AI-driven game animation and creative derivative product intelligent design system

Abstract

This invention relates to the field of intelligent design technology for cultural and creative derivative products, specifically an AI-driven intelligent design system for game and animation cultural and creative derivative products, including: element analysis, potential assessment, concept generation, prototype construction, and solution decision-making modules. The element analysis module imports stylized digital assets of the target work, and forms a preliminary design element pool through element stripping and decomposition; the potential assessment module evaluates and predicts the element pool and marks core design elements; the concept generation module performs retrieval and distillation based on a multi-source knowledge graph to generate a derivative product concept blueprint that integrates diverse needs; the prototype construction module drives related models to iterate and output a high-fidelity prototype; and the solution decision-making module selects the optimal design solution and production parameters through multi-dimensional simulation. This system realizes intelligent derivative product design, accurately selects effective design elements, and ensures that the concept solution takes into account cultural, functional, and manufacturing needs, improving the adaptability of the design solution for implementation.

Description

Technical Field

[0001] This invention relates to the field of intelligent design technology for cultural and creative products, and in particular to an AI-driven intelligent design system for game and animation cultural and creative products. Background Technology

[0002] Currently, the design of game, animation, and cultural and creative derivative products largely relies on manual extraction of relevant design elements from target works. These elements are then compiled into a design element library based on subjective experience. Derivative product concepts are often developed based on the designer's personal aesthetics and scattered market information. Prototype production and solution selection also depend on manual iteration and offline evaluation. When manually organizing digital assets of works, only simple element breakdown is performed, without targeted analysis of the elements' subsequent development value. Derivative product concepts are generated based on only a single dimension of information, failing to integrate market, user, and production information.

[0003] The lack of unified evaluation standards in manually selecting design elements makes it impossible to accurately judge the aesthetic expansion potential and practical functional transformation possibilities of elements. A large number of elements without derivative value are mixed in the material library, affecting the efficiency of subsequent design. In the process of generating derivative product concepts, the needs of cultural narrative, practical function, and production and manufacturing cannot be integrated simultaneously. The generated conceptual solutions are prone to problems such as being disconnected from the core of the work, having unreasonable functions, and being difficult to mass-produce.

[0004] To address the lack of standardized methods for evaluating the aesthetic adaptability and functional convertibility of design elements, and the inability to accurately label core design elements, a standardized method for element evaluation and labeling is needed. Furthermore, to address the lack of multi-dimensional knowledge support for derivative product concept generation, and the inability to integrate diverse needs to generate adaptable conceptual solutions, a knowledge system integrating multiple types of information needs to be built, enabling precise information extraction and solution generation. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing an AI-driven intelligent design system for game and animation cultural and creative derivative products.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an AI-driven intelligent design system for game and animation cultural and creative derivative products, comprising: The element parsing module imports the stylized digital assets of the target work and performs narrative element stripping and physical element decomposition on the stylized digital assets to form a primary design element pool. The potential assessment module performs an aesthetic plasticity assessment and a functional convertibility prediction on the primary design element pool. Based on the assessment and prediction results, it marks the core design elements with derivative potential in the primary design element pool. The concept generation module establishes a multi-source knowledge graph containing historical market data, user community interaction corpus, and production supply chain path knowledge. Based on the core design elements, it performs path retrieval and knowledge distillation in the graph to generate a derivative product concept blueprint that integrates cultural narrative, practical functions, and manufacturing constraints. The prototype building module, based on the aforementioned derivative concept blueprint, drives the collaborative work of the generative adversarial network architecture and the neural radiation field model to perform multiple rounds of stylized rendering, 3D reconstruction, and structural optimization iterations on the core design elements, outputting a high-fidelity derivative prototype set. The solution decision module performs manufacturability simulation, user interaction simulation, and market response prediction on the set of high-fidelity derivative prototypes. Based on the multi-dimensional performance data obtained from the simulation, modeling, and prediction, it filters, sorts, and integrates the data, and finally outputs the optimal derivative design scheme and its associated production parameters.

[0007] As a further aspect of the present invention, the stylized digital assets are subjected to narrative element stripping and materialization element decomposition, including: The primary design element pool includes symbol features, scene features, and character features; The stylized digital assets are analyzed using scene graph analysis technology to identify and extract the entity objects, relationships between objects, and event sequences contained therein, thus forming an initial narrative structure; For the entity objects in the initial narrative structure, the visual features, text descriptions, and semantic tags of the entity objects are decoupled using an attribute classification model and assigned to visual feature vectors, text description vectors, and semantic tag vectors, respectively. Establish a cross-modal association mapping between the visual feature vector, text description vector, and semantic tag vector of the entity object, and filter out the association groups that represent the key narrative identity of the entity object based on the mapping strength; The feasibility of materializing the association group is determined by assessing the achievable boundaries of the association group in terms of form, texture, and volume, based on the material association database and the morphological structure database. The associated groups that are determined to be materializable, together with the encoded fragments with high emotional value in the event sequence, are encapsulated into independent design elements, and then aggregated to form the primary design element pool containing symbolic features, scene features, and character features.

[0008] As a further aspect of the present invention, the aesthetic plasticity assessment and functional convertibility prediction are performed on the primary design element pool, including: Extract the core visual attributes of each design element from the primary design element pool, and calculate the matching degree with the historical and future trend styles in the trend style library to obtain the aesthetic trend conformity of each design element; At the same time, the functional metaphors of each design element are deconstructed, and based on the common sense knowledge base, the potential physical or interactive functions that each design element can be given are inferred to generate a list of functional transformations. A joint evaluation matrix is ​​constructed, which integrates the aesthetic trend conformity with the number of functions and the complexity of function implementation in the functional transformation list, and calculates the comprehensive plasticity index of each design element. Set a plasticity threshold, and mark design elements whose comprehensive plasticity index exceeds the plasticity threshold as core design elements with derivation potential; The marked core design elements and their corresponding aesthetic trend conformity and functional transformation list are linked and stored to form the input index for subsequent knowledge retrieval.

[0009] As a further aspect of the present invention, path retrieval and knowledge distillation are performed in the graph based on the core design elements, including: Using the core design elements as query nodes, multi-hop path exploration is performed in the multi-source knowledge graph to find multiple knowledge paths connecting to market best-selling product category nodes, user preference tag nodes, and mature production process nodes; The confidence level and information abundance of the multiple knowledge paths are evaluated and ranked. The knowledge paths with confidence levels higher than a predetermined value and higher information abundance rankings are retained to form candidate knowledge subgraphs. An attention mechanism is used to reweight the entities and relationships in the candidate knowledge subgraph to distill out the cultural narrative fragments, practical function combinations, and manufacturing constraints that are most closely related to the core design elements. The distilled cultural narrative fragments, practical functional combinations, and manufacturing constraints are creatively reorganized and logically stitched together using the core design elements as a framework to generate a structured derivative concept description. The description of the derivative concept is formalized and parameterized, and finally transformed into a derivative concept blueprint that integrates cultural narrative, practical function, and manufacturing constraints.

