Reconfigurable artificial intelligence ecosystem

Abstract

The invention discloses a reconfigurable artificial intelligence (AI) ecological system. The ecological system comprises an edge device based on machine learning, a cloud operating system and a modular service platform. The cloud operating system communicates with the edge device and the modular service platform to manage service delivery. After a task request from user equipment is received, the ecological system selects a proper edge model, generates a corresponding machine learning edge module and deploys the machine learning edge module to edge equipment. These edge devices then collect task-specific data, trim deployed modules locally, and use improved models to provide responsive and adaptive services. The ecosystem employs an open architecture, supporting cross-distributed environments to generate and deploy dynamic AI services on demand.

Description

(19) State Intellectual Property Office (12) Invention Patent Application (10) Application Publication Number (43) Application Publication Date (21) Application Number 202510983111.7 (22) Application Date 2025.07.16 (30) Priority Data 63 / 671,787 2024.07.16 US (71) Applicant City University of Hong Kong Address Tat Chee Road, Kowloon, Hong Kong, China (72) Inventors Wu Dapeng, Wang Shiqi, Chen Peilin (74) Patent Agency Shenzhen Yibao Intellectual Property Agency (General Partnership) 44588 Patent Attorney Wang Qin, Cao Yucun (51) Int.Cl. G06N 20 / 00 (2019.01) (54) Invention Title Reconfigurable Artificial Intelligence Ecosystem (57) Abstract This invention discloses a reconfigurable artificial intelligence (AI) ecosystem. The ecosystem includes: machine learning-based edge devices, a cloud operating system, and a modular service platform. The cloud operating system communicates with the edge devices and the modular service platform to manage service delivery. Upon receiving a task request from a user device, the ecosystem selects an appropriate edge model, generates a corresponding machine learning edge module, and deploys it to edge devices. These edge devices then collect task-specific data, fine-tune the deployed modules locally, and use the improved model to provide responsive and adaptive services. The ecosystem employs an open architecture, supporting the on-demand generation and deployment of dynamic AI services across distributed environments. Claims (2 pages), Description (6 pages), Drawings (2 pages), CN 121352062 A, 2026.01.16, CN 1 21 35 20 62 A. 1. A reconfigurable artificial intelligence ecosystem, comprising: one or more machine learning-based edge devices; a modular service platform including one or more AI-based modules; a cloud operating system communicating with the one or more machine learning-based edge devices and the modular service platform; wherein the cloud operating system includes one or more processors and a memory, the memory communicating with the one or more processors and storing instructions, wherein when the instructions are executed by the one or more processors, the modular service platform is configured to: receive a task request from a user device; select one or more edge models according to the task request; generate one or more machine learning edge modules based on the one or more selected edge models respectively; and transmit the one or more generated machine learning edge modules respectively to the one or more machine learning-based edge devices; and wherein the one or more edge devices are configured to: collect data related to the task request; fine-tune the one or more edge modules using the collected data; and use the one or more fine-tuned edge modules to provide services in response to the task request.2. The reconfigurable AI ecosystem of claim 1, wherein the modular service platform includes a large language model module configured to analyze the task request to obtain a parsed task request. 3. The reconfigurable AI ecosystem of claim 2, wherein the large language model module is further configured to convert the parsed task request into one or more service creation subtasks and one or more service deployment subtasks. 4. The reconfigurable AI ecosystem of claim 3, wherein the large language model module is further configured to execute the one or more service creation subtasks and the one or more service deployment subtasks. 5. The reconfigurable AI ecosystem of claim 4, wherein the one or more service creation subtasks include at least one task for constructing an event dictionary related to the task request. 6. The reconfigurable AI ecosystem of claim 5, wherein the one or more service deployment subtasks include: at least one task for designing an event detection algorithm based on the constructed event dictionary; at least one task for selecting one or more edge models based on the designed event detection algorithm; and at least one task for generating one or more machine learning edge modules based on the selected one or more edge models. 7. The reconfigurable artificial intelligence ecosystem of claim 1, wherein each of the one or more machine learning-based edge devices includes one or more sensors for collecting physiological and environmental data related to the task request. 8. The reconfigurable artificial intelligence ecosystem of claim 7, wherein the one or more sensors include at least one temperature sensor, at least one camera, at least one blood pressure monitor, and at least one pulse oximeter. Claims 1 / 2 Page 2 CN 121352062 A 9. The reconfigurable artificial intelligence ecosystem of claim 7, wherein the one or more machine learning edge modules are fine-tuned to: receive physiological and environmental data from the one or more sensors; autonomously detect events related to the task request; and transmit the detected events along with an alarm to the user device. 10. The reconfigurable artificial intelligence ecosystem of claim 6, wherein the one or more machine learning edge modules operate via at least one central processing unit, at least one graphics processing unit, and / or at least one advanced simplified instruction set computer. Claims 2 / 2 Page 3 CN 121352062 A Reconfigurable Artificial Intelligence Ecosystem Technical Field

