Systems, methods, kits, and apparatus for generative artificial intelligence, graph neural networks, transformer models, and convergent technology stacks in value chain networks.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- STRONG FORCE VCN PORTFOLIO 2019 LLC
- Filing Date
- 2024-04-25
- Publication Date
- 2026-06-16
Smart Images
Figure 2026519403000001 
Figure 2026519403000002 
Figure 2026519403000003
Abstract
Claims
1. A generative artificial intelligence system for value chain networks, It includes one or more processors and one or more memories, and the one or more processors and the one or more memories are The aforementioned generative artificial intelligence system executes a generative artificial intelligence algorithm trained on value chain network data. Receiving input data that includes at least one of the following: images, videos, audio, text, program code, and data. The process involves using the generative artificial intelligence algorithm to process the input data and generate output content, wherein the output content includes at least one of the following: structured text, images, videos, audio content, software source code, formatted data, algorithms, definitions, and context-dependent structures. To generate the internal state of the generative artificial intelligence system, including a set of weights and / or biases as a result of past processing, The generated output content is provided to the user interface for presentation to the user. A generative artificial intelligence system characterized by being configured to perform operations including those mentioned above.
2. The generative artificial intelligence system according to claim 1, characterized in that the input data further includes natural language representations, single or multidimensional shapes or models, real-world and / or virtual scene representations, LiDAR point cloud representations, sensor inputs and / or outputs, vehicle and / or machine telemetry, geographic maps, authentication credentials, financial transactions, smart contracts, processing instructions, and device configurations.
3. The generative artificial intelligence system according to claim 1 is characterized in that it is configured to maintain contextual awareness throughout the interaction in order to facilitate the continuity of the interaction with the user.
4. The generative artificial intelligence system according to claim 1, characterized in that the generative artificial intelligence system is configured to support the interpretability and explainability of its output by providing an explanation of the basis for the output.
5. The generative artificial intelligence system according to claim 1, characterized in that the output includes recommendations for improving one or more of the value chain network, entities within the value chain network, assets within the value chain network, and the value chain network environment.
6. Furthermore, the generative artificial intelligence system according to claim 1 includes a digital twin interface for a digital twin, the digital twin interface is configured to enable access to the generative artificial intelligence system, and the digital twin represents at least one of the following: the value chain network, entities within the value chain network, assets within the value chain network, or environments within the value chain network.
7. Furthermore, the generative artificial intelligence system according to claim 1 is characterized by including a digital twin module configured to generate a digital twin representing at least one of a value chain network, entities within the value chain network, assets within the value chain network, and environments within the value chain network.
8. The generative artificial intelligence system according to claim 1, wherein the processor is further configured to communicate with robot systems in the value chain network, and the memory stores instructions that enable the generative artificial intelligence system to translate human commands into robot actions, thereby facilitating interaction between humans and robots.
9. The generative artificial intelligence system according to claim 1, characterized in that the output content is further adjusted based on the user's job role.
10. The generative artificial intelligence system according to claim 1, characterized in that the generated output includes 3D printing instructions for a set of 3D printers.
11. The generative artificial intelligence system according to claim 1, characterized in that the system is part of a dual-process artificial neural network (DPANN) architecture.
12. Furthermore, the generative artificial intelligence system according to claim 1 is characterized by including an augmented reality (AR) interface configured to superimpose the generated output content onto the user's field of view in the real world.
13. A method for operating a generative artificial intelligence system within a value chain network, The aforementioned generative artificial intelligence system executes a generative artificial intelligence algorithm trained on value chain network data. Receiving input data that includes at least one of the following: images, videos, audio, text, program code, and data. The process involves using the generative artificial intelligence system to process the input data and generate output content, wherein the output content includes at least one of the following: structured text, images, videos, audio content, software source code, formatted data, algorithms, definitions, and context-dependent structures. To generate the internal state of the generative artificial intelligence system, including a set of weights and / or biases as a result of past processing, The generated output content is provided to the user interface for presentation to the user. A method characterized by including the following.
14. The method according to claim 13, further comprising receiving additional input data including natural language representations, single or multidimensional shapes or models, real-world and / or virtual scene representations, LiDAR point cloud representations, sensor inputs and / or outputs, vehicle and / or machine telemetry, geographic maps, authentication credentials, financial transactions, smart contracts, processing instructions, and device configurations.
15. Furthermore, the method according to claim 13 is characterized by including maintaining contextual awareness throughout the interaction in order to facilitate the continuity of the interaction with the user.
16. The method according to claim 13, further comprising supporting the interpretability and explainability of the output by providing an explanation of the basis for the output.
17. The method according to claim 13, characterized in that the generated output includes recommendations for improving one or more of the value chain network, entities within the value chain network, assets within the value chain network, and environments within the value chain network.
