Robot adaptive learning system and method based on big data model

Abstract

The invention discloses a robot adaptive learning system and method based on a big data model, and relates to the technical field of big data analysis. Historical successful record data is extracted from a robot task execution recording system, a unique identifier is generated, an operation angle sub-reference is calculated, and a robot operation parameter reference is constructed; historical success record subsets of different scenes and different operation object materials are constructed respectively, corresponding influence coefficients are calculated, and a first influence coefficient library and a second influence coefficient library are constructed; the system collects real-time scene and material information, calls an influence coefficient to correct an operation parameter benchmark and outputs real-time operation parameters, and comprises a historical data extraction module, an operation parameter benchmark construction module, a first influence coefficient construction module, a second influence coefficient construction module and a real-time parameter correction module. Scene and material changes can be adapted in real time, and the operation stability and safety of the robot in an unstructured scene are improved.

Description

Technical Field

[0001] This invention relates to the field of big data analytics, specifically to a robot adaptive learning system and method based on big data models. Background Technology

[0002] Currently, with the continuous evolution of technology and the increasing demand for applications, the application scenarios of robot systems are becoming more and more complex. Robots have gradually moved away from structured industrial scenarios and towards complex and dynamic human-machine collaborative scenarios. Traditional robot control methods are difficult to adapt to the operation tasks in unstructured scenarios, which not only leads to a decline in their own performance, but may also cause system failure in severe cases, thereby affecting the overall control effect and creating safety hazards. Summary of the Invention

[0003] The purpose of this invention is to provide a robot adaptive learning system and method based on big data models to solve the problems raised in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a robot adaptive learning method based on a big data model, the robot adaptive learning method comprising the following steps: Step S1: Extract historical success record data from the robot's task execution record system and generate a unique identifier for each data entry. The historical success record data includes the scene, the material of the operation object, the volume of the operation object, and the robot's operation parameters. Step S2: Based on the historical successful record data, construct a baseline for robot operation parameters by calculation; Step S3: Construct a subset of historical success records under different scenarios, calculate and store the first influence coefficient of each scenario relative to the general benchmark, and construct a first influence coefficient library; Step S4: Construct a subset of historical success records for different operation object materials, calculate and store the second influence coefficient of each material relative to the general benchmark, and construct a second influence coefficient library; Step S5: Collect real-time scene and object material data, and use the first and second influence coefficient libraries to correct the robot operation parameter benchmarks in real time.

[0005] Furthermore, the specific steps of step S1 are as follows: Historical successful records are extracted from the robot's task execution record system, and a unique identifier is generated for each historical successful record. The unique identifier is randomly generated by the system. The material of the operation object includes density, physical structure and surface characteristics. The volume of the operation object is generated by the robot's preset iDP3 algorithm. The robot operation parameters include operation angle and robot actuator force.

[0006] Furthermore, the specific steps of step S2 are as follows: Step S2-1: Extract the operation parameters from the historical successful record data. Based on the operation angles in the operation parameters, sum all the operation angles and divide the sum by the total number of records in the historical successful record data to obtain the first robot operation parameter sub-benchmark. Step S2-2: Calculate the mass of the operation object based on the volume of the operation object in the historical successful record data. The method for calculating the mass of the operation object is to multiply the volume of the operation object by the density of the operation object material, and then bind the mass of the operation object to the corresponding unique identifier. Step S2-3: Extract all robot actuator forces from historically successfully recorded data, and associate them with the mass of the operated object based on the unique identifier to form a robot parameter set; Step S2-4: For the robot parameter set, divide the robot actuator force by the mass of the associated operation object, denoted as P, bind P with the corresponding unique identifier, sum all P values, and then divide the sum by the total number of historical successful records as the second robot operation parameter sub-benchmark. Step S2-5: Construct robot operation parameter benchmarks, which include a first robot operation parameter sub-benchmark and a second robot operation parameter sub-benchmark.

