Model building method and apparatus, storage medium, program product, electronic device, and computer device

Abstract

The present application relates to the technical field of three-dimensional modeling, and provides a model building method and apparatus, a storage medium, a program product, an electronic device, and a computer device. The model building method comprises: acquiring a target size parameter and component size parameters of at least two component models, wherein the component models are used for building a target object model; determining a target model combination on the basis of the target size parameter and the component size parameters, wherein the target model combination comprises at least two component models; configuring a kinematic pair between any two component models in the target model combination on the basis of the component type of each component model in the target model combination; and building the target object model by means of the target model combination.

Description

Model building method, device, storage medium, program product, electronic device and computer device

[0001] This application claims priority to the Chinese patent application filed with the State Intellectual Property Office of China on July 25, 2023, with application number "202310921909.X" and application name "Model building method, device, storage medium, program product and electronic device", the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present application belongs to the field of three-dimensional modeling technology, and specifically relates to a model construction method, device, storage medium, program product, electronic device and computer device. Background Art

[0003] With the continuous advancement of robotics technology, the application of robots has gradually expanded to non-deterministic tasks in complex scenarios in civilian homes. Due to the highly unstructured nature of home environments, large-scale training and testing in homes are necessary to improve the generalization of service robots' capabilities. When setting up simulation scenarios, different object models need to be configured into the simulation scenes.

[0004] Summary of the Invention

[0005] This application aims to solve at least one of the technical problems existing in the prior art.

[0006] To this end, the first aspect of this application proposes a model construction method.

[0007] A second aspect of the present application provides a model building device.

[0008] A third aspect of the present application provides a readable storage medium.

[0009] A fourth aspect of the present application proposes a computer program product.

[0010] A fifth aspect of the present application provides an electronic device.

[0011] A sixth aspect of the present application provides a computer device.

[0012] In view of this, the first aspect of the present application provides a model construction method, which includes: obtaining target size parameters and component size parameters of at least two component models, the component models are used to construct a target object model; determining a target model combination based on the target size parameters and the component size parameters, the target model combination including at least two component models; configuring a kinematic pair between any two component models in the target model combination based on the component type of each component model in the target model combination; and constructing a target object model through the target model combination.

[0013] The second aspect of the present application provides a model building device, which includes: an acquisition module for acquiring target size parameters and component size parameters of at least two component models, and the component models are used to build a target object model; a determination module for determining a target model combination based on the target size parameters and the component size parameters, and the target model combination includes at least two component models; a configuration module for configuring a kinematic pair between any two component models in the target model combination according to the component type of each component model in the target model combination; and a construction module for building a target object model through the target model combination.

[0014] A third aspect of the present application provides a readable storage medium having a program or instructions stored thereon. When executed by a processor, the program or instructions implement the steps of the model building method described in any of the above technical solutions. Therefore, the readable storage medium possesses the advantages of the model building method described in any of the above technical solutions and can achieve the technical effects achieved by the model building method described in any of the above technical solutions. To avoid repetition, the details are not described here.

[0015] A fourth aspect of the present application provides a computer program product that, when executed by a processor, implements the steps of the model building method described in any of the aforementioned technical solutions. Therefore, this computer program product possesses the advantages of the model building method described in any of the aforementioned technical solutions and can achieve the technical effects achieved by the model building method described in any of the aforementioned technical solutions. To avoid repetition, the details are not described here.

[0016] The fifth aspect of the present application provides an electronic device, which includes: the model building device in any of the above technical solutions; and / or the readable storage medium in any of the above technical solutions; and / or the computer program product in any of the above technical solutions. Therefore, the electronic device has the advantages of the model building device in any of the above technical solutions, and can achieve the technical effects that can be achieved by the model building device in any of the above technical solutions, and / or the electronic device has the advantages of the readable storage medium in the above technical solutions, and can achieve the technical effects that can be achieved by the readable storage medium in the above technical solutions, and / or the electronic device has the advantages of the computer program product in the above technical solutions, and can achieve the technical effects that can be achieved by the computer program product in the above technical solutions. To avoid repetition, it will not be repeated here.

[0017] The sixth aspect of the present application proposes a computer device, which includes: a processor and a memory, the memory storing programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the steps of the model building method of the first aspect are implemented, thereby having the technical effects that can be achieved by the above-mentioned model building method. To avoid repetition, they will not be repeated here.

[0018] The technical solution of the present application realizes the automatic generation of the target object model. Compared with the aforementioned related technologies, on the one hand, the model construction method defined in the present application eliminates the operation of manual modeling, liberates labor, and reduces the labor cost of the robot training and testing process. On the other hand, the target object model generated by the model construction method defined in the present application has high generation efficiency, which is conducive to reducing the time cost of the robot training and testing process. On the other hand, the target object model generated by the model construction method defined in the present application is random. Compared with manual modeling, the simulation training scenarios involved are wider and the training and testing coverage is wider, which is conducive to improving the reliability of robot training and testing. In this way, the technical effect of optimizing the model construction method, improving the model generation rate and model generation randomness, improving the reliability of automated training and automated testing, and reducing the cost of automated testing and automated training is achieved.

