Modular skill library and reliability testing method and system for multifunctional robot arm

Abstract

The application provides a modular operation skill library and reliability test method and system for a multifunctional robot arm. It belongs to the technical field of intelligent robot software architecture, modular control and automated test. The method comprises the following steps: atomizing standard units of operation tasks of the multifunctional robot arm, generating standardized operation unit data; constructing a unified skill model according to the standardized operation unit data, and then forming a modular operation skill library; and configuring a visual recognition module, a voice instruction module and a large model module based on the modular operation skill library. By constructing the modular operation skill library, the operation tasks are decoupled into standardized atomic units, the development efficiency of the multifunctional robot arm operation is greatly improved, new tasks do not need to be developed from zero, and the existing units can be directly reused for quick combination, thereby saving a large amount of time and labor cost.

Description

A modular operational skill library and reliability testing method and system for multifunctional robotic arms Technical Field

[0001] This invention proposes a modular operation skill library and reliability testing method and system for multifunctional robotic arms, belonging to the interdisciplinary technical fields of intelligent robot software architecture, modular control and automated testing. Background Technology

[0002] Driven by the wave of industrial automation and intelligentization, multi-functional robotic arms are being used more and more widely in many fields, becoming key equipment for improving production efficiency and quality. However, the current operation control of multi-functional robotic arms faces many bottlenecks, making it difficult to meet the growing demand for diversified and flexible operations.

[0003] In existing technologies, operational logic is often hard-coded using a single script or state machine, resulting in tight coupling of skill implementations. Common "grab-move-place" workflows, for example, cannot be decomposed and reused. Adding new tasks requires development from scratch, leading to extremely low development efficiency. Furthermore, the lack of a unified skill model results in significant differences in skill interfaces, parameter formats, and error handling mechanisms across different tasks, making integration difficult and severely hindering the expansion and optimization of robotic arm functionality.

[0004] In terms of invocation methods, it only supports manually specifying task IDs, which cannot be integrated with advanced technologies such as visual recognition, voice commands, or large model (VLA), thus limiting the robotic arm's autonomous decision-making capabilities. Moreover, the reliability verification process is lacking, and the skills rely heavily on manual trial runs before going live, making it difficult to systematically test its success rate and accuracy maintenance capabilities under disturbances such as force control deviation, positioning error, and load changes.

[0005] Against this backdrop, there is an urgent need for a new modular skill library and reliability testing method to break through existing limitations, achieve atomic decoupling of tasks, multimodal triggering and dynamic combination, and quantify skill reliability through a systematic testing framework, so as to provide strong support for the intelligent development of multifunctional robotic arms. Summary of the Invention

[0006] This invention provides a modular operational skill library and reliability testing method and system for multifunctional robotic arms, to solve the problems mentioned in the background art above:

[0007] This invention proposes a modular operational skill library and reliability testing method for a multi-functional robotic arm, the method comprising:

[0008] S1. Decouple the tasks of the multi-functional robotic arm into atomized standard units to generate standardized task unit data; construct a unified skill model based on the standardized task unit data, and then form a modular task skill library;

[0009] S2. Based on the modular job skill library, configure the visual recognition module, voice command module and large model module to realize the visual / voice / VLA multimodal triggering mechanism; receive external commands according to the multimodal triggering mechanism, and dynamically combine standardized job units from the modular job skill library to generate executable job task sequence data;

[0010] S3. Based on the executable task sequence data, set the operation parameters for the multi-functional robotic arm to obtain operation parameter data; at the same time, collect operation environment information to generate operation environment data; combine the operation parameter data and operation environment data to construct test scenario data containing three dimensions: parameters, environment, and execution.

[0011] S4. Based on the test scenario data, start the automated test framework to simulate the operation process of the multi-functional robotic arm under disturbance, and collect the actual execution data during the test process; compare and analyze the actual execution data with the preset success standards and accuracy requirements to generate quantitative assessment data of skill reliability;

[0012] S5. Based on the quantitative assessment data of skill reliability, optimize and adjust the standardized operation units in the modular operation skill library to generate an optimized modular operation skill library; based on the optimized modular operation skill library, perform operation control of the multi-functional robotic arm and generate operation execution feedback data; continuously iterate and update the modular operation skill library based on the operation execution feedback data.

[0013] This invention proposes a modular operational skill library and reliability testing system for a multi-functional robotic arm, comprising:

[0014] One or more processors;

[0015] Memory, used to store one or more programs;

[0016] Wherein, when the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method described in any one of the above.

