patterns as a parameter of the effect of the determined (measured, recorded as the first parameter data). The data characterizing the effect of the actuator can be determined, for example, with a microphone (for example a microphone of the robot or a microphone in the environment for detecting the environment) and / or a structure-borne sound sensor. A measure of the sound and / or structure-borne noise can be recorded with a structure-borne sound sensor of the robot or a microphone or structure-borne sound sensor in the vicinity of the robot, for example by a further first robot.

In an example that can be combined with further features of the invention, the evaluation is dependent on S3 at least one first parameter from one or more first robots 10 , 11th , 12th and from a measure of the risk that is acceptable for a specific application, the most favorable combination (according to predefined criteria) of perfectly executable (mutually combinable) performance features of a functionality of a second robot 20th , 21 , 22nd to be chosen. Combinations of performance features of the functionality can be based on a stored rule, e.g. as a prioritization of different combinations according to their (in advance specified or adaptively determined) value for the user. This has the advantage that the best possible functionality (e.g. based on the user's point of view) can be offered, which under given boundary conditions, with a risk reduced to a certain level and / or with a risk that does not exceed a limit value, e.g. a certain Standard-compliant, risk is feasible.

In a further example, there is an at least probable need for actuation of a specific actuator 2a , 2 B , 2c , 2d of the robot 20th , 21 , 22nd determined when this is remotely controlled and / or when there is no presence or attention of the person (e.g. user or dispatcher of the robot). In a simplest case, such a prediction can be dependent on an actuation of a specific action with the second robot or a functionality of the second robot 20th , 21 , 22nd or on a mobile user device 40 ' (this can also include a remote control device for the robot). A case can also be determined in which a human is in the vicinity of the robot 20th , 21 , 22nd is located, but (at least for the moment) insufficiently in the observation or operation, in particular control of the robot 20th , 21 , 22nd is involved, e.g. no attention to the robot 20th , 21 , 22nd , the environment of the robot 20th , 21 , 22nd or the action of the robot 20th , 21 , 22nd dedicates or has been winding in the last few seconds.

• - Ansteuern S5 des Aktors des zweiten Roboters 20, 21, 22 mit einem zweiten Stimulus-Steuersignal,
• - Ermitteln S6 eines oder mehrerer zweiten Parameter kennzeichnend eine Wirkung des Aktors des zweiten Roboters 20, 21, 22, ... in Reaktion auf das zweite Stimulus-Steuersignal,
• - Abgleich S7 einer Information auf Basis des einen oder mehreren zweiten Parameter mit der bereitgestellter Information auf Basis eines ersten Roboter 10, 11, 12 und daraufhin:
• - Beeinflussen S8 einer Funktionalität eines oder mehrerer zweiten Roboter 20, 21, 22 und/oder einer Aktion, insbesondere Mission eines oder mehrerer zweiten Roboter 20, 21, 22 abhängig von dem Ergebnis des Abgleichs.

Depending on the need for actuation of a specific actuator of the robot 20th , 21 , 22nd One or more entries from the following list can be made, especially in combination with the cases identified above:

• - Approach S5 of the actuator of the second robot 20th , 21 , 22nd with a second stimulus control signal,
• - Detect S6 one or more second parameters characterizing an effect of the actuator of the second robot 20th , 21 , 22nd , ... in response to the second stimulus control signal,
• - Adjustment S7 information based on the one or more second parameters with the information provided based on a first robot 10 , 11th , 12th and then:
• - Influence S8 a functionality of one or more second robots 20th , 21 , 22nd and / or an action, in particular a mission of one or more second robots 20th , 21 , 22nd depending on the result of the comparison.

For example, a section of the first data on the basis of the first parameters identifies one or more first robots 10 , 11th , 12th a certain undesirable effect (e.g. identified as one of 2, 4, 8, ... gradations) that occurs when several (specific) controls of at least one specific actuator, e.g. a manipulator of one or more robots 10 , 11th , 12th has revealed. The selected section of the first data is then evaluated. In a variant of the method, a change characteristic of the degree of the undesired effect can be determined in the course of several actuations of the actuator. For example, if it is determined that a change characteristic, e.g. a corresponding time curve of the undesired effect, has an unfavorable tendency, e.g. within a few weeks or months, the relevant functionality, which the actuation of the actuator needs, e.g. for a remote-controlled process, can be influenced or a (further) appropriate measures can be initiated.

