A robot morphology optimization method and apparatus

Abstract

The application provides a robot form optimization method and device, and relates to the technical field of artificial intelligence. The method comprises the following steps: obtaining a plurality of first form files to be optimized; wherein the first form file is an extensible markup language file; performing performance analysis on each first form file to obtain a text description corresponding to each first form file; wherein the text description indicates at least one of the advantages, disadvantages and improvement suggestions of the first form file; inputting at least one of the text description and a preset operation instruction and the first form file into a preset coding model to output at least one second form file. Through the generation of the text description and the construction of the preset operation instruction, the first form file in the form of the extensible markup language file can be directly optimized by using the preset coding model, and compared with the existing optimization mode by using an image model, the form optimization is unconstrained, efficient and directional.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method and apparatus for optimizing robot morphology. Background Technology

[0002] For robots, their functional performance fundamentally depends on their physical form; therefore, designing high-performance robot forms is a core challenge in robot development. Current robot form design methods primarily simplify the form file defining the robot's form and abstract it into an image model. The robot's form is then improved by adjusting the parameters of points or edges in the image. While this approach avoids the difficulty of directly modifying the original form file, it limits the expressive power and high degree of freedom of the original form file because it can only make slight perturbations to points or edges in the image. Therefore, the optimization of existing methods has limitations. Furthermore, optimization in image models is usually non-directional; that is, modifications to points or edges are randomly generated, and the success of the optimization can only be determined based on the robot's subsequent behavior, resulting in low optimization efficiency. Summary of the Invention

[0003] This invention provides a robot morphology optimization method and apparatus. By generating text descriptions and constructing preset operation instructions, it can directly optimize a first morphology file in the form of an extensible markup language file using a preset encoding model. Compared with the existing optimization method using image models, this method achieves unconstrained, efficient and directional morphology optimization.

[0004] This invention provides a robot morphology optimization method, comprising the following steps: acquiring multiple first morphology files to be optimized; wherein the first morphology files are Extensible Markup Language (XML) files; performing performance analysis on each first morphology file to obtain a text description corresponding to each first morphology file; wherein the text description indicates at least one of the advantages, disadvantages, and improvement suggestions of the first morphology file; inputting the text description and at least one of the preset operation instructions and the first morphology files into a preset encoding model, and outputting at least one second morphology file.

[0005] Optionally, the step of inputting at least one of the text description and preset operation instructions, along with the first form file, into a preset encoding model to output at least one second form file includes: inputting the text description and a single first form file into the preset encoding model to generate a second form file; and / or inputting the preset operation instructions and multiple first form files into the preset encoding model to generate a second form file; wherein the preset operation instructions indicate merging multiple first form files.

[0006] Optionally, obtaining the first morphological file to be morphologically optimized includes: obtaining multiple basic morphological files; selecting the first morphological file from the multiple basic morphological files based on the similarity between every two basic morphological files and / or the evaluation score of each basic morphological file; wherein the evaluation score is calculated based on an adaptive hierarchical algorithm.

[0007] Optionally, the method further includes: evaluating the fourth morphological file corresponding to the basic morphological file and the second morphological file respectively using an adaptive hierarchical algorithm to obtain a first evaluation score corresponding to the basic morphological file and a second evaluation score corresponding to the second morphological file; wherein the morphological performance of the fourth morphological file is better than that of the basic morphological file; and selecting a target morphological file from the second morphological file based on the first evaluation score and the second evaluation score.

[0008] Optionally, the first evaluation score includes: a first intermediate score obtained using the base controller and a third intermediate score obtained using the adaptive controller; the step of evaluating the base morphological file and the second morphological file using the adaptive hierarchical algorithm to obtain a first evaluation score corresponding to the base morphological file and a second evaluation score corresponding to the second morphological file includes: evaluating the fourth morphological file corresponding to the base morphological file and the second morphological file using the base controller to obtain a first intermediate score corresponding to the fourth morphological file and a second intermediate score corresponding to the second morphological file; selecting target second morphological files from the second morphological files whose second intermediate score is greater than a preset score; evaluating the fourth morphological file and the target second morphological file using the adaptive controller to obtain a third intermediate score corresponding to the base morphological file and a second evaluation score corresponding to the target second morphological file; wherein, the adaptive controller is obtained by training the base controller through reinforcement learning.

