A design method and related device for shoe outsoles

Abstract

This application discloses a design method and related apparatus for shoe outsoles, relating to the field of information processing. The method includes: receiving an initial design pattern for the shoe outsole; performing mechanical prediction on at least two intermediate design patterns generated by a preset reinforcement learning algorithm based on the initial design pattern using a preset generative adversarial network (GAN) model, obtaining a mechanical prediction result for each intermediate design pattern, wherein each intermediate design pattern differs from the initial design pattern in at least a portion; and determining a target design pattern for the shoe outsole based at least on the mechanical prediction results of each intermediate design pattern, wherein the mechanical prediction result of the target design pattern is superior to the mechanical prediction results of non-target design patterns. This application, based on a preset GAN model and a preset reinforcement learning algorithm, provides real-time mechanical performance evaluation of the design pattern during the design phase, eliminating the need to wait for sample production to complete the evaluation, thus improving the overall efficiency and performance of shoe outsole design.

Description

Technical Field

[0001] This application relates to the field of information processing, and in particular to a design method and related apparatus for shoe outsoles. Background Technology

[0002] In the modern footwear design industry, technological advancements have brought about several revolutionary solutions, among which parametric generation technology and mechanical analysis technology are the most noteworthy.

[0003] Parametric generation technology uses advanced software tools to allow designers to automatically generate sole designs by adjusting parameters such as sole thickness, protrusion shape, and layout.

[0004] Mechanical analysis techniques, especially finite element analysis (FEA), are also widely used in the field of shoe sole design. This technology helps designers identify potential weaknesses during the design phase by simulating and predicting the mechanical problems that materials may encounter in actual use, such as pressure distribution, tension, and compressive response.

[0005] However, the evaluation process of mechanical performance using the parametric generation technology and mechanical analysis technology mentioned above is quite complex. In the actual design process, designers cannot directly and in real time evaluate the mechanical performance of the product. They need to wait until the sample is made and then conduct an evaluation through physical testing. This not only increases the cost of product development but also prolongs the time from product design to market and reduces the overall market response speed. Summary of the Invention

[0006] The first aspect of this application provides a method for designing a shoe outsole, including:

[0007] Receive the initial design pattern for the shoe outsole;

[0008] By using a pre-defined generative adversarial network model, a pre-defined reinforcement learning algorithm performs mechanical prediction on at least two intermediate design patterns generated based on the initial design pattern, and obtains the mechanical prediction result for each intermediate design pattern, wherein each intermediate design pattern is at least partially different from the initial design pattern.

[0009] Based at least on the mechanical prediction results of each intermediate design pattern, a target design pattern for the outsole is determined, wherein the mechanical prediction results of the target design pattern are better than those of the non-target design patterns.

[0010] In one possible implementation, the step of using a preset generative adversarial network model to perform mechanical prediction on at least two intermediate design patterns generated by a preset reinforcement learning algorithm based on the initial design pattern, and obtaining the mechanical prediction result for each intermediate design pattern, includes:

[0011] By using a pre-defined generative adversarial network model, a mechanical prediction is performed on the first design pattern to obtain the mechanical prediction result of the first design pattern, which includes an initial design pattern and an intermediate design pattern.

[0012] If the mechanical prediction result of the first design pattern does not meet the requirements of the shoe outsole design, a portion of the first design pattern is adjusted according to a preset reinforcement learning algorithm to obtain a second design pattern. The second design pattern is then used as the new first design pattern. The process returns to the step of performing mechanical prediction on the first design pattern using a preset generative adversarial network model to obtain the mechanical prediction result of the first design pattern.

[0013] In one possible implementation, determining the target design pattern of the outsole based at least on the mechanical prediction results of each intermediate design pattern includes:

[0014] Based on the mechanical prediction results of the first design pattern, which meet the requirements of the shoe outsole design, the first design pattern is determined as the target design pattern.

[0015] In one possible implementation, determining the target design pattern of the outsole based at least on the mechanical prediction results of each intermediate design pattern includes:

[0016] Record the number of times the second design pattern is obtained;

[0017] If the number of times is the same as the preset number of times, and the mechanical prediction result obtained by using the second design pattern as the new first design pattern meets the requirements of the shoe outsole design, the second design pattern is determined as the target design pattern.

[0018] In one possible implementation, if the mechanical prediction result based on the first design pattern does not meet the shoe outsole design requirements, a portion of the first design pattern is adjusted according to a preset reinforcement learning algorithm to obtain a second design pattern, including:

[0019] Based on the mechanical prediction results of the first design pattern, at least the number of stress concentration points of the first design pattern shall be determined;

[0020] Since the number of stress concentration points in the first design pattern does not meet the target number corresponding to the shoe outsole design requirements, a preset reinforcement learning algorithm is used to assign a reward value to the first design pattern based on the number of stress concentration points in the first design pattern. The reward value is related to the number of stress concentration points in the first design pattern.

[0021] By using a preset reinforcement learning algorithm, a target adjustment strategy is determined based on the number of stress concentration points in the first design pattern and the reward value of the first design pattern. A portion of the first design pattern is then adjusted based on the target adjustment strategy to obtain a second design pattern.

[0022] In one possible implementation, prior to receiving the initial design pattern of the shoe outsole, the process further includes:

[0023] Based on the mechanical design requirements of the shoe outsole, a preset mechanical prediction model is obtained;

[0024] Based on the preset mechanical prediction model, a training sample set is determined, which includes several shoe outsole training patterns and the mechanical prediction results corresponding to the training patterns;

[0025] An initial generative adversarial network model is trained based on the training sample set to obtain a preset generative adversarial network model.

[0026] In one possible implementation, after determining the target design pattern of the outsole based at least on the mechanical prediction results of each intermediate design pattern, the method further includes:

[0027] Receive the layout parameters of the shoe sole pattern;

[0028] Based on the layout parameters and the target design pattern, the overall design pattern of the shoe outsole is generated.

[0029] A second aspect of this application provides a design device for a shoe outsole, comprising:

[0030] A receiving module is used to receive the initial design pattern of the shoe outsole;

[0031] The prediction module is used to perform mechanical prediction on at least two intermediate design patterns generated by a preset reinforcement learning algorithm based on the initial design pattern through a preset generative adversarial network model, and to obtain the mechanical prediction result of each intermediate design pattern, wherein each intermediate design pattern is at least partially different from the initial design pattern.

[0032] A determination module is used to determine a target design pattern for the outsole based at least on the mechanical prediction results of each intermediate design pattern, wherein the mechanical prediction results of the target design pattern are better than those of the non-target design patterns.