In a variant of the method, the predetermined condition can be determined on the basis of the second data. Such a predetermined condition can be adjusted continuously, e.g. once a week, or event-driven, e.g. in the event of a certain occurrence.

In an example that can be combined with further features of the invention, a (specific) risk, in particular a risk exceeding a threshold in connection with a specific behavior, or with a specific actuation of a remote-controlled and / or automated action or functionality of a further robot 20th , 21 , 22nd is determined or estimated in a specific application. For example, the degree of risk is assessed in appropriately chosen 2, 4, 8, ... gradations. The determination of the degree of risk can be estimated, for example, on the basis of the data from a sensor located outside the robot, the position of the person (the user, dispatcher or an uninvolved person) in remote-controlled processes. For example, an (at least rudimentary) algorithm for determining specific risks is executed for an upcoming application, and one or more determined, in particular predicted, risks, for example for a specific application of the second robot 20th , 21 , 22nd taken into account. A product of a probability and a consequence can be determined as a measure of the risk. A measure of functional reliability or operational safety can be taken into account for a specific application (on site, if all the boundary conditions for operation are known). If a probability measure for the execution of a certain remote-controlled movement of the robot 20th , 21 , 22nd , comprising an automated and / or remote-controlled actuation of an actuator 2a , 2 B , 2c , 2d of the robot 20th , 21 , 22nd , fulfills a predetermined condition and / or if the extent of the Risks in connection with a significant part of the functionality and / or action, e.g. with an upcoming and / or specific work step of the robot 20th , 21 , 22nd , or the degree of risk which is due to the proportion of the automated and / or remote-controlled actuation of the actuator 2a , 2 B , 2c , 2d occurs, exceeds a certain threshold (or a corresponding reciprocal value falls below a corresponding threshold), a comparison of the corresponding data can be initiated.

• - Einlesen eines Abschnitts der bereitgestellter Information auf Basis der Auswertung S3 durch den zweiten Roboter 20, 21, 22, und
• - Beeinflussen S8 einer Funktionalität des zweiten Roboters 20, 21, 22 abhängig von der eingelesenen Information; Optional kann dabei auch ein oder mehrere der folgenden Schritte ausgeführt werden:
• - Ansteuern S5 des Aktors 2a, 2b, 2c, 2d des zweiten Roboters 20, 21, 22 mit einem zweiten Stimulus-Steuersignal,
• - Ermitteln S6 eines oder mehrerer zweiten Parameter kennzeichnend eine Wirkung des Aktors des zweiten Roboters 20, 21, 22, ... in Reaktion auf das zweite Stimulus-Steuersignal.
• - Berücksichtigung des einen oder mehrerer zweiten Parameters beim Beeinflussen der Mission, des Verhaltens, einer Aktion und/oder einer Funktionalität des zweiten Roboters 20, 21, 22.

In this example, the following steps are described as an example:

• - Reading in a section of the information provided on the basis of the evaluation S3 by the second robot 20th , 21 , 22nd , and
• - Influence S8 a functionality of the second robot 20th , 21 , 22nd depending on the information read; Optionally, one or more of the following steps can also be carried out:
• - Approach S5 of the actuator 2a , 2 B , 2c , 2d of the second robot 20th , 21 , 22nd with a second stimulus control signal,
• - Detect S6 one or more second parameters characterizing an effect of the actuator of the second robot 20th , 21 , 22nd , ... in response to the second stimulus control signal.
• - Consideration of the one or more second parameters when influencing the mission, the behavior, an action and / or a functionality of the second robot 20th , 21 , 22nd .

A comparison can be made S7 information based on the one or more second parameters with the information provided based on a first robot 10 , 11th , 12th are executed.

The information to be provided (based on the data of a first robot 10 , 11th , 12th ) determined and / or for operating the at least one second robot 20th , 21 , 22nd used or taken into account, depending on a degree of similarity and / or dissimilarity between the area of use, mission and / or functionalities of the first robot and the second robot.

One or more steps of the method can be selected and / or configured, depending on the value range in which an expected level of risk, e.g. for a specific, e.g. upcoming mission, action, application (use of the robot, robot functionality, system of the robot) lies or is expected.

The evaluation S3 of the at least one parameter characterizing the determined effect of the actuator 1a , 1b , 1c , 1d of at least one first robot 10 , 11th , 12th , or the actuator 2a , 2 B , 2c , 2d of at least one second robot 20th , 21 , 22nd may (alternatively or additionally) include a comparison with a predetermined condition.