[0001] This invention relates to the field of artificial intelligence (AI), specifically to a green, reconfigurable AI ecosystem with an open architecture that enables computing hardware to provide task-oriented AI services. Background Art

[0002] Contemporary artificial intelligence (AI) systems are primarily built using a closed architecture design. This means they are tightly integrated with proprietary hardware and software, incompatible with components developed by third-party manufacturers or developers. This design approach severely limits the modularity and flexibility of the system. In practice, these AI systems function much like a "black box"—their internal workings are opaque to the user, and modification or customization is extremely difficult, if not entirely impossible. Therefore, end users are limited to predefined functions and features, with little room for personalization, expansion, or innovation based on evolving needs.

[0003] This rigidity in the context of dynamic task objectives and user-specific domains (e.g., healthcare applications) imposes significant limitations. In many cases, users (e.g., healthcare service providers) may need to tailor new functions or services for specific patient groups or cases. However, within a closed AI architecture, meeting this requirement would necessitate procuring entirely new hardware / software systems. This not only wastes resources and increases costs, but also exacerbates the generation of electronic waste. Given the increasing emphasis on sustainable development and environmental protection in society today, this outcome clearly runs counter to the goals of green development.

[0004] The proliferation of wearable devices and the growing interest in task-oriented AI services further exacerbate the problem. Wearable technologies, such as smartwatches, fitness trackers, and medical monitoring devices, generate large amounts of health-related data in real time. Effective utilization of this data requires adaptable, scalable, and continuously learning and adaptable AI systems. However, the closed nature of current AI architectures limits the seamless integration of such data and the system's ability to evolve in response to new inputs or task objectives.

[0005] Therefore, there is an urgent need for green computing solutions that not only address the technical limitations of current systems but also prioritize adaptability, interoperability, and autonomy. Green computing solutions refer to the design and deployment of energy-efficient, sustainable systems that can operate for extended periods without frequent human intervention or system overhauls. It also requires the development of open, modular architectures that allow hardware and software components from different vendors to integrate and operate seamlessly.

[0006] The rapid growth of wearable devices and the growing interest in task-oriented artificial intelligence (AI) services have further exacerbated existing problems. Wearable technologies such as smartwatches, fitness trackers, and medical monitoring devices can generate large amounts of health-related data in real time. To effectively utilize this data, there is a pressing need for highly adaptable, scalable, and continuously learning and dynamically adjusting AI systems. However, the closed nature of existing AI architectures limits the seamless integration of such data and also reduces…The system's ability to respond and evolve based on new inputs or task objectives is weakened. Therefore, it is necessary to break through the current system's technical bottlenecks, design and deploy energy-saving and sustainable systems in accordance with the concept of green computing, and prioritize the system's adaptability, interactivity, and autonomy. These systems should be able to work stably for a long time without frequent human intervention or large-scale system maintenance. In addition, such solutions should also promote the development of open modular architectures, enabling hardware and software components from different vendors to be integrated and operate efficiently and collaboratively.

[0007] Furthermore, it must be recognized that many AI-supported services in the current healthcare field still rely heavily on human expertise and labor. For example, in patient care scenarios, nurses play a key role in data interpretation, decision-making, and taking action based on AI-generated insights. While this human-machine collaborative approach helps ensure the safety and personalization of services, it still places a heavy burden on labor costs and resource allocation. In high-demand environments, this reliance can lead to staff burnout and decreased efficiency, further highlighting the urgent need for AI systems with higher autonomy and intrinsic system intelligence. The transition from a closed, monolithic architecture to an open, flexible, and sustainable system is not only a necessary step for technological progress but also a strategic core for future development, especially in key areas such as healthcare.

