A method for assembly path planning

Abstract

The present invention relates to the field of artificial intelligence-assisted design technology, and in particular to an assembly path planning method, including: receiving an assembly path planning instruction; processing a three-dimensional assembly digital model, distinguishing parts in the three-dimensional assembly digital model by using different colors; extracting assembly scenes; planning an assembly path for parts of specific colors; performing detailed design of assembly path planning; controlling the three-dimensional assembly digital model to move according to detailed information of assembly path planning; updating the three-dimensional assembly digital model; extracting assembly scenes; and evaluating the result after the assembly path moves. Through this method, the problem of low accuracy of the planned assembly path in the prior art can be effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence-assisted design, and in particular to an assembly path planning method. Background Art

[0002] With the emergence of pre-trained natural language processing models based on the Transformer architecture, such as ChatGPT, it is possible to complete a variety of tasks in the form of questions and answers, such as answering technical questions, translation, summarization, generating text, etc. Through multiple rounds of questions and answers with the generative big model, it is possible to have an interactive discussion on a certain technical issue to obtain better results. At the same time, the generative big model further has multimodal capabilities, can understand and process images, videos, etc., and can plan and recommend assembly paths based on assembly scene pictures.

[0003] Limited by the capabilities of the generative large model itself, the existing technology can only give a rough assembly path direction when using the generative large model for assembly path planning. At the same time, the planned assembly path has a low accuracy rate, and collisions and interferences may occur during the assembly process. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes an assembly path planning method, which effectively solves the problem that when the prior art uses a generative large model for assembly path planning, it can only give a rough assembly path direction, and the planned assembly path has a low accuracy rate, and collision and interference may occur during the assembly process.

[0005] The present invention is achieved by adopting the following technical solutions:

[0006] An assembly path planning method comprises the following steps:

[0007] Step S1. Receiving assembly path planning instructions;

[0008] Step S2. Processing the three-dimensional assembly digital model, adding a coordinate system to the three-dimensional assembly digital model, and distinguishing the parts in the three-dimensional assembly digital model by different colors;

[0009] Step S3. Extracting the assembly scene;

[0010] Step S4. Calling the generative large model API interface, according to the assembly scene obtained in step S3, converting the object names in the assembly path planning instructions into color-coded names, and performing assembly path planning with colored parts;

[0011] Step S5. Calling the generative large model API interface, and performing detailed design of assembly path planning according to the assembly path planning information output in step S4;

[0012] Step S6. Generate control instructions based on the detailed assembly path planning information output in step S5 to control the three-dimensional assembly digital model to move according to the detailed assembly path planning information;

[0013] Step S7. According to the processing result of step S6, updating the three-dimensional assembly digital model;

[0014] Step S8. Extracting the assembly scene;

[0015] Step S9. Call the generative large model API to evaluate the result after the assembly path is moved; determine whether the verification is completed. If so, store the assembly path information and end the process; if not, determine whether an abnormality occurs. If no abnormality occurs, go to step S6 to continue processing. If an abnormality occurs, go to step S4 to optimize and update based on the current path.

[0016] It also includes generating a color matching database to match the names and colors of parts in the three-dimensional assembly digital model. When planning an assembly path with colored parts, the object names in the assembly path planning instructions are converted into color-coded names by retrieving the color matching database.

[0017] Extracting assembly scenes specifically refers to: converting the 3D assembly digital model into pictures and videos that can be understood by the generative large model, capturing views from several angles from the 3D assembly digital model, and recording a video of the 3D assembly digital model rotating around the coordinate axis.

[0018] The size of the coordinate system is consistent with the maximum three-dimensional size of the assembly scene.

[0019] In step S4, during the assembly path planning process, a search is performed from a knowledge base, in which relevant background knowledge and information about the three-dimensional assembly scene are pre-stored.

[0020] The assembly path planning information outputted in step S4 includes a corresponding distance of movement around the coordinate axis and / or a corresponding angle of rotation around the coordinate axis.

[0021] The assembly path planning refinement information is 1 / N of the assembly path planning information; wherein N is a natural number greater than 2.

[0022] The abnormal situations include interference and exceeding the safety distance.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0024] 1. The present invention uses a color matching method to accurately specify assembly path planning objects and improve the accuracy of assembly path planning in complex scenarios.