[0017] The beneficial effects of this invention are as follows: By constructing a modular task skill library, tasks are decoupled into standardized atomic units, significantly improving the development efficiency of multi-functional robotic arms. New tasks do not need to be developed from scratch; existing units can be reused and quickly combined, saving considerable time and manpower costs. Simultaneously, the unified skill model reduces the integration difficulty of different tasks, enhancing the scalability and compatibility of the robotic arm's functions, enabling it to easily adapt to diverse operational scenarios. The multi-modal triggering and dynamic combination mechanism allows the robotic arm to respond precisely to visual commands, flexibly adjust its movements according to voice requirements, and interact with large models to achieve autonomous decision-making, greatly improving operational flexibility and intelligence. The automated testing framework covers parameters, environment, and execution dimensions, systematically quantifying skill reliability, reducing the uncertainty and errors of manual trial runs, and avoiding problems such as low success rates and decreased accuracy in actual operations due to incomplete testing. This ensures the stability and high-quality completion of tasks, providing solid and reliable technical support for the intelligent development of general-purpose robots. Attached Figure Description

[0018] Figure 1 is a flowchart of the method steps described in this invention. Detailed Implementation

[0019] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0020] As shown in FIG1, one embodiment of the present invention provides a modular operational skill library and reliability testing method for a multifunctional robotic arm, the method comprising:

[0021] S1. Decouple the tasks of the multi-functional robotic arm by atomic standard unit (ASU) to generate standardized task unit data; construct a unified skill model based on the standardized task unit data. The unified skill model includes standardized skill interfaces, parameter formats and error handling mechanisms, thereby forming a modular task skill library.

[0022] S2. Based on the modular job skill library, configure the visual recognition module, voice command module, and large model (VLA) module to realize the visual / voice / VLA multimodal triggering mechanism; receive external commands according to the multimodal triggering mechanism, and dynamically combine standardized job units from the modular job skill library to generate executable job task sequence data;

[0023] S3. Based on the executable task sequence data, set the operation parameters for the multi-functional robotic arm to obtain operation parameter data; at the same time, collect operation environment information to generate operation environment data; combine the operation parameter data and operation environment data to construct test scenario data containing three dimensions: parameters, environment, and execution.

[0024] S4. Based on the test scenario data, start the automated test framework to simulate the operation process of the multi-functional robotic arm under disturbances such as force control deviation, positioning error, and load change, and collect the actual execution data during the test process; compare and analyze the actual execution data with the preset success standard and accuracy requirements to generate quantitative assessment data of skill reliability, which includes success rate data and accuracy maintenance capability data.

[0025] S5. Based on the quantitative assessment data of skill reliability, optimize and adjust the standardized operation units in the modular operation skill library to generate an optimized modular operation skill library; based on the optimized modular operation skill library, perform operation control of the multi-functional robotic arm and generate operation execution feedback data; continuously iterate and update the modular operation skill library based on the operation execution feedback data to form a verifiable, combinable, and evolvable intelligent skill ecosystem, providing intelligent skill infrastructure for general-purpose robots.

[0026] The working principle and effects of the above technical solution are as follows: By constructing a modular skill library through atomic decoupling, the reusability of skill units and the efficiency of task combination are greatly improved, reducing the skill development cycle and cost in new work scenarios; the multimodal triggering mechanism enhances the adaptability of the robotic arm to different external commands, avoiding the operational limitations caused by a single triggering method; the combination of three-dimensional testing scenarios and automated testing can accurately capture operational problems under disturbances, prevent operational failures caused by parameter mismatch with the environment, and significantly improve skill reliability; continuous iterative optimization allows the skill library to evolve continuously, which can not only adapt to diverse operational needs, but also ensure long-term stable operational accuracy, effectively improving the basic intelligent operation capabilities of general robots and improving the operational experience in practical applications.

[0027] In one embodiment of the present invention, S1 includes:

[0028] S11. Collect all the operation tasks of the multi-functional robotic arm in multiple scenarios, including industrial assembly, logistics sorting and precision operation, and generate the original dataset of operation tasks in multiple scenarios.

[0029] S12. Based on the Atomic Standard Unit (ASU) decoupling rule, the original dataset of multi-scenario task is decomposed into actions, logic, and priority, and standardized task unit data is generated. The standardized task unit data includes core information such as action parameters, execution logic, and safety thresholds.

[0030] S13. Based on standardized work unit data, construct a unified skill model architecture, define cross-scenario common skill interface protocols, parameter encapsulation formats and exception feedback handling mechanisms, and generate a unified skill model prototype.

[0031] S14. Conduct compatibility testing, redundancy optimization, and scalability upgrades on the unified skill model prototype, and integrate all standardized work unit data to form a modular work skill library. The modular work skill library supports dynamic addition and removal of units and rapid adaptation to scenarios.