[0008] One objective of this invention is to provide a practical AI ecosystem with an open architecture that can be reconfigured for various real-world scenarios, thereby enabling the efficient and reliable delivery of services such as healthcare monitoring. Users can change functionality without purchasing new hardware. This AI ecosystem can mitigate environmental pollution caused by non-recyclable components in old electronic devices and reduce carbon emissions from manufacturing new hardware, benefiting users in various scenarios.

[0009] According to one aspect of the invention, a reconfigurable artificial intelligence ecosystem is provided. The ecosystem includes: one or more machine learning-based edge devices; a modular service platform including one or more AI-based modules; and a cloud operating system communicating with the one or more machine learning-based edge devices and the modular service platform. The cloud operating system includes one or more processors and a memory. The memory communicates with the one or more processors and stores instructions. When the instructions are executed by the one or more processors, the modular service platform is configured to: receive a task request from a user device; select one or more edge models according to the task request; generate one or more machine learning edge modules based on the one or more selected edge models; and transmit the one or more generated machine learning edge modules to the one or more machine learning-based edge models respectively.Edge devices. The one or more edge devices are configured to: collect data related to the task request; fine-tune the one or more edge modules using the collected data; and use the one or more fine-tuned edge modules to provide services in response to the task request.

[0010] Preferably, the modular service platform includes a large language model module configured to analyze the task request to obtain a parsed task request.

[0011] Preferably, the large language model module is further configured to convert the parsed task request into one or more service creation subtasks and one or more service deployment subtasks.

[0012] Preferably, the large language model module is further configured to execute the one or more service creation subtasks and the one or more service deployment subtasks.

[0013] Preferably, the one or more service creation subtasks include at least one task for constructing an event dictionary related to the task request.

[0014] Preferably, the one or more service deployment subtasks include: at least one task for designing an event detection algorithm based on the constructed event dictionary; at least one task for selecting one or more edge models based on the designed event detection algorithm; and at least one task for generating one or more machine learning edge modules based on the selected one or more edge models.

[0015] Preferably, each of the one or more machine learning-based edge devices includes one or more sensors for collecting physiological and environmental data related to the task request.

[0016] Preferably, the one or more sensors include at least one temperature sensor, at least one camera, at least one blood pressure monitor, and at least one pulse oximeter.

[0017] Preferably, the one or more machine learning edge modules are fine-tuned to: receive physiological and environmental data from the one or more sensors; autonomously detect events related to the task request; and transmit the detected events along with alarms to the user device.

[0018] Preferably, the one or more machine learning edge modules operate via at least one central processing unit, at least one graphics processing unit, and / or at least one advanced simplified instruction set computer.

[0019] The AI ​​ecosystem provided in this application adopts the concept of green computing and aims to provide adaptive and ubiquitous intelligent services according to different user needs and environmental changes. The system can be seamlessly integrated with a variety of wearable devices and supports a sustainable operating mode. It possesses a high degree of autonomy, capable of automatically performing tasks such as event dictionary construction, model selection, code generation, event description, and pre-training. Unlike traditional systems that can only complete predefined tasks, this AI ecosystem can automatically perform tasks based on user needs.The system is designed to dynamically adjust its functionality (e.g., flexibly switching between childcare and elderly care) without requiring hardware replacement, thus achieving greater flexibility, convenience, and cost-effectiveness.

[0020] Driven by a machine learning service platform, the system can be flexibly scaled on public cloud infrastructure to handle larger datasets and more complex tasks. Its green design ensures that new functions are implemented on existing hardware, minimizing environmental impact and reducing e-waste generation. Through event description generation and continuous learning mechanisms, the system transmits only critical data, thereby reducing deployment scale and enhancing user privacy. The system also has continuous and targeted data collection capabilities, enabling early identification before potential problems escalate into serious events, effectively supporting proactive and preventative interventions.

[0021] The system's architecture is designed to support ubiquitous computing services, possessing excellent environmental adaptability, covering everything from indoor spaces such as living rooms and bedrooms to various outdoor environments. The versatility of the AI ​​ecosystem is reflected in its ability to consistently provide stable and reliable computing services under varying environmental conditions and with different user needs. The provided AI ecosystem represents an important step in combining green computing with universal service delivery, offering scalable, adaptable, and environmentally friendly computing solutions.