[0010] As a further aspect of the present invention, the driving generative adversarial network architecture works in conjunction with the neural radiation field model, including: Using the aforementioned derivative concept blueprint as a control condition, the generator module of the generative adversarial network is input to guide it in generating a set of two-dimensional concept sketches with various styles and perspectives that conform to the semantics of the aforementioned derivative concept blueprint. The sketch with the highest quality score is selected from the set of two-dimensional concept sketches and used as the input to the neural radiation field model. At the same time, combined with the structural constraint parameters in the derivative concept blueprint, the neural radiation field model is used to reconstruct the preliminary three-dimensional geometry and surface material. A joint discrimination task is constructed in the discriminator module of the generative adversarial network to simultaneously judge the artistic quality of the two-dimensional sketches output by the generator module, as well as the realism and consistency of the three-dimensional images rendered from different perspectives by the neural radiation field model. The feedback signal from the joint discrimination task is simultaneously backpropagated to the generator module of the generative adversarial network and the neural radiation field model for end-to-end joint optimization. After multiple rounds of joint optimization iterations, a set of high-fidelity derivative prototypes that meet the requirements in both two-dimensional artistic expression and three-dimensional structural entities is finally obtained.

[0011] As a further aspect of the present invention, the core design elements undergo multiple rounds of stylized rendering, 3D reconstruction, and structural optimization iterations, including: In each iteration, the generator of the generative adversarial network first receives the three-dimensional view rendering provided by the previous neural radiation field model as a style reference, and combines it with the derivative concept blueprint to output a new round of two-dimensional concept sketches that are more consistent in three dimensions. The neural radiation field model receives a new round of two-dimensional conceptual sketches and integrates structural priors from the three-dimensional geometry generated in the previous round to optimize the details and surface properties of the three-dimensional geometry and perform three-dimensional reconstruction. After 3D reconstruction, automatic structural mechanics analysis and supportability testing are performed on the obtained geometry to identify weak or non-physically manufactured geometric parts. Based on the results of the structural mechanics analysis and supportability test, structural optimization suggestions are generated, and the structural optimization suggestions are transformed into a geometric constraint loss function that can be understood by the neural radiation field model. The geometric constraint loss function is incorporated into the training objective of the neural radiation field model, driving it to synchronously optimize the geometric structure in the next round of reconstruction to meet manufacturing requirements, thus completing a single round of structural optimization iteration.

[0012] As a further aspect of the present invention, manufacturability simulation is performed on the set of high-fidelity derivative prototypes, including: For each prototype in the high-fidelity derivative prototype set, automatically match the candidate production process that best matches the manufacturing constraints in the derivative concept blueprint, and load the corresponding process parameterization model from the digital process library; The geometric and material data of each prototype are input into the matching process parameterization model to simulate the entire manufacturing process from raw material processing, forming, machining to surface treatment. In the simulation process, material utilization rate, estimated processing time, predicted yield of each process and tool path conflict detection results are calculated and recorded in real time to form a manufacturing feasibility report. For prototypes with tool path conflicts or low yield, an automatic geometry repair program is initiated to fine-tune the prototype based on the constraints of the process parameterization model, and the repair effect is verified in a simulation environment. Finally, the manufacturing feasibility reports of all prototypes after simulation and repair are compiled to generate a simulation results dataset that includes quantitative manufacturing costs, working hours, and technical risk indicators.

[0013] As a further aspect of the present invention, user interaction simulation and market response prediction are performed on the high-fidelity derivative prototype set, including: Construct a virtual test environment containing different user profiles, and import the 3D model and estimated physical properties of each prototype in the simulation result dataset into the virtual test environment; In the virtual testing environment, a standardized interactive task flow is designed for each user profile, including visual search, retrieval, use, and storage steps, and the task completion time, operation path, and virtual gaze point heatmap data are recorded. Collect real-time sentiment computing data of user profiles generated during simulated interaction and post-interview virtual interview texts. Quantify the emotional experience value and perceived value brought by each prototype through sentiment analysis models and text mining models. The emotional experience value, perceived value, and aesthetic feature vector of each prototype are correlated with historical market sales data and social media volume data to train a market response prediction model. Using the market response prediction model, the potential market share, price sensitivity, and word-of-mouth dissemination index of each prototype in the high-fidelity derivative prototype set are predicted to form market prediction data.

[0014] As a further aspect of the present invention, the step of filtering, sorting, and fusing the multidimensional performance data obtained from simulation, modeling, and prediction includes: A multi-dimensional decision-making space is established, whose dimensions include the technical implementation dimension from the manufacturing feasibility report, the user experience dimension from the user interaction simulation, and the business value dimension from the market forecast data; For each prototype in the set of high-fidelity derivative prototypes, calculate its coordinates in the multidimensional decision space, and calculate the point cloud distribution formed by the coordinate points of all prototypes. A multi-objective optimization algorithm is applied to find the Pareto optimal front in the point cloud distribution, and the prototypes located on the Pareto optimal front are selected as the winning candidate set. The prototypes in the winning candidate set are weighted and ranked according to their performance in different decision-making dimensions, with the weights dynamically adjusted by the strategic objectives set by the project. The features of the top-ranked prototypes are deconstructed, and their advantageous features in a specific dimension are algorithmically fused to generate a fused prototype scheme that theoretically outperforms all individual prototypes, which serves as the final candidate.

[0015] As a further aspect of the present invention, the final output of the optimal derivative design scheme and its associated production parameters includes: The final round of simulation and modeling of the fusion prototype scheme was conducted to verify the stability of its comprehensive performance indicators; Extract the geometric model, material texture, and color scheme of the fusion prototype solution that are confirmed in the final verification, and generate a final design file that can be directly edited by 3D software; From the verification process of the fusion prototype solution, the optimized process parameters, recommended raw material list, and detailed assembly sequence are extracted to generate standardized production process guidance documents; The final design documents and the production process guidance documents, along with their corresponding performance verification reports and cost accounting lists, are packaged into a complete derivative design data package. The derivative design data package is the optimal derivative design scheme, and all executable files and parameter lists contained therein constitute the associated production parameters.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: An aesthetic plasticity assessment and functional convertibility prediction are performed on the initial design element pool. Based on the assessment and prediction results, core design elements with derivative potential are marked in the initial design element pool. Redundant elements that do not have derivative development value can be screened out, freeing the selection of design elements from the limitations of subjective human judgment. The element screening results are in line with the aesthetic expression and functional transformation needs of game and animation cultural and creative derivative product development. The element sorting stage in the early stage of design is more in line with the actual development direction of derivative products, and the use of element materials is more targeted.

