AI edge calculation method and device, electronic equipment and storage medium

Abstract

The invention provides an AI edge calculation method and device, electronic equipment and a storage medium. The AI edge calculation method comprises the steps that an edge server obtains a first model set corresponding to a processing service type of the edge server; selecting a corresponding first business processing terminal according to the task plan and the data demand of the business processing terminal; obtaining terminal attributes and service request data; generating a second model set by using the terminal attribute data, a model adjustment algorithm and the first model set; determining a business processing party; when it is determined that the service processing party is the first service processing terminal, selecting a second model from the second model set and sending the second model to the terminal; acquiring the current environment and operation data of the terminal, and sending the data and the control instruction to the terminal to control the terminal to adjust the second model according to the data and a preset algorithm to generate a third model; and processing business data by using the model, and executing corresponding operation. Through the scheme, the performance of edge calculation can be improved, and more efficient and more intelligent business processing can be realized.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an AI edge computing method, device, electronic device and storage medium. Background Art

[0002] Edge computing is a new computing system and technology that moves computing power from the cloud to the edge of the network to achieve real-time business, efficient data processing, application intelligence, and security and privacy protection. At present, with the development of artificial intelligence technology, artificial intelligence technology is gradually involved in all aspects of people's work and life, but artificial intelligence technology has very high requirements for computing resources and storage resources, and is relatively inefficient and not smart enough when performing edge computing. Summary of the invention

[0003] Based on the above problems, the present invention proposes an AI edge computing method, device, electronic device and storage medium. Through the solution of the present invention, not only the performance of AI edge computing can be improved, but also more efficient and intelligent business processing can be achieved in various application scenarios.

[0004] In view of this, one aspect of the present invention proposes an AI edge computing method, comprising: The edge server obtains a plurality of first model sets corresponding to the business types of the business processed by the edge server, wherein the first model sets are generated by the cloud server according to the specific business types and business contents; The edge server selects a corresponding first business processing terminal from the business processing terminals for connection according to the task plans and data processing requirements of the multiple business processing terminals; The edge server obtains terminal attribute data and service request data of the first service processing terminal; The edge server generates a second model set according to the terminal attribute data, a preset first model adjustment algorithm and the first model set; The edge server determines the service processor according to the terminal attribute data, its own service processing capability and resource occupancy rate; When it is determined that the service processor is the first service processing terminal, the edge server determines a corresponding second model from the second model set according to the service request data, and sends the second model to the first service processing terminal; The edge server obtains current environment data of the environment in which the first service processing terminal is located and current operation data of the first service processing terminal; The edge server sends the current environment data, the current operation data and the first control instruction to the first business processing terminal, so as to control the first business processing terminal to adjust the second model to obtain a third model according to the current environment data, the current operation data and a preset second model adjustment algorithm; The edge server controls the first service processing terminal to process the service data using the third model, and controls the first service processing terminal to perform corresponding operations according to the processing result.

[0005] Optionally, the edge server obtains a plurality of first model sets corresponding to the business types of the business processed by itself, and the first model sets are generated by the cloud server according to the specific business types and business contents, including: The edge server sends a service type request to the cloud server, wherein the service type request includes: unique identification information of the edge server; hardware configuration information of the edge server, including processor type, memory size and storage capacity; geographic location information of the area served by the edge server; a list of service types currently processed by the edge server, each service type including a specific service type identification and service content description; The cloud server performs the following operations according to the business type request: confirm the authorization status of the edge server according to the unique identification information of the edge server; filter out the model type that is adapted to the hardware conditions according to the hardware configuration information of the edge server; select the training data set of the corresponding area according to the geographical location information of the area served by the edge server; for each business type: select the corresponding basic model from the model library according to the business type identification; determine the specific parameter configuration of the model according to the business content description; use the selected training data set to train and optimize the basic model; generate an optimized model adapted to the business type; The cloud server packages the generated multiple optimization models into a first model set, and sends a model distribution request to the edge server, wherein the model distribution request includes: metadata information of each optimization model, including model version, applicable business type and resource requirements; model deployment configuration information; model update strategy; The edge server receives the model distribution request and performs the following operations: verifying the integrity and security of the model set; creating a model operating environment according to the model deployment configuration information; storing the first model set in a preset model library; and setting a regular inspection and update mechanism according to the model update strategy.

[0006] Optionally, the step of the edge server selecting a corresponding first business processing terminal from the business processing terminals for connection according to task plans and data processing requirements of the multiple business processing terminals includes: The edge server establishes a terminal management table, which includes: the device identification of each business processing terminal, the network connection status of each business processing terminal, the task plan information of each business processing terminal, and the data processing requirement information of each business processing terminal; wherein the task plan information includes: task type identification task execution time period, task priority and task plan status; the data processing requirement information of each business processing terminal includes: processing type, data scale, real-time requirements and accuracy requirements; The edge server performs terminal classification according to the terminal management table, specifically including: grouping terminals according to task type; grouping terminals according to data processing type; within each group, sorting terminals according to task priority; marking terminals with multi-task processing capabilities; The edge server selects the first service processing terminal according to the following conditions, including: judging the task time overlap and selecting a terminal with matching time resources; judging the matching degree between the terminal's data processing capability and the demand; evaluating the stability of the network connection state; and calculating the task priority comprehensive score; wherein the calculation formula of the task priority comprehensive score is: score = w1×time matching degree + w2×processing capability matching degree + w3×network stability + w4×task priority, where w1, w2, w3, and w4 are preset weight coefficients; The edge server establishes a connection with the selected first business processing terminal, including: sending a connection request to the first business processing terminal; waiting for the terminal to respond and verifying the connection status; establishing a secure communication channel; synchronizing the task plan and data processing configuration; updating the connection status in the terminal management table; The edge server establishes a task execution monitoring mechanism, including: regularly checking the connection status; monitoring the task execution progress; recording data processing performance indicators; and dynamically adjusting the terminal selection strategy according to the monitoring results.