[0019] Additional aspects and advantages of the present application will become apparent in the following description or may be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0021] FIG1 shows a schematic flow chart of a model building method according to an embodiment of the present application;

[0022] FIG2 shows a schematic flow chart of a model building method according to an embodiment of the present application;

[0023] FIG3 shows a flow chart of a model building method according to an embodiment of the present application;

[0024] FIG4 shows a flow chart of a model building method according to an embodiment of the present application;

[0025] FIG5 shows a schematic flow chart of a model building method according to an embodiment of the present application;

[0026] FIG6 shows a flow chart of a model building method according to an embodiment of the present application;

[0027] FIG7 shows a schematic flow chart of a model building method according to an embodiment of the present application;

[0028] FIG8 shows a schematic flow chart of a model building method according to an embodiment of the present application;

[0029] FIG9 shows a schematic flow chart of a model building method according to an embodiment of the present application;

[0030] FIG10 shows a schematic diagram of multi-rigid-body dynamics of an object according to an embodiment of the present application;

[0031] FIG11 shows a schematic diagram of target model combination generation according to an embodiment of the present application;

[0032] FIG12 shows a structural block diagram of a model building device according to an embodiment of the present application.

[0033] 10 and 11 , the correspondence between the reference numerals and the component names is as follows: 1010 first rigid body, 1012 first local coordinate system, 1030 kinematic pair, 1032 first reference coordinate system, 1034 second reference coordinate system, 1020 second rigid body, 1022 second local coordinate system, 1002 global coordinate system, 1110 component model, and 1120 target model combination. DETAILED DESCRIPTION

[0034] In order to more clearly understand the above-mentioned objectives, features and advantages of the present application, the present application is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the features of this embodiment and the embodiments can be combined with each other.

[0035] In the following description, many specific details are set forth to facilitate a full understanding of the present application. However, the present application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present application is not limited to the specific embodiments disclosed below.

[0036] The following describes the model building method, apparatus, storage medium, program product, electronic device, and computer device according to some embodiments of the present application with reference to Figures 1 to 12.

[0037] In related technologies, the object models required for building simulation scenes include models with movable connectors such as wardrobes, cabinets, refrigerators, and microwave ovens. Such models require configuration of multi-body dynamic parameters, so they mostly rely on manual modeling, and it is difficult to obtain a large number of such models to build simulation scenes.

[0038] Therefore, one embodiment of the present application proposes a model construction method, which includes:

[0039] Step 102: obtaining target size parameters and component size parameters of at least two component models, where the component models are used to construct a target object model;

[0040] Step 104: determining a target model combination according to the target size parameter and the component size parameter, wherein the target model combination includes at least two component models;

[0041] Step 106 , configuring a kinematic pair between any two component models in the target model combination according to the component type of each component model in the target model combination;

[0042] Step 108: construct a target object model by combining target models.

[0043] In this embodiment, a model building method is defined, which can be used to build a simulation scenario model. The simulation scenario model can be used for large-scale training and testing of robots, so that the application of robots can be extended to non-deterministic tasks in complex scenarios in civilian households, or the application of robots can be extended to non-deterministic tasks in complex commercial scenarios.

[0044] Specifically, the robot can be a household sweeping robot, a shopping mall guide robot, a restaurant delivery robot, etc. Robots also include various types of service robots, such as household service robots.

[0045] In the process of building the model, the target size parameters and the component size parameters of the component model are first obtained, where the target size parameters are the size of the target object model to be built, and the component size of the component model is the size of each component that assembles the target object model. The number of component models in the target object model is at least two, and there can be a motion relationship between the two component models, that is, the shape of the target object model can be changed through motion.

[0046] For example, as shown in FIG11 , in a home scenario, the target model combination 1120 is a wardrobe, and the component model 1110 of the wardrobe includes drawers, doors, and handles, wherein the target size parameters may include at least one of the length, width, and height of the wardrobe, and may also include at least one of the number of rows and columns of the wardrobe, and the component size parameters are the length, width, and height of the drawers, doors, and handles, respectively.

[0047] In a shopping mall scenario, the target object model is a container, and its component models include a glass door, a handle, and a cargo box. The target size parameter can include at least one of the container's length, width, and height, and can also include at least one of the number of rows or columns. The component size parameters are the length, width, and height of the glass door, handle, and cargo box, respectively.

[0048] After obtaining the target size parameters of the target object model and the component size parameters of the component model, at least one target model combination having a size matching the target size parameters and including a corresponding component model is determined based on the target size parameters and the component size parameters.

[0049] For example, as shown in FIG11 , the target model combination 1120 that matches the target size parameters may include six drawers, four cabinet doors, and 10 handles, or may include six cabinet doors and six handles.

[0050] The assembled container that matches the target size parameters may include four cargo boxes and four cargo box lids, and may also include two glass doors and two handles.

[0051] After determining at least one target model combination, a kinematic pair is established between any two component models in the target model combination according to the type of each component model in the target model combination, and the kinematic pair can reflect the relative motion relationship between the two component models.

[0052] For example, the relative motion between two drawers in a wardrobe tends to be sliding, and the relative motion between a cargo box and its lid in a container tends to be rotating.

[0053] Finally, after completing the kinematic pairs between any two component models in the target model combination, a target object model is constructed whose overall dimensions match the target dimension parameters and includes the component model types, number of component models, and kinematic pairs between the component models. This allows the robot to be trained and tested on a large scale using the target object model.