[0025] 2. The present invention extracts assembly scenes, provides multimodal data input for the generative large model, uses the generative large model to assist in assembly path planning, and further improves the accuracy of assembly path planning.

[0026] 3. The present invention includes two-level assembly path planning, namely, assembly path planning and assembly path planning refinement design, which can improve the accuracy of the assembly path planning process.

[0027] 4. The present invention also gradually evaluates the assembly path planning results to improve the accuracy of the assembly path planning process. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, wherein:

[0029] Figure 1 It is a schematic diagram of the process of the present invention;

[0030] Figure 2 is a structural diagram of the present invention;

[0031] Figure 3 It is the original three-dimensional assembly model diagram in the present invention;

[0032] Figure 4 It is a three-dimensional assembly model diagram after color processing in the present invention;

[0033] Figure 5 A three-dimensional assembly model diagram after the assembly path is moved in the present invention;

[0034] Figure 6 A three-dimensional assembly model diagram for evaluating abnormalities after the assembly path is moved in the present invention;

[0035] Figure 7 The three-dimensional assembly model diagram of the assembly path movement after correction in the present invention;

[0036] Figure 8 A three-dimensional assembly model diagram after the assembly path in the present invention has been moved;

[0037] Markings in the figure:

[0038] 1. Rod-shaped object, 2. Hole, 3. Mounting plate. DETAILED DESCRIPTION

[0039] Example 1

[0040] As a basic embodiment of the present invention, the present invention includes an assembly path planning method, comprising the following steps:

[0041] Step S1. Receive assembly path planning instructions.

[0042] Step S2. Processing the three-dimensional assembly digital model: adding a coordinate system to the three-dimensional assembly digital model, and distinguishing the components in the three-dimensional assembly digital model by using different colors.

[0043] Step S3: Extract assembly scenes.

[0044] Step S4. Call the generative large model API interface, convert the object names in the assembly path planning instructions into color-coded names according to the assembly scenario obtained in step S3, perform assembly path planning with colored parts, and generate assembly path planning information.

[0045] Step S5. Call the generative large model API interface, perform assembly path planning refinement design according to the assembly path planning information output in step S4, and obtain assembly path planning refinement information.

[0046] Step S6. Generate control instructions based on the assembly path planning refinement information outputted in step S5, and control the three-dimensional assembly digital model to move according to the assembly path planning refinement information.

[0047] Step S7. Update the three-dimensional assembly digital model according to the processing result of step S6.

[0048] Step S8: Extract assembly scene.

[0049] Step S9. Call the generative large model API to evaluate the result after the assembly path is moved. Determine whether the verification is completed. If it is completed, store the assembly path information and the process ends. If the verification is not completed, determine whether an abnormality occurs. If no abnormality occurs, go to step S6 to continue processing. If an abnormality occurs, go to step S4 to optimize and update based on the current path.

[0050] Example 2

[0051] As a preferred embodiment of the present invention, the present invention includes an assembly path planning method, comprising the following steps:

[0052] Step S1. Receive assembly path planning instructions.

[0053] Step S2. Processing the three-dimensional assembly digital model: adding a coordinate system to the three-dimensional assembly digital model, and distinguishing the components in the three-dimensional assembly digital model by using different colors.

[0054] Step S3. Generate a color matching database to match the names and colors of the components in the three-dimensional assembly digital model.

[0055] Step S4. Extract assembly scene: The 3D assembly digital model is converted into pictures and videos that can be understood by the generative large model, and views of several angles are captured from the 3D assembly digital model, and a video of the 3D assembly digital model rotating around a specific coordinate axis is recorded. The specific coordinate axis refers to one of the X axis, Y axis, and Z axis in the coordinate axis. Alternatively, a video of the 3D assembly digital model rotating around the X axis, Y axis, and Z axis of the coordinate axis is recorded in sequence.

[0056] Step S5. Call the generative large model API interface, according to the assembly scene obtained in step S4, convert the object name in the assembly path planning instruction into a color coding name by searching the color matching database, perform assembly path planning with color parts, and obtain assembly path planning information.

[0057] Step S6. Call the generative large model API interface, perform assembly path planning refinement design according to the assembly path planning information output in step S5, and obtain assembly path planning refinement information. The assembly path planning refinement information is 1 / N of the assembly path planning information; wherein N is a natural number greater than 2.