[0007] Optionally, the step of generating the second model set by the edge server according to the terminal attribute data, a preset first model adjustment algorithm and the first model set includes: The edge server parses and classifies the terminal attribute data to determine the terminal hardware attributes, terminal software attributes and terminal function attributes; wherein the terminal hardware attributes include: processor architecture type, processor performance parameters, memory capacity and type, and whether there is a dedicated AI accelerator; the terminal software attributes include: operating system type and version, AI framework support and device driver compatibility; the terminal function attributes include: supported sensor types, actuator types and communication interface types; The edge server performs model analysis on each model in the first model set to obtain model structure information, model operation requirements and model application scenarios; wherein the model structure information includes: number and type of network layers, parameter scale and computational complexity; model operation requirements include: minimum computing resource requirements, memory usage requirements and response time requirements; model application scenarios include: applicable task types, input data requirements and output format requirements; The edge server executes a first model adjustment algorithm, specifically including: adjusting the implementation of the model calculation layer according to the terminal processor architecture; optimizing the parameter storage structure according to the terminal memory capacity; adjusting the operator implementation according to whether there is an AI accelerator; selecting the corresponding quantization accuracy from a predefined quantization accuracy level set according to the terminal performance, performing weight quantization, and performing activation value quantization; calculating parameter importance scores, removing low-importance parameters, and retraining the pruned model structure; The edge server performs adaptability verification on the optimized model, including: simulating the terminal hardware environment and configuring the target software environment; testing computing performance, memory usage and response delay, and testing various input scenarios; verifying output accuracy; and checking the exception handling mechanism; The edge server integrates the optimized models that have passed verification into a second model set, including: generating a model description file; packaging the model file; configuring model loading parameters; and setting version control information.

[0008] Optionally, the step of the edge server determining the service processor according to the terminal attribute data, its own service processing capability and resource occupancy rate includes: The edge server establishes a resource evaluation index system, which includes computing resource indicators, network resource indicators and storage resource indicators; wherein the computing resource indicators include: CPU utilization and available core number, GPU / NPU usage, memory utilization and available capacity; network resource indicators include: network bandwidth occupancy, network latency level, connection stability; storage resource indicators include: storage space utilization, I / O performance indicators, cache hit rate; The edge server constructs a processing capability evaluation model, including: Evaluate its own processing capabilities: calculate the current task queue length of the edge server; calculate the average processing delay; predict short-term load trends; Evaluate the first service processing terminal: analyze the computing capability of the first service processing terminal; evaluate the resource margin of the first service processing terminal; calculate the task adaptability; Evaluation of other processing terminals: obtaining real-time status of other terminals; evaluating collaborative processing capabilities; calculating load distribution ratios; The edge server performs multi-dimensional decision calculations, including: Calculate the local processing cost: Cost_local = q1 × computing resource usage + q2 × storage resource usage; Calculate the terminal processing cost: Cost_terminal = q3 × terminal resource usage + q4 × network transmission overhead; Calculate the cost of cooperative processing: Cost_cooperative = q5 × task decomposition overhead + q6 × coordination communication overhead; Computing processing benefits, including: computing processing time benefits; computing resource utilization benefits; computing system stability benefits; Generate decision scores, including: Score = benefit index - cost index; apply constraint adjustment; calculate the final decision score; The edge server establishes a dynamic load balancing mechanism, including: Real-time monitoring of load status, specifically: monitoring resource changes of each processor; recording task processing performance; analyzing load distribution; Perform load adjustment, specifically: set load threshold; calculate load deviation; trigger load migration; Update the processing strategy, specifically: record the processing effect; optimize the decision parameters; adjust the weight coefficient; The edge server determines the final service processor according to the following rules, including: When Score_local is the highest and meets the threshold requirement, the edge server is selected as the processor; the edge server allocates and locks a specific number of processor resources, memory resources, and storage resources from the available computing resource pool according to the resource requirements of the business processing, forms an independent resource space, and initializes the processing environment in the resource space; When Score_terminal is the highest and meets the threshold requirement, the first service processing terminal is selected as the processor; task parameters are configured for the first service processing terminal, and a monitoring mechanism is established; When Score_cooperative is the highest and meets the threshold requirements, the cooperative processing mode is selected, the task boundaries are divided, and the cooperative mechanism is established.

[0009] Optionally, when it is determined that the service processor is the first service processing terminal, the step of the edge server determining the corresponding second model from the second model set according to the service request data, and sending the second model to the first service processing terminal includes: The edge server parses the service request data, including: Analyze business type information, specifically: identify the main business type; identify the subtask type; determine the task priority; Analyze data feature information, specifically: analyze data scale; identify data format; determine data quality requirements; Analyze performance requirement information, specifically: obtain real-time requirements; obtain accuracy requirements; obtain resource constraints; The edge server selects candidate models from the second model set, including: Perform feature matching, specifically: matching business type labels; matching data format requirements; matching performance indicator requirements; Evaluate the model's suitability, specifically: calculate feature matching; evaluate performance satisfaction; calculate resource adaptability; Generate a list of candidate models, specifically: sort by matching degree; filter models that do not meet the constraints; record model scores; The edge server performs model combination optimization, including: Analyze task dependencies, specifically: build a task dependency graph; identify critical paths; determine the execution order; Calculate resource constraints, specifically: count available resources; calculate resource requirements; evaluate resource matching; Generate the best combination plan, specifically: calculate the combination benefits; evaluate the feasibility of the combination; select the best plan; The edge server performs a model packaging operation, including: Prepare model files, specifically: organize model parameters; package model structure; generate configuration files; Create deployment instructions, specifically: record dependencies; set deployment sequence; configure run parameters; Perform integrity checks, specifically: verify file integrity; check configuration correctness; generate verification information; The edge server performs model distribution, including: Establish a secure transmission channel, specifically: negotiate encryption method; establish data channel; verify connection security; Perform batch transfers, specifically: set transfer priority; control transfer rate; monitor transfer status; Verify the distribution results, specifically: check the reception integrity; verify the model availability; confirm the deployment success.