18. Interface connection to a digital twin via a digital twin interface, wherein the digital twin interface is configured to facilitate access to the generative artificial intelligence system, and The digital twin represents one or more of the following: the value chain network, entities within the value chain network, assets within the value chain network, and environments within the value chain network. The method according to claim 13, further comprising:
19. To generate a digital twin through a digital twin module that represents at least one of the following: a value chain network, entities within the value chain network, assets within the value chain network, and environments within the value chain network. The method according to claim 13, further comprising:
20. To communicate with one or more robot systems within the aforementioned value chain network, Converting human commands into robot actions, and To facilitate interaction between humans and robots in the value chain network, thereby enabling the robot system to perform tasks in response to human commands derived from the generated output content. The method according to claim 13, further comprising:
21. Adjusting the output content based on the user's job role within the organization, The method according to claim 13, further comprising:
22. The method according to claim 13, characterized in that the generated output includes 3D printing instructions for a set of 3D printers.
23. The method according to claim 13, characterized in that the generative artificial intelligence system operates as part of a dual-process artificial neural network (DPANN) architecture.
24. Connecting to an augmented reality (AR) interface, and The AR interface overlays the generated output content onto the user's field of view in the real world. The method according to claim 13, further characterized by including
25. A system for improving value chain networks using graph neural networks, It includes one or more processors and one or more memories, and the one or more processors and the one or more memories are Receiving an input graph dataset representing a value chain network, wherein the input graph dataset includes nodes representing entities in the value chain and edges representing transactions or relationships between entities. Based on the aforementioned input graph dataset, generate embeddings specific to the value chain context. The process involves applying multiple graph neural network layers to the aforementioned embedding to generate an output graph dataset, wherein the output graph dataset includes updates to nodes and edges that reflect changes in the value chain network. Perform clustering analysis on the nodes or edges of the output graph dataset to identify clusters representing highly interconnected segments of the value chain, and Output an optimized value chain network graph dataset that includes updated node characteristics, edge characteristics, or graph characteristics. A system characterized by being configured to perform operations including those mentioned above.
26. The system according to claim 25, further configured to perform dimensionality reduction of the graph dataset based on a similarity matrix in order to facilitate the clustering analysis.
27. The system according to claim 25, further configured to apply a spectral clustering method during the clustering analysis.
28. The system according to claim 25, further configured to generate an output graph dataset including a graphlet degree vector representing an iterative structure within the input graph dataset.
29. The system according to claim 25, further configured to perform inter-graph transformations to generate an output graph dataset that represents a graph dataset different from the input graph dataset.
30. The system according to claim 25, further configured to utilize instance-level descriptions based on gradients, features, perturbations, or decompositions in order to provide interpretability of the output graph dataset.
31. The system according to claim 25, further configured to perform a hyperparameter search process to improve the graph neural network based on user feedback regarding the output graph dataset.
32. The system according to claim 25, wherein the value chain network entity includes at least two of the following: products, suppliers, producers, manufacturers, retailers, companies, owners, operators, operating facilities, customers, consumers, workers, mobile devices, wearable devices, distributors, resellers, supply chain infrastructure facilities, supply chain processes, logistics processes, reverse logistics processes, demand forecasting processes, demand management processes, demand aggregation processes, machinery, ships, barges, warehouses, seaports, airports, air routes, waterways, roads, railways, bridges, tunnels, online retailers, e-commerce sites, demand factors, supply factors, distribution systems, floating assets, origin points, destinations, storage points, usage points, networks, information technology systems, software platforms, distribution centers, fulfillment centers, containers, container handling facilities, customs, export control, border control, drones, robots, robotic transport systems, 3D printers, vehicles, autonomous vehicles, transport facilities, waterways, and port infrastructure facilities.
33. This is a computer-implemented method for optimizing value chain networks using graph neural networks. A computer device receives an input graph dataset representing a value chain network, wherein the input graph dataset includes nodes representing entities in the value chain and edges representing transactions or relationships between entities. Based on the aforementioned input graph dataset, generate embeddings specific to the value chain context. The process involves applying multiple graph neural network layers to the aforementioned embedding to generate an output graph dataset, wherein the output graph dataset includes updates to nodes and edges that reflect changes in the value chain network. Perform clustering analysis on the nodes or edges of the output graph dataset to identify clusters representing highly interconnected segments of the value chain, and Output an optimized value chain network graph dataset that includes updated node characteristics, edge characteristics, or graph characteristics. A computer-implemented method characterized by including the following.
34. Furthermore, the computer-implemented method according to claim 33 is characterized by including dimensionality reduction of the graph dataset based on a similarity matrix in order to facilitate the clustering analysis.