[0032] The working principle and effects of the above technical solution are as follows: Full collection of multi-scenario task data ensures comprehensive coverage of the skill library, avoiding insufficient skill adaptation due to missing scenarios; standardized task units are generated through atomic decoupling, significantly improving unit reuse rate and reducing repetitive development work in different scenarios; a unified skill model architecture standardizes interface and parameter formats, enhancing cross-scenario adaptability and preventing integration obstacles caused by interface incompatibility; compatibility testing and optimization further improve the stability of the skill library, supporting dynamic addition and removal of units, enabling rapid response to new scenario requirements while reducing skill library maintenance costs.

[0033] In one embodiment of the present invention, S12 includes:

[0034] Based on the decoupling rules of atomic standard units, we formulate action decomposition dimensions, logic decomposition standards and priority judgment indicators, and generate ASU decoupling execution details.

[0035] Based on the ASU decoupling execution rules, the original dataset of multi-scenario job tasks is decomposed into the smallest action unit for each task, and independent action modules are separated to generate the original dataset of action units.

[0036] Logical correlation analysis is performed on the original dataset of action units to sort out the pre-dependencies and parallel collaboration relationships between action units, complete the logical decomposition, and generate a set of action units with logical correlations.

[0037] Based on the urgency of the task, the efficiency requirements, and the safety priority, the logically related action unit sets are classified into priority levels to generate action unit sets with priority labels;

[0038] Extract the core information of each unit in the action unit set with priority identification. The core information includes action parameters, execution logic and safety thresholds. Encapsulate the information in a standardized format to generate standardized job unit data.

[0039] The working principle and effects of the above technical solution are as follows: First, establish clear decoupling execution rules to ensure that action decomposition and logical breakdown are systematic and avoid arbitrariness and non-standardization in the decomposition process, thereby significantly improving the accuracy of task decomposition; decompose each task into the smallest action unit, making the function of each unit purer, significantly enhancing unit reusability, and reducing the amount of repetitive development work in different work scenarios; sort out the logical relationships between action units, clarify dependencies and collaboration relationships in advance, avoid logical confusion or conflict during subsequent combination, and improve the efficiency of work unit combination; classify priority levels and standardize the encapsulation of core information, so that work units have a clear execution orientation, which can not only ensure the priority execution of high-priority tasks and work safety, but also unify the unit format and reduce the integration cost of subsequent integration into the skill library.

[0040] In one embodiment of the present invention, S2 includes:

[0041] S21. Based on the interface specifications of the modular operation skill library, select and adapt the visual recognition module, voice command module and large model module. The visual recognition module includes 3D visual positioning and object feature extraction functions. The voice command module supports dialect recognition and noise environment adaptation. The large model module has intent understanding and scene reasoning capabilities, and generates a multimodal module adaptation list.

[0042] S22. Integrate and configure the visual recognition module, voice command module and large model module through the bus protocol, develop the data interaction channel between modules and the trigger priority determination logic, and realize the visual / voice / VLA multimodal triggering mechanism.

[0043] S23. Receive externally input job requirement instructions using a multimodal triggering mechanism. The job requirement instructions include visual images, voice instructions, text instructions, etc., and generate multimodal instruction raw data.

[0044] S24. The VLA module performs intent parsing, ambiguity resolution, and task decomposition on the raw data of multimodal instructions to generate standardized task requirement instructions.

[0045] S25. Based on standardized task requirement instructions, match suitable standardized task units from the modular task skill library, and generate executable task sequence data by serializing, scheduling in parallel, and avoiding conflicts through logical combination algorithms.

[0046] The working principle and effects of the above technical solution are as follows: Selecting suitable multimodal modules ensures precise matching between each module and the skill library interface, avoiding compatibility issues during integration and significantly reducing debugging costs during module integration; integrating configurations and building data interaction channels through bus protocols enables a multimodal triggering mechanism, significantly enhancing the robotic arm's responsiveness to different forms of external commands and avoiding operational limitations caused by single triggering methods; large-model parsing instructions effectively eliminates ambiguity, accurately decomposes tasks, prevents operational errors caused by misunderstandings of instructions, and improves the accuracy of task requirement transformation; matching and combining work units according to standardized instructions, combined with algorithms to avoid conflicts, can quickly generate executable task sequences, improving work planning efficiency, while ensuring the rigor of sequence logic and reducing the risk of failures in subsequent work execution.

[0047] In one embodiment of the present invention, S25 includes:

[0048] Based on standardized task requirement instructions, the core elements of the task are extracted, including job type, accuracy requirements, and execution timing constraints, and a list of unit matching conditions is generated.

[0049] Based on the unit matching condition list, the modular operation skill library is retrieved, standardized operation units with functional adaptability and parameter compatibility are selected, and a candidate operation unit set is generated.