[0017] By establishing a multi-source knowledge graph that includes historical market data, user community interaction corpora, and production supply chain knowledge, path retrieval and knowledge distillation are performed on the graph based on core design elements. This generates a derivative product concept blueprint that integrates cultural narrative, practical functions, and manufacturing constraints. Market rules, user needs, and production conditions can be integrated into the concept design process simultaneously, ensuring that the derivative product concept aligns with the cultural narrative of the target work, practical functions are adapted to the end-user's usage scenarios, and manufacturing constraints match the processing conditions of the actual production chain. This results in better adaptability of the concept solution for implementation, and multi-dimensional needs are simultaneously considered during the concept generation stage. Attached Figure Description

[0018] Figure 1This is a sequence diagram of the AI-driven intelligent design system for game and animation cultural and creative derivative products described in this invention. Figure 2 A flowchart for assessing aesthetic plasticity and predicting functional convertibility; Figure 3 A flowchart illustrating how adversarial network architectures and neural radiation field models work together. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0020] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0021] See Figure 1 This invention provides an AI-driven intelligent design system for game and animation cultural and creative derivative products. The system specifically includes: The element analysis module imports stylized digital assets from the target game or animation, performs narrative element stripping and physical element decomposition on these assets, separating them into three categories: symbols, scenes, and characters, forming a preliminary design element pool. The potential assessment module extracts the core visual attributes and functional metaphors of each design element in the preliminary pool, performs aesthetic plasticity assessment and functional convertibility prediction, calculates a comprehensive plasticity index through a joint assessment matrix, and marks core design elements exceeding a threshold. The concept generation module pre-constructs a multi-source knowledge graph containing historical market data, user community interaction corpora, and production supply chain knowledge, using core... The core design elements serve as query nodes, retrieving paths within the graph and performing knowledge distillation to reconstruct a derivative product concept blueprint that integrates cultural narrative, practical functionality, and manufacturing constraints. The prototype construction module uses the derivative product concept blueprint as control conditions to drive the collaborative iteration of generative adversarial networks and neural radiation field models, performing multiple rounds of stylized rendering, 3D reconstruction, and structural optimization on the core design elements, outputting a high-fidelity derivative product prototype set. The solution decision module performs manufacturability simulation, user interaction simulation, and market response prediction on the high-fidelity derivative product prototype set, filtering, sorting, and merging prototypes in a multi-dimensional decision space, ultimately outputting the optimal derivative product design solution and associated production parameters.

[0022] In one embodiment of the present invention, the element parsing module imports stylized digital assets from games or animations, analyzes the assets using scene graph parsing technology, identifies and extracts entity objects, relationships between objects, and event sequences to form an initial narrative structure; the attribute classification model decouples the visual features, text descriptions, and semantic tags of entity objects in the initial narrative structure into visual feature vectors, text description vectors, and semantic tag vectors, establishes a cross-modal association mapping among the three, and filters out association groups that represent the key narrative identities of entity objects based on the mapping strength; the materialization feasibility determination module calls the material association database and the morphological structure database to evaluate the achievable boundaries of association groups in terms of form, texture, and volume, encapsulates the association groups determined to be materializable and high-emotional-value fragments in the event sequences into independent design elements, and aggregates them into a primary design element pool containing symbolic features, scene features, and character features.

[0023] In practice, the stylized digital assets of the target game and animation works imported by the element analysis module can be two-dimensional original paintings or three-dimensional model resources containing characters, props, and scene layouts. The stylized digital assets are accompanied by relevant plot text descriptions or official setting materials. The scene graph analysis technology analyzes the visual content and associated text of the stylized digital assets, identifies the entity objects in the picture such as the protagonist's weapon, landmark buildings, and auxiliary creatures, and extracts the spatial relationships and interaction relationships between entity objects, as well as the event sequence composed of timelines or plot nodes, to form the initial narrative structure. The attribute classification model is based on a pre-trained visual encoder and text encoder. It encodes the pixel-level features and contour texture information of entity objects in the initial narrative structure into visual feature vectors, and encodes the names, attribute descriptions, and functional descriptions of entity objects into text description vectors. It also encodes the category identifiers and narrative character tags extracted from official settings or community tags into semantic tag vectors. Cross-modal association mapping calculates the cosine similarity between each pair of visual feature vectors, text description vectors, and semantic tag vectors. A similarity threshold is set to filter out vector groups with high association strength. For example, the association group between the unique shape visual feature vector of the protagonist's weapon and its "Sword of Inheritance" text description vector and "Key Item" semantic tag vector is determined to be the association group that represents the key narrative identity of the entity object.

[0024] In some embodiments, the materialization feasibility determination module calls a material association database, which stores parameter ranges for common materialization materials such as plastics, alloys, and fabrics in terms of gloss, hardness, and weight. At the same time, it calls a morphology database, which records the complexity classification of materialized forms, minimum wall thickness constraints, and requirements for suspended structure support. For the selected association groups, the module evaluates whether the form meets the manufacturable size, whether the texture conforms to the material characteristics range, and whether the volume is suitable for handheld or desktop placement scenarios when it is transformed into a physical entity, based on the material association database and the morphology database. If the evaluation results meet the preset feasible boundaries, the association group is marked as materializable. High-emotional-value segments in the event sequence are identified using a sentiment analysis model. This model analyzes emotional keywords, plot turning points, and high-frequency emotional expressions in user comments within the event sequence text, extracting event segments with emotional scores above a threshold, such as the ritual scene where the protagonist first obtains a weapon or the crucial confrontation moment when the antagonist appears. The materializable association groups and the encoding of high-emotional-value event segments are encapsulated into independent design elements. Each design element contains vector information of the association group, materialization evaluation parameters, and emotional tags for the event segment. All design elements are aggregated to form a primary design element pool containing symbolic features, scene features, and character features.