[0007] Furthermore, the specific steps of step S3 are as follows: Step S3-1: From the historical successful record data, retain historical successful records that are consistent with the scenario to form a subset of historical successful records under a certain scenario; Step S3-2: For a subset of historical successful records in a certain scenario, extract all its operation angles by unique identifier, perform summation, and then divide the summation result by the total number of historical successful records in the subset of a certain scenario, denoted as M. Divide M by the robot operation parameter benchmark, denoted as the first influence coefficient. Step S3-3: Calculate the subset of historical success records for all scenarios according to S3-2 to form the first influence coefficient library.

[0008] Furthermore, the specific steps of step S4 are as follows; Step S4-1: From the historical successful record data, retain historical successful records that are consistent with the material of the operation object, forming a subset of historical successful records under a certain material of the operation object; Step S4-2: For a subset of historical successful records under a certain operation object material, extract all of its P by unique identifier, perform summation, and then divide the summation result by the total number of historical successful records in a certain scenario, denoted as N. Divide N by the robot operation parameter benchmark, denoted as the second influence coefficient. Step S4-3: Calculate the subset of historical successful records under the material of all objects to form the second influence coefficient library according to S4-2.

[0009] Furthermore, the specific steps of step S5 are as follows: Step S5-1: Collect the real-time scene and real-time operation object material under the current operation scenario; Step S5-2: Based on the real-time scenario, query the corresponding first influence coefficient from the first influence coefficient library, and based on the material of the operation object, query the corresponding second influence coefficient from the second influence coefficient library; Step S5-3: Extract the robot operation parameter benchmark, use the first influence coefficient to correct the operation angle to form the first correction result, use the second influence coefficient to correct the robot actuator force and multiply it by the mass of the operated object to form the second correction result, and output the robot operation parameters for this operation based on the first correction result and the second correction result.

[0010] Furthermore, the robot adaptive learning system includes a historical data extraction module, an operation parameter benchmark construction module, a first influence coefficient construction module, a second influence coefficient construction module, and a real-time parameter correction module; The historical data extraction module is used to extract historical successful record data from the robot task execution record system and generate a unique identifier for each data entry; the operation parameter benchmark construction module is used to calculate the operation angle sub-benchmark, the quality of the operation object, and the P-value based on the historical successful record data, and integrate the sub-benchmarks to construct the robot operation parameter benchmark; the first influence coefficient construction module is used to filter a subset with consistent scenarios from the historical successful record data, calculate the first influence coefficient corresponding to the scenario, and construct a first influence coefficient library; the second influence coefficient module is used to filter a subset with consistent materials from the historical successful record data, calculate the second influence coefficient corresponding to the material, and construct a second influence coefficient library; the real-time parameter correction module is used to collect the current operation scenario and object material information, query the corresponding influence coefficients, and correct the robot operation parameter benchmark to output the operation parameters; The output of the historical data extraction module is electrically connected to the input of the operation parameter benchmark construction module; the output of the historical data extraction module is electrically connected to the input of the first influence coefficient construction module; the output of the historical data extraction module is electrically connected to the input of the second influence coefficient construction module; the output of the operation parameter benchmark construction module is electrically connected to the input of the real-time parameter correction module; the output of the first influence coefficient construction module is electrically connected to the input of the real-time parameter correction module; and the output of the second influence coefficient construction module is electrically connected to the input of the real-time parameter correction module.

[0011] Furthermore, the historical data extraction module includes a historical data extraction unit and a unique identifier generation unit; The historical record extraction unit is used to extract historical successful record data from the robot task execution record system; the unique identifier generation unit is used to generate a unique identifier for each historical successful record data.

[0012] Furthermore, the operating parameter benchmark construction module includes a parameter calculation unit and a benchmark integration unit; The parameter calculation unit is used to calculate the operating angle sub-reference, the quality of the operating object, and the P-value based on historical successful record data.

[0013] The reference integration unit is used to integrate the first robot operation parameter sub-reference and the second robot operation parameter sub-reference to construct the robot operation parameter reference. The first influence coefficient construction module includes a scenario subset construction unit and a first coefficient library generation unit; The scenario subset construction unit is used to filter records with consistent scenarios from historical successful record data to form a subset, and calculate the M value corresponding to the subset; the first coefficient library generation unit is used to calculate the first influence coefficient corresponding to each scenario, and construct the first influence coefficient library.