[0010] Optionally, the step in which the edge server sends the current environment data, the current operation data and the first control instruction to the first business processing terminal to control the first business processing terminal to adjust the second model according to the current environment data, the current operation data and a preset second model adjustment algorithm to obtain a third model includes: The edge server collects and pre-processes environmental data, including: Collect physical environment data such as temperature, humidity, light, vibration, noise, spatial position and movement status; Collect business environment data such as the distribution of peripheral equipment, network environment conditions, and electromagnetic interference conditions; Performing data preprocessing on the physical environment data and the business environment data, specifically: data standardization processing; outlier detection and processing; data time series alignment; The edge server organizes the current operation data, which includes resource usage status, network transmission status, and hardware operation parameters; wherein the resource usage status includes: CPU, GPU usage, memory occupancy, and storage space usage; the network transmission status includes: real-time bandwidth utilization, network delay data, and data packet loss rate; the hardware operation parameters include: processor temperature, power consumption data, and the working status of each component; The edge server generates a first control instruction, which includes a model adjustment strategy, a resource allocation strategy, and an execution control strategy; wherein the model adjustment strategy includes: determining an adjustment target, setting an adjustment range, and defining an adjustment step; the resource allocation strategy includes: allocating computing resources, setting memory limits, and controlling energy consumption levels; the execution control strategy includes: setting execution priority, defining a timeout mechanism, and configuring exception handling; The edge server integrates and sends the data packets, including: Data packaging, specifically: merging environmental data; integrating operational data; including control instructions; Data compression, specifically: selecting a compression algorithm; performing data compression; generating a checksum; Send in batches, specifically: determine the sending order; control the sending rate; monitor the sending status; The edge computing module of the first service processing terminal performs model adjustment, including: Parse received data, specifically: decompress data packets; verify data integrity; classify and store data; Evaluate the necessity of adjustment, specifically: analyze the degree of environmental change; evaluate the degree of performance impact; calculate the benefits of adjustment; Performing model adjustment, specifically: applying the second model adjustment algorithm; fine-tuning parameters; optimizing the model structure; The edge computing module verifies the adjusted third model, including: Perform performance testing, specifically: test computing performance; test response time; test resource usage; Verify functional correctness, specifically: test model accuracy; verify output stability; check for abnormal responses; Generate a verification report, specifically: record the adjustment effect; compile statistics on performance indicators; and provide optimization suggestions.

[0011] Another aspect of the present invention provides an AI edge computing device for executing an AI edge computing method, comprising: a data acquisition module and a control processing module; wherein: The data acquisition module is configured to: acquire a plurality of first model sets corresponding to the business type of the business processed by the module, wherein the first model sets are generated by the cloud server according to the specific business type and business content; The control processing module is configured to: select a corresponding first business processing terminal from the business processing terminals for connection according to the task plans and data processing requirements of the multiple business processing terminals; The data acquisition module is further configured to: acquire terminal attribute data and service request data of the first service processing terminal; The control processing module is also configured to: Generate a second model set according to the terminal attribute data, a preset first model adjustment algorithm and the first model set; Determine the service processor according to the terminal attribute data, its own service processing capability and resource occupancy rate; When it is determined that the service processor is the first service processing terminal, determining a corresponding second model from the second model set according to the service request data, and sending the second model to the first service processing terminal; The data acquisition module is further configured to: acquire current environment data of the environment in which the first service processing terminal is located and current operation data of the first service processing terminal; The control processing module is also configured to: Sending the current environment data, the current operation data and a first control instruction to the first business processing terminal to control the first business processing terminal to adjust the second model to obtain a third model according to the current environment data, the current operation data and a preset second model adjustment algorithm; The first business processing terminal is controlled to process the business data using the third model, and the first business processing terminal is controlled to perform corresponding operations according to the processing results.

[0012] Another aspect of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of an AI edge computing method when executing the computer program.

[0013] Another aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of an AI edge computing method when executed by a processor.

[0014] According to the technical solution of the present invention, the AI ​​edge computing method includes: the edge server obtains multiple first model sets corresponding to the business type of the business processed by itself, and the first model set is generated by the cloud server according to the specific business type and business content; the edge server selects the corresponding first business processing terminal from the business processing terminals for connection according to the task plans and data processing requirements of multiple business processing terminals; the edge server obtains the terminal attribute data and business request data of the first business processing terminal; the edge server generates a second model set according to the terminal attribute data, a preset first model adjustment algorithm and the first model set; the edge server determines the business processing party according to the terminal attribute data, its own business processing capability and resource occupancy rate; when the business processing party is determined to be the first business processing terminal, the The edge server determines the corresponding second model from the second model set according to the business request data, and sends the second model to the first business processing terminal; the edge server obtains the current environment data of the environment in which the first business processing terminal is located and the current operation data of the first business processing terminal; the edge server sends the current environment data, the current operation data and the first control instruction to the first business processing terminal to control the first business processing terminal to adjust the second model according to the current environment data, the current operation data and the preset second model adjustment algorithm to obtain a third model; the edge server controls the first business processing terminal to process the business data using the third model, and controls the first business processing terminal to perform corresponding operations according to the processing results. Through the scheme of the present invention, the edge server can use computing resources more efficiently and reduce unnecessary calculations and delays by dynamically selecting appropriate models according to the task plan and data processing requirements of the business processing terminal; the edge server adjusts the model according to the terminal attribute data and the current environment data, so that the model can adapt to different business scenarios and environmental changes in real time, thereby improving the accuracy and reliability of processing; by real-time monitoring of the computing resource occupancy rate and network transmission rate of the terminal, the edge server can intelligently allocate resources to avoid resource waste and ensure the efficiency of business processing; the implementation of edge computing makes data processing closer to the data source, reduces the time delay of data transmission, and is particularly suitable for application scenarios with high real-time requirements, such as autonomous driving, industrial monitoring, etc.; by obtaining the attribute data and business request data of the terminal, the edge server can provide more personalized services to meet the specific needs of different users and businesses; by dynamically adjusting the model and real-time monitoring of the system status, it can effectively respond to emergencies and environmental changes, and improve the overall robustness and stability of the system. The scheme of the present invention can not only improve the performance of AI edge computing, but also realize more efficient and intelligent business processing in a variety of application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a flow chart of an AI edge computing method provided by an embodiment of the present invention; Figure 2 It is a schematic block diagram of an AI edge computing device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0016] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0017] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0018] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0019] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0020] Refer to the following Figure 1 to Figure 2 To describe an AI edge computing method, device, electronic device and storage medium provided according to some embodiments of the present invention.