[0058] Step S7: Generate control instructions based on the assembly path planning refinement information outputted in step S6 to control the three-dimensional assembly digital model to move according to the assembly path planning refinement information.

[0059] Step S8. Update the three-dimensional assembly digital model according to the processing result of step S7.

[0060] Step S9: Extracting the assembly scene.

[0061] Step S 10 . Call the generative big model API to evaluate the result after the assembly path is moved. Determine whether the verification is completed. If it is completed, store the assembly path information and the process ends. If the verification is not completed, determine whether an abnormality occurs. If no abnormality occurs, go to step S7 to continue processing. If an abnormality occurs, go to step S5 to optimize and update based on the current path.

[0062] Example 3

[0063] As the best embodiment of the present invention, the present invention includes an assembly path planning method, referring to the attached specification Figure 2 , is based on a three-dimensional assembly digital model, an assembly scene extraction module, a color matching database, a knowledge base, a first decision module, a second decision module, an assembly path verification module, and an assembly path storage module. For more details, refer to the attached manual. Figure 1 , the assembly path planning method comprises the following steps:

[0064] Step S1. Receive assembly path planning instructions.

[0065] Step S2. Three-dimensional assembly digital model processing. The three-dimensional assembly digital model is used to display specific scenarios that require assembly path planning. A coordinate system is added to the three-dimensional assembly digital model. The size of the coordinate system is consistent with the maximum three-dimensional size of the assembly scene. At the same time, the parts in the three-dimensional assembly digital model are distinguished by different colors. Each part is assigned an independent and non-repetitive color, which is encoded according to the RGB color gamut 24-bit color. A total of 256*256*256 (166777216) different colors can be represented, which can meet the needs of distinguishing and displaying different parts in the assembly scene.

[0066] Step S3. Generate a color matching database. Generate a color matching database according to the color processing result of the three-dimensional assembly digital model, and match the names and colors of the parts in the three-dimensional assembly digital model. The colors are encoded using RGB 256 bits, and the names of the parts and colors are matched one by one. The color matching database can be used for retrieval by the first decision module when planning the assembly path.

[0067] Step S4. Extract the assembly scene based on the assembly scene extraction module. Convert the 3D assembly digital model into pictures and videos that can be understood by the generative large model, capture several angles of view from the 3D assembly digital model, and record the video of the 3D assembly digital model rotating around a specific coordinate axis as the input of the first decision module.

[0068] Step S5. The first decision module processes. The generative large model API interface is called, and according to the assembly scene obtained in step S4, i.e., the images and videos of each view of the three-dimensional assembly digital model, the object names in the assembly path planning instructions are converted into color-coded names by searching the color matching database, and assembly path planning with color parts is performed to obtain assembly path planning information. During the assembly path planning process, a search is performed from the knowledge base to obtain the three-dimensional assembly scene information in order to better perform assembly path planning. The knowledge base stores relevant background knowledge and information about the three-dimensional assembly scene.

[0069] The assembly path planning information given by the first decision module is expressed in the following format:

[0070] ①Move a specific distance around a specific axis;

[0071] ②Rotate a specific angle around a specific axis.

[0072] The specific axis refers to one of the X axis, Y axis and Z axis in the coordinate axis. The specific distance refers to the distance that the three-dimensional assembly digital model needs to move, and the specific angle refers to the angle that the three-dimensional assembly digital model needs to rotate.

[0073] and combinations of the two.

[0074] Step S6. Processing by the second decision module: Call the generative large model API interface, perform assembly path planning refinement design according to the assembly path planning information output by the first decision module in step S5, subdivide the assembly path planning information output by the first decision module into shorter steps, and obtain assembly path planning refinement information.

[0075] Typically, the assembly path planning refinement information in the second decision module is 1 / N of the assembly path planning information in the first decision module, where N is a natural number greater than 2.

[0076] The assembly path planning refinement information given by the second decision module is expressed in the following format:

[0077] ①Move a specific distance around a specific axis;

[0078] ②Rotate a specific angle around a specific axis.

[0079] and combinations of the two.

[0080] Step S7. Assembly path verification module processing. According to the assembly path planning refinement information output by the second decision module in step S6, a control instruction is generated to control the three-dimensional assembly digital model to move according to the assembly path planning refinement information. Only one path movement is performed each time, and the movement content can be:

[0081] ①Move a specific distance around a specific axis;

[0082] ②Rotate a specific angle around a specific axis.