[0014] Furthermore, the second influence coefficient construction module includes a material subset construction unit and a second coefficient library generation unit; The material subset construction unit is used to filter records with consistent materials from historical successful record data to form a subset, and calculate the N value corresponding to the subset; the second coefficient library generation unit is used to calculate the second influence coefficient corresponding to each material, and construct the second influence coefficient library; The real-time parameter correction module includes a real-time information acquisition unit and a parameter correction output unit; The real-time information acquisition unit is used to acquire the real-time scene and the material of the real-time operation object under the current operation scenario; the parameter correction output unit is used to correct the robot operation parameter benchmark with the first influence coefficient and the second influence coefficient, and output the robot operation parameters for this operation.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention improves the rationality and stability of the initial setting of robot operating parameters by extracting historical successful data of the robot and generating a unique identifier, and by combining the operating angle, object mass and actuator force to construct an operating parameter benchmark. The benchmark construction method with historical successful data as the core improves the rationality and stability of the initial setting of robot operating parameters.

[0016] 2. By constructing a first influence coefficient library related to the scene and a second influence coefficient library related to the material, this invention can accurately quantify the influence of different scenes and different operation objects and materials on operation parameters, enabling the robot to call the corresponding coefficients for specific scenes and materials, improving the pertinence of parameter adjustment, avoiding performance degradation or system failure due to insufficient adaptation to working conditions, and reducing safety hazards.

[0017] 3. This invention solves the problem that traditional methods are difficult to adapt to unstructured scenes by collecting real-time data on the current scene and the material of the object being operated on, calling the corresponding influence coefficient to correct the benchmark of the operating parameters, and outputting operating parameters that are adapted to the real-time working conditions. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the robot adaptive learning method based on a big data model according to the present invention. Figure 2 This is a schematic diagram of the robot adaptive learning system based on a big data model according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1: As Figure 1 As shown, this invention provides a technical solution: a robot adaptive learning method based on a big data model. The robot adaptive learning method includes the following steps: Step S1: Extract historical success record data from the robot's task execution record system and generate a unique identifier for each data entry. The historical success record data includes the scene, the material of the operation object, the volume of the operation object, and the robot's operation parameters. The specific steps of step S1 are as follows: Historical successful record data is extracted from the robot's task execution record system, and a unique identifier is generated for each historical successful record data. The unique identifier is randomly generated by the system. The material of the operation object includes density, physical structure and surface characteristics. The volume of the operation object is generated by the robot's preset iDP3 algorithm. The robot operation parameters include operation angle and robot actuator force. Step S2: Based on the historical successful record data, construct a baseline for robot operation parameters by calculation; The specific steps of step S2 are as follows: Step S2-1: Extract the operation parameters from the historical successful record data. Based on the operation angles in the operation parameters, sum all the operation angles and divide the sum by the total number of records in the historical successful record data to obtain the first robot operation parameter sub-benchmark. Step S2-2: Calculate the mass of the operation object based on the volume of the operation object in the historical successful record data. The method for calculating the mass of the operation object is to multiply the volume of the operation object by the density of the operation object material, and then bind the mass of the operation object to the corresponding unique identifier. Step S2-3: Extract all robot actuator forces from historically successfully recorded data, and associate them with the mass of the operated object based on the unique identifier to form a robot parameter set; Step S2-4: For the robot parameter set, divide the robot actuator force by the mass of the associated operation object, denoted as P, bind P with the corresponding unique identifier, sum all P values, and then divide the sum by the total number of historical successful records as the second robot operation parameter sub-benchmark. Step S2-5: Construct robot operation parameter benchmarks, which include a first robot operation parameter sub-benchmark and a second robot operation parameter sub-benchmark; Step S3: Construct a subset of historical success records under different scenarios, calculate and store the first influence coefficient of each scenario relative to the general benchmark, and construct a first influence coefficient library; The specific steps of step S3 are as follows: Step S3-1: From the historical successful record data, retain historical successful records that are consistent with the scenario to form a subset of historical successful records under a certain scenario; Step S3-2: For a subset of historical successful records in a certain scenario, extract all its operation angles by unique identifier, perform summation, and then divide the summation result by the total number of historical successful records in the subset of a certain scenario, denoted as M. Divide M by the robot operation parameter benchmark, denoted as the first influence coefficient. Step S3-3: Calculate the subset of historical success records for all scenarios according to S3-2 to form the first influence coefficient library; Step S4: Construct a subset of historical success records for different operation object materials, calculate and store the second influence coefficient of each material relative to the general benchmark, and construct a second influence coefficient library; The specific steps of step S4 are as follows; Step S4-1: From the historical successful record data, retain historical successful records that are consistent with the material of the operation object, forming a subset of historical successful records under a certain material of the operation object; Step S4-2: For a subset of historical successful records under a certain operation object material, extract all of its P by unique identifier, perform summation, and then divide the summation result by the total number of historical successful records in a certain scenario, denoted as N. Divide N by the robot operation parameter benchmark, denoted as the second influence coefficient. Step S4-3: Calculate the subset of historical successful records under all materials of a certain operation object according to S4-2 to form the second influence coefficient library; Step S5: Collect real-time scene and object material data, and correct the robot operation parameter benchmark in real time using the first and second influence coefficient libraries; The specific steps of step S5 are as follows: Step S5-1: Collect the real-time scene and real-time operation object material under the current operation scenario; Step S5-2: Based on the real-time scenario, query the corresponding first influence coefficient from the first influence coefficient library, and based on the material of the operation object, query the corresponding second influence coefficient from the second influence coefficient library; Step S5-3: Extract the robot operation parameter benchmark, use the first influence coefficient to correct the operation angle to form the first correction result, use the second influence coefficient to correct the robot actuator force and multiply it by the mass of the operated object to form the second correction result, and output the robot operation parameters for this operation based on the first correction result and the second correction result.