[0050] Dependency analysis is performed on the candidate job unit set to identify the pre-execution requirements and parallel collaboration possibilities between units, and a draft of unit combination logic is generated.

[0051] The draft unit combination logic is input into the logic combination algorithm to complete the unit serial sorting and parallel task grouping. At the same time, related issues, including detecting and avoiding action timing conflicts and resource competition, are addressed to generate a draft of the job task sequence.

[0052] The initial draft of the task sequence is subjected to execution feasibility verification. The feasibility verification is used to match the hardware performance of the robotic arm and the constraints of the operation scenario, optimize and adjust the unit execution parameters and timing intervals, and generate executable task sequence data.

[0053] The working principle and effects of the above technical solution are as follows: Extracting core task elements clarifies the matching criteria, avoiding unit adaptation deviations caused by blind searching, and significantly improving the matching accuracy of standardized work units; searching and filtering candidate units with compatible functions and parameters reduces the time spent filtering invalid units, improving the early efficiency of work unit combination; clarifying the dependencies and collaboration possibilities between units, clarifying the execution logic in advance, and avoiding timing chaos or collaboration contradictions in subsequent combinations; using algorithms to sort and group units and avoid conflicts, and then optimizing parameters and timing through feasibility verification to prevent sequence execution problems caused by hardware incompatibility or scene mismatch; the entire process can quickly generate logically rigorous and highly adaptable work task sequences, while reducing adjustment costs during subsequent execution, significantly improving the efficiency and reliability of robotic arm operation planning.

[0054] In one embodiment of the present invention, S3 includes:

[0055] S31. Based on the executable task sequence data, extract the execution parameter requirements for each task unit. The execution parameter requirements include movement speed, force control threshold, and positioning accuracy requirements. Generate a parameter configuration requirement list.

[0056] S32. Combining the hardware performance parameters of the robotic arm and the constraints of the operation scenario, the list of parameter configuration requirements is checked for rationality and optimized to obtain accurate operation parameter data.

[0057] S33. Collect environmental information of the work site through environmental sensors, including temperature and humidity sensors, obstacle detectors, and light sensors, and generate work environment data, which includes environmental interference factors, spatial layout, and material status.

[0058] S34. Perform data fusion processing on the operation parameter data and the operation environment data. The data fusion processing includes data cleaning, format unification and weight allocation, and supplements the dynamic variable dimensions in the operation execution process. The dynamic variable dimensions include material loss and equipment fatigue coefficient.

[0059] S35. Based on the fusion of data from three dimensions—parameters, environment, and execution—a diversified test scenario data covering extreme working conditions, complex interference, and routine operations is constructed. The diversified test scenario data supports parameterized adjustment and dynamic generation of the scenarios.

[0060] The working principle and effects of the above technical solution are as follows: Extracting the parameter requirements of the work unit and combining them with hardware and scenario verification optimization makes the work parameters more closely match the actual execution conditions, avoiding work failures caused by unreasonable parameters and significantly improving the accuracy of parameter configuration; comprehensively collecting environmental information through sensors to supplement dynamic variable dimensions, making data coverage more complete and enhancing the realism and comprehensiveness of the test scenario; data fusion processing standardizes data format and improves data quality, avoiding the impact of data clutter or missing data on the scenario construction effect; constructing diversified test scenarios based on 3D fusion data covers both normal and extreme working conditions, and supports dynamic adjustment, preventing reliability assessment deviations caused by a single test scenario; the entire process not only provides realistic scenario support for subsequent reliability testing, but also reduces repeated adjustments in the test preparation stage, significantly improving test efficiency and evaluation accuracy.

[0061] In one embodiment of the present invention, S34 includes:

[0062] S341. Extract the core fields, data types, and numerical ranges of the operation parameter data and the operation environment data, and generate a feature comparison table of the two types of data.

[0063] S342. Based on the feature comparison table, perform outlier removal, missing value completion, and duplicate data removal on the two types of data to generate a clean dataset.

[0064] S343. Based on the robotic arm operation data specifications, the parameter data and environmental data in the clean data set are converted to unify units, align timestamps, and standardize field formats to generate a standardized fusion data source.

[0065] S344. Based on the key influencing factors of the task, set the weight coefficients of the task parameters and environmental data. For example, the weight of the positioning accuracy parameter is higher than the weight of the temperature and humidity environment. Generate a weight allocation table.

[0066] S345. Based on the weight allocation table, the standardized fusion data source is integrated using a weighted fusion algorithm to generate preliminary fusion data; dynamic variables in the operation execution process, such as material loss and equipment fatigue coefficient, are filtered, and real-time calculation data of the corresponding variables are collected and added to the preliminary fusion data to generate complete fusion data. The complete fusion data includes static parameters, environmental information and dynamic variables.

