Generating a modular workflow for configuring industrial equipment using an ai pipeline tool

Abstract

The disclosure relates to a method for generating a workflow for performing a configuration task of at least one industrial equipment. The method comprises obtaining a query that is indicative of the configuration task. The method further comprises generating, based on the query and using an artificial intelligence, AI, pipeline tool, the workflow. Generating the workflow comprises obtaining an alternative out of at least one potential alternative for performing a respective sub-task in a chain of sub-tasks for the configuration task. Further, generating the workflow comprises obtaining a plurality of intermediate validation-points for validating a respective obtained alternative. Moreover, generating the workflow comprises validating, based on the obtained plurality of intermediate validation-points, the respective obtained alternative. The present disclosure further relates to a workflow, a data processing apparatus, a computer program, a computer-readable storage medium, and a system.

Description

[0001] ABB Schweiz AG 10.02.2025

[0002] A19408WO

[0003] Generating a modular workflow for configuring industrial equipment using an Al pipeline tool

[0004] FIELD OF THE INVENTION

[0005] The invention relates to generating a modular workflow for configuring industrial equipment using an Al pipeline tool.

[0006] BACKGROUND

[0007] A typical configuration workflow for industrial equipment such as a robotic system is generated manually. A user generally needs to study the handbooks, usually several of them, to understand the options and parameters to be configured

[0008] • for a given robot,

[0009] • for the additional components, e.g., a gripper,

[0010] • for the components programming languages, and

[0011] • for the communication between them.

[0012] Then the user interacts with several configuration tools or simulation software, such as Robotstudio for the robot, to configure the device via clicks and typing based on information from expert knowledge, user manuals or internet forums.

[0013] There have been workflows for Generative Al to execute complex tasks, such as Chain of Thoughts, Tree of Thoughts and Graph of thoughts. However, as the task gets more complicated, the chain I tree I graph of thoughts may grow significantly in size and it will be more difficult, and even impossible, for the Gen Al-based solution to arrive at the correct output because there are too many intermediate steps and options. In particular, if the task is to combine a robot and its peripherical devices to a complete and functional system. For example, the probability of hallucination increases as the number of steps increases. The reliability and quality of the intermediate steps will pose challenges to the Gen Al-based solution when addressing complex tasks. Without a modular structure of the Al system itself, validation becomes impossible. Further, the absence of a modular structure forces the feedback loop from the configuring equipment to be very large, andP240686W001 -2 - 10.02.2025 discrete feedback at the end of a workflow in the form of “works” or “doesn’t work” does not help to improve.

[0014] OBJECTIVE OF THE INVENTION

[0015] It is therefore an objective of the present invention to generate a modular workflow for configuring industrial equipment using an Al pipeline tool.

[0016] This objective is at least partially achieved by the subject matter of the independent claims of the present disclosure, wherein further examples are incorporated in the dependent claims.

[0017] SUMMARY

[0018] According to a first aspect, there is provided a method for generating a workflow for performing a configuration task of at least one industrial equipment, the method comprising: obtaining a query that is indicative of the configuration task; and generating, based on the query and using an artificial intelligence, Al, pipeline tool, the workflow by: obtaining an alternative out of at least one potential alternative for performing a respective sub-task in a chain of sub-tasks for the configuration task; obtaining a plurality of intermediate validation-points for validating a respective obtained alternative; and validating, based on the obtained plurality of intermediate validation-points, the respective obtained alternative.

[0019] Herein, the term “workflow” may refer to a graph representing or indicative of a process. The generated workflow may be a fully validated workflow and may comprise steps to be taken, decisions to be made, and / or paths to be followed for the configuration task.

[0020] The query may be obtained from a user, e.g., a domain expert of the industrial equipment to be configured. The query may be a sentence, e.g., an instruction, semantics of which can be processed using the Al pipeline tool. For instance, the user may request “please configure a robot system with a robotic arm, e.g., GoFa, and a XYZ gripper that can open and close. Define the I / O signals and then write a Rapid routine". Such a sentence can be processed and used to generate the workflow according to any one of the exemplary methods disclosed herein. In another example, the configuration tool may provide anP240686W001 - 3 - 10.02.2025 interactive button to the user, using which the user may selectively prompt the Al system to gather context and generate the workflow.