[0009] Optionally, multiple files of the first form may have the same machine characteristics.

[0010] The present invention also provides a robot morphology optimization device, comprising the following modules: The acquisition module is used to acquire multiple first morphology files to be morphologically optimized; wherein, the first morphology file is an Extensible Markup Language file; An analysis module is used to perform performance analysis on each of the first form files to obtain a text description corresponding to each of the first form files; wherein the text description indicates at least one of the advantages, disadvantages and improvement suggestions of the first form file; The generation module is used to input at least one of the text description and preset operation instructions, as well as the first form file, into a preset encoding model, and output at least one second form file.

[0011] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the robot morphology optimization method as described above.

[0012] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the robot morphology optimization method as described above.

[0013] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the robot morphology optimization method as described above.

[0014] The present invention provides a robot morphology optimization method and apparatus, which can directly optimize a first morphology file in the form of an extensible markup language file by generating text descriptions and constructing preset operation instructions. Compared with the existing optimization method through image models, it achieves unconstrained, efficient and directional morphology optimization. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 This is one of the flowcharts illustrating the robot morphology optimization method provided by the present invention.

[0017] Figure 2 This is a schematic diagram of the process for obtaining the first form file provided by the present invention.

[0018] Figure 3 This is the second flowchart of the robot morphology optimization method provided by the present invention.

[0019] Figure 4 This is a schematic diagram of the process for obtaining the first evaluation score and the second evaluation score provided by the present invention.

[0020] Figure 5 This is a schematic diagram of the robot morphology optimization device provided by the present invention.

[0021] Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention.

[0022] Figure 7 This is a comparative diagram of the robot morphology optimization method provided by the present invention and the robot morphology optimization method of the prior art. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0024] Figure 1 This is one of the flowcharts illustrating the robot morphology optimization method provided by the present invention, such as... Figure 1 As shown, the method includes the following: Step 101: Obtain multiple first morphology files to be optimized; wherein, the first morphology file is an Extensible Markup Language file; Extensible Markup Language (XML) files are structured files used by robots to define rigid bodies, joints, actuators, and their corresponding attribute configurations; they are also known as the robot's genotype.

[0025] In an optional embodiment, the first morphological file of this invention can be obtained by filtering a basic morphological file obtained according to existing technology (e.g., the image model method described in the background art). Specifically, the process of obtaining the first morphological file in step 101 is as follows: Figure 2 As shown, it includes: Step 201: Obtain multiple basic morphology files; Step 202: Select the first morphological file from multiple basic morphological files based on the similarity between every two basic morphological files and / or the evaluation score of each basic morphological file; wherein the evaluation score is calculated based on the adaptive hierarchical algorithm.

[0026] It should be noted that, in order to facilitate the verification of the robot morphology optimization method provided in the embodiments of the present invention, the morphology file of the current optimal solution is not selected as the basic morphology file in the embodiments of the present invention. Instead, the morphology file of the second-best solution is used as the first morphology file. In this way, after the optimization method of the embodiments of the present invention, the obtained second morphology file can be compared with the morphology file of the optimal solution obtained by optimization through the prior art, so as to prove in practice that the morphology optimization method provided in the embodiments of the present invention can obtain a better morphology file.

[0027] Specifically, in step 202, the first form file can be obtained in several ways. In one optional embodiment, to ensure the diversity of the first form file, text similarity can be calculated for every two basic form files, and the two basic form files with the highest text similarity can be used as the first form file. In another optional embodiment, the basic form files can be filtered based on their evaluation scores. Specifically, each basic form file can be evaluated using an adaptive hierarchical algorithm to obtain an evaluation score, which is then sorted, and the basic form files are filtered according to their scores. In yet another optional embodiment, the first form file can be obtained by randomly selecting from the basic form files. Here, the evaluation score can be understood as the Euclidean distance traveled from the starting point within a fixed time range.

[0028] Furthermore, in an optional embodiment, multiple first-form files possess the same machine characteristics. This is because robots typically possess different characteristics depending on the type of task they perform, their skeletal structure, or their physical properties. Optimizing with robots possessing the same characteristics allows for more targeted optimization, ensuring that the machine characteristics do not disappear during the optimization process and that a second-form file with those machine characteristics is always obtained. Therefore, in this embodiment of the invention, multiple first-form files with the same machine characteristics are preferentially selected for optimization. It is understood that after obtaining a second-form file with better performance in a single characteristic, further overall file optimization can be performed on multiple second-form files with different machine characteristics. That is, multiple second-form files with different machine characteristics can be used as new multiple first-form files, and subsequent steps 102 and 103 can be executed to achieve morphological optimization of multiple individual robots with multiple machine characteristics.