[0021] like Figure 1 As shown, one embodiment of the present invention provides an AI edge computing method, including: The edge server obtains a plurality of first model sets corresponding to the business types of the business processed by the edge server, wherein the first model sets are generated by the cloud server according to the specific business types and business contents; The edge server selects corresponding (one or more) first business processing terminals from the business processing terminals for connection according to the task plans (such as navigation, inspection, translation, etc.) and data processing requirements (such as image recognition, voice recognition, etc.) of the multiple business processing terminals; The edge server obtains terminal attribute data and service request data of the first service processing terminal; It can be understood that in this step, first, the edge server needs to define the required terminal attribute data type, such as device model, operating system version, hardware configuration, network status, etc.; the edge server requests the terminal's attribute data through the communication interface (such as REST API or WebSocket) with the first business processing terminal, and the terminal can actively send this data to the edge server periodically or when a specific event occurs (such as startup, status change); ensure that the acquired data is transmitted in a standard format (such as JSON or XML) to facilitate parsing and storage by the edge server; the edge server specifies the content of the business request data, including request type, request parameters, timestamp and other information; the edge server can set a monitoring mechanism to receive business request data from the first business processing terminal in real time, which can be achieved by establishing a persistent connection (such as WebSocket) or using a polling mechanism; whenever business request data is received, the edge server records and stores it in a local database for subsequent processing and analysis; the edge server integrates the terminal attribute data and the business request data to form a comprehensive data set. This process can be achieved through data processing algorithms to ensure data consistency and integrity; the integrated data can be stored in the database of the edge server for subsequent query and analysis; the edge server can regularly request updated terminal attribute data and business request data from the first business processing terminal to ensure the timeliness of the data; the edge server can feedback data processing results or status information to the terminal so that the terminal can make corresponding adjustments based on the feedback. Through this step, the edge server can effectively obtain the terminal attribute data and business request data of the first business processing terminal, and provide necessary information support for subsequent business processing; by obtaining terminal attributes and business request data in real time, the edge server can ensure the accuracy and timeliness of the data; the acquisition and processing of real-time data enables the system to quickly respond to user requests and improve user experience; by analyzing terminal attribute data, the edge server can better optimize resource allocation and improve overall system performance.

[0022] The edge server generates a second model set according to the terminal attribute data, a preset first model adjustment algorithm and the first model set; The edge server determines the service processor according to the terminal attribute data, its own service processing capability and resource occupancy rate (i.e., selects the service processor from the edge server, the first service processing terminal or other processing terminals); When it is determined that the service processor is the first service processing terminal, the edge server determines the corresponding (one or more) second models from the second model set according to the service request data, and sends the second models to the first service processing terminal; The edge server obtains current environment data of the environment in which the first service processing terminal is located and current operation data of the first service processing terminal (including computing resource occupancy rate, network transmission rate, operation parameters of each component, etc.); It is understandable that the edge server obtains the current environmental data of the environment where the first business processing terminal is located through sensors or API interfaces (these data may include environmental parameters such as temperature, humidity, light intensity, noise level, etc., which can be collected by the first business processing terminal itself or other sensors); the edge server monitors the operating status of the first business processing terminal and obtains the current operating data, which includes: computing resource occupancy rate (by monitoring the usage of CPU, memory and storage, the edge server can evaluate the computing resource occupancy rate of the terminal), network transmission rate (through network monitoring tools, the edge server can obtain the current network bandwidth usage and data transmission rate), operating parameters of each component (the edge server can obtain the operating status and performance parameters of each component (such as sensors, actuators, etc.) through the terminal's monitoring system); the edge server integrates the collected environmental data and operating data to form a comprehensive data set. This data set can be used for subsequent analysis and decision support; the edge server updates the environmental data and operating data regularly or according to event trigger mechanisms (such as data changes, task requests, etc.) to ensure that the acquired information is the latest; the collected data can be transmitted to the edge server through a secure communication protocol (such as HTTPS) and stored in a local database for subsequent access and analysis. Through this step, the edge server can effectively obtain the environmental data and operating data of the first business processing terminal, thereby providing support for subsequent business processing and decision-making; it can obtain and monitor the environment and operating status of the terminal in real time to improve the system's responsiveness; by analyzing and calculating resource occupancy and network transmission rate, the edge server can optimize resource allocation and improve overall system performance; the integrated environmental data and operating data provide an important basis for business decision-making, helping the system to better adapt to changing environments and needs.

[0023] The edge server sends the current environment data, the current operation data and the first control instruction to the first business processing terminal to control the first business processing terminal (the first business processing terminal is configured with an edge computing module) to adjust the second model according to the current environment data, the current operation data and a preset second model adjustment algorithm to obtain a third model; The edge server controls the first service processing terminal to process the service data using the third model, and controls the first service processing terminal to perform corresponding operations according to the processing result.