[0025] In practical implementation, the feasibility of materializing associated groups can be determined through a probabilistic model. This model, based on parameter constraints in the material association database and morphological structure database, calculates the probability of converting associated groups into physical entities. The formula for calculating the feasibility probability is as follows: in: This represents the probability of the entityization feasibility of the associated group. This indicates the degree of matching between the associated group's material parameters and the material association database. This indicates the degree of conformity between the morphological parameters of the associated group and the morphological structure database. This indicates the emotional value weight of the event segments corresponding to the associated groups. , , These are adjustment coefficients for material matching, form conformity, and emotional weight, respectively; when When the threshold is exceeded, the associated group is determined to be materializable. Optionally, when encapsulating materializable associated groups, a material preference recommendation list and form simplification suggestions will be attached. The material preference recommendation list is generated based on a material association database and lists the material types that match the visual characteristics of the associated group. The form simplification suggestions are generated based on a form structure database and propose structural optimization directions without affecting narrative recognizability.

[0026] It can be understood that the symbolic features of the primary design element pool correspond to iconic graphics or patterns with cultural symbolic significance, the scene features correspond to architectural, topographical, or atmospheric elements with spatial narrative functions, and the character features correspond to character images, equipment, or accompanying props that carry the plot development. Each design element is stored in the primary design element pool in a structured data format, including an index of visual feature vectors, an index of text description vectors, an index of semantic tag vectors, a result of entityization feasibility determination, an encoding of associated event fragments, and a corresponding feature classification label. In specific implementations, the data structure of the primary design element pool can be configured in JSON or Protocol Buffers format to facilitate data reading and exchange in subsequent modules. The index information of the design elements is associated with the metadata of the original stylized digital assets to ensure the traceability of the design elements' origin.

[0027] In one embodiment of the present invention, see [reference] Figure 2The potential assessment module extracts the core visual attributes of each design element in the primary design element pool, matches them with the historical and future trend styles in the trend style library, and calculates the aesthetic trend conformity of each design element; the functional metaphor deconstruction module infers the potential physical or interactive functions that the design elements can be given based on the common sense knowledge base, and generates a function transformation list; the joint evaluation matrix integrates the aesthetic trend conformity with the number of functions and implementation complexity in the function transformation list, calculates the comprehensive plasticity index, and marks those exceeding the set threshold as core design elements, and stores their aesthetic trend conformity and function transformation list in association. The concept generation module uses core design elements as query nodes and performs multi-hop path exploration in a multi-source knowledge graph, connecting nodes of best-selling product categories, user preference tags, and mature production processes. It sorts the paths by confidence and information abundance, retaining candidate knowledge subgraphs with high confidence and high abundance. The attention mechanism reweights the entities and relationships in the candidate subgraphs, distilling out the knowledge most closely related to cultural narrative fragments, practical function combinations, and manufacturing constraints. It then reorganizes these knowledge into structured derivative concept descriptions with core design elements as the skeleton, and formally encodes them into derivative concept blueprints.

[0028] In practical implementation, the potential assessment module extracts the core visual attributes of each design element in the primary design element pool. These core visual attributes include shape complexity, main color distribution, texture density, and line style characteristics. The trend style library stores visual feature vectors of historical trend styles and future trend style vectors predicted based on industry reports. The aesthetic trend conformity is calculated by comparing the cosine similarity between the core visual attribute vector of the design element and each style vector in the trend style library, and the highest similarity is taken as the aesthetic trend conformity value. The common sense knowledge base contains functional classifications of everyday items, human-computer interaction modes, and physical usage scenarios. The functional metaphor deconstruction module analyzes the semantic label vectors and text description vectors of the design elements, matches them with functional analogy rules in the common sense knowledge base, and infers the potential physical or interactive functions that the design elements can be given. For example, the "glowing pattern" design element is deconstructed into a night light illumination function and a signage function, generating a functional transformation list that includes function type, applicable scenarios, and implementation complexity.

[0029] In some embodiments, the dimensions of the joint evaluation matrix include aesthetic trend conformity, the number of functions in the functional transformation list, and functional implementation complexity; functional implementation complexity is divided into low, medium, and high levels based on the number of structural parts, circuit modules, and assembly difficulty required for the function and quantified into numerical values. The formula for calculating the comprehensive plasticity index is as follows:

[0030] in: Indicates the comprehensive plasticity index, Indicates the degree of conformity to aesthetic trends. Indicates the number of functions. Indicates the complexity of function implementation. , , These are the weighting coefficients for the corresponding dimensions; the plasticity threshold is set to a fixed value or a dynamically adjusted value according to project requirements. Design elements whose comprehensive plasticity index exceeds the plasticity threshold are marked as core design elements. The aesthetic trend conformity, functional transformation list, and design element ID of the core design elements are stored in the index database.

[0031] In practical implementation, the multi-source knowledge graph of the concept generation module includes node types such as IP topic nodes, best-selling product category nodes, user preference tag nodes, and production process nodes. Edge relationships include types such as "belongs to," "related to," "convertible," and "compatible." Using core design elements as query nodes, the module performs two to four hop path exploration within the multi-source knowledge graph to find multiple knowledge paths connecting to best-selling product category nodes (e.g., "blind box toys," "smart speakers"), user preference tag nodes (e.g., "collectible value," "utilitarianism"), and mature production process nodes (e.g., "injection molding," "3D printing"). The confidence level of these knowledge paths is assessed based on the source authority and data timeliness of the relationships within the path. Information richness is calculated based on the number of nodes and associated attributes within the path. Knowledge paths with a confidence level higher than 0.8 and ranking in the top 20% of information richness are retained to form candidate knowledge subgraphs.

[0032] Optionally, the attention mechanism assigns weights to entities and relationships in the candidate knowledge subgraph. Weight calculation is based on the semantic relevance of entities to core design elements and the frequency of relationship occurrence in historical success cases. The mechanism distills the cultural narrative fragments with the highest weights, such as "mythological heritage stories," practical function combinations, such as "lighting + sound," and manufacturing constraints, such as "minimum wall thickness 1.5mm." These distilled cultural narrative fragments, practical function combinations, and manufacturing constraints are then reorganized into a structured derivative concept description using the core design elements as a framework through template filling or sequence generation models. For example, it might be described as "a practical function combination modeled with [core design elements], carrying [cultural narrative fragments], and satisfying [manufacturing constraints]." The structured derivative concept description is formally encoded into a JSON-LD format derivative concept blueprint, containing semantic fields, functional parameters, and manufacturing constraint key-value pairs.

[0033] It is understandable that the retention strategy of candidate knowledge subgraphs can be adjusted according to the system resource configuration. When computing resources are sufficient, the number of retained paths can be increased to improve knowledge coverage, while when response speed requirements are high, the number of paths can be reduced to accelerate the distillation process. The parameterized encoding of the derivative concept blueprint supports direct calling of subsequent modules. For example, manufacturing constraint parameters can be directly imported into the process parameterized model, and cultural narrative fragments can be converted into prompt words to input the generative adversarial network.