[0021] For example: Extract historical records of successful sorting without component damage from the robot task execution record system; the system randomly generates a 12-bit unique identifier, and binds a unique identifier to each historical successful record.

[0022] For example: In a typical assembly workshop, the object is made of ABS plastic, and its volume is 6cm³, generated by the robot's preset iDP3 algorithm. The robot's operating parameters are an operating angle of 45° and an actuator force of 12N. In a cleanroom precision workshop, the object is made of aluminum alloy, and its volume is 8cm³, generated by the iDP3 algorithm. The robot's operating parameters are an operating angle of 50° and an actuator force of 25N. In another typical assembly workshop, the object is made of aluminum alloy, and its volume is 8cm³, generated by the iDP3 algorithm. The robot's operating parameters are an operating angle of 48° and an actuator force of 23N. The dimensions of the remaining historically successfully recorded data are consistent with the above examples, including the scene, object material, object volume, and robot operating parameters.

[0023] Extract robot operation parameters from historical successful records, focusing on the operation angle parameter. Sum all operation angles to obtain the total angle. Then divide the sum by the total number of historical successful records to calculate the first robot operation parameter sub-benchmark as 48°.

[0024] Based on the volume of the operation object in the historical successful record data, and combined with the density of the operation object's material, the mass of the operation object is calculated. Specifically, the operation object's volume is multiplied by its density. If the operation object's volume is 6 cm³ and its density is 1.1 g / cm³, its mass is 6 cm³ × 1.1 g / cm³ = 6.6 g. The calculated mass of the operation object for each record is then bound to its corresponding unique identifier.

[0025] Extract all robot actuator forces from historical successful records, associate them with the mass of the corresponding operated object based on unique identifiers, and form a set of robot parameters.