[0024] In this embodiment, by dynamically selecting a suitable model according to the task plan and data processing requirements of the business processing terminal, the edge server can use computing resources more efficiently and reduce unnecessary calculations and delays; the edge server adjusts the model according to the terminal attribute data and the current environment data, so that the model can adapt to different business scenarios and environmental changes in real time, thereby improving the accuracy and reliability of processing; by real-time monitoring of the computing resource occupancy rate and network transmission rate of the terminal, the edge server can intelligently allocate resources to avoid resource waste and ensure the efficiency of business processing; the implementation of edge computing makes data processing closer to the data source, reduces the time delay of data transmission, and is particularly suitable for application scenarios with high real-time requirements, such as autonomous driving, industrial monitoring, etc.; by obtaining the attribute data and business request data of the terminal, the edge server can provide more personalized services to meet the specific needs of different users and businesses; by dynamically adjusting the model and real-time monitoring of the system status, it can effectively respond to emergencies and environmental changes, and improve the overall robustness and stability of the system. The scheme of the present invention can not only improve the performance of AI edge computing, but also realize more efficient and intelligent business processing in a variety of application scenarios.

[0025] In some possible implementations of the present invention, the edge server obtains a plurality of first model sets corresponding to the business types of the business processed by itself, and the first model sets are generated by the cloud server according to the specific business types and business contents, including: The edge server sends a service type request to the cloud server, wherein the service type request includes: unique identification information of the edge server; hardware configuration information of the edge server, including processor type, memory size and storage capacity; geographic location information of the area served by the edge server; a list of service types currently processed by the edge server, each service type including a specific service type identification and service content description; The cloud server performs the following operations according to the business type request: confirm the authorization status of the edge server according to the unique identification information of the edge server; filter out the model type that is adapted to the hardware conditions according to the hardware configuration information of the edge server; select the training data set of the corresponding area according to the geographical location information of the area served by the edge server; for each business type: select the corresponding basic model from the model library according to the business type identification; determine the specific parameter configuration of the model according to the business content description; use the selected training data set to train and optimize the basic model; generate an optimized model adapted to the business type; The cloud server packages the generated multiple optimization models into a first model set, and sends a model distribution request to the edge server, wherein the model distribution request includes: metadata information of each optimization model, including model version, applicable business type and resource requirements; model deployment configuration information; model update strategy; The edge server receives the model distribution request and performs the following operations: verifying the integrity and security of the model set; creating a model operating environment according to the model deployment configuration information; storing the first model set in a preset model library; and setting a regular inspection and update mechanism according to the model update strategy.

[0026] The solution of this embodiment ensures that the generated model set can adapt to the actual application scenario to the greatest extent by considering the hardware conditions, geographical location and specific business needs of the edge server; filters the model based on the actual configuration of the edge server to avoid waste or shortage of resources; ensures the reliability and security of the model distribution process through a complete request-response mechanism; and establishes a version control and update mechanism for the model to ensure that the model always remains in the optimal state.

[0027] In some possible implementations of the present invention, the step of the edge server selecting a corresponding first business processing terminal from the business processing terminals for connection according to the task plans and data processing requirements of the multiple business processing terminals includes: The edge server establishes a terminal management table, which includes: the device identification of each business processing terminal, the network connection status of each business processing terminal, the task plan information of each business processing terminal, and the data processing requirement information of each business processing terminal; wherein the task plan information includes: task type identification task execution time period, task priority and task plan status (not started, executing, completed); the data processing requirement information of each business processing terminal includes: processing type (such as image recognition, voice recognition, etc.), data scale, real-time requirements and accuracy requirements; The edge server performs terminal classification according to the terminal management table, specifically including: grouping terminals according to task type; grouping terminals according to data processing type; within each group, sorting terminals according to task priority; marking terminals with multi-task processing capabilities; The edge server selects the first service processing terminal according to the following conditions, including: judging the task time overlap and selecting a terminal with matching time resources; judging the matching degree between the terminal's data processing capability and the demand; evaluating the stability of the network connection state; and calculating the task priority comprehensive score; wherein the calculation formula of the task priority comprehensive score is: score = w1×time matching degree + w2×processing capability matching degree + w3×network stability + w4×task priority, where w1, w2, w3, and w4 are preset weight coefficients; The edge server establishes a connection with the selected first business processing terminal, including: sending a connection request to the first business processing terminal; waiting for the terminal to respond and verifying the connection status; establishing a secure communication channel; synchronizing the task plan and data processing configuration; updating the connection status in the terminal management table; The edge server establishes a task execution monitoring mechanism, including: regularly checking the connection status; monitoring the task execution progress; recording data processing performance indicators; and dynamically adjusting the terminal selection strategy according to the monitoring results.

[0028] The solution of this embodiment achieves optimal allocation of computing resources through terminal classification and priority scoring mechanism; improves the overall processing efficiency of the system by identifying terminals with multi-tasking processing capabilities; ensures dynamic balance of system load through real-time monitoring and policy adjustment; and ensures stability and reliability of business processing through a complete connection establishment and monitoring mechanism.