[0022] Referring to the accompanying drawings, embodiments of the present invention will be described in more detail below, wherein:

[0023] FIG1 shows a schematic block diagram of an AI ecosystem based on an open architecture according to an embodiment of the present invention.

[0024] FIG2 shows an exemplary flow of an ecosystem for performing AI service creation, service deployment, and service execution according to an embodiment of the present invention. Detailed Description

[0025] In the following description, details of the present invention will be described by way of preferred embodiments. Modifications, additions, or substitutions to the invention will be apparent to those skilled in the art without departing from the spirit and scope thereof. Certain specific details omitted in this description are intended to avoid redundant description and obscuring the core content of the invention. However, the purpose of this disclosure is to enable those skilled in the art to understand and practice the technical solutions disclosed herein without excessive experimentation.

[0026] FIG1 shows a schematic block diagram of an AI ecosystem 100 based on an open architecture according to an embodiment of the present invention. The artificial intelligence ecosystem 100 consists of three layers: the first layer includes one or more machine learning-based edge devices (see page 3 / 6 of CN 121352062 A 101); the second layer includes a cloud operating system (OS) 102 for operating and managing the one or more edge devices; and the third layer includes a modular service platform 103, which includes one or more artificial intelligence (AI) modules running on the cloud operating system to support flexible migration and efficient scalability.

[0027] The cloud operating system 102 includes one or more processors and a memory that communicate with and store instructions for the service platform 103, which, when executed by the one or more processors, configure the service platform 103 to: receive a task request from a user device; select one or more edge models based on the task request; generate one or more machine learning edge modules based on the one or more selected edge models respectively; and transmit the one or more generated machine learning edge modules to one or more machine learning-based edge devices.

[0028] One or more edge devices 101 are configured to: collect data related to the task request; fine-tune one or more edge modules using the collected data; and use one or more fine-tuned edge modules to provide services in response to the task request.

[0029] One or more AI modules in the service platform can be trained to provide artificial intelligence services such as elderly care, child care, patient care, and security monitoring. Edge devices can be equipped with any type of sensor, including but not limited to temperature sensors, image sensors (cameras), blood pressure monitors, and pulse oximeters. Edge devices also include any type of computing hardware for executing machine learning modules. Computing hardware can include, but is not limited to, central processing units (CPUs), graphics processing units (GPUs), and advanced reduced instruction set computing (RISC) computers (e.g., ARM).

[0030] The AI ​​Ecosystem 100 features an open architecture, allowing any of its hardware / software components to be replaced by third-party components. Under this open architecture, if a user wants new AI functionality, they can reuse existing hardware and reconfigure its functionality to meet the new requirements, thus saving the cost of purchasing new hardware. The hardware system can be equipped with various data acquisition devices, capable of capturing multiple forms of data, and can dynamically reconfigure its functionality according to needs to support the execution of different tasks, making it suitable for various application scenarios and maximizing its benefits.

[0031] The AI ​​Ecosystem 100, as a pioneering solution, follows green computing principles, providing highly adaptable and ubiquitous computing services to dynamically respond to user needs and environmental conditions. The provided AI ecosystem aims to achieve seamless integration with heterogeneous wearable devices, supporting sustainable computing models and avoiding frequent hardware replacements. The system's design philosophy not only allows for flexible switching between scenarios such as elderly care and patient care based on user needs, but also continuously meets the computing demands arising from the evolving nature of various task-oriented services. This system is particularly adept at providing continuous, real-time services in key areas such as patient care. In one embodiment, an AI ecosystem is integrated into the ventilator system to support patient monitoring in a highly reliable manner. A key feature of this ventilator system is its ability to provide multi-tiered services, from basic status updates (nurse-level services) to comprehensive analysis and diagnosis (physician-level services), all through the integration of multiple...Various sensing modes are implemented to meet the needs of different clinical roles.

[0032] In addition, the AI ​​ecosystem realizes the creation and deployment of intelligent services in a self-organizing manner, enabling hardware to provide services autonomously. Specifically, the creation and deployment of autonomous services are realized through automatic event dictionary construction, model selection, code generation and pre-training.