[0021] The Al pipeline may refer to a series of steps designed to automate and streamline the process of building, training, evaluating and deploying machine learning models. The Al pipeline may perform any one or more of the steps of Al or machine learning (ML) lifecycle. The Al pipeline may perform at least one of: data (pre-)processing; modeling; learning; training, evaluating, and modifying the model; and deploying the model. The Al pipeline tool may refer to any tool, process, algorithm, or functional unit capable of performing any of the aforementioned functionalities of the Al pipeline. The Al pipeline tool may comprise or be a natural language model. Specifically, the natural language model may comprise or be a natural language processing, generating, and / or understanding model. The natural language model may comprise or be a machine learning model, specifically a generative Al, more specifically a multi modal large language model (LLM) or an LLM. The Al pipeline may be capable of processing the query and using the processed information for generating the workflow. Alternatively or additionally, the Al pipeline may be capable of dividing, separating, segmenting, decomposing, or modularizing the configuration task into a plurality of sub-tasks, and further capable of ordering, sorting, prioritizing, or organizing to form a chain of subtasks. Herein, the term “sub-task” refers to a partial task of the configuration task. A subtask may be performed independent of another sub-task. The sub-tasks are serially interconnected in a chain manner, and the intermediate validation-points are situated between, thereby connecting, the two consecutive sub-tasks in the chain of sub-tasks. Consequently, the configuration task forms a chain of process with the sub-tasks and the intermediate-validation points connected in an interleaving manner. Alternatively or additionally, the Al pipeline may be capable of creating, producing, generating, or determining alternative(s), solution(s), or implementable solution(s) for each sub-tasks in the chain of sub-tasks. Alternatively or additionally, the Al pipeline may be capable of obtaining, e.g., choosing, an alternative, or a solution, among the generated alternative(s), wherein the obtained, e.g., chosen, solution is used to perform a sub-task. Alternatively or additionally, the Al pipeline may be capable of generating, creating, providing, determining, deriving, defining intermediate validation-points, and / or retrieving information about the system itself based on a natural language prompt from the user. The natural language prompt may be “what are the joint limits of the robots?” and / or “which I / O devices does the system have?”. Herein, the term “obtaining” may refer to,P240686W001 - 4 - 10.02.2025 choosing, selecting, determining, computing, or receiving. For instance, any one of the plurality of intermediate validation points can be obtained, e.g., received, from the user, and / or any one of the plurality of intermediate points can be obtained, e.g., determined, using the Al pipeline tool. Obtaining the plurality of intermediate validation-points may be performed sequentially, i.e. , a subsequent intermediate validation-point may be obtained after validating the at least one alternative for the current validation-point. Two or more intermediate validation-points may be obtained simultaneously. The plurality of intermediate validation points may be obtained before, during, or after obtaining the potential alternatives to be obtained for the sub-tasks. Alternatively or additionally, the Al pipeline may be capable of validating the obtained alternative(s) for the respective subtask. Herein, the phrase “validating the obtained alternative” may refer to evaluating or determining whether the obtained, e.g., chosen, alternative satisfies at least one predefined criterion. Alternatively or additionally, the Al pipeline may be capable of augmenting, executing, and / or modifying intermediate validation-points. Herein, the term “intermediate validation-point” may be interchangeably used with a “checkpoint”

[0022] The method according to any example of the first aspect advantageously modularizes a big and / or complex task, which in turn enables the Al pipeline to work with more complex tasks that require many intermediate steps. Further, the method also reduces the risk and associated computational cost, specifically for the GenAI-based solution, because, in case some mistakes are made at a very early stage of a workflow, the workflow can return to an intermediate validation-point which had been lastly validated and does not need to start all over again from the very beginning. Moreover, the differences in configuration, programming and debugging strategies on different devices can be handled by the split into meaningful modules (e.g. part of the graph between checkpoints) taking the devices particular error codes into account during the validation step(s). These advantages in turn reduces a computational load and improves a computational efficiency.

[0023] According to an example, the method further comprises configuring, based on the generated workflow, the at least one industrial equipment.

