A six-joint robot pose correlation knowledge graph construction method

Abstract

The application discloses a kind of six-joint robot pose association knowledge graph construction methods, comprising: first, according to the ontology structure of robot and the relationship between the concept of modal parameter such as robot pose, analysis robot knowledge graph mode layer, secondly, the modal parameter identification of robot is obtained by the method of OMA and it is according to the mode layer that has been built and imported into Neo4j graphics database software to build graphical data layer, finally, the data layer that has been filled with data is fused, and the data layer of knowledge graph is automatically updated.Pattern layer and data layer jointly constitute the knowledge graph.The application has the advantages that: based on the constructed knowledge graph, a variety of query, analysis, prediction, recommendation algorithm can be realized in the field.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot intelligent manufacturing, and particularly relates to a six-joint robot pose correlation knowledge graph construction method. BACKGROUND

[0002] Robot milling: large curved surface components are widely used in aviation, aerospace and marine industries, and robot processing has the characteristics of flexible structure, large operation space and fast reconfiguration compared with machine tool processing, which can reach inaccessible areas of complex parts and is expected to become an important processing means in the field of large and complex component processing. Robot milling is still considered an open problem. One of the main problems is the inherent weak stiffness of the robot serial structure. The relatively low pose correlation structural dynamic stiffness of the robot often reduces the robot processing precision and surface integrity. The structural stiffness is spatial pose-dependent within the working range, and the stiffness characteristics are more sensitive to the spatial pose, especially when the robot is at the limit pose of the boundary region of the working range. The high spatial dimension characteristics of the pose correlation of the robot, the existing conventional finite discrete measurement method, cannot be implemented for the acquisition of the structural stiffness characteristics of the robot within the working range. For example, for a 6-DOF robot, each joint angle motion interval is roughly divided into 10 parts, and there are 1000000 poses for 6 joints. However, even within the robot working range space, the determination of such a large number of pose points, the robot structural stiffness information associated with these poses, will make the structural stiffness partition granularity too large because of the large interval ten-part division in the pose space, so even such a large data set cannot meet the demand for the acquisition of the structural stiffness characteristics of the continuous robot within the working space. At the same time, for such a large data set, the existing data storage method cannot achieve a reasonable structured storage and fast query.

[0003] OMA (Operational Modal Analysis) is a method of identifying the modal characteristics of the structure by using the excitation generated by the reciprocating motion of the robot during actual processing to obtain vibration data, without the need for conventional modal experiments such as hammering and excitation to identify the modal, which can ensure the automatic identification of the structural dynamics parameters associated with any pose within the robot working space.

[0004] Knowledge graph (Knowledge Graph) is an important branch technology of artificial intelligence, which is a structured semantic knowledge base for describing concepts and their relationships in the physical world in symbolic form. Its basic composition unit is a "entity-relation-entity" triple, as well as an entity and its related attribute-value pair. Entities are connected to each other through relationships, forming a network-like knowledge structure. Data is stored in the knowledge graph through entities and attribute values to form graph structure data, forming a graph database. This graph database has the advantages of data search and auxiliary big data analysis. SUMMARY

[0005] The present application aims at the defects of the prior art, and provides a six-joint robot pose correlation knowledge graph construction method, which graphically stores and automatically updates a large amount of data generated by the robot in a high-dimensional space, so as to facilitate subsequent data query, reasoning and processing pose recommendation.

[0006] In order to achieve the above application purposes, the technical solutions adopted by the present application are as follows:

[0007] A six-joint robot pose correlation knowledge graph construction method comprises the following steps:

[0008] S1. Constructing a mode layer of a robot knowledge graph according to the relationship between the robot pose and the modal parameter concept based on the ontology structure of the robot;

[0009] S2. Obtaining structured data by modal parameter identification of the robot through the OMA method and importing the data into a Neo4j graphical database software according to the mode layer that has been constructed to construct a graphical data layer.

[0010] S3. Knowledge fusion is performed on the data layer that has been filled with data, fusion rules are established, including entity linking, quality evaluation, knowledge updating, redundant fusion and error deletion, and the data layer of the knowledge graph is automatically updated.

[0011] Further, the specific steps of S1 are as follows:

[0012] S11. Analyzing the robot pose correlation dynamics in a top-down manner, determining the entities as the robot, the rod, the joint angle, the angle parameter, the modal parameter, the natural frequency and the damping ratio, and taking them as the nodes of the graph.