[0026] For the robot parameter set, the robot actuator force in each record is divided by the associated mass of the manipulated object to obtain a value P, and P is then bound to a corresponding unique identifier. For example, the P value is 12N ÷ 6.6g ≈ 1.82N / g, and the P values ​​for the remaining records are calculated in the same way. All P values ​​are summed to obtain a total P value of 168N / g. This summation is then divided by the total number of historical successful records to calculate the second robot operation parameter sub-benchmark as 1.4N / g.

[0027] By integrating the first robot operation parameter sub-reference and the second robot operation parameter sub-reference, a robot operation parameter reference is constructed. Finally, it is determined that the reference includes the operation angle reference of 48° and the P value reference of 1.4N / g.

[0028] From historical success records, records consistent with the specified scenario are retained, forming subsets of historical success records for different scenarios. Specifically, records from ordinary assembly workshops constitute the subset of historical success records for ordinary assembly workshops; records from cleanroom precision workshops constitute the subset of historical success records for cleanroom precision workshops.

[0029] For a subset of historical successful records in a typical assembly workshop scenario, the operation angles of all records in the subset are extracted by a unique identifier. These operation angles are summed to obtain the total angle. The sum is then divided by the total number of records in the subset to obtain the average angle M1, which is 49°. M1 is then divided by the operation angle benchmark of 48° in the robot operation parameter benchmark to calculate the first influence coefficient K1 corresponding to the typical assembly workshop scenario, which is approximately 1.02.

[0030] For a subset of historical successful records in a cleanroom precision workshop setting, the operating angles of all records in the subset are extracted by unique identifiers, summed to obtain the total angle, divided by the total number of records in the subset, and the average angle M2 is 46°. Then, M2 is divided by the operating angle benchmark of 48°, and the first influence coefficient K2 corresponding to the cleanroom precision workshop setting is calculated to be approximately 0.96.

[0031] S3-3: Constructing the First Influence Coefficient Library The first influence coefficient library contains first influence coefficients for two scenarios: 1.02 for the ordinary assembly workshop scenario and 0.96 for the cleanroom precision workshop scenario.

[0032] From the historical successful operation data, records with consistent material are retained based on the material of the operation object, forming subsets of historical successful operation records under different materials. Specifically, records with ABS plastic shells constitute the subset of historical successful operation records under ABS plastic shells; records with aluminum alloy metal shells constitute the subset of historical successful operation records under aluminum alloy metal shells.

[0033] For the subset of historical successful records under the ABS plastic shell material, the P-values ​​of all records in the subset are extracted by unique identifier. These P-values ​​are summed to obtain the total P-value. The sum is then divided by the total number of records in the subset to obtain the average P-value N1. Then, N1 is divided by the P-value benchmark of 1.4 N / g in the robot operation parameter benchmark to calculate the second influence coefficient K3 corresponding to the ABS plastic shell material as 1.0.

[0034] For the subset of historical successful records under the aluminum alloy metal shell material, the P value of all records in the subset is extracted by the unique identifier, and the sum of the P values ​​is obtained. Dividing the sum of the P values ​​by the total number of records in the subset, the average P value N2 is 1.26 N / g. Then, N2 is divided by the P value benchmark of 1.4 N / g, and the second influence coefficient K4 corresponding to the aluminum alloy metal shell material is calculated to be 0.9.

[0035] The second influence coefficient library contains second influence coefficients for two materials, with the second influence coefficient for ABS plastic shell material being 1.0 and the second influence coefficient for aluminum alloy metal shell material being 0.9.

[0036] The robot uses a pre-set visual sensor to collect real-time scene data and the material of the object being operated on. The real-time scene is identified as a cleanroom precision workshop, and the material of the object being operated on is an aluminum alloy metal shell. The robot's pre-set iDP3 algorithm calculates the real-time volume of the object to be operated on as 7.5 cm³.

[0037] Based on the collected real-time scenario being a cleanroom precision workshop, the corresponding first influence coefficient K2, calculated to be 0.96, was found in the first influence coefficient database. Based on the collected real-time operation object material being an aluminum alloy metal shell, the corresponding second influence coefficient K4, calculated to be 0.9, was found in the second influence coefficient database.