[0024] The generated workflow may comprise all or a portion of the potential and / or obtained alternative(s) for the respective sub-task. Alternatively or additionally, the generated workflow may comprise all or portion of decision paths taken while generating theP240686W001 - 5 - 10.02.2025 workflow. Alternatively the generated workflow may comprise only the evaluated and / or validated alternative(s) for the respective sub-task. Alternatively or additionally, the generated workflow indicates the validated alternative(s) for the respective sub-task. Alternatively or additionally, the generated workflow may be transformed, converted, compiled, etc., into a file of any form, specifically a configurational file, which may be used to configure the at least one industrial equipment.

[0025] According to an example, validating the obtained alternative is based on at least one of: a manual; a user input; and a feedback from the at least one industrial equipment performing the respective sub-task using the obtained alternative.

[0026] Validating the obtained alternative may comprise evaluating the obtained alternative with respect to at least one criterion. The at least one criterion, with respect to which the obtained alternative is evaluated, may be obtained from or using at least one of: the manual; the user input; and the feedback from the at least one industrial equipment performing the respective sub-task using the obtained alternative. For instance, the at least one criterion may relate to any electrical or mechanical parameter of the at least one industrial equipment, or of any part of the at least one industrial equipment, to be configured using the workflow. The user input may comprise or be a high-level intermediate target(s) for accomplishing the configuration task provided by the user. Alternatively or additionally, the obtained alternative can be communicated to and be directly performed on the at least one industrial equipment or on any part of the at least one industrial equipment. Any resulting data therefrom can be fed back and used by the Al pipeline tool for validating the obtained alternative. Alternatively or additionally, the user may be prompted, e.g., through a graphical user interface (GUI), to approve or provide input regarding the obtained alternative.

[0027] According to an example, generating the workflow comprises re-iterating, upon determining that the obtained alternative for a sub-task of the chain of sub-tasks is invalidated, the validation for the sub-task by returning to a preceding intermediate validation-point of the plurality of validation-points and selecting a different alternative for the sub-task.

[0028] Herein, the obtained alternative may be, or may be determined or marked as, invalidated, when the obtained alternative fails to meet or satisfy the at least one pre-defined criterion.P240686W001 -6 - 10.02.2025 The re-iteration of the validation may be repeated or persist until at least one alternative is validated, i.e., meets the at least one pre-defined criterion. In case none of the obtained and / or evaluated alternative(s) can be validated, generating the workflow may be halted and / or any one of previously obtained or evaluated alternative(s) for the concerned subtask may be obtained, based on for instance a performance ranking, by the Al pipeline tool or the user. In this regard, the user may be prompted to provide input, e.g., through a GUI.

[0029] The at least one industrial equipment may be a robotic system, a robot, or a part of a robot.

[0030] According to an example, generating the workflow comprises proceeding, upon determining that the obtained alternative for a sub-task of the chain of sub-tasks is validated, to a succeeding sub-task of the chain of sub-tasks.

[0031] According to an example, generating the workflow comprises storing and / or including in the workflow: the at least one potential alternative, the chain of sub-tasks, results of validations performed on the intermediate validation-points, a re-iteration path indicative of the re-iterating step, and / or a proceeding path indicative of the proceeding step.

[0032] The generated workflow may comprise all or a portion of the path or decisions taken while generating the workflow and / or the path to be followed for configuring the at least one industrial equipment. The path to be followed may comprise the validated alternatives for the sub-tasks.

[0033] According to an example, the method further comprises obtaining a further query that is indicative of a further configuration task; and generating, based on the further query and using the Al pipeline tool, a further workflow, the further workflow by: obtaining a further alternative out of at least one further potential alternative for performing a respective subtask in a chain of further sub-tasks for the further configuration task; obtaining a plurality of intermediate validation-points for validating a respective obtained further alternative; and validating, based on the obtained plurality of intermediate validation-point, the respective obtained further alternative.P240686W001 - 7 - 10.02.2025 According to an example, generating the further workflow comprises determining whether at least one of further potential alternatives corresponds to the previously validated obtained alternative.

[0034] According to an example, generating the further workflow comprises performing, upon determining that at least one of the further potential alternatives corresponds to the previously validated obtained alternative: selecting the determined at least one of the further potential alternatives as the obtained further alternative for the respective subtask of the chain of further sub-tasks; and skipping the validation for the obtained further alternative corresponding to the previously validated obtained alternative.