[0054] It can be seen that through the above process, the automatic generation of the target object model is achieved. Compared with the aforementioned related technologies, on the one hand, the model construction method defined in this application eliminates the operation of manual modeling, liberates labor, and reduces the labor cost of the robot training and testing process. On the other hand, the target object model generated by the model construction method defined in this application has high generation efficiency, which is conducive to reducing the time cost of the robot training and testing process. On the other hand, the target object model generated by the model construction method defined in this application is random. Compared with manual modeling, the simulation training scenarios involved are wider and the training and testing coverage is wider, which is conducive to improving the reliability of robot training and testing. In this way, the technical effect of optimizing the model construction method, improving the model generation rate and model generation randomness, improving the reliability of automated training and automated testing, and reducing the cost of automated testing and automated training is achieved.

[0055] As shown in FIG2 , in the above embodiment, the step of determining the target model combination according to the target size parameter and the component size parameter includes:

[0056] Step 202 , determining at least one group of first model combinations based on the target size parameter and the component size parameter, wherein the sum of the component size parameters of the component models in each group of first model combinations matches the target size parameter;

[0057] Step 204: extracting a target model combination from at least one group of first model combinations through a random algorithm.

[0058] In this embodiment, in the process of determining the target model combination based on the target size parameters and the component size parameters, at least one group of first model combinations is first determined within the framework defined by the target size parameters. That is, the sum of the component size parameters of the component models included in each group of first model combinations matches the target size parameters. In the specific calculation process, the length dimensions, width dimensions, and height dimensions of the multiple component models in the first model combination can be accumulated, and finally the three accumulated values ​​are compared with the target size parameters to determine whether they are compatible. After the generation of at least one group of first model combinations is completed, the target model combination is extracted from the at least one group of first model combinations using a random algorithm to construct it into the target object model.

[0059] By limiting the above process, the automatic combination of component models is achieved, so that the first model combination can cover a variety of possible combinations that meet the target size parameters. At the same time, the process also ensures the randomness of the target object model finally generated, thereby achieving the technical effect of expanding the scenario coverage of robot automated testing and automated training, and improving the reliability of automated training and automated testing.

[0060] As shown in FIG3 , in some embodiments, optionally, the step of extracting a target model combination from at least one set of first model combinations by a random algorithm includes:

[0061] Step 302, obtaining a preset number of model combinations in the target object model;

[0062] Step 304, determining a target quantity based on a preset quantity;

[0063] Step 306: extracting a target number of target model combinations from at least one group of first model combinations through a random algorithm.

[0064] In this embodiment, in the process of extracting target model combinations from at least one set of first model combinations using a random algorithm, a preset number of first model combinations included in the target object model is first obtained, where the preset number corresponds to the training and testing requirements. After obtaining the preset number, a corresponding target number is determined based on the properties of the preset number, and this target number serves as a benchmark for screening the first model combinations. Ultimately, the random algorithm is used to extract the target number of target model combinations from at least one set of first model combinations, thereby completing the random generation of target model combinations.

[0065] By limiting the above process, the automatic combination of component models is achieved, so that the first model combination can cover a variety of possible combinations that meet the target size parameters. At the same time, the process also ensures the randomness of the target object model finally generated, thereby achieving the technical effect of expanding the scenario coverage of robot automated testing and automated training, and improving the reliability of automated training and automated testing.

[0066] In some embodiments, optionally, constructing a target object model using target model combinations includes: based on the target quantity being the same as a preset quantity, arranging the target number of target model combinations to obtain the target object model. In this embodiment, after determining that the preset quantity is the target quantity, the target number of target model combinations can be directly randomly extracted, and the target object model can be constructed based on the target model combinations of the target data volume.

[0067] Since some furniture has multiple structural components and the multiple structural components are different, a target number of target model combinations are directly extracted, and the target object model is constructed by arranging the target model combinations.

[0068] Through the above method, a corresponding target object model can be constructed for furniture including multiple different structural components, so that the structures of multiple target model combinations in the target object model are different. This improves the rationality of the automatically generated target object model while taking into account the random generation requirements of the model, thereby achieving the technical effect of improving the reliability of robot automated testing and automated training.

[0069] In some embodiments, optionally, the step of determining the target quantity based on the preset quantity includes:

[0070] Based on the preset number N being an even number, the target number is M, where M=N / 2;

[0071] Based on the preset number N being an odd number, the target number is M, where

[0072] In this embodiment, in the process of determining the target number according to the preset number, if the preset number N is an even number, the target number M is calculated as: M=N / 2; if the preset number N is an odd number, the target number M is calculated as: This means that M is N / 2 rounded up.

[0073] Because fixtures like wardrobes and containers often have symmetrical structures, by limiting the target number determination method described above, a structurally symmetrical target object model can be generated by combining the target number of first models. This ensures that the resulting target object model resembles a real product on the market. This improves the rationality of the automatically generated target object model while also taking into account the need for random model generation, thereby achieving the technical effect of enhancing the reliability of automated robot testing and automated training.

[0074] As shown in FIG4 , in some embodiments, optionally, the step of constructing a target object model by combining target models includes:

[0075] Step 402: Based on the preset number N being an even number, the M target model combinations are arranged in sequence according to a preset arrangement direction;

[0076] Step 404: mirror-copy the M target model combinations to obtain M second model combinations;

[0077] Step 406 : construct a target object model by combining the M target models and the M second models.