[0067] The working principle and effects of the above technical solution are as follows: It generates a data feature comparison table, providing a clear basis for processing operational parameters and environmental data, avoiding data chaos caused by blind processing, and significantly improving the standardization of data processing; it effectively improves data quality by cleaning up anomalies and filling in missing data, preventing inferior data from interfering with subsequent fusion results; it unifies data units and aligns timestamps to complete standardized conversion, reducing compatibility barriers when integrating data of different formats and improving data fusion efficiency; it sets weight coefficients based on key factors, making the fused data more aligned with the core needs of the operation and enhancing the data's relevance; it supplements dynamic variables to generate complete fused data, avoiding the omission of changes in the operation process by static data, and making the data coverage more comprehensive; the entire process not only provides accurate and reliable data support for the subsequent construction of test scenarios but also reduces repeated scenario construction due to data problems, significantly reducing the cost of the test preparation phase.

[0068] In one embodiment of the present invention, S345 includes:

[0069] Read the weight coefficients of each data dimension in the weight allocation table, combine them with the field attributes of the standardized fusion data source, determine the input parameters and calculation rules of the weighted fusion algorithm, and generate the algorithm configuration parameter set;

[0070] The standardized fusion data source and the algorithm configuration parameter set are input into the weighted fusion algorithm. The parameter data and environmental data are dimensionally aligned and weighted summation is performed to generate preliminary fusion data.

[0071] By combining the entire process of robotic arm operation, the types of dynamic variables that affect the operation performance are analyzed, and core dynamic variables are screened out. The core dynamic variables include material loss and equipment fatigue coefficient, and a dynamic variable screening list is generated.

[0072] Based on the dynamic variable screening list, real-time calculation data of each dynamic variable is obtained through sensor acquisition, equipment operation log analysis, etc., such as the remaining amount of material and the fatigue coefficient derived from the cumulative running time of the equipment, to generate a dynamic variable dataset.

[0073] The format and range of the dynamic variable dataset are validated to generate standardized dynamic variable data. The standardized dynamic variable data is then added to the preliminary fusion data to complete the dimensional integration of static parameters, environmental information and dynamic variables, generating complete fusion data containing three types of core information.

[0074] The working principle and effects of the above technical solution are as follows: First, the algorithm input parameters and calculation rules are clearly defined, making the weighted fusion systematic and avoiding parameter chaos during algorithm execution, thus significantly improving the accuracy of data fusion. Data dimension alignment and weighted calculation are completed according to weight coefficients, making the initial fused data more aligned with the core needs of the task, enhancing the data's relevance and effectiveness; core dynamic variables are screened to avoid interference from irrelevant variables, reduce data redundancy, and improve data processing efficiency; real-time data of dynamic variables is collected and verified to ensure the reliability of dynamic data and prevent inferior data from affecting the fusion results. Finally, static parameters, environmental information, and dynamic variables are integrated to make the fused data more comprehensive, avoiding the omission of real-time changes during the operation process by static data. The entire process generates high-quality, complete fused data, providing solid support for subsequent test scenario construction, and reduces repeated scenario adjustments due to incomplete data, significantly lowering test preparation costs.

[0075] In one embodiment of the present invention, step S4 includes:

[0076] S41. Based on diversified test scenario data, configure test rules for the automated test framework. The test rules include test process, data collection frequency, and termination conditions to generate customized test solutions.

[0077] S42. Start the automated testing framework to simulate typical disturbance scenarios, including force control deviation (±5%~±20%), positioning error (0.1mm~1mm), and load change (5%~30% of rated load), and drive the multi-functional robotic arm to execute the work task sequence;

[0078] S43. Through a high-precision data acquisition module, including a force sensor, a position encoder, and a vision tracking system, the actual execution data during the test process is collected in real time, such as the actual force control value, positioning deviation, and task completion time.

[0079] S44. Perform noise reduction, feature extraction, and standardization transformation on the actual execution data to generate standardized execution data that can be compared and analyzed; retrieve the preset job success criteria (e.g., task completion rate ≥ 95%) and accuracy requirements (e.g., positioning error ≤ 0.3mm) to construct a comparative analysis model;

[0080] S45. Input the standardized execution data into the comparative analysis model, perform success rate statistics, accuracy deviation calculation and stability assessment, and generate skill reliability quantitative assessment data. The skill reliability quantitative assessment data includes single unit success rate, sequence execution success rate and long-term accuracy maintenance capability data.