[0034] In one embodiment of the present invention, see [reference] Figure 3 The prototype construction module takes the derivative concept blueprint as the control condition input to the generator of the generative adversarial network (GAN), generating a set of 2D concept sketches in various styles and perspectives. The sketch with the highest quality score is selected and input into the neural radiation field model. Combined with the structural constraint parameters in the blueprint, the initial 3D geometry and surface materials are reconstructed. The discriminator of the GAN constructs a joint discrimination task to judge the artistic quality of the 2D sketches and the realism and consistency of the multi-view 3D images rendered by the neural radiation field model. The feedback signal is backpropagated to the generator and the neural radiation field model for end-to-end joint optimization. In each iteration, the generator receives the previous round's 3D perspective rendering image from the neural radiation field model as a style reference and outputs a new round of 2D concept sketches in combination with the blueprint. The neural radiation field model receives the new sketches and integrates the prior geometric structure of the previous round to optimize 3D details and surface properties. It performs automatic structural mechanics analysis and supportability detection to identify weak parts, generates structural optimization suggestions, and converts them into a geometric constraint loss function to be added to the training objective, driving the next round of reconstruction to simultaneously optimize the geometric structure.

[0035] In the specific implementation, the prototype construction module receives the derivative concept blueprint, which includes semantic descriptions of core design elements, target style tags, and structural constraint parameters. The generator module of the generative adversarial network takes the derivative concept blueprint as conditional input. The generator architecture adopts the conditional generative network of the StyleGAN series and outputs a set of 2D concept sketches with a resolution of 512×512 pixels, covering two styles: Chinese trendy cyberpunk and retro hand-painted, and including frontal 45-degree and side 90-degree perspectives. The set capacity is set to 50 sketches. Each sketch in the 2D concept sketch set is evaluated for artistic quality by an independent scoring network. The scoring network is trained based on an aesthetic evaluation dataset and outputs a normalized quality score. The neural radiation field model uses the Instant-NGP framework, inputting the 2D concept sketch with the highest quality score, and simultaneously loading the structural constraint parameters from the derivative concept blueprint, including a maximum outer diameter of 300 mm and a minimum wall thickness of 1.2 mm, to reconstruct the preliminary 3D geometry and diffuse material texture, as shown in Table 1.

[0036] Table 1: Comparison of Quality Indicators for Different Iteration Rounds 1 0.72 0.15 62% 2 0.78 0.11 75% 3 0.82 0.09 85% In its implementation, the discriminator module of the generative adversarial network constructs a joint discrimination task. The discriminator's input includes a 2D sketch output by the generator and 3D images rendered by the neural radiation field model from eight uniformly distributed viewpoints. The discriminator's branch for the 2D sketch outputs scores for artistic style matching and compositional rationality, while its branch for the 3D images outputs scores for single-view realism and consistency across multiple viewpoints. The feedback signal from the joint discrimination task updates the generator's generation weights and the voxel density and color network parameters of the neural radiation field model simultaneously through gradient backpropagation. At the beginning of each iteration, the generator receives the 3D viewpoint rendering provided by the neural radiation field model from the previous iteration as a style reference. The 3D viewpoint rendering has a resolution of 256×256 pixels and includes lighting and material appearance information. The generator then combines the derivative concept blueprint to generate a new 2D concept sketch. The pixel resolution of the new sketch remains unchanged, but it maintains a higher consistency with the 3D reference sketch in terms of texture details.

[0037] In some embodiments, the neural radiation field model receives a new round of two-dimensional conceptual sketches as input, while incorporating the structural priors of the three-dimensional geometry generated in the previous round. The structural priors are stored in the form of point clouds, containing vertex positions and normal information. The coordinates of the sampling points of the neural radiation field model are input into the MLP network after position encoding, and the output is density and color values. Optimized three-dimensional geometric details and surface properties are generated through volume rendering. The structural mechanics analysis and supportability detection module runs on a simplified version of the finite element method, with a mesh generation accuracy of 2 mm. The detection content includes the deflection deformation of the cantilever structure and the ratio of the bottom support contact area, identifying weak or non-physically compliant geometric parts, such as unsupported horizontal arms exceeding 100 mm in length or unstable bases with a bottom contact area less than 20% of the projected area.

[0038] The structural optimization suggestion generation module transforms detected problems into geometric adjustment instructions, such as increasing the thickness of stiffeners or increasing the diameter of the base. The formula for calculating the geometric constraint loss function is as follows:

[0039] in: Represents the geometric constraint loss function. Indicates the first The current position of each vertex. Indicates the optimized target position. The total number of vertices. This represents the minimum wall thickness threshold. Indicates the average wall thickness of the region. , To balance the performance, a geometric constraint loss function is added to the total training loss of the neural radiation field model, simultaneously optimizing the geometry to meet manufacturing requirements during the next round of 3D reconstruction. Optionally, the collaborative training cycle of the generative adversarial network and the neural radiation field model is set to three rounds, with the generator updating 2000 steps and the neural radiation field model updating 5000 steps in each round, and the batch size set to 4. After multiple rounds of joint optimization iterations, the final output set of high-fidelity derivative prototypes contains five prototype models that meet the standards in both 2D artistic expression and 3D structural entities. The prototype format is an OBJ file with UV texture mapping. It can be understood that the mesh accuracy for structural mechanics analysis can be adjusted according to the prototype size. When the overall prototype size is less than 100 mm, the mesh accuracy is increased to 1 mm; when the size exceeds 500 mm, the accuracy is reduced to 5 mm to balance computational efficiency and accuracy.

[0040] In one embodiment of the present invention, the solution decision module matches each prototype in the high-fidelity derivative prototype set with candidate production processes corresponding to the manufacturing constraints in the derivative concept blueprint. It loads a process parameterization model from a digital process library, inputs the prototype's geometry and material data into the model to simulate the entire manufacturing process, and calculates material utilization, processing time, process yield, and tool path conflicts in real time to generate a manufacturing feasibility report. For prototypes with conflicts or low yields, an automatic geometry repair program is initiated. After fine-tuning based on process constraints, the effect is verified in a simulation environment. The reports of all prototypes are summarized to generate a simulation result dataset containing cost, time, and technical risks. A virtual test environment with different user profiles is constructed, importing the prototype's 3D model and estimated physical properties. Standardized interactive task flows are designed for user profiles, recording task completion time, operation path, and gaze point heatmaps. Sentimental computing data and virtual interview texts from simulated interactions are collected. Sentiment analysis and text mining are used to quantify emotional experience value and perceived value. Historical market data is correlated to train a market response prediction model to predict the market share, price sensitivity, and word-of-mouth dissemination index of each prototype.