[0033] A third aspect of this application provides a computer program product including computer-readable instructions that, when executed on an electronic device, cause the electronic device to implement the shoe outsole design method of the first aspect or any implementation thereof.

[0034] A fourth aspect of this application provides an electronic device, including at least one processor and a memory connected to the processor, wherein:

[0035] The memory is used to store computer programs;

[0036] The processor is used to execute the computer program so that the electronic device can implement the shoe outsole design method of the first aspect or any implementation thereof.

[0037] The fifth aspect of this application provides a computer storage medium carrying one or more computer programs that, when executed by an electronic device, enable the electronic device to implement the shoe outsole design method described in the first aspect or any implementation thereof. Attached Figure Description

[0038] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0039] Figure 1 This is a schematic flowchart illustrating a shoe outsole design method provided in an embodiment of this application;

[0040] Figure 2 This is a schematic diagram of the initial design pattern provided in the embodiments of this application;

[0041] Figure 3 These are intermediate design patterns and schematic diagrams of mechanical prediction results provided in the embodiments of this application;

[0042] Figure 4 This is a schematic diagram of the target design pattern provided in the embodiments of this application;

[0043] Figure 5 This is a flowchart illustrating how a preset generative adversarial network model performs mechanical prediction on at least two intermediate design patterns generated by a preset reinforcement learning algorithm based on the initial design pattern, thereby obtaining the mechanical prediction result for each intermediate design pattern, according to an embodiment of this application.

[0044] Figure 6 A schematic diagram of a generative adversarial network model provided in the embodiments of this application;

[0045] Figure 7 This is a flowchart illustrating the process of determining the target design pattern of the shoe outsole based on the mechanical prediction results of at least each intermediate design pattern, as provided in the embodiments of this application.

[0046] Figure 8This is a flowchart illustrating how the mechanical prediction results based on the first design pattern provided in this application embodiment do not meet the shoe outsole design requirements, and how the second design pattern is obtained by adjusting a portion of the first design pattern according to a preset reinforcement learning algorithm.

[0047] Figure 9 This is a schematic diagram of the reward value curve for reinforcement learning provided in the embodiments of this application;

[0048] Figure 10 This is a schematic diagram of the process for obtaining a preset generative adversarial network model provided in an embodiment of this application;

[0049] Figure 11 This is a schematic diagram illustrating the process of applying the target design pattern provided in this application embodiment to the outsole of a shoe to obtain the overall design pattern;

[0050] Figure 12 This is a schematic diagram of the overall design pattern of the shoe outsole provided in the embodiments of this application. Figure 1 ;

[0051] Figure 13 This is a schematic diagram of the overall design pattern of the shoe outsole provided in the embodiments of this application. Figure 2 ;

[0052] Figure 14 This is a schematic diagram of the structure of a shoe outsole design device provided in an embodiment of this application;

[0053] Figure 15 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0054] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is for explaining specific embodiments only and is not intended to limit the scope of this application.

[0055] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0056] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0057] Reference Figure 1 , Figure 1 This is a flowchart illustrating a shoe outsole design method provided in an embodiment of this application, as shown below. Figure 4 As shown in the embodiment of this application, a shoe outsole design method may include steps 101 to 103, which are described in detail below.

[0058] 101. Receive the initial design pattern for the shoe outsole;

[0059] The initial design pattern is the original design pattern created by the designers based on the requirements.

[0060] As an example, the initial design pattern could be a rhombus, trapezoid, or other shape.

[0061] The initial design pattern is the main pattern applied to the outsole of the shoe, and multiple patterns can be set in the outsole.

[0062] Figure 2 This is a schematic diagram of the initial design pattern provided in the embodiments of this application. The schematic diagram shows two initial design patterns: initial design pattern 201 is a square and initial design pattern 202 is a scale shape.

[0063] 102. Using a pre-defined generative adversarial network model, a pre-defined reinforcement learning algorithm performs mechanical prediction on at least two intermediate design patterns generated based on the initial design pattern, and obtains the mechanical prediction result for each intermediate design pattern, wherein each intermediate design pattern is at least partially different from the initial design pattern;

[0064] The electronic device implementing this method has a preset generative adversarial network model and a preset reinforcement learning algorithm. Through the preset reinforcement learning algorithm, multiple intermediate design patterns are obtained by adjusting the initial design pattern. The preset generative adversarial network model can predict the mechanical prediction results of these multiple intermediate design patterns and obtain the mechanical prediction results of each intermediate design pattern.

[0065] As an example, if the mechanical prediction is a stress prediction, then stress prediction is performed for each intermediate design pattern to obtain the corresponding stress prediction result.

[0066] The mechanical prediction results can be presented in the form of images. Of course, the mechanical prediction results can be displayed on the screen of an electronic device or not, and can be set according to the actual situation. This application does not impose any restrictions.

[0067] The Generative Adversarial Network (GAN) model is a neural network model that can be trained.

[0068] In one possible implementation, a large number of training samples can be determined based on a mechanics prediction model, and the original model can be trained using these training samples to obtain a pre-defined generative adversarial network model. This process is followed by... Figure 10 The details are explained in the text.

[0069] The intermediate design pattern is obtained by making partial adjustments to the initial design pattern, and the various intermediate design patterns are not exactly the same.

[0070] Figure 3 This is a schematic diagram of the intermediate design pattern and mechanical prediction results provided in the embodiments of this application. The intermediate design pattern in the schematic diagram is consistent with... Figure 2 The schematic diagram shows two intermediate design patterns 301 and 302 corresponding to the initial design pattern 201. The two intermediate design patterns are mechanically predicted using a pre-set generative adversarial network model, and the corresponding mechanical prediction results 303-304 are obtained. The mechanical prediction result schematic diagram 303 corresponds to the intermediate design pattern 301, and the mechanical prediction result 304 corresponds to the intermediate design pattern 302.

[0071] In one possible implementation, a pre-defined generative adversarial network (GAN) model can be used to perform mechanical prediction on a first design pattern to obtain the corresponding mechanical prediction result. If the mechanical prediction result of the first design pattern does not meet the design requirements, a second design pattern is generated according to a pre-defined reinforcement learning algorithm. This second design pattern is then processed as a new first design pattern to obtain the mechanical prediction result of the new first design pattern. Subsequent... Figure 5 The paper details the process by which a preset generative adversarial network model performs mechanical predictions on at least two intermediate design patterns generated by a preset reinforcement learning algorithm based on the initial design pattern, and obtains the mechanical prediction results for each intermediate design pattern.

[0072] 103. Based at least on the mechanical prediction results of each intermediate design pattern, determine the target design pattern of the outsole, wherein the mechanical prediction results of the target design pattern are better than those of the non-target design patterns.