[0029] Step 102: Perform performance analysis on each first form file to obtain a text description corresponding to each first form file; wherein the text description indicates at least one of the advantages, disadvantages and improvement suggestions of the first form file; Specifically, the first-form file and preset prompts can be input into a preset semantic model to output a text description corresponding to each first-form file. For example, the preset prompts may include: setting the role of the preset semantic model as a robot expert, requiring a text description that conforms to robot physiology and XML file syntax requirements, and the text description contains specific and actionable improvement suggestions.

[0030] Step 103 inputs at least one of the text description and preset operation instructions, as well as the first form file, into the preset encoding model, and outputs at least one second form file.

[0031] Among them, the preset encoding model can be a large language model, such as the ChatGPT large model developed by OpenAI, the Doubao large model Yuanbao, etc.

[0032] In this step, the embodiments of the present invention provide two methods for generating the second-form file, which are described in detail below: In one optional embodiment, a text description and a single first-form file are input into a preset encoding model to generate a second-form file. It is understood that since the text description indicates at least one of the advantages, disadvantages, and improvement suggestions of the first-form file, after simultaneously inputting the text description and the first-form file into the preset encoding model, the first-form file can be directly improved based on the advantages, disadvantages, and improvement suggestions in the text description to obtain an optimized second-form file. This differs from existing technologies, where optimization is undirected; that is, perturbations to points or edges in the image are random, and whether the optimized version is better than the original is determined by the robot's subsequent actions. In this embodiment, however, by performing performance analysis on the first-form file beforehand, the specific direction of optimization can be obtained, enabling further optimization of the first-form file based on the text description to generate the corresponding second-form file.

[0033] In another optional embodiment, a preset operation instruction and multiple first-form files are input into a preset encoding model to generate a second-form file; wherein the preset operation instruction instructs the merging of multiple first-form files. For example, the preset operation instruction may be text describing the high-quality merging of the advantages of multiple first-form files, so that the generated second-form file can integrate the advantages of multiple first-form files, thereby possessing better morphological characteristics.

[0034] The following is based on Figure 7 Taking an example, the specific process of the robot morphology optimization method provided in this embodiment of the invention is compared and explained with that of the prior art robot morphology optimization method. Figure 7As shown in (a), in this embodiment of the invention, the first form file is first subjected to performance analysis using a preset semantic model, and at least one of the text description obtained from the performance analysis and preset operation instructions is input together with the first form file into a preset encoding model to obtain a second form file. Figure 7 As shown in (b), the prior art first converts the first morphological file into an image model composed of points and edges, and then through multiple iterations of optimization, makes small-scale modifications to the points and edges, and finally converts the resulting graphic into a second morphological file.

[0035] In this embodiment of the invention, combining a text description with a single first-form file can be understood as a mutation of the genotype, while merging multiple first-form files can be understood as a crossover of genotypes. In other words, this embodiment of the invention achieves the generation of the second-form file based on two different dimensions.

[0036] In a further optional embodiment, the present invention also performs relevant verification on the obtained second morphology file to ensure that the robot morphology optimization method provided by the present invention is better than the optimization method of the prior art, such as... Figure 3 As shown, the robot morphology optimization method provided in this embodiment of the invention includes: Step 301: Obtain multiple basic morphology files; Step 302: Based on the similarity between every two basic morphology files and / or the evaluation score of each basic morphology file, select the first morphology file from multiple basic morphology files; wherein, the first morphology file is an Extensible Markup Language file; Step 303: Perform performance analysis on each first form file to obtain a text description corresponding to each first form file; wherein the text description indicates at least one of the advantages, disadvantages and improvement suggestions of the first form file; Step 304: Input at least one of the text description and preset operation instructions, as well as the first form file, into the preset encoding model, and output at least one second form file; Step 305: Use the adaptive hierarchical algorithm to evaluate the fourth morphological file and the second morphological file corresponding to the basic morphological file respectively, and obtain the first evaluation score corresponding to the basic morphological file and the second evaluation score corresponding to the second morphological file; wherein, the morphological performance of the fourth morphological file is better than that of the basic morphological file; For example, the fourth morphological file is the morphological file of the optimal solution obtained by optimizing the image model in the prior art, and the basic morphological file is the morphological file of the suboptimal solution obtained by optimizing the image model in the prior art.