[0029] In some possible implementations of the present invention, the step of generating the second model set by the edge server according to the terminal attribute data, the preset first model adjustment algorithm and the first model set includes: The edge server parses and classifies the terminal attribute data to determine the terminal hardware attributes, terminal software attributes and terminal function attributes; wherein the terminal hardware attributes include: processor architecture type (such as ARM, x86, etc.), processor performance parameters (such as core number, main frequency, etc.), memory capacity and type, and whether there is a dedicated AI accelerator; terminal software attributes include: operating system type and version, AI framework support and device driver compatibility; terminal function attributes include: supported sensor types, actuator types and communication interface types; The edge server performs model analysis on each model in the first model set to obtain model structure information, model operation requirements and model application scenarios; wherein the model structure information includes: number and type of network layers, parameter scale and computational complexity; model operation requirements include: minimum computing resource requirements, memory usage requirements and response time requirements; model application scenarios include: applicable task types, input data requirements and output format requirements; The edge server executes a first model adjustment algorithm, specifically including: adjusting the implementation of the model calculation layer according to the terminal processor architecture; optimizing the parameter storage structure according to the terminal memory capacity; adjusting the operator implementation according to whether there is an AI accelerator; selecting the corresponding quantization accuracy from a predefined quantization accuracy level set according to the terminal performance, performing weight quantization, and performing activation value quantization; calculating parameter importance scores, removing low-importance parameters, and retraining the pruned model structure; It can be understood that in this step, according to the terminal performance indicators, the corresponding quantization precision (such as a specific numerical value representing the bit width) is selected from a predefined quantization precision level set, and the quantization precision level set includes but is not limited to 8-bit integer, 4-bit integer, and 2-bit integer quantization precision, wherein the quantization precision of lower bit width corresponds to lower computing resource requirements but may result in higher precision loss, and the aforementioned selection operation is performed based on the balance between the preset terminal processing capabilities and the model accuracy requirements.

[0030] The edge server performs adaptability verification on the optimized model, including: simulating the terminal hardware environment and configuring the target software environment; testing computing performance, memory usage and response delay, and testing various input scenarios; verifying output accuracy; and checking the exception handling mechanism; The edge server integrates the optimized models that have passed verification into a second model set, including: generating a model description file; packaging the model file; configuring model loading parameters; and setting version control information.

[0031] The solution of this embodiment ensures that the generated model can run efficiently on the target terminal by comprehensively considering the hardware, software and functional attributes of the terminal; significantly reduces the model resource usage while ensuring functionality through model structure optimization, quantization and pruning; ensures the correctness and stability of the optimized model through a complete verification mechanism; and can generate differentiated optimization models according to the characteristics of different terminals.

[0032] In some possible implementations of the present invention, the step of constructing a first model adjustment algorithm is further included, specifically including: Construct a model evaluation index system, including performance indicators (such as computing delay time, memory usage, energy consumption level, etc.), precision indicators (such as task accuracy, error range, recall rate, etc.), resource utilization indicators (such as CPU utilization, memory usage efficiency, bandwidth occupancy, etc.); Establish a model structure analysis mechanism, including: Analyze the model computation graph: identify compute-intensive nodes; mark memory-intensive nodes; calculate data dependencies between nodes; Evaluate operator characteristics: count the usage frequency of various operators; calculate operator resource consumption; analyze operator optimization space; Build an adaptive optimization strategy generator, including: According to the terminal processor architecture characteristics: generate instruction set optimization strategy; design parallel computing strategy; formulate storage access strategy; According to the terminal memory capacity: generate data cache strategy; design parameter compression strategy; formulate dynamic loading strategy; According to the terminal functional characteristics: generate task decomposition strategy; design resource scheduling strategy; formulate load balancing strategy; Implement a model converter, including: Design quantization conversion module: implement dynamic quantization algorithm; implement mixed precision quantization; implement automatic adjustment of quantization parameters; Design pruning conversion module: implement structural importance evaluation; implement iterative pruning algorithm; implement automatic weight compensation; Design structure reorganization module: realize layer fusion optimization; realize branch merging optimization; realize computational graph reordering; Build an optimization effect verification mechanism, including: Establish a benchmark test set: input data of different scales; test cases in different scenarios; boundary condition testing; Realize performance comparison analysis: calculate the index changes before and after optimization; evaluate the stability of optimization effect; generate optimization report; Establish an optimization strategy feedback mechanism: record optimization historical data; analyze optimization effect rules; update optimization strategy parameters; Implement online dynamic adjustment mechanism, including: Build a real-time monitoring module: monitor terminal resource status; track model operation indicators; detect performance anomalies; Implement dynamic optimization triggers: set optimization trigger thresholds; evaluate the necessity of optimization; control optimization frequency; Establish an optimization strategy selector: predict optimization effect; evaluate optimization cost; select the optimal strategy.

[0033] In this embodiment, the intelligent adjustment of the model is achieved through an adaptive optimization strategy generator; the performance of the optimized model is ensured through a complete evaluation and verification mechanism; efficient use of terminal resources is achieved through multi-dimensional optimization strategies; and the model can adapt to changes in the terminal operating environment through an online dynamic adjustment mechanism.

[0034] In some possible implementations of the present invention, the step of the edge server determining the service processor according to the terminal attribute data, its own service processing capability and resource occupancy rate includes: The edge server establishes a resource evaluation index system, which includes computing resource indicators, network resource indicators and storage resource indicators; wherein the computing resource indicators include: CPU utilization and available core number, GPU / NPU usage, memory utilization and available capacity; network resource indicators include: network bandwidth occupancy, network latency level, connection stability; storage resource indicators include: storage space utilization, I / O performance indicators, cache hit rate; The edge server constructs a processing capability evaluation model, including: Evaluate its own processing capabilities: calculate the current task queue length of the edge server; calculate the average processing delay; predict short-term load trends; Evaluate the first service processing terminal: analyze the computing capability of the first service processing terminal; evaluate the resource margin of the first service processing terminal; calculate the task adaptability; Evaluation of other processing terminals: obtaining real-time status of other terminals; evaluating collaborative processing capabilities; calculating load distribution ratios; The edge server performs multi-dimensional decision calculations, including: Calculate the local processing cost: Cost_local = q1 × computing resource usage + q2 × storage resource usage; Calculate the terminal processing cost: Cost_terminal = q3 × terminal resource usage + q4 × network transmission overhead; Calculate the cost of cooperative processing: Cost_cooperative = q5 × task decomposition overhead + q6 × coordination communication overhead; Computing processing benefits, including: computing processing time benefits; computing resource utilization benefits; computing system stability benefits; Generate decision scores, including: Score = benefit index - cost index; apply constraint adjustment; calculate the final decision score; It can be understood that q1 to q6 are preset weight values. Applying constraint adjustment is the process of considering the impact of various constraints on the score and making corresponding adjustments based on the original decision score. The types of constraints include hard constraints and soft constraints; hard constraints are conditions that must be met, such as: computing resource usage cannot exceed the maximum capacity of the system, network bandwidth cannot exceed physical limitations, response time must meet real-time requirements, memory usage must be within a safe range, etc.; soft constraints are conditions that are expected to be met but can be violated appropriately, such as: processor utilization is best kept within a certain range, energy consumption levels are best controlled below the target value, and load balancing is best maintained within an ideal range.