[0073] Among them, based on the mechanical prediction results of each intermediate design pattern, the one with a mechanical prediction result that is better than the other intermediate design patterns is determined as the target design pattern.

[0074] In one possible implementation, the mechanical prediction result of the target design pattern is better than the mechanical prediction result of the non-target design pattern, and the mechanical prediction result of the target design pattern meets the requirements of the shoe outsole design. The process of determining the target design pattern of the shoe outsole is described in detail in the following embodiments.

[0075] Figure 4 This is a schematic diagram of the target design pattern provided in the embodiments of this application. The target design pattern 401 in the schematic diagram is... Figure 2 The target design pattern, corresponding to the initial design pattern 202, is obtained by making partial adjustments to the initial design pattern. The overall shape of the target design pattern is consistent with the initial design pattern, but the internal details are different.

[0076] In one possible implementation, after determining the target design pattern of the shoe outsole, the arrangement of that target design pattern within the outsole can also be set, and subsequently... Figure 11 The process of applying the target design pattern to the outsole of a shoe to obtain the overall design pattern is explained in detail.

[0077] In this embodiment, an initial design pattern for the shoe outsole is obtained. A preset generative adversarial network (GAN) model is used to perform mechanical prediction on multiple intermediate design patterns generated by a preset reinforcement learning algorithm based on the initial design pattern. The mechanical prediction result for each intermediate design pattern is obtained, and each intermediate design pattern differs from the initial design pattern in at least a portion. Based on the mechanical prediction results of these intermediate design patterns, an intermediate design pattern with a superior mechanical prediction result is selected as the target design pattern. This process integrates the preset GAN model and the preset reinforcement learning algorithm, generating multiple intermediate design patterns and performing mechanical prediction on each of them. Mechanical prediction of the design pattern can be achieved during the design phase, enabling real-time mechanical performance evaluation of the design without waiting for the sample to be manufactured. This improves the overall efficiency and performance of the shoe outsole design.

[0078] Figure 5This application provides a flowchart illustrating how a preset generative adversarial network model performs mechanical prediction on at least two intermediate design patterns generated by a preset reinforcement learning algorithm based on the initial design pattern, thereby obtaining the mechanical prediction result of each intermediate design pattern. The flowchart may include steps 501 to 502, which are described in detail below.

[0079] 501. By using a pre-set generative adversarial network model, a mechanical prediction is performed on the first design pattern to obtain the mechanical prediction result of the first design pattern, which includes an initial design pattern and an intermediate design pattern.

[0080] The preset generative adversarial network model is capable of making mechanical predictions on the input pattern and obtaining the mechanical prediction results.

[0081] In one possible implementation, the mechanical prediction result can be represented in the form of a mechanical distribution diagram.

[0082] As an example, the mechanical prediction is a prediction of the stress for the first design pattern, and the resulting mechanical prediction is a stress diagram of the first design pattern.

[0083] The first design pattern can be an initial design pattern or an intermediate design pattern obtained by adjusting the initial design pattern.

[0084] The pre-defined generative adversarial network is a neural network model that may include a generator and a discriminator.

[0085] The generator is responsible for creating a corresponding mechanical distribution map from a given shoe sole design image, while the discriminator is tasked with evaluating the authenticity of these mechanical distribution maps, i.e., the similarity between the mechanical distribution map and the actual mechanical distribution map.

[0086] In one possible implementation, the generator of the pre-defined generative adversarial network employs a modified Unet structure, capable of generating high-resolution, richly detailed mechanical distribution maps.

[0087] The generator utilizes a deep symmetric Unet structure, which enables efficient feature extraction and image reconstruction. In the downsampling layers, multiple convolutional layers progressively capture high-level features, with each convolution potentially followed by batch normalization and LeakyReLU activation to enhance the model's non-linear processing capabilities. In the upsampling layers, transposed convolutional layers progressively amplify the feature maps, restoring image detail and resolution. Furthermore, the dropout mechanism in the upsampling layers improves the model's generalization ability. Skip connections preserve deep features, crucial for accurate reconstruction of stress images.

[0088] In one possible implementation, the pre-defined discriminator of the generative adversarial network (GAN) employs a PatchGAN design, enabling local realism evaluation of each small region in the stress image, which significantly improves sensitivity to details. By merging the target image and the generated image, features are extracted layer by layer, and finally, a specific convolutional operation is used to provide an evaluation of the region's realism.

[0089] In one possible implementation, the loss function of the generative adversarial network is pre-defined, combining GAN loss and L1 loss. GAN loss ensures that the generated images are visually difficult to distinguish from real images, while L1 loss emphasizes pixel-level consistency, ensuring that the generated stress images are highly consistent with the actual images in terms of detail.

[0090] Figure 6 The schematic diagram of the generative adversarial network model provided in this embodiment includes a generator 601 and a discriminator 602; wherein, the generator 601 generates a plurality of images 603, which include real images and fake images, and the images are respectively input into the discriminator 602, which judges the input images to obtain prediction results.

[0091] In this embodiment, the image generated by the generator is a mechanical distribution map generated based on the input first design pattern. The discriminator discriminates the mechanical distribution maps to obtain a prediction result. The prediction result may include a label for each mechanical distribution map. The label may be true or false. The mechanical distribution map with the label "true" may be output as the mechanical distribution map predicted by the generative adversarial network model for the first design pattern.

[0092] The pre-defined generative adversarial network is trained based on an adversarial training strategy, and subsequently... Figure 10 The training process is described in detail in the document.

[0093] 502. If the mechanical prediction result of the first design pattern does not meet the requirements of the shoe outsole design, adjust part of the first design pattern according to the preset reinforcement learning algorithm to obtain a second design pattern. Use the second design pattern as the new first design pattern and return to execute the step of performing mechanical prediction on the first design pattern through the preset generative adversarial network model to obtain the mechanical prediction result of the first design pattern.

[0094] The outsole design requirement can be the mechanical requirement of the outsole. It can be determined whether the mechanical prediction result of the first design pattern meets the outsole design requirement. If not, the first design pattern is adjusted to obtain the second design pattern.

[0095] The first design pattern can be adjusted according to a preset chemical strengthening algorithm to obtain the second design pattern.

[0096] The reinforcement learning algorithm can employ the Q-learning reinforcement algorithm.

[0097] If the mechanical prediction result of the first design pattern does not meet the requirements of the shoe outsole design, the first design pattern is adjusted according to the preset reinforcement learning algorithm to obtain a second design pattern. The second design pattern is then used as a new first design pattern, and a preset generative adversarial network model is used to perform mechanical prediction on the new first design pattern to obtain a new mechanical prediction result. This process is repeated.