[0035] The further configuration task may comprise or be a re-configuration task for reconfiguring the at least one industrial equipment, or at least a part thereof, which has already been configured, e.g., based on the workflow that has been previous generated according to any one of the methods disclosed herein. For instance, the configuration task may have been to configure a robotic arm with a gripper attached thereto for gripping an object, and the further configuration task may be to configure the same robotic arm with the gripper attached thereto for releasing the grip. In such case, one or more of further sub-tasks defined based on the further configuration task may at least partially overlap with the one or more of the sub-tasks defined based on the earlier configuration task. The earlier workflow may be stored on and retrieved from a storage unit. Upon determining, e.g., by the Al pipeline tool, that such overlap(s) exists, it may also be determined whether at least one of the further potential alternatives for any of the further sub-tasks corresponds to any of the previously validated obtained alternatives in the earlier workflow. In determining that such correspondence exists, such further potential alternative(s) may be obtained and / or the validation for such obtained further potential alternative(s) may be skipped, i.e., the validation step is not performed. That is, the method according to the above example facilitates the recycling or re-using of existing at least part of the previously validated solutions, which have been enabled by modularizing a complex configuration task, i.e., breaking down or splitting the configuration task into sub-tasks and validating the solution obtained for each of the subtasks. Such example advantageously achieves an improved computational efficiency and a reduced computational load by avoiding the need of starting the entirety of the further configuration task all over again to update, modify, and validate all of the intermediate steps therein.P240686W001 - 8 - 10.02.2025

[0036] According to an example, generating the further workflow comprises storing and / or including in the further workflow: the determined at least one of the further potential alternatives, and a skipping path indicative of the skipping step.

[0037] The generated further workflow may comprise all or a portion of the path or decisions taken while generating the further workflow and / or the path to be followed for further configuring the at least one industrial equipment. The path to be followed may comprise the validated further alternatives for the sub-tasks.

[0038] According to an example, the least one potential alternative is represented using a chain of thoughts, a multiple chain of thoughts, a tree of thoughts, or a graph of thoughts.

[0039] The chain of thoughts, the multiple chain of thoughts, the tree of thoughts, and the graph of thoughts may be various approaches to solving problems using the Al pipeline tool, specifically using a large language model (LLM). The term “thought” therein refers to a coherent language sequence that serves as an intermediate step toward problem solving. Herein, any one of the thoughts in the chain of thoughts, the multiple chain of thoughts, the tree of thoughts, or the graph of thoughts may correspond to an alternative which may be chosen for performing a sub-task. For instance, a tree of thoughts approach may generate multiple initial thought processes, assess the generated paths, then iterate the process upon chosen path(s) until arriving at the final thought, which may be a candidate for solving a problem. The final thought may correspond to a chosen alternative for performing a respective sub-task in the chain of sub-task as described herein. When validated, another tree of thoughts may be generated, evaluated, and leads to the final thought for the next validation. It is understood that different approaches may be used for different sub-tasks. When invalidated, the paths may be re-assessed to find another final thought for the validation.

[0040] According to an example, the at least one industrial equipment comprises or is a part of a robot, a robot, or a robotic system.

[0041] According to an example, the Al pipeline tool comprises or is a natural language model.P240686W001 - 9 - 10.02.2025 According to a second aspect, there is provided a workflow for configuring at least one industrial equipment, the workflow being generated according to a method of any one the examples of the first aspect disclosed herein.

[0042] The generated workflow may comprise all or a portion of the path or decisions taken while generating the workflow and / or the path to be followed for configuring the at least one industrial equipment. The path to be followed may comprise the validated alternatives for the respective sub-task in the chain of sub-tasks.

[0043] According to a third aspect, there is provided data processing apparatus comprising means for carrying out the method according to any one of the examples of the first aspect disclosed herein.

[0044] According to a fourth aspect, there is provided a computer program comprising instructions which, when the computer program is executed by a computer, cause the computer to carry out the method according to any one of the examples of the first aspect disclosed herein.

[0045] According to a fifth aspect, there is provided a computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to carry out the method according to any one of the examples of the first aspect disclosed herein.

[0046] According to a sixth aspect, there is provided a system for configuring at least one industrial equipment, the system comprising: the at least one industrial equipment; and a data processing apparatus according to any one of the examples of the second aspect disclosed herein.