35. Furthermore, the computer-implemented method according to claim 33 is characterized by including the application of a spectral clustering technique during the clustering analysis.
36. Furthermore, the computer-implemented method according to claim 33 is characterized by comprising generating an output graph dataset that includes a graphlet degree vector representing the iterative structure in the input graph dataset.
37. Furthermore, the computer-implemented method according to claim 33 is characterized by including performing an inter-graph transformation to generate an output graph dataset that represents a graph dataset different from the input graph dataset.
38. Furthermore, the computer-implemented method according to claim 33 is characterized by including the use of instance-level descriptions based on gradients, features, perturbations, or decompositions in order to provide interpretability of the output graph dataset.
39. Furthermore, the computer-implemented method according to claim 33 is characterized by including a hyperparameter search process for improving the graph neural network based on user feedback regarding the output graph dataset.
40. The computer-implemented method according to claim 33, characterized in that the value chain network entity includes at least two of the following: products, suppliers, producers, manufacturers, retailers, companies, owners, operators, operating facilities, customers, consumers, workers, mobile devices, wearable devices, distributors, resellers, supply chain infrastructure facilities, supply chain processes, logistics processes, reverse logistics processes, demand forecasting processes, demand management processes, demand aggregation processes, machinery, ships, barges, warehouses, seaports, airports, air routes, waterways, roads, railways, bridges, tunnels, online retailers, e-commerce sites, demand factors, supply factors, distribution systems, floating assets, origin points, destinations, storage points, usage points, networks, information technology systems, software platforms, distribution centers, fulfillment centers, containers, container handling facilities, customs, export control, border control, drones, robots, robotic transport systems, 3D printers, vehicles, autonomous vehicles, transport facilities, waterways, and port infrastructure facilities.
41. A system for improving value chain networks (VCNs) using a transformer model, It includes one or more memory and one or more processors, and the one or more memory and the one or more processors are Receiving an input dataset representing a VCN including nodes and edges, Encode the input dataset using an encoder from a transformer model to generate a contextual representation. The contextual representation is decoded using the decoder of the transformer model, and an output dataset demonstrating the improvement of the VCN is generated, and To facilitate decision-making in the aforementioned VCN, the improved VCN dataset is output. A system characterized by being configured to perform operations including those mentioned above.
42. The system according to claim 41, further configured to apply a multi-head attention mechanism to determine the relationships between nodes and edges within the VCN.
43. The system according to claim 41, further configured to perform autoregressive prediction of a VCN configuration using the output dataset.
44. The system according to claim 41, further configured to utilize position encoding in order to maintain the sequence order of the VCN data within the transformer model.
45. The system according to claim 41, further configured to perform layer normalization within the transformer model.
46. The system according to claim 41, further configured to adjust a set of transformer model parameters based on a set of training data representing past VCN performance.
47. The system according to claim 41, further configured to generate predictions of VCN resource allocation based on the output dataset.
48. A computer-implemented method for improving a value chain network (VCN) using a transformer model, A computer device receives an input dataset representing a VCN including nodes and edges. Encode the input dataset using an encoder from a transformer model to generate a contextual representation. The contextual representation is decoded using the decoder of the transformer model, and an output dataset showing VCN optimization is generated, and To facilitate decision-making in the aforementioned VCN, the improved VCN dataset is output. A computer-implemented method characterized by including the following.
49. Furthermore, the computer-implemented method according to claim 48 is characterized by including the application of a multi-head attention mechanism to determine the relationships between nodes and edges within the VCN.
50. Furthermore, the computer-implemented method according to claim 48 is characterized by further comprising performing autoregressive prediction of a VCN configuration using the output dataset.
51. Furthermore, the computer-implemented method according to claim 48 is characterized by including the use of position encoding to maintain the sequence order of the VCN data within the transformer model.
52. Furthermore, the computer-implemented method according to claim 48 is characterized by including performing layer normalization within the transformer model.
53. Furthermore, the computer-implemented method according to claim 48 is characterized by further comprising adjusting a set of transformer model parameters based on a set of training data representing past VCN performance.
54. Furthermore, the computer-implemented method according to claim 48 is characterized by comprising generating a prediction of VCN resource allocation based on the output dataset.
55. A dynamic visual system for mobile devices, Variable focus liquid lens optical assembly, A control system configured to adjust one or more optical parameters and data collected from the variable focus liquid lens optical assembly in real time, A processing system that dynamically learns based on results, parameters, and a training set of data collected from the variable focus liquid lens optical assembly, and trains one or more machine learning models for recognizing objects related to the mobile device, A dynamic visual system characterized by including the following.
56. The dynamic vision system according to claim 55, characterized in that the variable focus liquid lens assembly is continuously adjusted by a control system based on environmental factors and feedback from the processing system to generate an object concept.