[0013] S12. Determining the relationship between the entities and taking it as the edge of the graph.

[0014] S13. Connecting all entities in the form of node-edge-node to construct the mode layer of the graph.

[0015] Further, the specific steps of S2 are as follows: the following method is used to construct the data layer:

[0016] S21. Identifying the modal parameters of the robot at different poses by the OMA method.

[0017] S22. Preprocessing the joint rotation angle and its corresponding modal parameter into a set of structured data.

[0018] S23. Use Neo4j graph database to convert a set of structured data according to the schema layer into a graph structured data, that is, a knowledge graph block. A plurality of knowledge graph blocks jointly constitute a knowledge graph data layer.

[0019] Further, the specific steps of S3 are as follows:

[0020] S31. Entity linking: link the newly generated node data with the existing data, and aggregate the nodes of the same sample together.

[0021] S32. Quality evaluation: evaluate the nodes that have been aggregated through fusion rules, and merge the nodes with the same attribute values into the same node.

[0022] S33. Knowledge update: use the merge instruction to create the nodes that need to be redundantly fused during each data import, and dynamically update the data layer.

[0023] Compared with the prior art, the advantages of the present application are:

[0024] (1) The pose associated high-dimensional data of the six-joint robot can be stored in a graph structure;

[0025] (2) The automatic update and construction of the data layer can be realized;

[0026] (3) The knowledge graph of other forms of robots can be constructed, including but not limited to the six-joint robot;

[0027] (4) Strong scalability, a variety of query, analysis, prediction, recommendation algorithms can be realized in the field based on the knowledge graph. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 is a flowchart of the six-joint robot pose associated knowledge graph construction method of the embodiment of the present application;

[0029] Figure 2 is a schematic diagram of the robot structure provided by the embodiment of the present application;

[0030] Figure 3 is a schematic diagram of the knowledge graph schema layer provided by the embodiment of the present application;

[0031] Figure 4 is a schematic diagram of the structured data provided by the embodiment of the present application;

[0032] Figure 5 is a schematic diagram of the structured data file format conversion provided by the embodiment of the present application;

[0033] Figure 6 is a schematic diagram of the data layer construction code provided by the embodiment of the present application;

[0034] Figure 7 is a visual vector diagram of a knowledge graph data layer provided by an embodiment of the present application;

[0035] Figure 8 is a visual vector diagram of a data layer after fusion provided by an embodiment of the present application;

[0036] Figure 9 is an actual construction flowchart provided by an embodiment of the present application. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical scheme and advantages of the present application more clear and explicit, the present application is further described in detail below according to the drawings and embodiments.

[0038] The present application takes ABB 6660 robot as the implementation object, as shown in the figure, the specific experimental steps are as follows: Figure 1

[0039] 1. According to the body structure of the robot and the related theoretical analysis of the relationship between the concepts of robot pose, modal parameters and other concepts, the mode layer of the robot knowledge graph is constructed;

[0040] (1) The top-down method is used to analyze the dynamics characteristics related to the robot pose, and the robot structure is as shown in the figure, the entities of the robot are determined as robot, rod, joint angle, angle parameter, modal parameter, natural frequency and damping ratio, which are taken as the nodes of the graph. Figure 2

[0041] (2) The relationship between the entities is determined as the edge of the graph, and the entity relationship is expressed by triplets and divided into two kinds, one is attribute value, and the other is object attribute. Attribute value is used to assign value to entity, indicating the internal attribute of entity such as (10Hz, value, first order natural frequency), and the corresponding two nodes are usually an entity and a string. While the object attribute is used to represent the external association between two entities, such as (first order natural frequency, subclass, natural frequency), and the corresponding two nodes of the triplet are usually two entities.

[0042] (3) All entities are connected in the form of node-edge-node to construct the mode layer of the graph, as shown in the figure. Figure 3

[0043] 2. The modal parameters of the robot are identified by the method of OMA to obtain structured data, and the structured data is imported into the Neo4j graph database software according to the mode layer built to construct the graphical data layer.

[0044] ​​​(1) The method using OMA identifies the modal parameters of the robot in different poses. The ABB robot is programmed and controlled by RobotStudio to determine the space traversal mode and self-excitation mode; the acceleration sensor is arranged on the robot, and the collected data is analyzed by the LMS vibration data acquisition and analysis platform to obtain the pose-related modal parameters.