[0098] In one possible implementation, the subsequent process of determining the target design pattern can be based on the mechanical prediction results of the first design pattern meeting the shoe outsole design requirements, thereby determining the first design pattern as the target design pattern.

[0099] As an example, if the outsole design requirement is a stress concentration point requirement, then the number of stress concentration points in the mechanical prediction result of the first design pattern can be determined, and then it can be determined whether the outsole design requirement is met based on the number. If the number of stress concentration points in the mechanical prediction result meets the outsole design requirement, the first design pattern is determined as the target design pattern; otherwise, a second design pattern is generated based on the first design pattern.

[0100] In one possible implementation, stress concentration points can be determined based on the mechanical prediction results of the first design pattern. A pre-defined reinforcement learning algorithm is then used to determine an adjustment strategy for the image based on these stress concentration points. The first design pattern is then adjusted according to this strategy to obtain a second design pattern. Subsequent... Figure 8 The Chinese side will provide a detailed explanation of this process.

[0101] In one possible implementation, a preset number of iterations can be set. Based on whether the preset number of iterations is used and whether the mechanical prediction results of the first design pattern meet the shoe outsole design requirements, a target design pattern can be determined, followed by... Figure 7 The process is explained in detail in the document.

[0102] The process of adjusting the first design pattern to obtain the second design pattern is an optimization process of the design pattern, so that the mechanical prediction result of the final target design pattern can meet the design requirements of the shoe outsole.

[0103] In this embodiment, a pre-set generative adversarial network is used to perform mechanical prediction on the first design pattern to obtain the corresponding mechanical prediction result. This enables real-time mechanical prediction of the design pattern. Based on the mechanical prediction result, it is determined whether the shoe outsole design requirements are met. If not, a portion of the first design pattern is adjusted according to a pre-set reinforcement learning algorithm to obtain a second design pattern. This second design pattern is then used as the new first design pattern for mechanical prediction. This process is repeated until a design pattern that meets the shoe outsole design requirements is obtained. This allows for the determination of a design pattern that meets the shoe outsole design requirements during the design stage, eliminating the need to wait for the sample to be made before conducting mechanical evaluation, thus improving the overall efficiency and performance of the shoe outsole design.

[0104] Figure 7 This is a flowchart illustrating the process of determining the target design pattern of the shoe outsole based on the mechanical prediction results of at least each intermediate design pattern, as provided in the embodiments of this application. It may include steps 701 to 702, which are described in detail below.

[0105] 701. Record the number of times the second design pattern is obtained;

[0106] The number of times the second design pattern is obtained each time can be used as the current iteration number.

[0107] When the first design pattern is the initial design pattern, the second design pattern generated based on the initial design pattern is the first intermediate design pattern. At this time, the number of times the second design pattern is obtained is recorded as 1. When the second design pattern is generated subsequently, the recorded number is accumulated to record the number of times the second design pattern is obtained.

[0108] 702. If the number of times is the same as the preset number of times, and the mechanical prediction result obtained by using the second design pattern as the new first design pattern meets the requirements of the shoe outsole design, the second design pattern is determined as the target design pattern.

[0109] The preset number of iterations is the set number of iterations.

[0110] The number of iterations is set to balance the mechanical performance of the target design pattern with the preservation of the original pattern. Fewer iterations preserve the shape of the original pattern better, while more iterations improve mechanical performance, but excessive iterations can lead to the loss of the original pattern shape. The final target design pattern is the optimal one obtained through a finite number of iterations.

[0111] If the number of iterations performed so far is the same as the preset number, it indicates that the number of iterations performed has reached the set number of iterations, and the mechanical test results of the second design pattern meet the requirements of the shoe outsole design, then the second design pattern is determined as the target design pattern.

[0112] If the number of iterations does not meet the preset number of iterations, then the current iteration count has not reached the set number of iterations to be executed, and the aforementioned step 502 continues to be executed.

[0113] In this embodiment, the number of times the second design pattern is obtained is recorded each time. If the number of times is the same as the preset number of times, the mechanical prediction result obtained based on the second design pattern is determined to meet the design requirements of the shoe outsole. Then, the second design pattern is determined as the target design pattern. By setting the number of optimization iterations, a design pattern with better performance is obtained while retaining the shape of the initial design pattern.

[0114] Figure 8 The flowchart of the second design pattern obtained by adjusting a portion of the first design pattern according to a preset reinforcement learning algorithm when the mechanical prediction result based on the first design pattern provided in the embodiments of this application does not meet the design requirements of the shoe outsole can include steps 801 to 803. These steps are described in detail below.

[0115] 801. Based on the mechanical prediction results of the first design pattern, determine at least the number of stress concentration points of the first design pattern;

[0116] In this embodiment, stress is used as a factor in mechanical performance analysis, and the stress is used to optimize the pattern.

[0117] The mechanical prediction results are analyzed and calculated to obtain the number and distribution of stress concentration points in the first design pattern.

[0118] 802. Since the number of stress concentration points in the first design pattern does not meet the target number corresponding to the shoe outsole design requirements, a reward value is assigned to the first design pattern based on the number of stress concentration points in the first design pattern through a preset reinforcement learning algorithm. The reward value is related to the number of stress concentration points in the first design pattern.

[0119] The first design pattern can be converted into a two-dimensional matrix of a specific size. The elements in the matrix are 0 or 1. An element value of 1 indicates that there is material at that position, while an element value of 0 indicates that the position is hollow (no material). In this embodiment, the adjustment of the first design pattern is to adjust the elements in the two-dimensional matrix from 1 to 0, so as to reduce the number of stress concentration points in the design pattern and reduce material, thereby achieving the effect of performance enhancement and material reduction.

[0120] As an example, the two-dimensional matrix can be 32×32. Of course, depending on the actual situation, other sizes of two-dimensional matrices can be set, and this application does not impose any restrictions.

[0121] Specifically, the design pattern is adjusted by using a pre-set reinforcement learning algorithm, Q-learning, to obtain a design pattern with better mechanical properties, thereby realizing an iterative optimization process for the design pattern.

[0122] The preset reinforcement learning algorithm can optimize the design pattern based on the mechanical test results of the first design pattern.

[0123] Q-learning is an algorithm in the field of reinforcement learning that does not rely on a model of the environment and learns the optimal policy directly through interaction with the environment.

[0124] The core of Q-Learning is learning a value function called the Q-function, which assigns a reward value (Q-value) to each pair of states (s) and actions (a). This value represents the sum of expected future rewards that can be obtained if action a is performed in state s and the optimal policy is followed afterwards.