[0078] In this embodiment, when the preset number N is an even number, the number M of target model combinations is half of N. On this basis, in the process of constructing the target object model through the target model combinations, the M target model combinations are first arranged in sequence according to the preset arrangement direction to generate half of the target object model. Thereafter, the M target model combinations that have been arranged are mirrored according to the preset reference to obtain M second model combinations, so as to ensure that each of the M second model combinations has a symmetrical structure. Finally, the target object model is constructed by combining the M target model combinations and the M second model combinations to ensure that the target object model matches the preset number N.

[0079] By limiting the above process, a target object model with a mirror-symmetrical structure can be generated when the preset number N is an even number, so that the final generated target object model is closer to the products available on the market, ensuring that the simulation scenes used to test and train robots are closer to real scenes, thereby achieving the technical effect of improving the reliability of robot automated testing and automated training.

[0080] As shown in FIG5 , in some embodiments, optionally, the step of constructing a target object model by combining target models includes:

[0081] Step 502: extracting M-1 third model combinations from the M target model combinations based on the preset number N being an odd number;

[0082] Step 504: mirror-copy the M-1 third model combinations to obtain M-1 fourth model combinations;

[0083] Step 506: construct a target object model through M-1 third model combinations, M-1 fourth model combinations and fifth model combinations. The fifth model combination is located between the M-1 third model combinations and the M-1 fourth model combinations. The fifth model combination is the other model combinations in the M target model combinations except the M-1 third model combinations.

[0084] In this embodiment, when the preset number N is an odd number, the number M of target model combinations is N divided by 2 and then obtained upward. On this basis, in the process of constructing the target object model through the target model combination, M-1 third model combinations are first extracted from the M target model combinations, and the third model combination serves as a symmetrical structure on the target object model. Thereafter, the M-1 third model combinations are mirror-copied to obtain M-1 fourth model combinations. Finally, a fifth model combination is added between the M-1 third model combinations and the M-1 fourth model combinations to construct the target object model. That is, the target object model finally generated includes a non-mirror area in the middle and mirror areas on both sides. Among them, in the process of screening the fifth model combination, it is necessary to first eliminate M-1 third model combinations from the M target model combinations, and then determine the corresponding number of fifth model combinations from the remaining model combinations to ensure that the non-mirror area in the middle of the target object model finally generated does not overlap with the non-mirror areas on both sides.

[0085] By limiting the above process, a target object model with an independent middle structure and mirror-symmetrical structures on both sides can be generated with a preset number N as the base number, so that the final generated target object model is closer to the products available on the market, ensuring that the simulation scenes used to test and train robots are closer to real scenes, thereby achieving the technical effect of improving the reliability of robot automated testing and automated training.

[0086] In a specific embodiment of this application, using a wardrobe as an example, the solution first requires dividing the wardrobe's components into distinct component models based on kinematic pairs. These components are then modeled separately using 3D modeling software. Next, the wardrobe's main components, such as drawers and doors, are coded by height. This height corresponds to the component size parameters of the component model, e.g., small drawer = 1, large drawer = small door = 2, and large door = 3.

[0087] Then, we can input a digital combination of two integers at the beginning of the program, which corresponds to the target size parameter, such as (3, 2), which means that the wardrobe has two columns and the height is 3. Then, we can use a computer algorithm to list all the first model combinations of a column of wardrobes with a height of 3, including but not limited to recursive depth search algorithms, such as (1, 1, 1), (1, 2), (2, 1), (3), etc.

[0088] We then select the first model combination corresponding to that column based on the code. For example, if it's 1, we choose the small drawer model. If it's 2, which corresponds to multiple models, we randomly select one from the first model combination. For the second column, to ensure rationality, we can no longer use a deep search algorithm to prevent the unreasonable spatial asymmetry of the resulting items. Therefore, we can randomly copy the combination from the first column or, if the combination in the first column includes a door, mirror the door's kinematic pair to achieve a left-right opening.

[0089] When there are more than two columns, for an even number of columns N, we can randomly shuffle columns 1 to N / 2, and then update columns N / 2+1 to N using the method described above. For an odd number of columns, we first shuffle columns 1 to ceil(N / 2), where "ceil()" represents rounding down, then shuffle the middle column separately, and then copy columns ceil(N / 2)+2 to N.

[0090] As shown in FIG6 , in some embodiments, optionally, after the step of constructing the target object model by combining the target models, the following steps are further included:

[0091] Step 602, obtaining a scaling factor of the target object model;

[0092] Step 604: scaling the target object model according to the scaling factor.

[0093] In this embodiment, after the target object model is constructed based on the target model combination, a scaling factor for the target object model is first obtained. This scaling factor corresponds to the virtual scene construction requirements, robot testing requirements, and robot training requirements. After obtaining the scaling factor, the target object model is proportionally scaled according to the scaling factor and added to the virtual scene.

[0094] By limiting the above-mentioned scaling process, the size of the target model finally added to the virtual scene can be made to better match the actual environment or actual product, thereby ensuring that the simulation scene used to test and train the robot is closer to the target scene, thereby achieving the technical effect of improving the reliability of robot automated testing and automated training.

[0095] As shown in FIG7 , in some embodiments, optionally, the step of scaling the target object model according to the scaling factor includes:

[0096] Step 702: scaling each target model combination in the target object model according to the scaling factor;

[0097] Step 704: Obtain the reference coordinate system of each kinematic joint in the target model combination, as well as the local coordinate system corresponding to each kinematic joint, where the local coordinate system is the coordinate system of the component model connected to the kinematic joint;

[0098] Step 706: Update the reference coordinate system according to the local coordinate system and the scaling factor.