[0038] The established robot operation parameters are extracted as the operating angle of 48° and the P-value of 1.4 N / g. The operating angle is corrected using a first influence coefficient K2, calculated by multiplying the operating angle reference by K2 (48° × 0.96), resulting in a first correction of 46.08°. The robot actuator force is corrected using a second influence coefficient K4. First, the mass of the manipulated object is calculated as 7.5 cm³ × 2.7 g / cm³ = 20.25 g, based on its volume of 7.5 cm³ and material density of 2.7 g / cm³. Then, the P-value is multiplied by K4 to obtain a corrected P-value of 1.4 N / g × 0.9 = 1.26 N / g. Finally, the corrected P-value is multiplied by the object mass (1.26 N / g × 20.25 g) to obtain a second correction of approximately 25.52 N. Based on the first and second correction results, the robot operation parameters are output as an operating angle of 46.08° and an actuator force of 25.52 N. Example 2, as Figure 2 As shown, the present invention provides a robot adaptive learning system based on a big data model. The robot adaptive learning system includes a historical data extraction module, an operation parameter benchmark construction module, a first influence coefficient construction module, a second influence coefficient construction module, and a real-time parameter correction module. The historical data extraction module is used to extract historical successful record data from the robot task execution record system and generate a unique identifier for each data entry; the operation parameter benchmark construction module is used to calculate the operation angle sub-benchmark, the quality of the operation object, and the P-value based on the historical successful record data, and integrate the sub-benchmarks to construct the robot operation parameter benchmark; the first influence coefficient construction module is used to filter a subset with consistent scenarios from the historical successful record data, calculate the first influence coefficient corresponding to the scenario, and construct a first influence coefficient library; the second influence coefficient module is used to filter a subset with consistent materials from the historical successful record data, calculate the second influence coefficient corresponding to the material, and construct a second influence coefficient library; the real-time parameter correction module is used to collect the current operation scenario and object material information, query the corresponding influence coefficients, and correct the robot operation parameter benchmark to output the operation parameters; The output of the historical data extraction module is electrically connected to the input of the operation parameter benchmark construction module; the output of the historical data extraction module is electrically connected to the input of the first influence coefficient construction module; the output of the historical data extraction module is electrically connected to the input of the second influence coefficient construction module; the output of the operation parameter benchmark construction module is electrically connected to the input of the real-time parameter correction module; the output of the first influence coefficient construction module is electrically connected to the input of the real-time parameter correction module; and the output of the second influence coefficient construction module is electrically connected to the input of the real-time parameter correction module.

[0039] The historical data extraction module includes a historical data extraction unit and a unique identifier generation unit; The historical record extraction unit is used to extract historical successful record data from the robot task execution record system; the unique identifier generation unit is used to generate a unique identifier for each historical successful record data.

[0040] The operational parameter benchmark construction module includes a parameter calculation unit and a benchmark integration unit; The parameter calculation unit is used to calculate the operating angle sub-reference, the quality of the operating object, and the P-value based on historical successful record data.

[0041] The reference integration unit is used to integrate the first robot operation parameter sub-reference and the second robot operation parameter sub-reference to construct the robot operation parameter reference. The first influence coefficient construction module includes a scenario subset construction unit and a first coefficient library generation unit; The scenario subset construction unit is used to filter records with consistent scenarios from historical successful record data to form a subset, and calculate the M value corresponding to the subset; the first coefficient library generation unit is used to calculate the first influence coefficient corresponding to each scenario, and construct the first influence coefficient library.

[0042] The second influence coefficient construction module includes a material subset construction unit and a second coefficient library generation unit; The material subset construction unit is used to filter records with consistent materials from historical successful record data to form a subset, and calculate the N value corresponding to the subset; the second coefficient library generation unit is used to calculate the second influence coefficient corresponding to each material, and construct the second influence coefficient library; The real-time parameter correction module includes a real-time information acquisition unit and a parameter correction output unit; The real-time information acquisition unit is used to acquire the real-time scene and the material of the real-time operation object under the current operation scenario; the parameter correction output unit is used to correct the robot operation parameter benchmark with the first influence coefficient and the second influence coefficient, and output the robot operation parameters for this operation.