[0081] The working principle and effects of the above technical solution are as follows: Customized testing schemes based on diversified scenario configurations make testing rules more aligned with actual needs, avoiding evaluation biases caused by general testing schemes and significantly improving the relevance of testing; simulating typical disturbance scenarios to drive robotic arm operations comprehensively verifies performance under complex working conditions, avoiding incomplete reliability assessments caused by single conventional scenario testing; high-precision modules collect data in real time, ensuring the authenticity and integrity of actual execution data and reducing the interference of data distortion on evaluation results; noise reduction and standardization of data improve data quality, making comparative analysis more accurate; quantitative evaluation data is generated through comparative models, making skill reliability assessment more intuitive and objective, avoiding errors from subjective judgment; the entire process efficiently completes skill reliability verification and provides accurate basis for subsequent skill library optimization, significantly reducing the blindness of optimization adjustments.

[0082] In one embodiment of the present invention, S42 includes:

[0083] The customized test plan is loaded into the automated test framework, the association rules between the scene trigger nodes and the robotic arm actions in the test process are parsed, and the test execution configuration file is generated.

[0084] The parameter ranges and classification standards of three typical disturbances—force control deviation, positioning error, and load change—are analyzed, and a disturbance scenario parameter table is generated.

[0085] Based on the disturbance scenario parameter table, the scenario simulation module of the automated testing framework generates virtual disturbance environment data corresponding to the disturbance intensity. The virtual disturbance environment data includes force feedback simulation data, position offset data, and load change curves.

[0086] The virtual disturbance environment data is temporally correlated with the job task sequence to determine the timing and duration of each disturbance scenario, and to generate a task execution script with disturbance time sequence markers.

[0087] The execution engine of the automated testing framework is started, and the task execution script with disturbance timing mark is sent to the multi-functional robotic arm control system, driving the robotic arm to execute the task sequence according to the script and apply the preset disturbance simultaneously.

[0088] The working principle and effects of the above technical solution are as follows: It parses the test process to generate an execution configuration file, providing a clear basis for test execution, avoiding confusion between scenario triggering and robotic arm movements, and significantly improving the accuracy of test configuration; it organizes disturbance parameters to generate parameter tables, standardizing the intensity of different disturbances, avoiding arbitrariness in disturbance simulation, and enhancing the consistency and comparability of disturbance scenarios; it generates virtual disturbance environment data, accurately reproducing complex disturbance conditions of different intensities, avoiding the high cost and uncontrollable problems of actually building disturbance scenarios; it uses time-series correlation to determine the timing and duration of disturbance application, preventing test distortion caused by the disconnect between disturbance and task, and improving the realism of test scenarios; it automatically distributes scripts to synchronously apply disturbances, reducing errors caused by manual intervention, improving test execution efficiency, and comprehensively verifying the real-time response capability of the robotic arm under disturbance, avoiding omissions in testing key conditions, and providing more comprehensive actual execution data support for subsequent reliability assessment.

[0089] In one embodiment of the present invention, step S5 includes:

[0090] S51. Determine the threshold and locate the shortcomings in the quantitative assessment data of skill reliability, identify standardized operation units with low success rate and large accuracy fluctuation in the modular operation skill library, and generate an optimization target list.

[0091] S52. For the work units in the optimization target list, combined with the disturbance impact data during the test process, adjust the action parameter thresholds, optimize the execution logic, or add exception handling branches to generate unit optimization schemes.

[0092] S53. Apply the unit optimization scheme to the modular operation skill library, complete the unit iterative update, and generate the optimized modular operation skill library;

[0093] S54. Based on the optimized modular operation skill library, issue operation control commands to the multi-functional robotic arm to drive it to perform operation tasks in the actual scene and collect operation execution feedback data, including operation completion quality, execution efficiency, failure rate, etc.

[0094] S55. Establish a correlation mapping mechanism between job execution feedback data and skill library optimization, and set an iterative update threshold, such as triggering a second optimization if the failure rate is ≥3%;

[0095] S56. Based on the association mapping mechanism and iterative update threshold, continuously adjust and add standardized operation units in the modular operation skill library to generate an iteratively upgraded modular operation skill library; through multiple rounds of iteration, form a verifiable (supports full-scenario testing and verification), combinable (supports cross-task unit recombination), and evolvable (supports autonomous optimization and upgrading) intelligent skill ecosystem.

[0096] The working principle and effects of the above technical solution are as follows: By accurately locating the bottleneck units in the skill library through threshold determination, blind optimization work is avoided, significantly improving the targeting and efficiency of skill library optimization; targeted optimization schemes are formulated based on perturbation data, adjusting parameters, optimizing logic, or supplementing abnormal branches, effectively improving the success rate and accuracy stability of the working units and reducing the occurrence of failures in actual operations; feedback data is collected and verified in actual scenarios, making the optimization direction more aligned with real application needs and avoiding ineffective adjustments caused by optimization deviating from reality; an association mapping mechanism and iterative thresholds are established to achieve continuous dynamic upgrading of the skill library, enhancing its adaptability and evolution; the resulting intelligent skill ecosystem can provide standardized and highly reliable skill support for general-purpose robots, adapt to the evolution of diverse working scenarios, and significantly reduce subsequent skill development and maintenance costs.