[0033] First, a large language model (LLM) module will be used to build an event dictionary related to the user's requested task. For example, if the task is advanced care, the relevant events include blood glucose spikes and heart rate abnormalities. Then, a model will be selected from the model library and code for detecting events in the dictionary will be automatically generated. The system will also automatically search for relevant datasets to pre-train the selected model. The pre-trained model will then be deployed in the sensor as a base model. The sensor will further fine-tune the base model using its local dataset. It is worth mentioning that these stages are performed in a self-organizing manner by the AI ​​ecosystem of the present invention without human intervention, ensuring that various user requests can be quickly served in a highly adaptable and ubiquitous manner through intelligent services and deployment.

[0034] Figure 2 illustrates an exemplary process of AI service creation, deployment, and execution for healthcare task requests within an ecosystem according to an embodiment of the present invention.

[0035] First, a user sends a request for a personalized task (e.g., elderly care) to the service platform via their user device. Then, an AI-based parser module within the service platform automatically processes the request, configured to interpret the user's input and extract relevant parameters. A subsequent planner module converts the interpreted request into a series of executable subtasks related to service creation and deployment. These subtasks may include, but are not limited to, building an event dictionary specific to the personalized task and designing corresponding event detection algorithms. For example, in an elderly care application, the event dictionary may define critical events such as accidents (e.g., falls) and medical conditions (e.g., diabetes, cardiac arrest, and sleep apnea), enabling the system to monitor and respond to such events promptly and automatically.

[0036] The design of the event detection algorithm includes selecting a suitable computational model and generating executable code through an automated process. After generating the model and corresponding code, they are transmitted to an edge device (also known as an edge device or information provider) for localized execution. The edge device is then configured to execute the received computational model and executable code for real-time monitoring and event detection, thereby enabling distributed and responsive healthcare service delivery.

[0037] The edge device may include an infrared temperature sensor for monitoring the body temperature of elderly individuals and a camera for capturing real-time video footage. Furthermore, the edge device may also be equipped with a medical processor configured to utilize data from the service platform.The received model and executable program code are used to analyze temperature data and video streams. Based on the historical health database of the elderly, the medical processor can perform lifetime alignment analysis to identify deviations or patterns related to individual health status; perform attribute decoupling to isolate relevant indicators; and generate structured event descriptions reflecting the analysis results. Subsequently, the generated event descriptions and corresponding alarms will be transmitted to the user device for timely response and intervention.

[0038] The functional units and modules of the reconfigurable artificial intelligence ecosystem according to the embodiments disclosed herein can be implemented using computing devices, computer processors, or electronic circuits, including but not limited to application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), microcontrollers, and other programmable logic devices configured or programmed according to the teachings of this disclosure. The computer instructions or software code running in the computing device, computer processor, or programmable logic device can be easily prepared by a practitioner skilled in software or electronic arts based on the teachings of this disclosure.

[0039] All or part of the methods according to the embodiments can be executed in one or more computing devices, including mobile computing devices such as server computers, personal computers, laptops, smartphones, and tablets.

[0040] Embodiments may include computer storage media, transient and non-transient storage devices having computer instructions or software code stored therein, which can be used to program or configure computing devices, computer processors, or electronic circuits to perform any of the processes of the present invention. Storage media, transient and non-transient storage devices include, but are not limited to, floppy disks, optical disks, Blu-ray discs, DVDs, CD-ROMs and magneto-optical disks, ROMs, RAMs, flash memory devices, or any type of media or device suitable for storing instructions, code, and / or data.

[0041] Each functional unit and module according to various embodiments may also be implemented in a distributed computing environment and / or cloud computing environment, wherein all or part of the machine instructions are executed in a distributed manner by one or more processing devices interconnected by communication networks, such as intranets, wide area networks (WANs), local area networks (LANs), the Internet, and other forms of data transmission media.

[0042] While this disclosure has been described and illustrated with reference to specific embodiments thereof, such descriptions and illustrations are not limiting. Illustrations are not necessarily drawn to scale. Due to manufacturing processes and tolerances, there may be differences between the illustrations in this disclosure and actual devices. Other embodiments of this disclosure may not be specifically described. Modifications may be made to adapt specific circumstances, materials, composition, methods, or processes to the objectives and scope of this disclosure. All such modifications are intended to be incorporated into this agreement.Within the scope of the appended claims. While the methods disclosed herein have been described with reference to specific actions performed in a particular order, it is understood that these actions may be combined, subdivided, or reordered to form equivalent methods without departing from the doctrine of this disclosure. Therefore, unless specifically stated herein, the order and grouping of actions are not intended to be limiting.