[0043] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A robot adaptive learning method based on a big data model, characterized by: The robot adaptive learning method includes the following steps: Step S1: Extract historical success record data from the robot's task execution record system and generate a unique identifier for each data entry. The historical success record data includes the scene, the material of the operation object, the volume of the operation object, and the robot's operation parameters. Step S2: Based on the historical successful record data, construct a baseline for robot operation parameters by calculation; Step S3: Construct a subset of historical success records under different scenarios, calculate and store the first influence coefficient of each scenario relative to the general benchmark, and construct a first influence coefficient library; Step S4: Construct a subset of historical success records for different operation object materials, calculate and store the second influence coefficient of each material relative to the general benchmark, and construct a second influence coefficient library; Step S5: Collect real-time scene and object material data, and use the first and second influence coefficient libraries to correct the robot operation parameter benchmarks in real time.

2. The robot adaptive learning method based on a big data model according to claim 1, characterized in that: The specific steps of step S1 are as follows: Historical successful records are extracted from the robot's task execution record system, and a unique identifier is generated for each historical successful record. The unique identifier is randomly generated by the system. The material of the operation object includes density, physical structure and surface characteristics. The volume of the operation object is generated by the robot's preset iDP3 algorithm. The robot operation parameters include operation angle and robot actuator force.

3. The robot adaptive learning method based on a big data model according to claim 1, characterized in that: The specific steps of step S2 are as follows: Step S2-1: Extract the operation parameters from the historical successful record data. Based on the operation angles in the operation parameters, sum all the operation angles and divide the sum by the total number of records in the historical successful record data to obtain the first robot operation parameter sub-benchmark. Step S2-2: Calculate the mass of the operation object based on the volume of the operation object in the historical successful record data. The method for calculating the mass of the operation object is to multiply the volume of the operation object by the density of the operation object material, and then bind the mass of the operation object to the corresponding unique identifier. Step S2-3: Extract all robot actuator forces from historically successfully recorded data, and associate them with the mass of the operated object based on the unique identifier to form a robot parameter set; Step S2-4: For the robot parameter set, divide the robot actuator force by the mass of the associated operation object, denoted as P, bind P with the corresponding unique identifier, sum all P values, and then divide the sum by the total number of historical successful records as the second robot operation parameter sub-benchmark. Step S2-5: Construct robot operation parameter benchmarks, which include a first robot operation parameter sub-benchmark and a second robot operation parameter sub-benchmark.

4. The robot adaptive learning method based on a big data model according to claim 1, characterized in that: The specific steps of step S3 are as follows: Step S3-1: From the historical successful record data, retain historical successful records that are consistent with the scenario to form a subset of historical successful records under a certain scenario; Step S3-2: For a subset of historical successful records in a certain scenario, extract all its operation angles by unique identifier, perform summation, and then divide the summation result by the total number of historical successful records in the subset of a certain scenario, denoted as M. Divide M by the robot operation parameter benchmark, denoted as the first influence coefficient. Step S3-3: Calculate the subset of historical success records for all scenarios according to S3-2 to form the first influence coefficient library.

5. The robot adaptive learning method based on a big data model according to claim 1, characterized in that: The specific steps of step S4 are as follows; Step S4-1: From the historical successful record data, retain historical successful records that are consistent with the material of the operation object, forming a subset of historical successful records under a certain material of the operation object; Step S4-2: For a subset of historical successful records under a certain operation object material, extract all of its P by unique identifier, perform summation, and then divide the summation result by the total number of historical successful records in a certain scenario, denoted as N. Divide N by the robot operation parameter benchmark, denoted as the second influence coefficient. Step S4-3: Calculate the subset of historical successful records under the material of all objects to form the second influence coefficient library according to S4-2.