[0097] One embodiment of the present invention provides a modular operational skill library and reliability testing system for a multifunctional robotic arm, comprising:

[0098] One or more processors;

[0099] Memory, used to store one or more programs;

[0100] Wherein, when the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method described in any one of the above.

[0101] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A modular operational skill library and reliability testing method for a multi-functional robotic arm, characterized in that, The method includes: S1. Decoupling the multi-functional robotic arm's tasks into atomized standard units to generate standardized task unit data; constructing a unified skill model based on the standardized task unit data to form a modular task skill library; S2. Configuring a visual recognition module, a voice command module, and a large model module based on the modular task skill library to implement a vision / voice / VLA multimodal triggering mechanism; receiving external commands based on the multimodal triggering mechanism and dynamically combining standardized task units from the modular task skill library to generate executable task sequence data; S3. Setting task parameters for the multi-functional robotic arm based on the executable task sequence data to obtain task parameter data; simultaneously collecting task environment information to generate task environment data; combining task parameter data and task environment data to construct test scenario data containing three dimensions: parameters, environment, and execution; S4. Based on the test scenario data, starting an automated testing framework to simulate the multi-functional robotic arm's operation process under disturbance, collecting actual execution data during the test process; comparing and analyzing the actual execution data with preset success standards and accuracy requirements to generate skill data. The process involves several steps: S5, optimizing and adjusting standardized work units in the modular work skill library based on the quantitative assessment data of skill reliability, generating an optimized modular work skill library; controlling the operation of the multi-functional robotic arm based on the optimized modular work skill library, and generating work execution feedback data; continuously iterating and updating the modular work skill library based on the work execution feedback data; S4, which includes: S41, configuring test rules for the automated test framework based on diversified test scenario data, and generating a customized test plan; S42, starting the automated test framework, simulating typical disturbance scenarios, and driving the multi-functional robotic arm to execute a sequence of work tasks; S43, collecting actual execution data in real time during the test process through a high-precision data acquisition module; S44, performing noise reduction, feature extraction, and standardization conversion on the actual execution data to generate standardized execution data that can be compared and analyzed; retrieving preset work success standards and accuracy requirements to construct a comparative analysis model; and S45, inputting the standardized execution data into the comparative analysis model to perform success rate statistics, accuracy deviation calculation, and stability evaluation, generating quantitative assessment data of skill reliability.

2. The modular operation skill library and reliability testing method for a multifunctional robotic arm according to claim 1, characterized in that, S1 includes: S11, collecting all the operation tasks of the multi-functional robotic arm in multiple scenarios to generate a raw dataset of operation tasks in multiple scenarios; S12, based on the atomic standard unit decoupling rules, performing action decomposition, logical decomposition, and priority division on the raw dataset of operation tasks in multiple scenarios to generate standardized operation unit data; S13, based on the standardized operation unit data, constructing a unified skill model architecture, defining a cross-scenario common skill interface protocol, parameter encapsulation format, and anomaly feedback handling mechanism to generate a unified skill model prototype; S14, performing compatibility testing, redundancy optimization, and scalability upgrades on the unified skill model prototype, and integrating all standardized operation unit data to form a modular operation skill library.

3. The modular operation skill library and reliability testing method for a multifunctional robotic arm according to claim 2, characterized in that, S12 includes: based on the atomic standard unit decoupling rules, formulating action decomposition dimensions, logical decomposition standards, and priority judgment indicators to generate ASU decoupling execution rules; based on the ASU decoupling execution rules, decomposing the original dataset of multi-scenario job tasks into the smallest action units for each task, separating independent action modules, and generating the original dataset of action units; performing logical association analysis on the original dataset of action units, sorting out the pre-dependencies and parallel collaboration relationships between action units, completing logical decomposition, and generating a set of action units with logical associations; combining the urgency, execution efficiency requirements, and safety priority of the job tasks, classifying the set of action units with logical associations into priority levels, and generating a set of action units with priority labels; extracting the core information of each unit in the set of action units with priority labels, encapsulating it in a standardized format, and generating standardized job unit data.