[0037] In this step, in an optional embodiment, the adaptive hierarchical algorithm may include two parts: an evaluation score obtained using the base controller and an evaluation score obtained using the adaptive controller. Therefore, the first evaluation score also includes two parts: a first intermediate score obtained using the base controller and a third intermediate score obtained using the adaptive controller. Specifically, the process of obtaining the second evaluation score in step 305 can be as follows: Figure 4 As shown, it specifically includes: Step 401: Use the basic controller to evaluate the fourth form file and the second form file corresponding to the basic form file respectively, and obtain the first intermediate score corresponding to the fourth form file and the second intermediate score corresponding to the second form file. Step 402: Filter the target second form files from the second form files whose second intermediate score is greater than the preset score; Step 403: Use the adaptive controller to evaluate the fourth morphological file and the target second morphological file respectively, and obtain the third intermediate score corresponding to the basic morphological file and the second evaluation score corresponding to the target second morphological file; wherein, the adaptive controller is obtained by training the basic controller through reinforcement learning.

[0038] The basic controller is directly obtained based on the existing "Transform-Control Strategy" (Transform2Act), while the adaptive controller is trained using reinforcement learning based on the basic controller, application scenario, and corresponding training data. In this embodiment, the basic controller can perform preliminary screening of the second-form files based on rules, allowing only target second-form files with a second intermediate score greater than a preset score to be further evaluated, effectively improving evaluation efficiency. For the fourth-form files, both the basic controller and the adaptive controller are used for evaluation, resulting in more convincing comparison results. It should be noted that the evaluation method for the fourth-form files can be selected according to actual needs. Since the adaptive controller requires model training, consuming additional resource costs, while the basic controller can be directly obtained and used, when computational resources are sufficient, both the basic controller and the adaptive controller can be used to evaluate the fourth-form files. When computational resources are insufficient, only the basic controller can be used for evaluation. Steps 401 to 403 are only a preferred embodiment and do not represent the only embodiment of this solution.

[0039] Step 306: Based on the first evaluation score and the second evaluation score, select the target morphological file from the second morphological file.

[0040] The first evaluation score can be either the first median score or the third median score, or it can be the average of the first and third median scores. The target format file is then selected from the second format files based on whether the second evaluation score is greater than the first evaluation score.

[0041] It can be seen that by comparing the second form file obtained from the embodiments of the present invention with the fourth form file obtained from the optimal solution based on the prior art, the second form file can be further screened to obtain a better target form file.

[0042] The following uses Table 1 as an example to illustrate the evaluation results between the second form file obtained by the embodiments of the present invention and the fourth form file obtained by optimization using various methods of the prior art.

[0043] Table 1 Evaluation Results The existing optimization methods are based on existing literature. Specifically, the methods for Evolutionary Structure Search (ESS) can be found at https: / / arxiv.org / abs / 1706.06133, the methods for Random Graph Search (RGS) can be found at https: / / arxiv.org / abs / 1906.05370, the methods for Neural Graph Evolution (NGE) can be found at https: / / arxiv.org / abs / 1906.05370, and the methods for Transform-Control Policy (Transform2Act) can be found at [link to relevant documentation]. For details on the design of the symmetric sensing robot SARD (https: / / arxiv.org / abs / 2110.03659), the specific method for designing the robot RoboMorph (https: / / arxiv.org / abs / 2306.00036), the specific method for optimizing the robot structure RoboMorph (https: / / arxiv.org / abs / 2407.08626), and the specific method for collaborative robot design RoboMoRe (https: / / arxiv.org / abs / 2506.00276), please refer to the webpage https: / / arxiv.org / abs / 2407.08626. Furthermore, the robot structure generation methods MorphoGen1, MorphoGen2, and MorphoGen3 are robot morphology optimization methods in this embodiment of the invention. MorphoGen1, MorphoGen2, and MorphoGen3 represent three different results obtained using the robot morphology optimization methods provided in this embodiment of the invention. MorphoGen1 exhibits the best evaluation performance, followed by MorphoGen2 and MorphoGen3. As shown in Table 1, for each method in the embodiments of the present invention, at least one structure was set up, and evaluation scores were obtained using both a basic controller and an adaptive controller. Comparison of the data in Table 1 shows that, regardless of whether the evaluation is performed using the basic controller or the adaptive controller, the structure obtained by the robot morphology optimization method provided in the embodiments of the present invention moves a significantly greater Euclidean distance in the same time period than many existing optimization methods. This indicates that the morphology optimization method provided in the embodiments of the present invention can effectively optimize based on extensible markup language files and has good optimization performance.