[0035] The edge server establishes a dynamic load balancing mechanism, including: Real-time monitoring of load status, specifically: monitoring resource changes of each processor; recording task processing performance; analyzing load distribution; Perform load adjustment, specifically: set load threshold; calculate load deviation; trigger load migration; Update the processing strategy, specifically: record the processing effect; optimize the decision parameters; adjust the weight coefficient; The edge server determines the final service processor according to the following rules, including: When Score_local is the highest and meets the threshold requirement, the edge server is selected as the processor; the edge server allocates and locks a specific number of processor resources, memory resources, and storage resources from the available computing resource pool according to the resource requirements of the business processing, forms an independent resource space, and initializes the processing environment in the resource space; When Score_terminal is the highest and meets the threshold requirement, the first service processing terminal is selected as the processor; task parameters are configured for the first service processing terminal, and a monitoring mechanism is established; When Score_cooperative is the highest and meets the threshold requirements, the cooperative processing mode is selected, the task boundaries are divided, and the cooperative mechanism is established.

[0036] The solution of this embodiment achieves the optimal configuration of system resources through a multi-dimensional evaluation and decision-making mechanism; avoids resource overload or idleness through a dynamic load balancing mechanism; improves overall business processing efficiency through intelligent processing party selection; and ensures system operation stability through reasonable resource allocation and load control.

[0037] In some possible implementations of the present invention, when it is determined that the service processor is the first service processing terminal, the step of the edge server determining the corresponding second model from the second model set according to the service request data, and sending the second model to the first service processing terminal includes: The edge server parses the service request data, including: Analyze business type information, specifically: identify the main business type; identify the subtask type; determine the task priority; Analyze data feature information, specifically: analyze data scale; identify data format; determine data quality requirements; Analyze performance requirement information, specifically: obtain real-time requirements; obtain accuracy requirements; obtain resource constraints; The edge server selects candidate models from the second model set, including: Perform feature matching, specifically: matching business type labels; matching data format requirements; matching performance indicator requirements; Evaluate the model's suitability, specifically: calculate feature matching; evaluate performance satisfaction; calculate resource adaptability; Generate a list of candidate models, specifically: sort by matching degree; filter models that do not meet the constraints; record model scores; The edge server performs model combination optimization, including: Analyze task dependencies, specifically: build a task dependency graph; identify critical paths; determine the execution order; Calculate resource constraints, specifically: count available resources; calculate resource requirements; evaluate resource matching; Generate the best combination plan, specifically: calculate the combination benefits; evaluate the feasibility of the combination; select the best plan; The edge server performs a model packaging operation, including: Prepare model files, specifically: organize model parameters; package model structure; generate configuration files; Create deployment instructions, specifically: record dependencies; set deployment sequence; configure run parameters; Perform integrity checks, specifically: verify file integrity; check configuration correctness; generate verification information; The edge server performs model distribution, including: Establish a secure transmission channel, specifically: negotiate encryption method; establish data channel; verify connection security; Perform batch transfers, specifically: set transfer priority; control transfer rate; monitor transfer status; Verify the distribution results, specifically: check the reception integrity; verify the model availability; confirm the deployment success.

[0038] The solution of this embodiment ensures the selection of the most suitable model through multi-dimensional feature matching; achieves efficient use of resources through model combination optimization; ensures the reliability of model deployment through a complete packaging and distribution mechanism; and protects the security of model data through a secure transmission mechanism.

[0039] In some possible implementations of the present invention, the step in which the edge server sends the current environment data, the current operation data, and the first control instruction to the first business processing terminal to control the first business processing terminal to adjust the second model according to the current environment data, the current operation data, and a preset second model adjustment algorithm to obtain a third model includes: The edge server collects and pre-processes environmental data, including: Collect physical environment data such as temperature, humidity, light, vibration, noise, spatial position and movement status; Collect business environment data such as the distribution of peripheral equipment, network environment conditions, and electromagnetic interference conditions; Performing data preprocessing on the physical environment data and the business environment data, specifically: data standardization processing; outlier detection and processing; data time series alignment; The edge server organizes the current operation data, which includes resource usage status, network transmission status, and hardware operation parameters; wherein the resource usage status includes: CPU, GPU usage, memory occupancy, and storage space usage; the network transmission status includes: real-time bandwidth utilization, network delay data, and data packet loss rate; the hardware operation parameters include: processor temperature, power consumption data, and the working status of each component; The edge server generates a first control instruction, which includes a model adjustment strategy, a resource allocation strategy, and an execution control strategy; wherein the model adjustment strategy includes: determining an adjustment target, setting an adjustment range, and defining an adjustment step; the resource allocation strategy includes: allocating computing resources, setting memory limits, and controlling energy consumption levels; the execution control strategy includes: setting execution priority, defining a timeout mechanism, and configuring exception handling; The edge server integrates and sends the data packets, including: Data packaging, specifically: merging environmental data; integrating operational data; including control instructions; Data compression, specifically: selecting a compression algorithm; performing data compression; generating a checksum; Send in batches, specifically: determine the sending order; control the sending rate; monitor the sending status; The edge computing module of the first service processing terminal performs model adjustment, including: Parse received data, specifically: decompress data packets; verify data integrity; classify and store data; Evaluate the necessity of adjustment, specifically: analyze the degree of environmental change; evaluate the degree of performance impact; calculate the benefits of adjustment; Performing model adjustment, specifically: applying the second model adjustment algorithm; fine-tuning parameters; optimizing the model structure; The edge computing module verifies the adjusted third model, including: Perform performance testing, specifically: test computing performance; test response time; test resource usage; Verify functional correctness, specifically: test model accuracy; verify output stability; check for abnormal responses; Generate a verification report, specifically: record the adjustment effect; compile statistics on performance indicators; and provide optimization suggestions.