[0043] The foregoing description is provided for the purpose of illustrating and describing the invention and is not intended to be exhaustive or to limit the invention to the precise forms disclosed. Many modifications and variations will be apparent to those skilled in the art.

[0044] The above embodiments were chosen and described in order to best explain the principles of the invention and its practical application, thereby enabling others skilled in the art to understand the various embodiments of the invention and the various modifications suitable for the intended particular use. Description page 6 / 6 CN 121352062 A Figure 1 Description drawing 1 / 2 page 10 CN 121352062 A Figure 2 Description drawing 2 / 2 page 11 CN 121352062 A Abstract Abdominal ultrasound examination method, system and device The present invention discloses a reconfigurable artificial intelligence (AI) ecosystem. The ecosystem consists: machine learning-based edge devices, a cloud-based operating system, and a modular service platform. The cloud operating system communicates with edge devices and the modular service platform to manage service delivery. Upon receiving a mission request from a subscriber device, the ecosystem selects appropriate edge models, generates corresponding machine learning edge modules, and deploys them to the edge devices. The edge devicesthen collect mission- specific data, fine-tune the deployed modules locally, and use the refined models to deliver responsive and adaptive services. The ecosystem adopts an open architecture supporting on-demand generation and deployment of dynamic AI service across distributed environments.

Claims

1. A reconfigurable artificial intelligence ecosystem, comprising: one or more machine learning based edge devices; a modular service platform comprising one or more artificial intelligence based modules; a cloud operating system in communication with the one or more machine learning based edge devices and the modular service platform; wherein the cloud operating system comprises one or more processors and a memory in communication with the one or more processors and storing instructions that, when executed by the one or more processors, configure the modular service platform to: receive a task request from a user device; select one or more edge models based on the task request; generate one or more machine learning edge modules based on the one or more selected edge models, respectively; and transmit the one or more generated machine learning edge modules to the one or more machine learning based edge devices, respectively; and wherein the one or more edge devices are configured to: collect data related to the task request; fine-tune the one or more edge modules using the collected data; and provide a service responsive to the task request using the one or more fine-tuned edge modules.

2. The reconfigurable artificial intelligence ecosystem of claim 1, wherein, the modular service platform comprises a large language model module configured to analyze the task request to obtain a parsed task request.

3. The reconfigurable artificial intelligence ecosystem of claim 2, wherein, the large language model module is further configured to convert the parsed task request into one or more service creation sub-tasks and one or more service deployment sub-tasks.

4. The reconfigurable artificial intelligence ecosystem of claim 3, wherein, the large language model module is further configured to execute the one or more service creation sub-tasks and the one or more service deployment sub-tasks.

5. The reconfigurable artificial intelligence ecosystem of claim 4, wherein, the one or more service creation sub-tasks comprise at least one task for building an event dictionary related to the task request.

6. The reconfigurable artificial intelligence ecosystem of claim 5, wherein, the one or more service deployment sub-tasks comprise: at least one task for designing an event detection algorithm based on the built event dictionary; at least one task for selecting one or more edge models based on the designed event detection algorithm; and at least one task for generating one or more machine learning edge modules based on the selected one or more edge models, respectively.

7. The reconfigurable artificial intelligence ecosystem of claim 1, wherein, each of the one or more machine learning based edge devices comprises one or more sensors for collecting physiological data and environmental data related to the task request.

8. The reconfigurable artificial intelligence ecosystem of claim 7, wherein, the one or more sensors comprise at least one temperature sensor, at least one camera, at least one sphygmomanometer, and at least one pulse oximeter.

9. The reconfigurable artificial intelligence ecosystem of claim 7, wherein, the one or more machine learning edge modules are fine-tuned to: receive physiological and environmental data from the one or more sensors; autonomously detect an event related to the task request; and transmit the detected event to the user device along with an alert. the one or more machine learning edge modules are operated by at least one central processing unit, at least one graphics processing unit, and / or at least one advanced reduced instruction set computer.

10. The reconfigurable artificial intelligence ecosystem of claim 6, wherein, ​

Images