[0125] In one possible implementation, the initial value in the Q table is set to 0.

[0126] The process begins by creating an initial Q-table, where each row corresponds to a possible state and each column to a possible action. The intersection of the row and column represents the Q-value for that action in that state. Then, an action is selected at each time step; the Q-learning algorithm chooses an action for the current state. After executing the selected action, the next state and immediate reward are observed. Finally, the Q-table is updated based on the observed reward and the maximum Q-value for the next state, updating the values ​​corresponding to the current state and action in the Q-table.

[0127] The selection process can be implemented using an ε-greedy strategy, which selects the action with the highest Q value most of the time, but has a small probability of randomly selecting any action to ensure the exploratory nature of the algorithm.

[0128] Each time an action is performed, a new design pattern is obtained, and the Q table is updated based on the mechanical test results of the new design pattern.

[0129] The Q-table is updated according to the classic Bellman equation, with the goal of optimizing the Q-value to maximize long-term reward. The reward function is designed to be directly related to the reduction of stress concentration, meaning that a pattern with more uniform stress distribution receives a higher reward.

[0130] The selection action can be to randomly select a matrix element in the first design pattern and replace its corresponding element value from 1 to 0.

[0131] In this Q-learning, a reward is set during the learning process. The more stress concentration points there are, the lower the reward is, and the fewer stress concentration points there are, the higher the reward is. Therefore, the reinforcement learning algorithm will learn in the direction of fewer stress concentration points.

[0132] During the adjustment process, the design pattern will be optimized in a way that disperses stress and reduces the number of concentration points.

[0133] In one possible implementation, a reward value can be assigned to the first design pattern based on the number of stress concentration points; the more stress concentration points, the lower the reward; the fewer stress concentration points, the higher the reward.

[0134] In one possible implementation, the change in the number of stress concentration points of the first design pattern relative to the number of stress concentration points of the first design pattern in the previous iteration step can be correlated with the reward; the greater the reduction in the number of stress concentration points, the greater the reward value.

[0135] 803. Using a preset reinforcement learning algorithm, a target adjustment strategy is determined based on the number of stress concentration points in the first design pattern and the reward value of the first design pattern. A portion of the first design pattern is adjusted based on the target adjustment strategy to obtain a second design pattern.

[0136] Specifically, based on the number of stress concentration points and their reward values ​​in the first design pattern, a target adjustment strategy is determined. This target adjustment strategy can adjust a portion of the first design pattern, resulting in a second design pattern with a smaller number of stress concentration points than the first design pattern, thus achieving an optimization process for the design pattern.

[0137] This adjustment strategy corresponds to the action in the Q table.

[0138] Through multiple iterations, the system gradually learns and updates its adjustment strategy to generate an optimal pattern that reduces stress concentration.

[0139] In each optimization process, the action selection at each step is not only based on the current Q value to execute the best action (utilization), but also includes a certain probability of random action selection (exploration) to avoid local optima and enhance learning efficiency.

[0140] The pre-defined reinforcement learning algorithm updates and adjusts its strategy. In each iteration, the system generates a new, better design pattern based on the reward value of the current design pattern and the adjustment strategy. By combining the selection of random actions with the selection of the current best action, the reinforcement learning algorithm avoids getting trapped in local optima, thereby gradually optimizing the design pattern.

[0141] As an example, the design requirement for a shoe outsole with a design pattern is to distribute stress as evenly as possible under load, minimizing stress concentration points. First, an initial design pattern A is randomly generated. Calculations show that this design has 5 stress concentration points, resulting in a low reward value. In the first iteration, a new design pattern B is generated based on the current adjustment strategy. Calculations show that this design has 3 stress concentration points, with a higher reward value than design A. The reinforcement learning algorithm updates its adjustment strategy by comparing the reward values ​​of design A and design B, tending to generate patterns similar to design B. Next, a new design pattern C is generated. Calculations show that this design has 4 stress concentration points, with a reward value between design A and design B. The system combines the reward values ​​of design B and design C to further adjust the strategy. After multiple iterations, the system gradually optimizes the design patterns, finally generating design D, which has only 1 stress concentration point and obtains the highest reward value.

[0142] In this process, the action selection at each step is not only based on the current Q value to execute the best action, but also includes a certain probability of random action selection to avoid local optima and enhance learning efficiency.

[0143] Figure 9 This is a schematic diagram of the reward value curve for reinforcement learning provided in this application embodiment. The horizontal axis represents the number of iterations, and the vertical axis represents the reward value. Curve 901 is the reward value curve obtained by determining the action based on the current optimal Q value, and also includes the selection of random actions with a certain probability. Curve 902 is the reward value curve obtained by determining the action only using the current optimal Q value. Figure 9 As can be seen from the curves, the efficiency of the two curves is almost the same, while curve 902 will form the optimal layout solution. Therefore, the selection exploration method that combines the optimal Q value and random actions is more effective.

[0144] In this embodiment, based on the mechanical prediction results of the first design pattern, at least the number of stress concentration points in the first design pattern is determined. Since the number of stress concentration points in the first design pattern does not meet the target number required for the shoe outsole design, a preset reinforcement learning algorithm assigns a reward value to the first design pattern based on the number of stress concentration points. This reward value is related to the number of stress concentration points in the first design pattern. Using the preset reinforcement learning algorithm, a target adjustment strategy is determined based on the number of stress concentration points in the first design pattern and the reward value of the first design pattern. A portion of the first design pattern is adjusted based on this target adjustment strategy to obtain a second design pattern. By continuously adjusting and optimizing the strategy, an optimal design pattern with more uniform stress distribution and fewer stress concentration points is gradually generated, thereby achieving the design optimization goal.

[0145] Figure 10 This is a schematic diagram of the process of obtaining a preset generative adversarial network model provided in the embodiments of this application. Before receiving the initial design pattern of the shoe outsole, the process may include steps 1001 to 1003, which are described in detail below.

[0146] 1001. Based on the mechanical design requirements of the shoe outsole, obtain the preset mechanical prediction model;

[0147] Different mechanical prediction models are selected for different types of mechanical design requirements.

[0148] As an example, if the mechanical design requirement is a more uniform stress distribution and fewer stress concentration points, then a stress prediction model is selected.

[0149] In addition, the material properties of the shoe outsole can be combined to select a mechanical prediction model that meets the mechanical design requirements corresponding to the material properties.

[0150] In one approach, the material property is used as a parameter input to the mechanical prediction model. The material is a rubber material used in shoe soles. Its density and Young's modulus are obtained through tensile testing and measurement and are used as material properties input into the mechanical prediction model to provide more realistic simulation results.