57. The dynamic vision system according to claim 56, characterized in that the object concept includes contextual intelligence relating to the object and its environment, and provides excellent object recognition by the dynamic vision system.
58. The dynamic vision system according to claim 55, characterized in that a first machine learning model is used to optimize signal acquisition by the variable focus liquid lens optical assembly, and a second machine learning model is used to process the signal to achieve a desired visual result.
59. The dynamic vision system according to claim 55, characterized in that the processing system receives a real-time adjustable data stream from the variable focus liquid lens optical assembly to generate situational awareness or create an out-of-focus image of an object, thereby capturing rich metadata and contextual intelligence about the object and its environment.
60. The dynamic vision system according to claim 55, characterized in that the control system and the processing system are integrated with the variable focus liquid lens optical assembly.
61. The dynamic vision system according to claim 55, characterized in that the optical parameters adjusted by the control system include at least one of focal length, specular reflectance, color, environment, and lens shape.
62. The dynamic visual system according to claim 61, characterized in that adjusting the optical parameters changes at least one of spherical aberration, field curvature, coma aberration, chromatic aberration, distortion, vignetting, ghosting, flare, and diffraction.
63. The processing system is configured to learn from the results, parameters, and data set from the variable focus liquid lens optical assembly and derive the configuration of the liquid lens optical assembly. The dynamic vision system according to claim 55, characterized in that the configuration includes at least one of the material, geometry, shape, optical properties, performance, and design of the liquid lens.
64. The dynamic vision system according to claim 55, characterized in that the machine learning model is implemented on a semiconductor chip integrated into the mobile device housing the variable focus liquid lens optical assembly.
65. The aforementioned machine learning model is pre-trained in a separate system, such as a cloud computing environment, using a large training dataset of visual information and / or results, in order to perform a set of machine vision tasks. The dynamic vision system according to claim 55, characterized in that the pre-trained machine learning model is deployed in a device or system including the variable focus liquid lens optical assembly.
66. The dynamic vision system according to claim 55, characterized in that the mobile device includes at least one of a smartphone, a tablet, a smartwatch, an augmented reality (AR) device, and a portable game console.
67. A method for dynamically adjusting the visual system in a mobile device, Adjusting a variable focus liquid lens optical assembly, The control system controls one or more optical parameters and data collected from the variable focus liquid lens optical assembly in real time, and The processing system processes the results, parameters, and data set collected from the variable focus liquid lens optical assembly and trains one or more machine learning models for recognizing objects related to the mobile device. A method characterized by including the following.
68. The method according to claim 67, further comprising continuously adjusting the variable focus liquid lens assembly based on environmental factors and feedback from the processing system to generate an object concept.
69. The method according to claim 68, characterized in that generating the object concept includes incorporating contextual intelligence regarding the object and its environment, and providing superior object recognition by a dynamic visual system.
70. The method according to claim 67, further comprising using a first machine learning model to optimize signal acquisition by the variable focus liquid lens optical assembly and using a second machine learning model to process the signal to achieve a desired visual result.
71. The method according to claim 67, further comprising the processing system receiving a real-time adjustable data stream from the variable focus liquid lens optical assembly, generating situational awareness, or creating an out-of-focus image of the object to capture rich metadata and contextual intelligence about the object and its environment.
72. The method according to claim 67, characterized in that the control system and the processing system are integrated with the variable focus liquid lens optical assembly.
73. Furthermore, the method according to claim 67 is characterized in that the control system further includes adjusting optical parameters, including at least one of focal length, specular reflectance, color, environment, and lens shape.
74. The method according to claim 73, characterized in that the adjustment of the optical parameters changes at least one of spherical aberration, field curvature, coma aberration, chromatic aberration, distortion, vignetting, ghosting, flare, and diffraction.
75. The method according to claim 67, further comprising configuring the processing system to learn based on the results, parameters, and data set from the variable focus liquid lens optical assembly, and deriving a configuration of the liquid lens optical assembly that includes at least one of the material, geometry, shape, optical properties, performance, and design of the liquid lens.
76. The method according to claim 67, characterized in that the machine learning model is implemented on a semiconductor chip integrated into the mobile device housing the variable focus liquid lens optical assembly.
77. The method according to claim 67, further comprising pre-training the machine learning model on another system, such as a cloud computing environment, using a large training dataset of visual information and / or results, in order to perform a set of machine vision tasks, and deploying the pre-trained machine learning model to a device or system including the variable focus liquid lens optical assembly.
78. The method according to claim 67, characterized in that the mobile device includes at least one of a smartphone, a tablet, a smartwatch, an augmented reality (AR) device, and a portable game console.