[0083] and combinations of the two.

[0084] Step S8: Update the three-dimensional assembly digital model information. According to the processing result of the assembly path verification module in step S7, update the three-dimensional assembly digital model.

[0085] Step S9: Extracting the assembly scene.

[0086] Step S 10 . The second decision module processes. Call the generative large model API, and the second decision module evaluates the result after the assembly path moves. Determine whether the verification is completed. If it is completed, store the assembly path information and the process ends. If the verification is not completed, determine whether there are abnormal situations such as collision interference, exceeding the safety distance, etc. If there are no abnormalities, go to step S7 to continue processing. If there are abnormalities, go to step S5 and require the first decision module to optimize and update based on the current path. Among them, storing the assembly path information can be implemented based on the assembly path storage module.

[0087] Example 4

[0088] As another preferred embodiment of the present invention, the present invention includes an assembly path planning method. Taking the assembly path planning process of a rod-shaped object 1 as an example, the method includes the following steps:

[0089] Step S1. Receive assembly path planning instructions. Figure 3 In this embodiment, the rod-shaped object 1 is moved through the hole 2 on the mounting plate 3 .

[0090] Step S2. 3D assembly digital model processing. A coordinate system is added to the 3D assembly digital model. The size of the coordinate system is consistent with the maximum 3D size of the assembly scene. At the same time, the parts in the 3D assembly digital model are distinguished by different colors. Each part is assigned an independent and non-repetitive color and encoded according to the RGB color gamut 24-bit color. In this embodiment, the rod 1 is set to yellow, the RGB color code is #FFFF00, and the mounting plate 3 is set to blue, the RGB color code is #0000FF.

[0091] Step S3. Generate a color matching database, the database storage information is shown in Table 1 below:

[0092] Table 1

[0093]

[0094] Step S4. Extract the assembly scene, convert the 3D assembly model into pictures and videos that can be understood by the generative large model, capture several angles of view from the 3D assembly model, and record the video of the 3D assembly model rotating around a specific coordinate axis, as shown in the attached manual. Figure 4 shown.

[0095] Step S5. Call the generative large model API interface, and convert the object names in the assembly path planning instructions into color-coded names by searching the color matching database according to the view images and videos of the three-dimensional assembly digital model, and perform assembly path planning with color parts to obtain assembly path planning information. During the assembly path planning process, the knowledge base is searched to obtain the three-dimensional assembly scene information for better assembly path planning.

[0096] The converted assembly instruction information is: move the object with RGB color code #FFFF00 through hole 2 with color code #0000FF. In this way, objects in the assembly scene can be accurately specified to avoid ambiguity.

[0097] In this embodiment, the assembly path planning information is as follows:

[0098] ①Translate 10mm along the positive direction of the Y axis;

[0099] ②Rotate 75° around the Y axis;

[0100] ③Translate 30mm along the positive direction of the Y axis.

[0101] Step S6. Call the generative large model API interface, perform assembly path planning refinement design according to the assembly path planning information output in step S4, subdivide the output assembly path planning information into shorter steps, and obtain assembly path planning refinement information.

[0102] Typically, the assembly path planning refinement information is 1 / N of the assembly path planning information, where N is a natural number greater than 2, and is 10 in this embodiment.

[0103] Assembly path planning detailed information is expressed in the following format:

[0104] ①Translate 1mm along the positive direction of the Y axis;

[0105] ② Repeat step ① and move 9 times;

[0106] ③Rotate 7.5° around the Y axis;

[0107] ④Repeat step ③ and rotate 9 times;

[0108] ⑤Translate 3mm along the positive direction of the Y axis;

[0109] ⑥Repeat step ⑤ and move 9 times.

[0110] Step S7. According to the assembly path planning refinement information outputted in step S6, the three-dimensional assembly digital model is controlled to move according to the assembly path planning refinement information, and only one path movement is performed each time.

[0111] The instruction for the first 3D assembly model movement is:

[0112] ①Translate 1mm along the positive direction of the Y axis.

[0113] Step S8: Update the three-dimensional assembly digital model information. According to the processing result in step S7, update the three-dimensional assembly digital model.

[0114] Step S9: Extracting the assembly scene.