[0041] In practical implementation, the solution decision module performs manufacturability simulation for each prototype in the high-fidelity derivative prototype set. Each prototype has geometric and material data. The automatic matching process is based on the manufacturing constraints specified in the derivative concept blueprint, such as a maximum product size of 300 mm and a material type of ABS plastic. The candidate production process is injection molding. The digital process library loads the parametric model of injection molding, which includes mold structure parameters, injection pressure curves, and cooling time formulas. The geometric data of the prototype is input into the parametric model to simulate the entire process from plastic granule drying, melt injection, pressure holding and cooling to demolding. Material data is used to set the material shrinkage rate and flow characteristics. During the simulation, material utilization rate, estimated processing time, and predicted yield of each process are calculated in real time. The collision detection algorithm identifies the interference between the tool path and the mold cavity, generating a manufacturing feasibility report (see Table 2).

[0042] Table 2: Comparison of Manufacturing Feasibility Indicators for the Two Prototypes A001 88% 52 94% none A002 76% 68 87% Lateral concave region In practice, for prototypes with toolpath conflicts or a yield rate below 85%, an automatic geometry repair program is initiated. This program generates adjustment suggestions based on constraints from the process parameterization model, such as adjusting the concave angle from 5 degrees to 2 degrees or increasing the radius of sharp corner fillets to 1 mm. The manufacturing process is then rerun in the simulation environment to verify the repair effect until the conflict is eliminated and the yield rate reaches the threshold. A manufacturing feasibility report is compiled from all prototypes after simulation and repair, generating a simulation results dataset. This dataset includes the quantified manufacturing cost, total man-hours, and technical risk level for each prototype. Cost calculations are based on material unit prices and machine unit time rates, while the technical risk level is determined based on the yield rate fluctuation range and equipment dependence.

[0043] In some embodiments, a virtual test environment is constructed to simulate user interaction, comprising three user profiles: collectors, practical users, and gift-giving users. The virtual test environment is built using the Unity engine, importing the 3D models of prototypes A001 and A002 and the estimated weight and surface texture parameters from the simulation result dataset. Standardized interaction task flows are designed for each user profile. The interaction task flow for collectors includes observation in the display case, handheld appreciation, and return to its original position; for practical users, it includes desktop access, button operation, and storage; and for gift-giving users, it includes unpacking, display and sharing, and repackaging. Task completion time, operation path trajectory, and virtual gaze point heatmap data are recorded. The operation path trajectory is stored as a 3D coordinate sequence, and the virtual gaze point heatmap is generated based on a gaze projection algorithm.

[0044] Affective computing data is generated through simulation of facial expression changes and physiological parameters of virtual characters. Virtual interview texts are generated based on a template generation system. A affective analysis model maps the affective computing data to dimensions of pleasure, excitement, and satisfaction. A text mining model extracts keyword frequency and affective polarity from the virtual interview texts. The affective experience value and perceived value of each prototype are quantified. The formula for calculating affective experience value is:

[0045] in: Indicates the value of emotional experience. Indicates the first The intensity rating of each emotional dimension. This represents the weighting coefficient for the emotional dimension; perceived value is derived by weighting the user profile's ratings of the prototype's functionality and aesthetics. The correlation analysis module aligns the emotional experience value, perceived value, and aesthetic feature vectors with historical market sales data and social media buzz data. Historical market sales data includes the sales volume and price of similar derivative products, while social media buzz data includes the number of topic discussions and the proportion of positive comments. A market response prediction model based on random forest is trained, and the model outputs the potential market share, price sensitivity, and word-of-mouth dissemination index for each prototype.

[0046] Optionally, the number of user profiles in the virtual testing environment can be expanded to five, adding geek-type users and parent-child-type users, and the expanded interactive task flow covers more usage scenarios; the training data period for the market response prediction model can be selected from the most recent year or three years to adapt to the life cycle characteristics of different products. It is understandable that the emotional computing data from user interaction simulations can also be enhanced through EEG simulation modules, and virtual interview texts can employ larger-scale generative models to improve diversity, thereby more accurately quantifying the value of emotional experiences.

[0047] In one embodiment of the present invention, the solution decision module establishes a multi-dimensional decision space covering technical implementation, user experience, and commercial value. It calculates the coordinates of each prototype in the high-fidelity derivative prototype set within this space, applies a multi-objective optimization algorithm to find the Pareto optimal front in the point cloud distribution, and selects prototypes on the front as the winning candidate set. The module dynamically adjusts the dimensional weights according to the project's strategic objectives, weights and ranks the candidate prototypes, deconstructs the features of the top-ranked prototypes, and integrates the advantageous features in specific dimensions to generate a fused prototype solution with superior overall performance. The fused solution undergoes a final round of simulation to verify stability. Geometric models, material textures, and color schemes are extracted to generate directly editable final design files. Optimized process parameters, raw material lists, and assembly sequences are extracted to generate standardized production process guidance documents. The design files, process files, performance reports, and cost calculations are packaged into a complete derivative design data package, serving as the optimal design solution and associated production parameters.

[0048] In practical implementation, the solution decision-making module establishes a multi-dimensional decision space comprising three orthogonal dimensions: the technical implementation dimension based on indicators from the manufacturing feasibility report, the user experience dimension based on the output of user interaction simulation, and the commercial value dimension based on market forecast data. The coordinates of each prototype within the high-fidelity derivative prototype set are determined by its normalized scores across these three dimensions. The technical implementation dimension score is derived from a weighted combination of manufacturing cost and yield rate; the user experience dimension score is derived from the average of emotional experience value and perceived value; and the commercial value dimension score is derived from a linear combination of market share and word-of-mouth dissemination index. The coordinates of all prototypes form a three-dimensional point cloud distribution. The range of this point cloud distribution reflects the differences in technology, experience, and business among different prototypes. For example, some prototypes may score nearly full marks in the technical implementation dimension but be low in the commercial value dimension, while others may excel in the user experience dimension but face bottlenecks in technical implementation.

[0049] In some embodiments, the multi-objective optimization algorithm employs the NSGA-II framework to find the Pareto optimal front in a 3D point cloud distribution. A prototype on the Pareto optimal front is defined as a solution where increasing the score in any dimension necessarily leads to a decrease in the score in at least one other dimension. These prototypes are selected to form a winning candidate set, the size of which depends on the density of points on the front, typically retaining five to ten prototypes. A weighted score is applied to the prototypes in the winning candidate set, calculated using the following formula:

[0050] in: This represents the weighted total score. , , These represent the normalized scores of the prototype in the dimensions of technical implementation, user experience, and commercial value, respectively. , , These are the dynamic weights for the corresponding dimensions; the dynamic weights are adjusted according to the strategic goals set for the project, such as when the goal is mass production and promotion. and When set to a higher value for brand promotion purposes and Set to a higher value. The prototypes of the winning candidate set are sorted from highest to lowest weighted total score, and the top three are selected to enter the feature fusion stage.