[0151] The preset mechanical prediction model can predict the mechanical properties of an input image and obtain the corresponding mechanical prediction results.

[0152] The pre-defined mechanical prediction model transforms the different patterns of shoe sole particles into coarse-grained "atoms" through 3D modeling, connecting all adjacent coarse-grained "atoms". After mapping this model onto real-world dimensions, a model with a suitable level of detail can be obtained by adjusting the resolution.

[0153] In this embodiment, a sole particle corresponds to an element in the two-dimensional matrix.

[0154] The boundary conditions for the simulation were set as follows: the positions of the bottom atoms were fixed, the forces on the top atoms were fixed, and then pressure was applied from top to bottom.

[0155] In this study, under different pressures, an energy minimization algorithm is used to optimize the compressed structure and obtain the smallest possible internal stress.

[0156] Furthermore, because the pre-defined mechanical prediction model in this embodiment uses coarse-grained assumptions, the computational speed is improved, making it possible to prepare the dataset. Thus, by meticulously simulating the behavior of shoe sole materials under stress, a dataset containing high-quality, high-precision data on stress, potential energy, and other parameters is obtained. These datasets meticulously reflect the various mechanical responses that different designs may encounter in actual use, providing a foundation for training the generative adversarial network model. During the simulation, various physical properties of the material (such as elastic modulus and yield strength) and environmental conditions (such as temperature and humidity) are considered to ensure that the dataset accurately reflects real-world conditions.

[0157] 1002. Based on the preset mechanical prediction model, determine the training sample set, which includes several shoe outsole training patterns and the mechanical prediction results corresponding to the training patterns;

[0158] In this process, several shoe outsole training patterns are randomly generated, and the training patterns are input into the preset mechanical prediction model to obtain the corresponding mechanical prediction results. Each training pattern and its corresponding mechanical prediction result are used as a pair of training samples, and the several training patterns can be used to obtain a training sample set composed of several training sample pairs.

[0159] In one possible implementation, a series of randomly generated outsole patterns can be used to predict the stress on the outsole patterns using a preset mechanical prediction model, resulting in a dataset containing stress and potential energy data. The data in the dataset can then be normalized to obtain training samples.

[0160] 1003. Train the initial generative adversarial network model based on the training sample set to obtain the preset generative adversarial network model.

[0161] Specifically, the initial generative adversarial network model is trained based on each training sample in the training sample set to obtain the pre-trained generative adversarial network model.

[0162] The training process involves tuning the parameters of the generative adversarial network model. By using backpropagation and gradient descent methods, the parameters of the generator and discriminator are continuously adjusted to reduce the deviation between the predicted image and the actual image, thereby improving the prediction accuracy of the model.

[0163] The generative adversarial network model includes a generator and a discriminator.

[0164] The generator can perform several functions, including layer-by-layer feature extraction, feature compression and encoding, and feature reconstruction and output.

[0165] The layer-by-layer feature extraction function is designed so that the image is first input into the encoder and then undergoes a combination of multiple Conv2D, BatchNormalization, and LeakyReLU operations. Each Conv2D operation extracts image features through convolutional kernels, BatchNormalization normalizes activation values ​​to accelerate training and stabilize the model, and the LeakyReLU activation function introduces non-linearity and prevents the "neuron death" problem. As the number of layers increases, the spatial resolution gradually decreases while the number of feature channels gradually increases, thereby extracting increasingly higher-level image features.

[0166] The feature compression and encoding function involves the encoder layer by layer reducing the spatial size of the feature map while increasing the number of channels in the feature map, which helps to capture global features at low resolution.

[0167] As an example, the minimum feature map size is taken as (1, 1, 512), representing the global feature vector.

[0168] In this feature decoding and reconstruction process, the decoder uses Conv2D_transpose (deconvolution) operations to restore the spatial size of the feature map layer by layer, while reducing the number of channels. The deconvolution operation effectively restores the low-resolution feature map to a high-resolution one by learning upsampling weights. In the first few layers of the decoder, a Dropout layer is introduced to randomly discard the output of some neurons as a regularization method to prevent overfitting and improve the model's generalization ability.

[0169] In the training of this stable model, after each deconvolution layer of the decoder, BatchNormalization is applied to further normalize the activation values, ensuring the stability and efficiency of the training process.

[0170] In this feature reconstruction and output process, the decoder restores the size of the feature map layer by layer until it is restored to the same size as the input image. At the same time, it aggregates features by reducing the number of channels and finally outputs the feature-reconstructed image.

[0171] The training process of this discriminator involves training its ability to distinguish between real and fake images output by the generator. It can perform local authenticity assessments for each small region of the stress image, which greatly improves its sensitivity to details. By merging the target image and the generated image, features are extracted layer by layer, and finally, a region authenticity assessment is given through specific convolution operations.

[0172] Moreover, during training, the loss functions used are GAN loss and L1 loss to ensure that the images generated by the trained model are visually difficult to distinguish from real images, and that the generated stress images are highly consistent with the actual images in terms of detail.

[0173] In this embodiment, a corresponding preset mechanical prediction model is obtained based on the mechanical design requirements of the shoe outsole. A training sample set is determined based on this preset mechanical prediction model, which includes several shoe outsole training patterns and the corresponding mechanical prediction results. An initial generative adversarial network (GAN) model is trained based on this training sample set to obtain a preset GAN model. A training sample set is then generated based on the preset mechanical prediction model to train the preset GAN model. This trained preset GAN model can quickly process input images to obtain corresponding mechanical prediction results, providing a technical basis for improving the efficiency of shoe outsole design.

[0174] Figure 11 This is a schematic diagram of the process of applying the target design pattern to the outsole of a shoe to obtain the overall design pattern, according to an embodiment of this application. Before receiving the initial design pattern of the outsole, the process may include steps 1101 to 1102, which are described in detail below.

[0175] 1101. Receive the layout parameters of the shoe sole pattern;

[0176] The layout parameters can be set by the designer according to actual needs, can be obtained by adjusting the preset layout parameters, or can be parameter values ​​directly input by the designer.

[0177] The arrangement parameters may include the spacing and angle of the input points, as well as other sole-related parameters other than the target design pattern, such as the sole outline.

[0178] The input point is the location of the center of the target design pattern on the outsole of the shoe.

[0179] In one possible implementation, a web interface can be pre-defined, which provides designers with an interface to adjust the spacing, angle, and other data of the input points. This web interface has the functions of data input and result preview.