The predetermined condition can include, for example, one or more predefined mathematical relationships in the form of a regularity, rule, rule and / or relation. These can identify or represent relationships between one or more of the corresponding operating parameters (also to be understood as corresponding values of the parameters). The mathematical relationship can represent coefficients for one or more, in particular predetermined, equations, polynomials or a simulation. This is preferably an expediently defined mathematical relationship. The mathematical relationship particularly preferably comprises at least one function which characterizes a time dependency of a variable and / or a dependency in relation to an area of application of the robot. In particular, the mathematical relationship can characterize at least one function that determines a gradient of an operating parameter according to a space-related quantity (relative space-related quantities, e.g. characterizing distance, distance and / or absolute quantities, e.g. characterizing certain position sequence) and / or according to a time-related quantity (Relative quantities, for example time intervals, and / or absolute quantities, such as a sequence of specific points in time). The first data of one or more first operating parameters in the robot can be aggregated, e.g. statistically evaluated, data on the basis of the one or more first operating parameters, in particular on the basis of a longer period of time.

The proposed method enables the component status of the robots to be reliably monitored. It replaces the adjustment of a practically infinite variety of boundary conditions in the context of a test and a validation. Furthermore, a very precise, and therefore very valuable, prediction of a failure, a service requirement, the admissibility or non-admissibility of certain operating modes of the robot, certain robot components or robot functions can be made.

The invention can, inter alia, achieve an advantage in terms of safety, reliability, quality of robot operation and also a cost advantage. This also enables a quantitatively defined and / or dynamic release, restriction, blocking and / or change, e.g. an appropriate adaptation of at least one action, of a performance feature of the functionality of a (second) robot, in particular for one or more specific boundary conditions and / or areas of application are executed.

In doing so, (at least the second) robots can operate in a new or previously unknown environment, area of application, e.g. on other planets, taking into account the information provided and / or deal with a new mission or task much more successfully. Appropriate information, for example for operating a specific mission of a second robot, a second robot in a specific environment or under specific boundary conditions, or for a specific second robot, in particular as required, can be generated (by means of at least one first robot) will.

Claims (13)