[0051] In practical implementation, the features of the top-ranked prototypes are deconstructed, including geometric shape features, functional module configurations, and surface decoration schemes. The algorithm fusion module extracts the advantageous features of each prototype in a specific dimension. For example, prototype A has the highest score in the technical implementation dimension, and its advantageous feature is simplified structural design; prototype B has the highest score in the user experience dimension, and its advantageous feature is the interactive interface layout; prototype C has the highest score in the commercial value dimension, and its advantageous feature is visual distinctiveness. The advantageous features are fused through parametric methods. When merging the simplified structural design and the interactive interface layout, the assembly interface is adjusted. The visual distinctiveness features are integrated into the fused geometry through a texture transfer algorithm, generating a fused prototype solution that theoretically has a comprehensive performance exceeding that of a single prototype. Optionally, the fused prototype solution undergoes a final round of simulation and verification. The verification content includes re-simulation of the manufacturing process, re-simulation of user interaction, and re-prediction of market response. During the verification process, the stability of key indicators is monitored, such as controlling the fluctuation range of manufacturing costs within ±5%, the standard deviation of user satisfaction less than 0.1%, and the predicted range of market share not exceeding 10%. The final design files are extracted from the validated fusion prototype. The geometric model is saved in STEP format for compatibility with mainstream CAD software, and the material textures are saved as PNG format color maps and normal maps. The color scheme is recorded as Pantone color chart numbers and RGB values. The production process guidance document (PPD) relies on the parameters confirmed in the final validation. These parameters include injection temperature ranges, injection pressure segment settings, a recommended raw material list listing ABS plastic grades and additive ratios, and a detailed assembly sequence describing the component docking order and fastener specifications with illustrations and text. The final design files, PPD, performance validation report, and cost calculation list are packaged into a ZIP format derivative design data package. This package represents the optimal derivative design scheme and includes executable files such as STEP geometry files and STL printing files, as well as parameter lists including process parameter tables and raw material lists.

[0052] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An AI-driven game animation and creative derivative product intelligent design system, characterized in that, The system includes: The element parsing module imports the stylized digital assets of the target work and performs narrative element stripping and physical element decomposition on the stylized digital assets to form a primary design element pool. The potential assessment module performs an aesthetic plasticity assessment and a functional convertibility prediction on the primary design element pool. Based on the assessment and prediction results, it marks the core design elements with derivative potential in the primary design element pool. The concept generation module establishes a multi-source knowledge graph containing historical market data, user community interaction corpus, and production supply chain path knowledge. Based on the core design elements, it performs path retrieval and knowledge distillation in the graph to generate a derivative product concept blueprint that integrates cultural narrative, practical functions, and manufacturing constraints. The prototype building module, based on the aforementioned derivative concept blueprint, drives the collaborative work of the generative adversarial network architecture and the neural radiation field model to perform multiple rounds of stylized rendering, 3D reconstruction, and structural optimization iterations on the core design elements, outputting a high-fidelity derivative prototype set. The solution decision module performs manufacturability simulation, user interaction simulation, and market response prediction on the set of high-fidelity derivative prototypes. Based on the multi-dimensional performance data obtained from the simulation, modeling, and prediction, it filters, sorts, and integrates the data, and finally outputs the optimal derivative design scheme and its associated production parameters.

2. The AI-driven game animation and creative derivative product intelligent design system based on AI of claim 1, wherein The stylized digital assets are subjected to narrative element stripping and materialization element decomposition, including: The primary design element pool includes symbol features, scene features, and character features; The stylized digital assets are analyzed using scene graph analysis technology to identify and extract the entity objects, relationships between objects, and event sequences contained therein, thus forming an initial narrative structure; For the entity objects in the initial narrative structure, the visual features, text descriptions, and semantic tags of the entity objects are decoupled using an attribute classification model and assigned to visual feature vectors, text description vectors, and semantic tag vectors, respectively. Establish a cross-modal association mapping between the visual feature vector, text description vector, and semantic tag vector of the entity object, and filter out the association groups that represent the key narrative identity of the entity object based on the mapping strength; The feasibility of materializing the association group is determined by assessing the achievable boundaries of the association group in terms of form, texture, and volume, based on the material association database and the morphological structure database. The associated groups that are determined to be materializable, together with the encoded fragments with high emotional value in the event sequence, are encapsulated into independent design elements, and then aggregated to form the primary design element pool containing symbolic features, scene features, and character features. 3.The AI-driven game animation and creative derivative product intelligent design system based on AI of claim 1, wherein, An aesthetic plasticity assessment and functional convertibility prediction are performed on the aforementioned pool of primary design elements, including: Extract the core visual attributes of each design element from the primary design element pool, and calculate the matching degree with the historical and future trend styles in the trend style library to obtain the aesthetic trend conformity of each design element; At the same time, the functional metaphors of each design element are deconstructed, and based on the common sense knowledge base, the potential physical or interactive functions that each design element can be given are inferred to generate a list of functional transformations. A joint evaluation matrix is ​​constructed, which integrates the aesthetic trend conformity with the number of functions and the complexity of function implementation in the functional transformation list, and calculates the comprehensive plasticity index of each design element. Set a plasticity threshold, and mark design elements whose comprehensive plasticity index exceeds the plasticity threshold as core design elements with derivation potential; The marked core design elements and their corresponding aesthetic trend conformity and functional transformation list are linked and stored to form the input index for subsequent knowledge retrieval.

4. The AI-driven intelligent design system for game and animation cultural and creative derivative products as described in claim 3, characterized in that, Based on the core design elements, path retrieval and knowledge distillation are performed in the graph, including: Using the core design elements as query nodes, multi-hop path exploration is performed in the multi-source knowledge graph to find multiple knowledge paths connecting to market best-selling product category nodes, user preference tag nodes, and mature production process nodes; The confidence level and information abundance of the multiple knowledge paths are evaluated and ranked. The knowledge paths with confidence levels higher than a predetermined value and higher information abundance rankings are retained to form candidate knowledge subgraphs. An attention mechanism is used to reweight the entities and relationships in the candidate knowledge subgraph to distill out the cultural narrative fragments, practical function combinations, and manufacturing constraints that are most closely related to the core design elements. The distilled cultural narrative fragments, practical functional combinations, and manufacturing constraints are creatively reorganized and logically stitched together using the core design elements as a framework to generate a structured derivative concept description. The description of the derivative concept is formalized and parameterized, and finally transformed into a derivative concept blueprint that integrates cultural narrative, practical function, and manufacturing constraints.