[0044] In summary, the robot morphology optimization method provided by this invention can directly optimize the first morphology file in the form of an extensible markup language file by generating text descriptions and constructing preset operation instructions. Compared with the existing optimization method using image models, it achieves unconstrained, efficient and directional morphology optimization.

[0045] The robot morphology optimization device provided by the present invention is described below. The robot morphology optimization device described below and the robot morphology optimization method described above can be referred to in correspondence.

[0046] like Figure 5 As shown, the robot morphology optimization device 500 provided by the present invention includes: The acquisition module 501 is used to acquire multiple first morphology files to be morphologically optimized; wherein, the first morphology file is an Extensible Markup Language file; Analysis module 502 is used to perform performance analysis on each of the first form files to obtain a text description corresponding to each of the first form files; wherein the text description indicates at least one of the advantages, disadvantages and improvement suggestions of the first form file; The generation module 503 is used to input at least one of the text description and preset operation instructions, as well as the first form file, into a preset encoding model, and output at least one second form file.

[0047] In an optional embodiment of the present invention, the generation module 503 is further configured to input the text description and a single first form file into the preset encoding model to generate a second form file; and / or input the preset operation instruction and multiple first form files into the preset encoding model to generate a second form file; wherein the preset operation instruction indicates that multiple first form files are merged.

[0048] In an optional embodiment of the present invention, the acquisition module 501 is further configured to acquire multiple basic morphological files; and to select the first morphological file from the multiple basic morphological files based on the similarity between every two basic morphological files and / or the evaluation score of each basic morphological file; wherein the evaluation score is calculated based on an adaptive hierarchical algorithm.

[0049] In an optional embodiment of the present invention, the apparatus further includes an evaluation module, configured to evaluate the fourth morphological file corresponding to the basic morphological file and the second morphological file respectively using an adaptive hierarchical algorithm to obtain a first evaluation score corresponding to the basic morphological file and a second evaluation score corresponding to the second morphological file; wherein the morphological performance of the fourth morphological file is better than that of the basic morphological file; and to select a target morphological file from the second morphological file based on the first evaluation score and the second evaluation score.

[0050] In an optional embodiment of the present invention, the first evaluation score includes: a first intermediate score obtained using a base controller and a third intermediate score obtained using an adaptive controller; the evaluation module is further configured to: evaluate a fourth morphological file and a second morphological file corresponding to a base morphological file using the base controller to obtain a first intermediate score corresponding to the fourth morphological file and a second intermediate score corresponding to the second morphological file; filter target second morphological files from the second morphological files whose second intermediate score is greater than a preset score; evaluate the fourth morphological file and the target second morphological file using the adaptive controller to obtain a third intermediate score corresponding to the base morphological file and a second evaluation score corresponding to the target second morphological file; wherein the adaptive controller is obtained by training the base controller through reinforcement learning.

[0051] In an optional embodiment of the present invention, a plurality of first form files have the same machine characteristics.

[0052] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6 As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communication interface 620, and the memory 630 communicate with each other through the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute a robot morphology optimization method, which includes: acquiring multiple first morphology files to be optimized; wherein the first morphology files are Extensible Markup Language (XML) files; performing performance analysis on each first morphology file to obtain a text description corresponding to each first morphology file; wherein the text description indicates at least one of the advantages, disadvantages, and improvement suggestions of the first morphology file; inputting the text description and at least one of the preset operation instructions and the first morphology files into a preset encoding model, and outputting at least one second morphology file.