A method for operating one or more robots (20, 21, 22), in particular a robot system and / or a robot swarm, comprising: - inducing or stimulating (S1), a first behavior or a first change in a behavior of one or more first robots (10, 11, 12), - Determination (S2) of one or more first parameters characterizing a first effect that has occurred on the induced or stimulated behavior or the change in the behavior of the one or more first robots (10, 11, 12), - Training (S3) of at least one neural network (30) to operate at least one second robot (20, 21, 21) as a function of the determined first parameters. Procedure according to the Claim 1 , in which the one or more first robots (10, 11, 12) and one or more second robots (20, 21, 22) for at least partially the same or similar missions and / or for the same or similar goals, in particular at least one common mission and / or operate a common goal; and / or in which at least one first parameter of the first behavior or the first change in the behavior of the first robot (10, 11, 12) from the neural network (30), in particular depending on a mission and / or on a goal of at least a second Robot (20, 21, 22), determined, changed and / or initiated. Method according to one of the preceding claims, in which influencing (S4), in particular controlling a change in behavior, of at least one second robot (20, 21, 21) by means of the trained neural network (30) and / or a check pont, weight information, bias terms of the trained neural network (30) is performed. Method according to one of the preceding claims, wherein a parameter in particular a mode of interpreting the environment of the second robot (20, 21, 22) and / or a second object by the second robot (20, 21, 22) by means of the at least one trained neural network (30) and / or The basis of a checkpoint, bias term, weight information, of the trained neural network (30) is carried out; and or wherein the first effect is a change in an operating parameter of the first robot (10, 11, 12), a parameter of a specific action of the first robot (10, 11, 12), a parameter of the environment of the first robot (10, 11 , 12) and / or changing the state of a first object in the vicinity of the first robot (10, 11, 12). Method according to one of the preceding claims, wherein the first effect is a change in one or more entries from the following list: - A parameter characterizing a behavior of a first object in the vicinity of the first robot (10, 11, 12), and / or - A parameter characterizing an interaction between the first robot (10, 11, 12) and its environment, and / or - a parameter characterizing an interaction of the first robot (10, 11, 12) with a specific first object in a specific environment of the robot (10, 11, 12), in particular in a specific context and / or in a specific situation. Method according to one of the preceding claims, wherein the first object is an object with a natural and / or artificial intelligence and the first effect on the induced behavior or change in the behavior of the first robot (10, 11, 12) is an action in relation to the object and / or is a change in the behavior of the first object; and or wherein the first behavior or a first change in the behavior of the one or more first robots (10, 11, 12) is selected as a function of a random variable and / or using one or more specific conditions. Method according to one of the preceding claims, wherein the causing or stimulation (S1) of a first behavior or a change in a behavior of one or more first robots (10, 11, 12), and / or the determination (S2), in particular a selection of one or more first parameters to be determined and / or processed, of one or more first effects and / or the processing (S3) of one or more first parameters of the one or more first effects is carried out depending on: - a parameter of the environment of the at least one second robot (20, 21, 22) and / or - a parameter of a second object in the environment of the second robot (20, 21, 22), and / or - a short-term goal of the second robot (20, 21, 22), and / or - a long-term goal of the second robot (20, 21, 22), and / or - a solution path of the second robot (20, 21, 22) in relation to a goal of the second robot (20, 21, 22). Method according to one of the preceding claims, in which by means of the trained neural network (30) and / or on the basis of a checkpoint, bias terms, weight information of the trained neural network (30): - the behavior of the second robot (20, 21, 22) is influenced, and / or, - a desired behavior and / or undesired behavior of the second robot (20, 21, 22) is recognized; and / or with the means of the trained neural network (30) and / or on the basis of a checkpont, bias term, weight information of the trained neural network (30) a release, change and / or restriction of at least one performance feature of the functionality and / or one Operating parameters of the at least one second robot (20, 21, 22) is executed. Method according to one of the preceding claims, wherein causing or stimulating (S1) a first behavior or a first change in a behavior of one or more first robots (10, 11, 12), determining (S2) one or more first parameters characterizing one or a plurality of first effects, and / or the evaluation (S3) of one or more first parameters characterizing the one or more first effects on a plurality of identical or similar first behaviors or first changes in the behavior of the one or more first robots (10, 11 , 12), in connection with several different contexts and / or situations, and / or on a multiplicity of different first behaviors or first changes in the behavior of the one or more first robots (10, 11, 12), in connection with the same or similar ones Contexts and / or situations. A method according to any one of the preceding claims, comprising: - causing or stimulating (S5) a second behavior or a second change in a behavior of a second robot (20, 21, 22), - Determination (S6) of one or more second parameters characterizing a second effect that occurred on the first behavior or the second change in the behavior of the one or more second robots (20, 21, 22), - Comparison (S7) of information based on the one or more second parameters with information based on the one or more first parameters characterizing the first effects, and then: - Training (S8) the neural network (30) by means of information based on the one or more second parameters and information based on the one or more first parameters. System comprising at least a first robot (10, 11, 12) and / or a second robot (20, 21, 22), and a neural network (30), the first robot (10, 11, 12) and / or the second robot (20, 21, 22) and the neural network (30) are set up to execute an operative connection, the system being designed to: initiate or stimulate (S1) a first behavior or a first change in one or more behavior first robot (10, 11, 12), further designed to: determine (S2) one or more first parameters characterizing a first effect on the induced or stimulated behavior or the change in the behavior of the one or more first robots (10, 11, 12) has occurred and is designed to train (S3) the neural network (30) to operate at least one second robot (20, 21, 21) depending on the one or more determined first parameters. Robot system or robot swarm comprising one or more first robots (10, 11, 12) and one or more second robots (20, 21, 22), the robot system or robot swarm being set up by means of the at least one first robot (10, 11, 12) ) and / or the at least one second robot (20, 21, 22) to carry out a predefinable, in particular common, mission and / or a predefinable, in particular common, goal by means of the first robots (10, 11, 12) and / or the second robots (20, 21, 22), one or more of the first robots (10, 11, 12) being set up, a neural network (30) for operating the at least one second robot (10, 11, 12) by executing the Procedure according to one of the Claims 1 until 10 to train. Computer program, the computer program being designed, upon its execution one, in particular to the system according to Claim 11 or to the robot system or robot swarm Claim 12 according to the associated computing unit, in particular by means of a digital neural network (30), one or more steps of the method according to one or more Claims 1 until 10 to execute.

Images