[0047] BRIEF DESCRIPTION OF FIGURES

[0048] Examples of the disclosure will be described in the following with reference to the following drawings.

[0049] Figure 1 illustrates an exemplary embodiment of the system for configuring at least one industrial equipment.P240686W001 - 10 - 10.02.2025 Figure 2 illustrates an exemplary process of generating a workflow for configuring a gripper attached to a GoFa robot using the exemplary system of the present disclosure.

[0050] Figure 3 illustrates an exemplary embodiment of a part of the system shown in Figure 2.

[0051] Figure 4 illustrates an exemplary embodiment of the workflow generated according to an exemplary method disclosed herein.

[0052] Figure 5 illustrates an exemplary embodiment of a method for generating a workflow for performing a configuration task of at least one industrial equipment.

[0053] DETAILED DESCRIPTION

[0054] The Figures are merely schematic representations and serve only to illustrate examples of the disclosure. Identical or equivalent elements are in principle provided with the same reference signs.

[0055] Figure 1 illustrates an exemplary embodiment of a system for configuring at least one industrial equipment. Specifically, the system 100 comprises at least one industrial equipment 110 and a data processing apparatus 120. The data processing apparatus 120 comprises a data storage unit 130 and a data processing unit 160. The data storage unit 130 comprises a computer-readable storage medium 140. On the computer-readable storage medium 140, there is provided a computer program 150. The computer program 150 and, thus, also the computer-readable storage medium 140, comprise instructions which, when executed by the data processing unit 160, or, more generally speaking, a computer, cause the computer or the data processing unit 160 to carry out a method for generating a workflow for configuring at least one industrial equipment, according to any one of the examples of the first aspect of the invention as disclosed herein. The data storage unit 130 may store the generated workflow and / or the data processing unit 160 may implement or run the Al pipeline tool. Any module comprised in the system 100 may be communicatively coupled to another module comprised therein to perform the method according to any one of the examples of the first aspect of the invention as disclosed herein. A further detailed exemplary system overview is elaborated in reference to Figure 2.P240686W001 - 11 - 10.02.2025 Figure 2 illustrates an exemplary process of generating a workflow for configuring a gripper attached to a robotic arm, e.g., a GoFa robot, using the exemplary system of the present disclosure. The user 250 queries “please configure a robot system with GoFa and a XYZ gripper 280 that can open and close. Define the I / O signals and then with a Rapid routine" 250a. That is, an overall configuration task 250b is obtained by a global orchestrator 220. The orchestrator 220 is configured to supervise and / or coordinate any entity connected thereto including a graph of checkpoints database 210 and the Al-pipeline tool. The overall configuration task 250b may include or define the validationpoints, or herein also referred to as “checkpoints”, provided by the user 250.

[0056] In case the checkpoints are provided, the global orchestrator 220 searches the checkpoints or even the required graph of checkpoints (GoC) in the GoC database (DB) 210 based on the query 250a of the user 250. Herein, a GoC may be a workflow comprising checkpoints. In case the search is successful, the global orchestrator 220 retrieves 210a at least a part of one or more of stored GoC(s). The retrieved data may form a basis for generating a final GoC 220c, i.e. , the generated workflow. For instance, a part of the final GoC 220c may correspond to a part of the retrieved GoC, which comprises the solutions that had already been validated. Consequently, the corresponding part of the final GoC 220c need not be validated. In case the search is unsuccessful, or the checkpoints are not provided in the first place, the global orchestrator 220 uses the Al pipeline tool to generate the graph of checkpoints. The global orchestrator may also argument 220a the database 210 by updating the GOC DB 210 with existing GOCs or parts of them which are considered to be generic and reusable.

[0057] The following information are further provided to global orchestrator 220:

[0058] • Manuals and other documentation 270a: these comprise different instructions for configuration of Robot and also associated parts. This can be converted to different steps based on different conditions and form a graph of checkpoints using large language models (LLMs) 260.

[0059] • Actions and events in the configuration software 270b: these can be recorded and through process mining and LLMs 260, they can be converted into graph of checkpoints.

[0060] • Additionally, the graph of checkpoints can be stored from the past configuration generation for robots.P240686W001 - 12 - 10.02.2025 The checkpoints may also be suggested by the human expert.