[0180] In one possible implementation, the desired shoe sole outline and the desired layout design image can be input on the web interface. The image is then processed into a format that the device program can read.

[0181] In one possible implementation, the web interface provides a slider, which designers can use to adjust the horizontal and vertical spacing of the design images, as well as the angle and size parameters of the design images.

[0182] 1102. Based on the layout parameters and the target design pattern, generate the overall design pattern of the shoe outsole.

[0183] Specifically, the target design pattern in the outsole is arranged according to the arrangement parameters to generate the overall design pattern of the outsole.

[0184] In one possible implementation, the target design pattern determined in the preceding steps is arranged according to the input layout parameters and the image of the shoe sole outline, and finally the overall design pattern of the overall layout is given.

[0185] The overall design pattern can be saved as an SVG (Scalable Vector Graphics) file.

[0186] Figure 12 This is a schematic diagram of the overall design pattern of the shoe outsole provided in the embodiments of this application. Figure 1 In this schematic diagram, ○ represents the target design pattern, and several target design patterns are arranged in the outsole of the shoe.

[0187] Figure 13 This is a schematic diagram of the overall design pattern of the shoe outsole provided in the embodiments of this application. Figure 2 In this schematic diagram, the target design pattern corresponds to the aforementioned Figure 4 The target design pattern is scale-shaped. Several target design patterns are arranged on the outsole of the shoe. A character area 1301 is set in the center of the outsole. This area can be used to set the manufacturer's logo or other characters, but not the target design pattern. A symmetrical blank area 1302 is set in the foot area, and a blank area 1303 is set in the middle area. The plane of the blank area can be lower than the plane of the area where the target design pattern is located.

[0188] Among them, in Figure 13 In this application, blank areas are set only in the forefoot and middle areas, or blank areas can be set in the forefoot, heel, and middle areas. The location of the blank areas can be set according to the actual situation, and there are no restrictions in this application.

[0189] In one possible implementation, the target design drawing can be arranged on the outsole of the shoe according to preset layout parameters to obtain the overall design pattern.

[0190] Among them, mechanical prediction can be performed based on the overall design pattern of the generated outsole to obtain the optimal layout, which is convenient for designers to operate. The mechanical evaluation of the entire outsole can be carried out without waiting for the sample to be made, thus improving the overall efficiency and performance of the outsole design.

[0191] In this embodiment, the layout parameters of the sole pattern are received; based on the layout parameters and the determined target design pattern, the overall design pattern of the outsole is generated. Based on the designer's needs, the layout parameters can be adjusted to form a variety of point distributions, enabling the design of multiple point distribution scenarios and improving the flexibility and efficiency of outsole design.

[0192] The above describes a shoe outsole design method provided by the embodiments of this application. The following will describe the apparatus for performing the above shoe outsole design method.

[0193] Please see Figure 14 , Figure 14 This is a structural schematic diagram of a shoe outsole design device provided in an embodiment of this application. Figure 14 As shown, the shoe outsole design device 1400 includes: a receiving module 1401, a prediction module 1402, and a determination module 1403;

[0194] The receiving module 1401 is used to receive the initial design pattern of the shoe outsole;

[0195] The prediction module 1402 is used to perform mechanical prediction on at least two intermediate design patterns generated by a preset reinforcement learning algorithm based on the initial design pattern through a preset generative adversarial network model, and to obtain the mechanical prediction result of each intermediate design pattern, wherein each intermediate design pattern is at least partially different from the initial design pattern.

[0196] The determining module 1403 is used to determine the target design pattern of the outsole based at least on the mechanical prediction results of each intermediate design pattern, wherein the mechanical prediction results of the target design pattern are better than the mechanical prediction results of the non-target design patterns.

[0197] In one possible implementation, the prediction module includes:

[0198] The prediction unit is used to perform mechanical prediction on a first design pattern by using a preset generative adversarial network model, and obtain the mechanical prediction result of the first design pattern, which includes an initial design pattern and an intermediate design pattern.

[0199] The adjustment unit is used to adjust a portion of the first design pattern according to a preset reinforcement learning algorithm if the mechanical prediction result of the first design pattern does not meet the design requirements of the shoe outsole, to obtain a second design pattern, and to use the second design pattern as a new first design pattern. The unit then returns to execute the step of performing mechanical prediction on the first design pattern through a preset generative adversarial network model to obtain the mechanical prediction result of the first design pattern.

[0200] In one possible implementation, the determining module is specifically used for:

[0201] Based on the mechanical prediction results of the first design pattern, which meet the requirements of the shoe outsole design, the first design pattern is determined as the target design pattern.

[0202] In one possible implementation, the determining module includes:

[0203] A recording unit is used to record the number of times the second design pattern is obtained;

[0204] The determining unit is used to determine the second design pattern as the target design pattern if the number of times is the same as the preset number of times, based on the mechanical prediction result obtained by taking the second design pattern as the new first design pattern, which meets the shoe outsole design requirements.

[0205] In one possible implementation, the adjustment unit is specifically used for:

[0206] Based on the mechanical prediction results of the first design pattern, at least the number of stress concentration points of the first design pattern shall be determined;

[0207] Since the number of stress concentration points in the first design pattern does not meet the target number required for the shoe outsole design, a preset reinforcement learning algorithm is used to assign a reward value to the first design pattern based on the number of stress concentration points in the first design pattern. The reward value is related to the number of points in the first design pattern.

[0208] By using a preset reinforcement learning algorithm, a target adjustment strategy is determined based on the number of stress concentration points in the first design pattern and the reward value of the first design pattern. Based on the target adjustment strategy, a portion of the first design pattern is adjusted to obtain a second design pattern.

[0209] One possible implementation also includes:

[0210] The module is used to obtain a preset mechanical prediction model based on the mechanical design requirements of the outsole before receiving the initial design pattern of the outsole.

[0211] The sample determination module is used to determine a training sample set based on the preset mechanical prediction model. The training sample set includes several shoe outsole training patterns and the mechanical prediction results corresponding to the training patterns.

[0212] The training module is used to train an initial generative adversarial network model based on the training sample set, thereby obtaining a preset generative adversarial network model.

[0213] One possible implementation also includes:

[0214] The parameter receiving module is used to receive the layout parameters of the shoe sole pattern;

[0215] The generation module permanently generates the overall design pattern of the shoe outsole based on the layout parameters and the target design pattern.

[0216] It should be noted that the functional explanations of the various components in the shoe outsole design device provided in this application embodiment are as described in the foregoing method embodiment, and will not be repeated in this embodiment.