4. The modular operation skill library and reliability testing method for a multifunctional robotic arm according to claim 1, characterized in that, S2 includes: S21, based on the interface specifications of the modular job skill library, selecting suitable visual recognition modules, voice command modules, and large model modules to generate a multimodal module adaptation list; S22, integrating and configuring the visual recognition modules, voice command modules, and large model modules through a bus protocol, developing data interaction channels between modules and trigger priority determination logic to realize a visual / voice / VLA multimodal triggering mechanism; S23, using the multimodal triggering mechanism to receive externally input job requirement instructions and generate multimodal instruction raw data; S24, using the large model module to perform intent parsing, ambiguity elimination, and task decomposition on the multimodal instruction raw data to generate standardized task requirement instructions; S25, based on the standardized task requirement instructions, matching suitable standardized job units from the modular job skill library, and using a logical combination algorithm to perform unit concatenation, parallel scheduling, and conflict avoidance to generate executable job task sequence data.

5. The modular operation skill library and reliability testing method for a multi-functional robotic arm according to claim 1, characterized in that, S3 includes: S31, extracting the execution parameter requirements of each work unit based on the executable task sequence data, and generating a parameter configuration requirement list; S32, combining the hardware performance parameters of the robotic arm and the constraints of the work scenario, performing rationality verification and optimization adjustment on the parameter configuration requirement list to obtain accurate work parameter data; S33, collecting environmental information of the work site through environmental sensors to generate work environment data; S34, performing data fusion processing on the work parameter data and work environment data to supplement the dynamic variable dimension in the work execution process; S35, constructing diversified test scenario data covering extreme working conditions, complex interference, and routine operations based on the fused data of parameters, environment, and execution.

6. The modular operation skill library and reliability testing method for a multifunctional robotic arm according to claim 5, characterized in that, S34 includes: S341, extracting the core fields, data types, and numerical ranges of the operation parameter data and the operation environment data, and generating a feature comparison table for the two types of data; S342, based on the feature comparison table, performing outlier removal, missing value completion, and duplicate data removal on the two types of data, generating a clean data set; S343, based on the robotic arm operation data specifications, performing unit unification, timestamp alignment, and field format standardization conversion on the parameter data and environmental data in the clean data set, generating a standardized fusion data source; S344, combining the key influencing factors of the operation task, setting weight coefficients for the operation parameters and environmental data, and generating a weight allocation table; S345, based on the weight allocation table, integrating the standardized fusion data source through a weighted fusion algorithm, generating preliminary fusion data; filtering dynamic variables during the operation execution process, collecting real-time calculation data of the corresponding variables, and supplementing them into the preliminary fusion data to generate complete fusion data.

7. The modular operation skill library and reliability testing method for a multifunctional robotic arm according to claim 6, characterized in that, S345 includes: reading the weight coefficients of each data dimension in the weight allocation table, combining the field attributes of the standardized fusion data source, determining the input parameters and calculation rules of the weighted fusion algorithm, and generating an algorithm configuration parameter set; inputting the standardized fusion data source and the algorithm configuration parameter set into the weighted fusion algorithm, performing dimension alignment and weighted summation calculation on the parameter data and environmental data, and generating preliminary fusion data; combining the entire robotic arm operation process, analyzing the types of dynamic variables affecting the operation performance, filtering out core dynamic variables, and generating a dynamic variable screening list; based on the dynamic variable screening list, obtaining real-time calculation data of each dynamic variable through sensor acquisition and equipment operation log parsing, and generating a dynamic variable dataset; performing format verification and range rationality verification on the dynamic variable dataset, and generating standardized dynamic variable data; supplementing the standardized dynamic variable data into the preliminary fusion data, completing the dimensional integration of static parameters, environmental information, and dynamic variables, and generating complete fusion data containing three types of core information.

8. The modular operation skill library and reliability testing method for a multi-functional robotic arm according to claim 1, characterized in that, S5 includes: S51, determining thresholds and identifying shortcomings in the quantitative assessment data of skill reliability, identifying standardized work units with low success rates and large accuracy fluctuations in the modular work skill library, and generating an optimization target list; S52, for the work units in the optimization target list, adjusting the action parameter thresholds, optimizing the execution logic, or supplementing the abnormal handling branches based on the disturbance impact data during the testing process, generating a unit optimization scheme; S53, applying the unit optimization scheme to the modular work skill library, completing the unit iterative update, and generating an optimized modular work skill library; S54, based on the optimized modular work skill library, issuing work control commands to the multi-functional robotic arm, driving it to perform work tasks in the actual scenario, and collecting work execution feedback data; S55, establishing a correlation mapping mechanism between work execution feedback data and skill library optimization, and setting an iterative update threshold; S56, continuously dynamically adjusting and adding new standardized work units in the modular work skill library according to the correlation mapping mechanism and the iterative update threshold, generating an iteratively upgraded modular work skill library; forming an intelligent skill ecosystem through multiple rounds of iteration.

9. A modular operational skill library and reliability testing system for a multi-functional robotic arm, comprising: One or more processors; A memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method of any one of claims 1 to 8.

Images