[0061] The global orchestrator 220 communicates with the checkpoint orchestrator 230 to generate the final GoC 220c. The global orchestrator 220 transmits an intermediary GoC 220b with partially empty edges to the checkpoint orchestrator 230. The intermediary GoC 220b comprises checkpoints, potential solutions to be validated, and optionally validated solution(s). The checkpoint orchestrator 230 uses the Al pipeline tool to obtain alternatives, or “thoughts” in the chain / graph / tree of thoughts, for executing the subtasks that are required for a given checkpoint. The checkpoint orchestrator 230 will exercise the typical “chain I tree I graph of thoughts” and utilize the communicatively coupled various modules 240 to do so. Further, the checkpoint orchestrator 230 validates the chosen alternatives at the respective checkpoint. The checkpoint orchestrator 230 generates a validation feedback 230a using the LLM 260, and provides it to the global orchestrator 220. In case all of the performed validations succeed, the validation feedback 230a may be a graph or chain of thoughts connecting two consecutive checkpoints. In case at least one validation fails, the validation feedback 230a may comprise or be a notification of failure. In such case, the various modules 240 communicates the validation results with the checkpoint orchestrator 230, and / or regenerates the corresponding files. The validation loop continues until a correct configuration I program is generated or a threshold of retries is reached. If a threshold of retries is reached then a different path in the GoC may be required to be followed. This new path could start at certain branches in the chain I tree I graph of thoughts for a given checkpoint, or at the checkpoint itself. The new path will not go further back in the graph than the last checkpoint. The detailed interaction between the checkpoint orchestrator and the various modules are further elaborated in reference to Figure 3.

[0062] Figure 3 illustrates an exemplary embodiment of a part of the system shown in Figure 2. Specifically, an interaction between the checkpoint orchestrator and the various modules to reach a given checkpoint based on a given situation (partial configuration) is illustrated. The various modules 240 comprises the following modules:

[0063] ■ Configuration generation module:

[0064] The configuration generation module generates configuration files of the robot. The module is called by the checkpoint orchestrator to fulfill certain steps in the graph until a checkpoint is reached.

[0065] ■ Program generation module:P240686W001 - 13 - 10.02.2025 The program generation module generates program of the robot to e.g., execute certain tasks. The module is called by the checkpoint orchestrator to fulfill certain steps in the graph until a checkpoint is reached.

[0066] ■ Validation module:

[0067] The validation module validates the partial generated configuration I program. This could be done by generating additional test cases and / or a piece of test code with certain limitations. The module is called by the checkpoint orchestrator to validate certain generated configuration I program. Validation on the way towards the checkpoint could be optional and decided by the checkpoint orchestrator; validation at the checkpoint is mandatory. One example of the validation module could be a module that can upload the generated configuration I program to the robot system to validate them in the system.

[0068] The various modules 240 interact with the LLM 260 for generating configurations, programs, and validation procedures. The LLM 260 is provided with partial graph, desired checkpoint 324, and user manuals, configuration files, DB content, etc. 323. The various modules further interact with a digital twin 322 for testing configurations and programs and for returning feedback, e.g., error messages. The checkpoint orchestrator 230 generates validated configuration file(s) and / or program(s).

[0069] it is understood that the global orchestrator 220, the checkpoint orchestrator 230, and the various modules 240 described herein in reference to figures 2 and 3 may be comprised in and / or an implementation of the data processing unit 160 of Figure 1.

[0070] Figure 4 illustrates an exemplary embodiment of the workflow generated according to an exemplary method disclosed herein. The validated workflow starts at step 410 and ends at step 460. The workflow comprises N number of intermediate checkpoints 431, 432, to 43n for validating potential solutions for the respective sub-task in a chain of sub-tasks 421 and 422, wherein N is a natural number. For simplicity, further sub-tasks in the chain of sub-tasks are omitted. The potential solutions for a first sub-task 421 and a second sub-task 422 are represented using tree of thoughts, wherein each node represents a partial or full solution. The nodes are interconnected with arrows indicating a flow of information or process. The invalidated solutions are marked with “X” signs, whereas the validated solutions are connected to the subsequent checkpoint using arrows. For instance, three potential solutions 451, 452, and 453 exist for solving the second subtask 422. Firstly, at the second checkpoint 432, the first potential solution 451 isP240686W001 - 14 - 10.02.2025 assessed, and is determined to fail to satisfy a pre-determined criterion. Accordingly, the first potential solution 451 is marked as invalidated, the third potential solution 453 is selected, and the validation is repeated from the first checkpoint 431. The repetition is illustrated by means of a feedback path 441. Similarly, the third potential solution 453 is invalidated, and the validation process is once again repeated. This time, the second potential solution 452 is evaluated and validated. Then, the workflow continues with the validation process for the subsequent sub-task. The validation process continues until step 460.