[0099] In this embodiment, when scaling a target object model according to a scaling factor, each target model combination included in the target object model is first scaled according to the scaling factor. Subsequently, a reference coordinate system corresponding to each kinematic joint in the target model combination is obtained, as well as a local coordinate system for the component model corresponding to each kinematic joint. The local coordinate system is the coordinate system of the two component models connected by the kinematic joint. Finally, the reference coordinate system is updated using the local coordinate system and the scaling factor to ensure that the updated reference coordinate system matches the scaled target object model.

[0100] By limiting the above-mentioned reference coordinate system update process, it is possible to ensure that the motion simulation results of the target object model finally added to the virtual scene are accurate and effective, and to ensure that the robot can perform motion association with the target object model during the simulation training process and the simulation test process to simulate motion scenes such as collisions and scratches, thereby achieving the technical effect of improving the reliability of robot automated testing and automated training.

[0101] As shown in FIG8 , in some embodiments, optionally, after the step of constructing the target object model by combining the target models, the following steps are further included:

[0102] Step 802, configuring material information for the target object model;

[0103] Step 804: Configure physical parameters for the target object model based on the material information.

[0104] In this embodiment, after the target object model is constructed based on the target model combination, material information for the target object model is first configured. The material information may include material type information. After the material information is configured, the physical parameters of the target object model are configured based on the material information. The physical parameters correspond to the material information, and different material types correspond to different physical parameters.

[0105] By defining the material and physical parameter configuration processes described above, the target object model ultimately added to the virtual scene possesses the physical properties of the corresponding material. During robot simulation, these physical properties can be combined with the robot's inherent motion data to simulate collisions, scrapes, and other motion simulation data, thereby broadening the application scope of virtual scenes and improving the reliability of automated robot testing and training.

[0106] In some embodiments, optionally, the physical parameter includes at least one of the following: mass, inertia tensor.

[0107] As shown in FIG10 , in a specific embodiment of the present application, in order to adapt to the depth search algorithm, the length, width, and height of the component model are generally modeled according to fixed data. In order to improve the generalization of the model, the size of the component model can be changed by the scaling factor, thereby changing the size of the target model combination with the kinematic pair 1030. If the size of the target model combination with the kinematic pair 1030 changes due to scaling, the reference coordinate system of the kinematic pair 1030 will also be adjusted. The reference coordinate system of the kinematic pair 1030 includes a first reference coordinate system 1032 and two second reference coordinate systems 1034. Taking the first reference coordinate system 1032 as the first reference coordinate system, the reference coordinate system of the kinematic pair 1030 is the first reference coordinate system 1032 and the second reference coordinate system 1034. p The first rigid body 1010 is taken as an example, and the first local coordinate system 1012 of the first rigid body 1010 is O b , the scaling factor of the first rigid body 1010 is S. Then the origin O′ of the new first reference coordinate system 1032 is p The calculation formula is: p =O b +S×(O b -O p ).

[0108] Similarly, the second reference coordinate system 1034 is also updated by the second local coordinate system 1022 of the second rigid body 1020 and the scaling factor of the second rigid body 1020 , which will not be described again here to avoid repetition.

[0109] FIG10 also includes a global coordinate system 1002 .

[0110] Finally, we randomly search for texture maps in the texture library and add them to the generated wardrobe model. In addition to the image, the texture map also contains the average density of the material type. We can then obtain the relevant physical parameters based on the volume, center of mass, and mass vector pre-stored during geometric modeling and add them. These are shown in the following two formulas: m = ρ × V;

[0111] Where m is mass, ρ is density, V is volume, I is inertia, and r is offset. is the tensor product operation.

[0112] This method can generate models of target objects with various shapes and joints, and their individual rigid bodies can move based on their connections. Taking the wardrobe generation example, this method can generate a variety of wardrobe combinations, with movable joints that also meet the rules of multi-rigid-body dynamics.

[0113] When an external force acts on the wardrobe, the wardrobe can realize opening and closing and other movements based on real-time calculations of the simulator.

[0114] As shown in FIG9 , in a specific embodiment of the present application, the model building process is as follows:

[0115] Step 902: input row and column related parameters;

[0116] Among them, the row and column related parameters correspond to the target size parameters;

[0117] Step 904: Using a depth-first search method, search for all possible combinations based on the rigid body component codes.

[0118] Step 906: Add the corresponding component 3D model according to the code. If the code corresponds to multiple components, randomly select one of the multiple components to add.

[0119] Step 908: If there are multiple columns, assign a combination of the multiple columns based on rationality;

[0120] Step 910 , randomize the scaling of each component and adjust the position of the motion sub-coordinate;

[0121] Step 912, scaling the outer bounding box according to the combined row and column height, width, and depth and adding;

[0122] Step 914: Obtain physical parameters based on the randomly generated material and the volume and other dimensions of the model.

[0123] As shown in Figure 12, an embodiment of the present application proposes a model building device 1200, which includes: an acquisition module 1202, used to obtain target size parameters and component size parameters of at least two component models, and the component models are used to build a target object model; a determination module 1204, used to determine a target model combination based on the target size parameters and the component size parameters, and the target model combination includes at least two component models; a configuration module 1206, used to configure a kinematic pair between any two component models in the target model combination according to the component type of each component model in the target model combination; and a construction module 1208, used to build a target object model through the target model combination.