[0217] In this embodiment, an initial design pattern for the shoe outsole is obtained. A preset generative adversarial network (GAN) model is used to perform mechanical prediction on multiple intermediate design patterns generated by a preset reinforcement learning algorithm based on the initial design pattern. The mechanical prediction result for each intermediate design pattern is obtained, and each intermediate design pattern differs from the initial design pattern in at least a portion. Based on the mechanical prediction results of these intermediate design patterns, an intermediate design pattern with a superior mechanical prediction result is selected as the target design pattern. This process integrates the preset GAN model and the preset reinforcement learning algorithm, generating multiple intermediate design patterns and performing mechanical prediction on each of them. Mechanical prediction of the design pattern can be achieved during the design phase, enabling real-time mechanical performance evaluation of the design without waiting for the sample to be manufactured. This improves the overall efficiency and performance of the shoe outsole design.

[0218] This application also provides an electronic device in its embodiments. (See reference...) Figure 15 The diagram illustrates a structural schematic suitable for implementing the electronic device in the embodiments of this application. The electronic device in the embodiments of this application may include, but is not limited to, fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 15 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0219] like Figure 15As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 1501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1502 or a program loaded from a storage device 1508 into a random access memory (RAM) 1503. When the electronic device is powered on, the RAM 1503 also stores various programs and data required for the operation of the electronic device. The processing unit 1501, ROM 1502, and RAM 1503 are interconnected via a bus 1504. An input / output (I / O) interface 1505 is also connected to the bus 1504.

[0220] Typically, the following devices can be connected to I / O interface 1505: input devices 1506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 1507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1508 including, for example, memory cards, hard drives, etc.; and communication devices 1509. Communication device 1509 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 15 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0221] This application also provides a computer program product including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the shoe outsole design methods provided in this application.

[0222] This application also provides a computer-readable storage medium carrying one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the shoe outsole design methods provided in this application.

[0223] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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. In addition, in the accompanying drawings of the device embodiments provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0224] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, 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 is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods of the various embodiments of this application.

[0225] In the above embodiments, the implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, in the form of a computer program product.

[0226] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

Claims

1. A method for designing a shoe outsole, characterized in that, include: Receive the initial design pattern for the shoe outsole; By using a pre-defined generative adversarial network model, a mechanical prediction is performed on the first design pattern to obtain the mechanical prediction result of the first design pattern. The first design pattern includes an initial design pattern and intermediate design patterns, and each intermediate design pattern is at least partially different from the initial design pattern. Based on the mechanical prediction results of the first design pattern, at least the number of stress concentration points of the first design pattern shall be determined; Since the number of stress concentration points in the first design pattern does not meet the target number corresponding to the shoe outsole design requirements, a preset reinforcement learning algorithm is used to assign a reward value to the first design pattern based on the number of stress concentration points in the first design pattern. The reward value is related to the number of stress concentration points in the first design pattern. By using a preset reinforcement learning algorithm, a target adjustment strategy is determined based on the number of stress concentration points in the first design pattern and the reward value of the first design pattern. A portion of the first design pattern is adjusted based on the target adjustment strategy to obtain a second design pattern. The second design pattern is then used as a new first design pattern. The process returns to the step of performing mechanical prediction on the first design pattern using a preset generative adversarial network model to obtain the mechanical prediction result of the first design pattern. Based at least on the mechanical prediction results of each intermediate design pattern, a target design pattern for the outsole is determined, wherein the mechanical prediction results of the target design pattern are better than those of the non-target design patterns.

2. The shoe outsole design method according to claim 1, characterized in that, The determination of the target design pattern for the outsole, based at least on the mechanical prediction results of each intermediate design pattern, includes: Based on the mechanical prediction results of the first design pattern, which meet the requirements of the shoe outsole design, the first design pattern is determined as the target design pattern.

3. The shoe outsole design method according to claim 1, characterized in that, The determination of the target design pattern for the outsole, based at least on the mechanical prediction results of each intermediate design pattern, includes: Record the number of times the second design pattern is obtained; If the number of times is the same as the preset number of times, and the mechanical prediction result obtained by using the second design pattern as the new first design pattern meets the requirements of the shoe outsole design, the second design pattern is determined as the target design pattern.

4. The shoe outsole design method according to claim 1, characterized in that, Before receiving the initial design pattern of the shoe outsole, the following is also included: Based on the mechanical design requirements of the shoe outsole, a preset mechanical prediction model is obtained; Based on the preset mechanical prediction model, a training sample set is determined, which includes several shoe outsole training patterns and the mechanical prediction results corresponding to the training patterns; An initial generative adversarial network model is trained based on the training sample set to obtain a preset generative adversarial network model.

5. The shoe outsole design method according to claim 1, characterized in that, After determining the target design pattern of the outsole based at least on the mechanical prediction results of each intermediate design pattern, the method further includes: Receive the layout parameters of the shoe sole pattern; Based on the layout parameters and the target design pattern, the overall design pattern of the shoe outsole is generated.

6. A design device for a shoe outsole, characterized in that, include: A receiving module is used to receive the initial design pattern of the shoe outsole; The prediction module is used to perform mechanical prediction on a first design pattern by using a preset generative adversarial network model, and obtain the mechanical prediction result of the first design pattern. The first design pattern includes an initial design pattern and intermediate design patterns, and each intermediate design pattern is at least partially different from the initial design pattern. Based on the mechanical prediction results of the first design pattern, at least the number of stress concentration points of the first design pattern shall be determined; Since the number of stress concentration points in the first design pattern does not meet the target number corresponding to the shoe outsole design requirements, a preset reinforcement learning algorithm is used to assign a reward value to the first design pattern based on the number of stress concentration points in the first design pattern. The reward value is related to the number of stress concentration points in the first design pattern. By using a preset reinforcement learning algorithm, a target adjustment strategy is determined based on the number of stress concentration points in the first design pattern and the reward value of the first design pattern. A portion of the first design pattern is adjusted based on the target adjustment strategy to obtain a second design pattern. The second design pattern is then used as a new first design pattern. The process returns to the step of performing mechanical prediction on the first design pattern using a preset generative adversarial network model to obtain the mechanical prediction result of the first design pattern. A determination module is used to determine a target design pattern for the outsole based at least on the mechanical prediction results of each intermediate design pattern, wherein the mechanical prediction results of the target design pattern are better than those of the non-target design patterns.

7. A computer program product, characterized in that, Includes computer-readable instructions that, when executed on an electronic device, cause the electronic device to implement the shoe outsole design method as described in any one of claims 1 to 5.

8. An electronic device, characterized in that, It includes at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is used to execute the computer program to enable the electronic device to implement the shoe outsole design method as described in any one of claims 1 to 5.

Images