6. The robot adaptive learning method based on a big data model according to claim 1, characterized in that: The specific steps of step S5 are as follows: Step S5-1: Collect the real-time scene and real-time operation object material under the current operation scenario; Step S5-2: Based on the real-time scenario, query the corresponding first influence coefficient from the first influence coefficient library, and based on the material of the operation object, query the corresponding second influence coefficient from the second influence coefficient library; Step S5-3: Extract the robot operation parameter benchmark, use the first influence coefficient to correct the operation angle to form the first correction result, use the second influence coefficient to correct the robot actuator force and multiply it by the mass of the operated object to form the second correction result, and output the robot operation parameters for this operation based on the first correction result and the second correction result.

7. A robot adaptive learning system based on a big data model, applied to the robot adaptive learning method based on a big data model as described in any one of claims 1-6, characterized in that: The robot adaptive learning system includes a historical data extraction module, an operation parameter benchmark construction module, a first influence coefficient construction module, a second influence coefficient construction module, and a real-time parameter correction module; The historical data extraction module is used to extract historical successful record data from the robot task execution record system and generate a unique identifier for each data entry; the operation parameter benchmark construction module is used to calculate the operation angle sub-benchmark, the quality of the operation object, and the P-value based on the historical successful record data, and integrate the sub-benchmarks to construct the robot operation parameter benchmark; the first influence coefficient construction module is used to filter a subset of historical successful record data with consistent scenarios, calculate the first influence coefficient corresponding to the scenario, and construct a first influence coefficient library; the second influence coefficient module is used to filter a subset of historical successful record data with consistent materials, calculate the second influence coefficient corresponding to the material, and construct a second influence coefficient library; The real-time parameter correction module is used to collect the current operation scenario and object material information, query the corresponding influence coefficient, and correct the robot operation parameter benchmark to output the operation parameters. The output of the historical data extraction module is electrically connected to the input of the operation parameter benchmark construction module; the output of the historical data extraction module is electrically connected to the input of the first influence coefficient construction module; the output of the historical data extraction module is electrically connected to the input of the second influence coefficient construction module; the output of the operation parameter benchmark construction module is electrically connected to the input of the real-time parameter correction module; the output of the first influence coefficient construction module is electrically connected to the input of the real-time parameter correction module; and the output of the second influence coefficient construction module is electrically connected to the input of the real-time parameter correction module.

8. The robot adaptive learning system based on a big data model according to claim 7, characterized in that: The historical data extraction module includes a historical data extraction unit and a unique identifier generation unit; The historical record extraction unit is used to extract historical successful record data from the robot task execution record system; the unique identifier generation unit is used to generate a unique identifier for each historical successful record data.

9. The robot adaptive learning system based on a big data model according to claim 7, characterized in that: The operational parameter benchmark construction module includes a parameter calculation unit and a benchmark integration unit; The parameter calculation unit is used to calculate the operation angle sub-reference, the quality of the operation object, and the P-value based on historical successful record data. The reference integration unit is used to integrate the first robot operation parameter sub-reference and the second robot operation parameter sub-reference to construct the robot operation parameter reference. The first influence coefficient construction module includes a scenario subset construction unit and a first coefficient library generation unit; The scenario subset construction unit is used to filter records with consistent scenarios from historical successful record data to form a subset, and to calculate the M value corresponding to the subset; The first coefficient library generation unit is used to calculate the first influence coefficient corresponding to each scenario and construct the first influence coefficient library.

10. The robot adaptive learning system based on a big data model according to claim 7, characterized in that: The second influence coefficient construction module includes a material subset construction unit and a second coefficient library generation unit; The material subset construction unit is used to filter records with consistent materials from historical successful record data to form a subset, and calculate the N value corresponding to the subset; the second coefficient library generation unit is used to calculate the second influence coefficient corresponding to each material, and construct the second influence coefficient library. The real-time parameter correction module includes a real-time information acquisition unit and a parameter correction output unit; The real-time information acquisition unit is used to acquire the real-time scene and the material of the real-time operation object under the current operation scenario; the parameter correction output unit is used to correct the robot operation parameter benchmark with the first influence coefficient and the second influence coefficient, and output the robot operation parameters for this operation.

Images