[0071] Figure 5 illustrates an exemplary embodiment of a method for generating a workflow for performing a configuration task of at least one industrial equipment. Specifically, the method is for generating a workflow for performing a configuration task of at least one industrial equipment. The method starts at 510. At 520, a query that is indicative of the configuration task is obtained. At 530, the workflow is generated, based on the query and using an artificial intelligence (Al) pipeline tool, by: obtaining an alternative out of at least one potential alternative for performing a respective sub-task in a chain of subtasks for the configuration task; obtaining a plurality of intermediate validation-points for validating a respective obtained alternative; and validating, based on the obtained plurality of intermediate validation-points, the respective obtained alternative. The method ends at 540.

[0072] As used herein, the phrase “at least one,” in reference to a list of one or more entities should be understood to mean at least one entity selected from any one or more of the entities in the list of entities, but not necessarily including at least one of each and every entity specifically listed within the list of entities and not excluding any combinations of entities in the list of entities. This definition also allows that entities may optionally be present other than the entities specifically identified within the list of entities to which the phrase “at least one” refers, whether related or unrelated to those entities specifically identified. Thus, as a non-limiting example, “at least one of A and B” (or, equivalently, “at least one of A or B,” or, equivalently “at least one of A and / or B”) may refer, in one example, to at least one, optionally including more than one, A, with no B present (and optionally including entities other than B); in another example, to at least one, optionally including more than one, B, with no A present (and optionally including entities other than A); in yet another example, to at least one, optionally including more than one, A, and at least one, optionally including more than one, B (and optionally including other entities).P240686W001 - 15 - 10.02.2025 In other words, the phrases “at least one,” “one or more,” and “and / or” are open-ended expressions that are both conjunctive and disjunctive in operation. For example, each of the expressions “at least one of A, B, and C,” “at least one of A, B, or C,” “one or more of A, B, and C,” “one or more of A, B, or C,” and “A, B, and / or C” may mean A alone, B alone, C alone, A and B together, A and C together, B and C together, A, B, and C together, and optionally any of the above in combination with at least one other entity.

[0073] Other variations to the disclosed examples can be understood and effected by those skilled in the art in practicing the claimed disclosure, from the study of the drawings, the disclosure, and the appended claims. In the claims the word “comprising” does not exclude other elements or steps and the indefinite article “a” or “an” does not exclude a plurality. A single processor or other unit may fulfill the functions of several items or steps recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. A computer program may be stored / distributed on a suitable medium such as an optical storage medium or a solid-state medium supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems. Any reference signs in the claims should not be construed as limiting the scope of the claims.P240686W001 - 16 - 10.02.2025 LIST OF REFERENCE SIGNS

[0074] 100 System

[0075] 110 Industrial equipment

[0076] 120 Data processing apparatus

[0077] 130 Storage unit

[0078] 140 Storage medium

[0079] 150 Computer program

[0080] 160 Data processing unit

[0081] 210 Graph of Checkpoints database

[0082] 210a Data retrieval

[0083] 220 Global orchestrator

[0084] 220a Data upload

[0085] 220b Intermediary graph of checkpoints

[0086] 220c Transmitted data

[0087] 230 Checkpoint orchestrator

[0088] 230a Validation feedback

[0089] 240 Various modules

[0090] 250 User

[0091] 250a User query

[0092] 250b Configuration task

[0093] 260 Large language model

[0094] 270a User manuals

[0095] 270b Action events in existing systems

[0096] 321 Validated configuration file(s) and / or program(s)

[0097] 322 Digital twin

[0098] 323 User manuals, configuration files, DB content, and others.