[0115] Step S 10 . Call the generative large model API to evaluate the result after the assembly path is moved. Determine whether the verification is completed. If it is completed, store the assembly path information and the process ends. If the verification is not completed, determine whether there are abnormal situations such as collision interference, exceeding the safety distance, etc. If there are no abnormalities, go to step S7 to continue processing. If there are abnormalities, go to step S5 and require optimization and update based on the current path.

[0116] In this embodiment, steps S7 to S8 are repeated. 10, until the 3D assembly model completes the following movement, as shown in the attached manual Figure 5 As shown:

[0117] ①Translate 10mm along the positive direction of the Y axis.

[0118] Continue to repeat step S7 to step S 10 , until the 3D assembly model completes the following movement, as shown in the attached manual Figure 6 As shown:

[0119] ① Move 10mm in the positive direction along the Y axis;

[0120] ② Rotate 75° around the Y axis.

[0121] At this time, it is determined that an abnormality has occurred and the rod 1 cannot move through the hole 2 on the mounting plate 3, and the process goes to step S5, requiring optimization and updating based on the current path. According to the feedback of the abnormal result, optimization and updating are performed, and the assembly path planning information after optimization and updating is as follows:

[0122] ① Move 10mm in the positive direction along the Y axis;

[0123] ② Rotate 90° around the Y axis;

[0124] ③ Translate 30mm along the positive direction of the Y axis.

[0125] Repeat steps S6 to S 10 , until the assembly path planning is completed, as shown in the attached manual Figure 7 And the instruction manual Figure 8 shown.

[0126] In summary, after reading the present invention document, ordinary technicians in this field can make various other corresponding transformation schemes based on the technical scheme and technical concept of the present invention without creative mental labor, which all fall within the scope of protection of the present invention.

Claims

1. An assembly path planning method, characterized in that: The following steps are involved: Step S1. Receiving assembly path planning instructions; Step S2. Processing the three-dimensional assembly digital model, adding a coordinate system to the three-dimensional assembly digital model, and distinguishing the parts in the three-dimensional assembly digital model by different colors; Step S3. Extracting the assembly scene; Step S4. Calling the generative large model API interface, according to the assembly scene obtained in step S3, converting the object names in the assembly path planning instructions into color-coded names, and performing assembly path planning with colored parts; Step S5. Calling the generative large model API interface, and performing detailed design of assembly path planning according to the assembly path planning information outputted from step S4; the detailed design specifically refers to: subdividing the assembly path planning information outputted from step S4 into shorter steps to obtain detailed information of assembly path planning; the detailed information of assembly path planning is 1 / N of the assembly path planning information; wherein N is a natural number greater than 2; Step S6. Generate control instructions based on the detailed assembly path planning information output in step S5 to control the three-dimensional assembly digital model to move according to the detailed assembly path planning information; Step S7. According to the processing result of step S6, updating the three-dimensional assembly digital model; Step S8. Extracting the assembly scene; Step S9. Call the generative large model API to evaluate the result after the assembly path is moved; determine whether the verification is completed. If so, store the assembly path information and end the process; if not, determine whether an abnormality occurs. If no abnormality occurs, go to step S6 to continue processing. If an abnormality occurs, go to step S4 to optimize and update based on the current path.

2. An assembly path planning method according to claim 1, characterized in that: It also includes generating a color matching database to match the names and colors of parts in the three-dimensional assembly digital model. When planning an assembly path with colored parts, the object names in the assembly path planning instructions are converted into color-coded names by retrieving the color matching database.

3. An assembly path planning method according to claim 2, characterized in that: Extracting assembly scenes specifically refers to: converting the 3D assembly digital model into pictures and videos that can be understood by the generative large model, capturing views from several angles from the 3D assembly digital model, and recording a video of the 3D assembly digital model rotating around the coordinate axis.

4. An assembly path planning method according to claim 1 or 3, characterized in that: The size of the coordinate system is consistent with the maximum three-dimensional size of the assembly scene.

5. An assembly path planning method according to claim 4, characterized in that: In step S4, during the assembly path planning process, a search is performed from a knowledge base, in which relevant background knowledge and information about the three-dimensional assembly scene are pre-stored.

6. An assembly path planning method according to claim 5, characterized in that: The assembly path planning information outputted in step S4 includes a corresponding distance of movement around the coordinate axis and / or a corresponding angle of rotation around the coordinate axis.

7. An assembly path planning method according to claim 1, characterized in that: The abnormal situations include interference and exceeding the safety distance.

Images