[0124] In this embodiment, a model building device 1200 is defined, which can be used to build a simulation scenario model. The simulation scenario model can be used for large-scale training and testing of robots, so that the application of robots can be extended to non-deterministic tasks in complex scenarios of civilian households, or the application of robots can be extended to non-deterministic tasks in complex commercial scenarios.

[0125] Specifically, the robot can be a household sweeping robot, a shopping mall guiding robot, a restaurant delivery robot, etc.

[0126] The model building device 1200 includes an acquisition module 1202 , a determination module 1204 , a configuration module 1206 , and a construction module 1208 .

[0127] During the model construction process, the acquisition module 1202 first obtains the target size parameters and the component size parameters of the component model, where the target size parameters are the size of the target object model to be constructed, and the component size of the component model is the size of each component that assembles the target object model. The number of component models in the target object model is at least two, and there may be a motion relationship between the two component models, that is, the shape of the target object model can be changed through motion.

[0128] After obtaining the target size parameters of the target object model and the component size parameters of the component model, the determination module 1204 determines at least one target model combination whose size matches the target size parameters and includes a corresponding component model based on the target size parameters and the component size parameters.

[0129] After determining at least one target model combination, the configuration module 1206 establishes a kinematic pair between any two component models in the target model combination according to the type of each component model in the target model combination, and the kinematic pair can reflect the relative motion relationship between the two component models.

[0130] Finally, after completing the kinematic pairs between any two component models in the target model combination, the construction module 1208 constructs a target object model whose overall dimensions match the target dimension parameters and includes the component model types, number of component models, and kinematic pairs between the component models. This allows the robot to conduct large-scale training and testing using the target object model.

[0131] It can be seen that through the above process, the automatic generation of the target object model is achieved. Compared with the aforementioned related technologies, on the one hand, the model construction method defined in this application eliminates the operation of manual modeling, liberates labor, and reduces the labor cost of the robot training and testing process. On the other hand, the target object model generated by the model construction method defined in this application has high generation efficiency, which is conducive to reducing the time cost of the robot training and testing process. On the other hand, the target object model generated by the model construction method defined in this application is random. Compared with manual modeling, the simulation training scenarios involved are wider and the training and testing coverage is wider, which is conducive to improving the reliability of robot training and testing. In this way, the technical effect of optimizing the model construction method, improving the model generation rate and model generation randomness, improving the reliability of automated training and automated testing, and reducing the cost of automated testing and automated training is achieved.

[0132] In some embodiments, optionally, the determining module 1204 is further configured to determine, based on the target size parameter and the component size parameter, at least one group of first model combinations, wherein the sum of the component size parameters of the component models in each group of first model combinations matches the target size parameter;

[0133] The model building device 1200 further includes an extraction module, which is configured to extract a target model combination from at least one group of first model combinations through a random algorithm.

[0134] In an embodiment of the present application, by limiting the above-mentioned process, the automatic combination of component models is realized, so that the first model combination can cover a variety of combination possibilities that meet the target size parameters. At the same time, the process also ensures the randomness of the target object model finally generated, thereby achieving the technical effect of expanding the scenario coverage of robot automated testing and automated training, and improving the reliability of automated training and automated testing.

[0135] In some embodiments, optionally, the acquisition module 1202 is further configured to acquire a preset number of model combinations in the target object model;

[0136] The determination module 1204 is further configured to determine a target quantity based on a preset quantity;

[0137] The extraction module is also used to extract a target number of target model combinations from at least one group of first model combinations through a random algorithm.

[0138] In an embodiment of the present application, by limiting the above-mentioned process, the automatic combination of component models is realized, so that the first model combination can cover a variety of combination possibilities that meet the target size parameters. At the same time, the process also ensures the randomness of the target object model finally generated, thereby achieving the technical effect of expanding the scenario coverage of robot automated testing and automated training, and improving the reliability of automated training and automated testing.

[0139] In some embodiments, optionally, the determining module 1204 is further configured to determine, based on the preset number N being an even number, the target number being M, where M=N / 2;

[0140] The determination module 1204 is further configured to determine, based on the preset number N being an odd number and the target number being M,

[0141] In the embodiments of the present application, by limiting the target number determination method described above, a structurally symmetrical target object model can be generated by combining the target number of first models, so that the target object model ultimately constructed is close to a real product on the market. This improves the rationality of the automatically generated target object model while also taking into account the need for random model generation, thereby achieving the technical effect of improving the reliability of automated robot testing and automated training.

[0142] In some embodiments, optionally, the construction module 1208 is further configured to arrange the M target model combinations in sequence according to a preset arrangement direction based on the preset number N being an even number;

[0143] The model building device 1200 further includes:

[0144] A copy module, used to mirror the M target model combinations to obtain M second model combinations;

[0145] The construction module 1208 constructs a target object model by combining the M target models and the M second models.

[0146] In an embodiment of the present application, a target object model with a mirror-symmetrical structure can be generated when the preset number N is an even number, so that the final generated target object model is closer to the product available on the market, to ensure that the simulation scene used to test and train the robot is closer to the real scene, thereby achieving the technical effect of improving the reliability of robot automated testing and automated training.