[0053] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0054] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the robot morphology optimization method provided by the above methods. The method includes: acquiring a plurality of first morphology files to be optimized; wherein the first morphology files are Extensible Markup Language (XML) files; performing performance analysis on each first morphology file to obtain a text description corresponding to each first morphology file; wherein the text description indicates at least one of the advantages, disadvantages, and improvement suggestions of the first morphology file; inputting the text description and at least one of the preset operation instructions and the first morphology files into a preset encoding model, and outputting at least one second morphology file.

[0055] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a robot morphology optimization method provided by the above methods. The method includes: acquiring a plurality of first morphology files to be morphologically optimized; wherein the first morphology files are Extensible Markup Language (XML) files; performing performance analysis on each first morphology file to obtain a text description corresponding to each first morphology file; wherein the text description indicates at least one of the advantages, disadvantages, and improvement suggestions of the first morphology file; inputting the text description and at least one of preset operation instructions and the first morphology files into a preset encoding model, and outputting at least one second morphology file.

[0056] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0057] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for optimizing robot morphology, characterized in that, include: Obtain multiple first morphology files to be morphologically optimized; wherein, the first morphology file is an Extensible Markup Language file; A performance analysis is performed on each of the first form files to obtain a text description corresponding to each of the first form files; wherein the text description indicates at least one of the advantages, disadvantages and improvement suggestions of the first form file; The text description and at least one of the preset operation instructions, along with the first form file, are input into a preset encoding model to output at least one second form file.

2. The robot morphology optimization method according to claim 1, characterized in that, The step of inputting at least one of the text description and preset operation instructions, along with the first form file, into a preset encoding model and outputting at least one second form file includes: The text description and the single first form file are input into the preset encoding model to generate a second form file; And / or, The preset operation instruction and multiple first-form files are input into a preset encoding model to generate a second-form file; wherein, the preset operation instruction indicates that multiple first-form files are merged.

3. The robot morphology optimization method according to claim 1, characterized in that, The process of obtaining the first morphological file to be optimized includes: Obtain multiple basic morphology files; The first morphological file is obtained by filtering from a plurality of the base morphological files based on the similarity between every two base morphological files and / or the evaluation score of each base morphological file. The evaluation score is calculated based on an adaptive hierarchical algorithm.

4. The robot morphology optimization method according to claim 3, characterized in that, Also includes: An adaptive hierarchical algorithm is used to evaluate the fourth morphological file and the second morphological file corresponding to the basic morphological file, respectively, to obtain a first evaluation score corresponding to the basic morphological file and a second evaluation score corresponding to the second morphological file; wherein, the morphological performance of the fourth morphological file is better than that of the basic morphological file; Based on the first evaluation score and the second evaluation score, target morphological files are selected from the second morphological files.

5. The robot morphology optimization method according to claim 4, characterized in that, The first evaluation score includes: a first intermediate score obtained using the basic controller and a third intermediate score obtained using the adaptive controller; the step of evaluating the basic morphological file and the second morphological file respectively using an adaptive hierarchical algorithm to obtain a first evaluation score corresponding to the basic morphological file and a second evaluation score corresponding to the second morphological file includes: The basic controller is used to evaluate the fourth form file and the second form file corresponding to the basic form file respectively, to obtain a first intermediate score corresponding to the fourth form file and a second intermediate score corresponding to the second form file. Filter the second form files from the second form files to find target second form files whose second intermediate score is greater than the preset score; An adaptive controller is used to evaluate the fourth morphological file and the target second morphological file respectively, to obtain a third intermediate score corresponding to the basic morphological file and a second evaluation score corresponding to the target second morphological file; wherein, the adaptive controller is obtained by training the basic controller through reinforcement learning.

6. The robot morphology optimization method according to claim 4, characterized in that, Multiple files of the first form have the same machine characteristics.

7. A robot morphology optimization device, characterized in that, include: The acquisition module is used to acquire multiple first morphology files to be morphologically optimized; wherein, the first morphology file is an Extensible Markup Language file; An analysis module is used to perform performance analysis on each of the first form files to obtain a text description corresponding to each of the first form files; wherein the text description indicates at least one of the advantages, disadvantages and improvement suggestions of the first form file; The generation module is used to input at least one of the text description and preset operation instructions, as well as the first form file, into a preset encoding model, and output at least one second form file.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the robot morphology optimization method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the robot morphology optimization method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the robot morphology optimization method as described in any one of claims 1 to 6.

Images