5. The AI-driven intelligent design system for game and animation cultural and creative derivative products as described in claim 1, characterized in that, The driving generative adversarial network architecture works in conjunction with the neural radiation field model, including: Using the aforementioned derivative concept blueprint as a control condition, the generator module of the generative adversarial network is input to guide it in generating a set of two-dimensional concept sketches with various styles and perspectives that conform to the semantics of the aforementioned derivative concept blueprint. The sketch with the highest quality score is selected from the set of two-dimensional concept sketches and used as the input to the neural radiation field model. At the same time, combined with the structural constraint parameters in the derivative concept blueprint, the neural radiation field model is used to reconstruct the preliminary three-dimensional geometry and surface material. A joint discrimination task is constructed in the discriminator module of the generative adversarial network to simultaneously judge the artistic quality of the two-dimensional sketches output by the generator module, as well as the realism and consistency of the three-dimensional images rendered from different perspectives by the neural radiation field model. The feedback signal from the joint discrimination task is simultaneously backpropagated to the generator module of the generative adversarial network and the neural radiation field model for end-to-end joint optimization. After multiple rounds of joint optimization iterations, a set of high-fidelity derivative prototypes that meet the requirements in both two-dimensional artistic expression and three-dimensional structural entities is finally obtained. 6.The AI-driven game animation and creative derivative product intelligent design system based on AI of claim 5, wherein, The core design elements undergo multiple rounds of stylized rendering, 3D reconstruction, and structural optimization iterations, including: In each iteration, the generator of the generative adversarial network first receives the three-dimensional view rendering provided by the previous neural radiation field model as a style reference, and combines it with the derivative concept blueprint to output a new round of two-dimensional concept sketches that are more consistent in three dimensions. The neural radiation field model receives a new round of two-dimensional conceptual sketches and integrates structural priors from the three-dimensional geometry generated in the previous round to optimize the details and surface properties of the three-dimensional geometry and perform three-dimensional reconstruction. After 3D reconstruction, automatic structural mechanics analysis and supportability testing are performed on the obtained geometry to identify geometric parts that are structurally weak or do not conform to physical manufacturing. Based on the results of the structural mechanics analysis and supportability test, structural optimization suggestions are generated, and the structural optimization suggestions are transformed into a geometric constraint loss function that can be understood by the neural radiation field model. The geometric constraint loss function is incorporated into the training objective of the neural radiation field model, driving it to synchronously optimize the geometric structure in the next round of reconstruction to meet manufacturing requirements, thus completing a single round of structural optimization iteration.

7. The AI-driven intelligent design system for game and animation cultural and creative derivative products as described in claim 1, characterized in that, Manufacturability simulation of the aforementioned high-fidelity derivative prototype set includes: For each prototype in the high-fidelity derivative prototype set, automatically match the candidate production process that best matches the manufacturing constraints in the derivative concept blueprint, and load the corresponding process parameterization model from the digital process library; The geometric and material data of each prototype are input into the matching process parameterization model to simulate the entire manufacturing process from raw material processing, forming, machining to surface treatment. In the simulation process, material utilization rate, estimated processing time, predicted yield of each process and tool path conflict detection results are calculated and recorded in real time to form a manufacturing feasibility report. For prototypes with tool path conflicts or low yield, an automatic geometry repair program is initiated to fine-tune the prototype based on the constraints of the process parameterization model, and the repair effect is verified in a simulation environment. Finally, the manufacturing feasibility reports of all prototypes after simulation and repair are compiled to generate a simulation results dataset that includes quantitative manufacturing costs, working hours, and technical risk indicators. 8.The AI-driven game animation and creative derivative product intelligent design system based on AI of claim 7, wherein, User interaction simulation and market response prediction are performed on the aforementioned high-fidelity derivative prototype set, including: Construct a virtual test environment containing different user profiles, and import the 3D model and estimated physical properties of each prototype in the simulation result dataset into the virtual test environment; In the virtual testing environment, a standardized interactive task flow is designed for each user profile, including visual search, retrieval, use, and storage steps, and the task completion time, operation path, and virtual gaze point heatmap data are recorded. Collect real-time sentiment computing data of user profiles generated during simulated interaction and post-interview virtual interview texts. Quantify the emotional experience value and perceived value brought by each prototype through sentiment analysis models and text mining models. The emotional experience value, perceived value, and aesthetic feature vector of each prototype are correlated with historical market sales data and social media volume data to train a market response prediction model. Using the market response prediction model, the potential market share, price sensitivity, and word-of-mouth dissemination index of each prototype in the high-fidelity derivative prototype set are predicted to form market prediction data. 9.The AI-driven game animation and creative derivative product intelligent design system based on AI of claim 8, wherein, The process of filtering, sorting, and fusing multidimensional performance data obtained from simulation, modeling, and prediction includes: A multi-dimensional decision-making space is established, whose dimensions include the technical implementation dimension from the manufacturing feasibility report, the user experience dimension from the user interaction simulation, and the business value dimension from the market forecast data; For each prototype in the set of high-fidelity derivative prototypes, calculate its coordinates in the multidimensional decision space, and calculate the point cloud distribution formed by the coordinate points of all prototypes. A multi-objective optimization algorithm is applied to find the Pareto optimal front in the point cloud distribution, and the prototypes located on the Pareto optimal front are selected as the winning candidate set. The prototypes in the winning candidate set are weighted and ranked according to their performance in different decision-making dimensions, with the weights dynamically adjusted by the strategic objectives set by the project. The features of the top-ranked prototypes are deconstructed, and their advantageous features in a specific dimension are algorithmically fused to generate a fused prototype scheme that theoretically outperforms all individual prototypes, which serves as the final candidate.

10. The AI-driven game animation and creative derivative product intelligent design system based on AI of claim 9, wherein The final output of the optimal derivative design scheme and its associated production parameters includes: The final round of simulation and modeling of the fusion prototype scheme was conducted to verify the stability of its comprehensive performance indicators; Extract the geometric model, material texture, and color scheme of the fusion prototype solution that are confirmed in the final verification, and generate a final design file that can be directly edited by 3D software; From the verification process of the fusion prototype solution, the optimized process parameters, recommended raw material list, and detailed assembly sequence are extracted to generate standardized production process guidance documents; The final design documents and the production process guidance documents, along with their corresponding performance verification reports and cost accounting lists, are packaged into a complete derivative design data package. The derivative design data package is the optimal derivative design scheme, and all executable files and parameter lists contained therein constitute the associated production parameters.

Images