[0147] In some embodiments, optionally, the construction module 1208 is further configured to extract M-1 third model combinations from the M target model combinations based on the preset number N being an odd number;

[0148] A copy module, used for mirroring M-1 third model combinations to obtain M-1 fourth model combinations;

[0149] Construction module 1208 is used to construct a target object model through M-1 third model combinations, M-1 fourth model combinations and a fifth model combination, where the fifth model combination is located between the M-1 third model combinations and the M-1 fourth model combinations, and the fifth model combination is the other model combination in the M target model combinations except the M-1 third model combinations.

[0150] In the embodiment of the present application, by limiting the above process, a target object model with an independent middle structure and mirror-symmetric structures on both sides can be generated with a preset number N as the base number, so that the target object model finally generated is closer to the product existing on the market, so as to ensure that the simulation scene used for testing and training robots is closer to the real scene, thereby achieving the technical effect of improving the reliability of robot automated testing and automated training

[0151] In some embodiments, optionally, the acquisition module 1202 is further configured to acquire a scaling factor of the target object model;

[0152] The model building device 1200 further includes a scaling module, which is used to scale the target object model according to a scaling factor.

[0153] In an embodiment of the present application, the size of the target model finally added to the virtual scene can be made to better match the actual environment or actual product, so as to ensure that the simulation scene used to test and train the robot is closer to the target scene, thereby achieving the technical effect of improving the reliability of robot automated testing and automated training.

[0154] In some embodiments, optionally, the scaling module is further configured to scale each target model combination in the target object model according to the scaling factor;

[0155] The acquisition module 1202 is further used to obtain the reference coordinate system of each kinematic pair in the target model combination, and the local coordinate system corresponding to each kinematic pair, where the local coordinate system is the coordinate system of the component model connected by the kinematic pair;

[0156] The scaling module is further used to update the reference coordinate system according to the local coordinate system and the scaling factor.

[0157] In an embodiment of the present application, by limiting the above-mentioned reference coordinate system update process, it is possible to ensure that the motion simulation results of the target object model finally added to the virtual scene are accurate and effective, and to ensure that the robot can perform motion association with the target object model during the simulation training process and the simulation test process to simulate motion scenes such as collisions and scratches, thereby achieving the technical effect of improving the reliability of robot automated testing and automated training.

[0158] In some embodiments, optionally, the model building device 1200 further includes a configuration module, where the configuration module is used to configure material information for the target object model; and configure physical parameters for the target object model based on the material information.

[0159] In the embodiments of this application, by defining the aforementioned material configuration process and physical parameter configuration process, the target object model ultimately added to the virtual scene possesses the physical properties of the corresponding material. During robot simulation, these physical properties can be used in conjunction with the robot's inherent motion data to simulate collisions, scrapes, and other motion simulation data, thereby broadening the application scope of virtual scenes and improving the reliability of automated robot testing and training.

[0160] In some embodiments, optionally, the physical parameter includes at least one of the following: mass, inertia tensor.

[0161] One embodiment of the present application provides a readable storage medium having a program or instruction stored thereon, which, when executed by a processor, implements the steps of the model building method in any of the above technical solutions.

[0162] In this embodiment, a readable storage medium is provided. The readable storage medium stores a program or instructions. When executed by a processor, the program or instructions implement the steps of the model building method described in any of the above embodiments. Therefore, this readable storage medium possesses the advantages of the model building method described in any of the above embodiments and can achieve the technical effects achieved by the model building method described in any of the above embodiments. To avoid repetition, further description is omitted here.

[0163] One embodiment of the present application provides a computer program product, which, when executed by a processor, implements the steps of the model building method in any of the above technical solutions.

[0164] In this embodiment, a computer program product is provided. When executed by a processor, the computer program product can implement the steps of the model building method described in any of the above embodiments. Therefore, this computer program product possesses the advantages of the model building method described in any of the above embodiments and can achieve the technical effects achieved by the model building method described in any of the above embodiments. To avoid repetition, further description is omitted here.

[0165] One embodiment of the present application provides an electronic device, which includes: the model building device in any of the above technical solutions; and / or the readable storage medium in any of the above technical solutions; and / or the computer program product in any of the above technical solutions.

[0166] In this embodiment, an electronic device is proposed that includes the model building device in any of the above embodiments, and / or the readable storage medium in the above embodiments, and / or the computer program product in any of the above embodiments. Therefore, the electronic device has the advantages of the model building device in any of the above embodiments, and can achieve the technical effects that can be achieved by the model building device in any of the above embodiments, and / or the electronic device has the advantages of the readable storage medium in the above embodiments, and can achieve the technical effects that can be achieved by the readable storage medium in the above embodiments, and / or the electronic device has the advantages of the computer program product in the above embodiments, and can achieve the technical effects that can be achieved by the computer program product in the above embodiments. To avoid repetition, no further details will be given here.

[0167] One embodiment of the present application provides a computer device, which includes: a processor and a memory, the memory storing programs or instructions that can be run on the processor, and when the programs or instructions are executed by the processor, the steps of the model building method in any of the above embodiments are implemented, thereby having the technical effects that can be achieved by the above-mentioned model building method. To avoid repetition, they will not be described here.