[0045] (2) The joint rotation angle and the corresponding modal parameter are preprocessed as a set of structured data. The data processed by LMS is exported, and the modal data and the joint rotation angle data of the robot are preprocessed as a plurality of sets of structured data on the computer. According to the data batch import data format rule of the Neo4j database, the data is edited into Excel, as shown in Figure 4 , the first line is the class name of the node, and each line below is a set of structured data. The workbook file (.xlsx) is exported as a comma-separated value (.csv) file. As shown in Figure 5 .

[0046] (3) A set of structured data is converted into a graph structured data according to the mode layer using the Neo4j graph database. A plurality of sets of graph structured data collectively constitute a knowledge graph data layer. The conversion process uses the Cypher declarative graph query language used by the Neo4j database, and the code is written according to the mode layer, as shown in Figure 6 . The data layer is visualized as shown in Figure 7 .

[0047] 3. There are a large number of same redundant nodes (such as the same joint rotation angle and the same dynamic characteristics) in the automatically filled data layer. Fusion rules are established, including entity linking, quality evaluation, knowledge updating, redundant fusion, and implementation through the database instruction Merge().

[0048] (1) Entity linking: link the newly generated node data with the existing data, and the database will judge the same sample nodes and aggregate them together.

[0049] (2) Quality evaluation: The nodes after aggregation are evaluated by the fusion rule, and the nodes with the same attribute value are merged into the same node. As shown in Figure 8 , the specific fusion rule depends on which node is created using the merge instruction, and the fusion rule adopted in this embodiment is shown in Figure 6 specific code.

[0050] (3) Knowledge updating: In each data import, the nodes that need to be redundantly fused are created using the merge instruction, and the data layer is dynamically updated.

[0051] The actual overall process of the embodiment of the application is shown in Figure 9 .

[0052] This embodiment characterizes the dynamics characteristics of ABB 6660 robot using knowledge graph and uses Neo4j graph database as data layer. Graph database is a NoSQL database, which supports the storage and visualization of the relationship between data compared with traditional database, and has higher response speed suitable for real-time data access, which is conducive to the relationship storage of massive high-dimensional pose data and the selection of optimal machining pose of robots with the same end position, and lays the foundation for subsequent big data analysis based on graph neural network.

[0053] Those skilled in the art will appreciate that the embodiments described herein are presented for the purpose of helping the reader understand the implementation method of the present application, and should be understood as the protection scope of the present application is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations according to the technical inspiration disclosed in the present application without departing from the essence of the present application, and these modifications and combinations are still within the protection scope of the present application.

Claims

1.A method for constructing a pose association knowledge graph of a six-joint robot, characterized in that, Comprising the following steps: S1. Analyzing the relationship between the robot pose, modal parameter concept according to the body structure of the robot to build the mode layer of the robot knowledge graph; The specific steps of S1 are as follows: S11. Analyze the robot pose associated with the dynamic characteristics in a top-down manner, determine the entity as the robot, rod, joint angle, angle parameter, modal parameter, natural frequency, damping ratio, and take it as the node of the graph; S12. Determine the relationship between entities and take it as the edge of the graph; S13. Connect all entities in the form of node-edge-node to build the mode layer of the graph; S2. Obtain structured data by modal parameter identification of the robot through the OMA method and import it into the Neo4j graph database software according to the mode layer built to build the graphical data layer; The specific steps of S2 are as follows: S21. Identify the modal parameters of the robot at different poses using the OMA method; S22. Preprocess the joint rotation angle and its corresponding modal parameters into a set of structured data; S23. Convert a set of structured data into a graph structured data according to the mode layer using the Neo4j graph database, which is a knowledge graph block; multiple knowledge graph blocks together constitute the knowledge graph data layer; S3. Knowledge fusion is performed on the data layer filled with data, fusion rules are established, including entity linking, quality evaluation, knowledge updating, redundant fusion and error deletion, and the data layer of the knowledge graph is automatically updated. 2.The method of claim 1, wherein: The specific steps of S3 are as follows: S31. Entity linking: link the newly generated node data with the existing data, and aggregate the nodes of the same sample together; S32. Quality evaluation: evaluate the nodes after aggregation through fusion rules, and merge the nodes with the same attribute value into the same node; S33. Knowledge update: use the merge instruction to create the nodes that need to be redundantly fused at each data import, and dynamically update the data layer.

Images