[0099] 324 Partial graph and desired checkpoint

[0100] 400 Workflow

[0101] 410 Sta rt of wo rkf I ow

[0102] 421 First sub-task

[0103] 422 Second sub-task

[0104] 431 First validation-point

[0105] 432 Second validation-point

[0106] 43n nthvalidation-pointP240686W001 10.02.2025 441 Feedback path

[0107] 451 First potential alternative

[0108] 452 Second potential alternative

[0109] 453 Third potential alternative

[0110] 460 End of workflow

[0111] 510 Start of method

[0112] 520 First step of method

[0113] 530 Second step of method

[0114] 540 End of method

Claims

P240686W001 - 18 - 10.02.2025 Claims:

1. A method for generating a workflow for performing a configuration task of at least one industrial equipment, the method comprising:• obtaining a query that is indicative of the configuration task; and• generating, based on the query and using an artificial intelligence, Al, pipeline tool, the workflow by:obtaining an alternative out of at least one potential alternative for performing a respective sub-task in a chain of sub-tasks for the configuration task;obtaining a plurality of intermediate validation-points for validating a respective obtained alternative; andvalidating, based on the obtained plurality of intermediate validationpoints, the respective obtained alternative.

2. The method of claim 1, further comprising:• configuring, based on the generated workflow, the at least one industrial equipment.

3. The method of claim 1 or 2, wherein validating the obtained alternative is based on at least one of:• a manual;• a user input; and• a feedback from the at least one industrial equipment performing the respective sub-task using the obtained alternative.

4. The method of any one of the preceding claims, wherein generating the workflow comprises:• re-iterating, upon determining that the obtained alternative for a sub-task of the chain of sub-tasks is invalidated, the validation for the sub-task by returning to a preceding intermediate validation-point of the plurality of validation-points and selecting a different alternative for the sub-task;• proceeding, upon determining that the obtained alternative for a sub-task of the chain of sub-tasks is validated, to a succeeding sub-task in the chain of sub-tasks; and / orP240686W001 - 19 - 10.02.2025 • storing and / or including in the workflow: the at least one potential alternative, the chain of sub-tasks, results of validations performed on the intermediate validation-points, a re-iteration path indicative of the re-iterating step, and / or a proceeding path indicative of the proceeding step.

5. The method of any one of the preceding claims, further comprising:• obtaining a further query that is indicative of a further configuration task; and • generating, based on the further query and using the Al pipeline tool, a further workflow, the further workflow by:obtaining a further alternative out of at least one further potential alternative for performing a respective sub-task in a chain of further sub-tasks for the further configuration task;obtaining a plurality of intermediate validation-points for validating a respective obtained further alternative; andvalidating, based on the obtained plurality of intermediate validation-point, the respective obtained further alternative.

6. The method of claim 5, wherein generating the further workflow comprises:• determining whether at least one of further potential alternatives corresponds to the previously validated obtained alternative;• performing, upon determining that at least one of the further potential alternatives corresponds to the previously validated obtained alternative: o selecting the determined at least one of the further potential alternatives as the obtained further alternative for the respective subtask of the chain of further sub-tasks; ando skipping the validation for the obtained further alternative corresponding to the previously validated obtained alternative; and / or • storing and / or including in the further workflow: the determined at least one of the further potential alternatives, and a skipping path indicative of the skipping step.

7. The method of any one of the preceding claims, wherein the least one potential alternative is represented using a chain of thoughts, a multiple chain of thoughts, a tree of thoughts, or a graph of thoughts.P240686W001 -20 - 10.02.2025 8. The method of any one of the preceding claims, wherein the at least one industrial equipment comprises or is a part of a robot, a robot, or a robotic system.

9. The method of any one of the preceding claims, wherein the Al pipeline tool comprises or is a natural language model.

10. A workflow for configuring at least one industrial equipment, the workflow being generated according to a method of any one of claims 1 to 9.

11. A data processing apparatus comprising means for carrying out the method of any one of claims 1 to 9.

12. A computer program comprising instructions which, when the computer program is executed by a computer, cause the computer to carry out the method of any one of claims 1 to 9.

13. A computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to carry out the method of any one of claims 1 to 9.

14. A system for configuring at least one industrial equipment, the system comprising:• the at least one industrial equipment; and• a data processing apparatus according to claim 11.

Images