[0168] It should be clarified that in the claims, specification and drawings of this application, the term "plurality" refers to two or more. Unless otherwise clearly defined, the orientation or positional relationship indicated by the terms "upper" and "lower" is based on the orientation or positional relationship shown in the drawings. It is only for the purpose of more conveniently describing this application and making the description process simpler, and is not intended to indicate or imply that the device or element referred to must have the specific orientation described, be constructed and operated in a specific orientation. Therefore, these descriptions cannot be understood as limitations on this application. The terms "connect", "install", "fix" and the like should be understood in a broad sense. For example, "connection" can be a fixed connection between multiple objects, or a detachable connection between multiple objects, or an integral connection; it can be a direct connection between multiple objects, or an indirect connection between multiple objects through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood based on the specific circumstances of the above data.

[0169] In the claims, specification, and drawings of this application, the terms "one embodiment," "some embodiments," "a specific embodiment," and the like mean that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in the embodiment or example of this application. In the claims, specification, and drawings of this application, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0170] The above are merely preferred embodiments of the present application and are not intended to limit the present application. Those skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A model building method, wherein: include: Acquire target size parameters and component size parameters of at least two component models, wherein the component models are used to construct a target object model; Determine a target model combination according to the target size parameter and the component size parameter, wherein the target model combination includes at least two of the component models; According to the component type of each component model in the target model combination, configuring a kinematic pair between any two component models in the target model combination; The target object model is constructed by combining the target models.

2. The model building method according to claim 1, wherein: The step of determining a target model combination according to the target size parameter and the component size parameter includes: Determine at least one group of first model combinations according to the target size parameter and the component size parameter, wherein the sum of the component size parameters of the component models in each group of the first model combinations matches the target size parameter; The target model combination in the at least one group of first model combinations is extracted by a random algorithm.

3. The model building method according to claim 2, wherein: The step of extracting the target model combination from the at least one group of first model combinations by a random algorithm comprises: Obtaining a preset number of model combinations in the target object model; Based on the preset quantity, determining a target quantity; By using the random algorithm, the target number of target model combinations is extracted from the at least one group of first model combinations.

4. The model building method according to claim 3, wherein: The target object model is constructed by combining the target models, including: Based on the target quantity being the same as the preset quantity, the target model combinations of the target quantity are arranged to obtain the target object model.

5. The model building method according to claim 3, wherein: The step of determining a target quantity based on the preset quantity includes: Based on the preset number N being an even number, the target number is M, where M=N / 2; Based on the fact that the preset number N is an odd number, the target number is M, wherein, 6. The model building method according to claim 5, wherein: The target object model is constructed by combining the target models, including: Based on the fact that the preset number N is an even number, the M target model combinations are arranged in sequence according to a preset arrangement direction; Mirroring the M target model combinations to obtain M second model combinations; The target object model is constructed by combining the M target models and the M second models.

7. The model building method according to claim 5 or 6, wherein: The target object model is constructed by combining the target models, including: Based on the preset number N being an odd number, extracting M-1 third model combinations from the M target model combinations; Mirroring the M-1 third model combinations to obtain M-1 fourth model combinations; The target object model is constructed by the M-1 third model combinations, the M-1 fourth model combinations and the fifth model combination, the fifth model combination is located between the M-1 third model combinations and the M-1 fourth model combinations, and the fifth model combination is the other model combination among the M target model combinations except the M-1 third model combinations.

8. The model building method according to any one of claims 1 to 7, wherein: After constructing the target object model by combining the target models, the method further includes: Obtaining a scaling factor of the target object model; The target object model is scaled according to the scaling factor.

9. The model building method according to claim 8, wherein: The scaling process of the target object model according to the scaling factor includes: Scaling each target model combination in the target object model according to the scaling factor; Acquire a reference coordinate system of each kinematic pair in the target model combination, and a local coordinate system corresponding to each kinematic pair, wherein the local coordinate system is a coordinate system of the component model connected to the kinematic pair; The reference coordinate system is updated according to the local coordinate system and the scaling factor.

10. The model building method according to any one of claims 1 to 9, wherein: After constructing the target object model by combining the target models, the method further includes: Configuring material information for the target object model; Based on the material information, physical parameters are configured for the target object model.

11. The model building method according to claim 10, wherein: The physical parameter includes at least one of the following: mass, inertia tensor.

12. A model building device, wherein: include: An acquisition module, used for acquiring target size parameters and component size parameters of at least two component models, wherein the component models are used for constructing a target object model; A determination module, configured to determine a target model combination according to the target size parameter and the component size parameter, wherein the target model combination includes at least two of the component models; A configuration module, configured to configure a kinematic pair between any two of the component models in the target model combination according to the component type of each of the component models in the target model combination; The construction module is used to construct a target object model by combining the target models.

13. A readable storage medium having a program or instruction stored thereon, wherein: When the program or instruction is executed by a processor, the steps of the model building method according to any one of claims 1 to 11 are implemented.

14. A computer program product, wherein: When the computer program product is executed by a processor, the steps of the model building method according to any one of claims 1 to 11 are implemented.

15. An electronic device, wherein: include: The model building device as claimed in claim 12; and / or The readable storage medium according to claim 13; and / or The computer program product of claim 14.

16. A computer device, wherein: include: A processor and a memory, wherein the memory stores a program or instruction that can be run on the processor, and when the program or instruction is executed by the processor, the steps of the model building method according to any one of claims 1 to 11 are implemented.

Images