[0040] This embodiment scheme enables the model to adapt to environmental changes through the collection and processing of real-time environmental data; improves the model operation efficiency through the analysis and adjustment of operation data; ensures the reliability of model adjustment through a complete data processing and verification mechanism; and realizes real-time adjustment of the model through an efficient data transmission and processing mechanism.

[0041] See also Figure 2Another embodiment of the present invention provides an AI edge computing device for executing an AI edge computing method, including: a data acquisition module and a control processing module; wherein, The data acquisition module is configured to: acquire a plurality of first model sets corresponding to the business type of the business processed by the module, wherein the first model sets are generated by the cloud server according to the specific business type and business content; The control processing module is configured to: select a corresponding first business processing terminal from the business processing terminals for connection according to the task plans and data processing requirements of the multiple business processing terminals; The data acquisition module is further configured to: acquire terminal attribute data and service request data of the first service processing terminal; The control processing module is also configured to: Generate a second model set according to the terminal attribute data, a preset first model adjustment algorithm and the first model set; Determine the service processor according to the terminal attribute data, its own service processing capability and resource occupancy rate; When it is determined that the service processor is the first service processing terminal, determining a corresponding second model from the second model set according to the service request data, and sending the second model to the first service processing terminal; The data acquisition module is further configured to: acquire current environment data of the environment in which the first service processing terminal is located and current operation data of the first service processing terminal; The control processing module is also configured to: Sending the current environment data, the current operation data and a first control instruction to the first business processing terminal to control the first business processing terminal to adjust the second model to obtain a third model according to the current environment data, the current operation data and a preset second model adjustment algorithm; The first business processing terminal is controlled to process the business data using the third model, and the first business processing terminal is controlled to perform corresponding operations according to the processing results.

[0042] It should be known that Figure 2 The block diagram of the AI ​​edge computing device shown is for illustration only, and the number of modules shown does not limit the scope of protection of the present invention. The AI ​​edge computing device provided in this embodiment can be used to execute the various embodiments of the corresponding AI edge computing method. For the specific implementation process, please refer to the description of the embodiments of each method, which will not be repeated here.

[0043] Another embodiment of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of an AI edge computing method when executing the computer program.

[0044] Another embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of an AI edge computing method.

[0045] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0046] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0047] In the several embodiments provided in the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of the above-mentioned units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0048] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0049] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0050] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a memory, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the above-mentioned methods of each embodiment of the present application. The aforementioned memory includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or CD-ROM and other media that can store program codes.

[0051] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable memory, and the memory can include: a flash drive, a read-only memory (English: Read-Only Memory, abbreviated as: ROM), a random access memory (English: Random Access Memory, abbreviated as: RAM), a magnetic disk or an optical disk, etc.

[0052] The embodiments of the present application are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

[0053] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions without departing from the spirit and scope of the present invention, and can make various changes and modifications, including the combination of the above-mentioned different functions and implementation steps, including software and hardware implementation methods, all of which are within the scope of protection of the present invention.

Claims

1. An AI edge computing method, characterized in that: include: The edge server obtains a plurality of first model sets corresponding to the business types of the business processed by the edge server, wherein the first model sets are generated by the cloud server according to the specific business types and business contents; The edge server selects a corresponding first business processing terminal from the business processing terminals for connection according to the task plans and data processing requirements of the multiple business processing terminals; The edge server obtains terminal attribute data and service request data of the first service processing terminal; The edge server generates a second model set according to the terminal attribute data, a preset first model adjustment algorithm and the first model set; The edge server determines the service processor according to the terminal attribute data, its own service processing capability and resource occupancy rate; When it is determined that the service processor is the first service processing terminal, the edge server determines a corresponding second model from the second model set according to the service request data, and sends the second model to the first service processing terminal; The edge server obtains current environment data of the environment in which the first service processing terminal is located and current operation data of the first service processing terminal; The edge server sends the current environment data, the current operation data and the first control instruction to the first business processing terminal, so as to control the first business processing terminal to adjust the second model to obtain a third model according to the current environment data, the current operation data and a preset second model adjustment algorithm; The edge server controls the first service processing terminal to process the service data using the third model, and controls the first service processing terminal to perform corresponding operations according to the processing results.

2. The AI ​​edge computing method according to claim 1, characterized in that: The edge server obtains a plurality of first model sets corresponding to the business types of the business processed by the edge server, wherein the first model sets are generated by the cloud server according to the specific business types and business contents, including: The edge server sends a service type request to the cloud server, wherein the service type request includes: unique identification information of the edge server; hardware configuration information of the edge server, including processor type, memory size and storage capacity; geographic location information of the area served by the edge server; a list of service types currently processed by the edge server, each service type including a specific service type identification and service content description; The cloud server performs the following operations according to the business type request: confirm the authorization status of the edge server according to the unique identification information of the edge server; filter out the model type that is adapted to the hardware conditions according to the hardware configuration information of the edge server; select the training data set of the corresponding area according to the geographical location information of the area served by the edge server; for each business type: select the corresponding basic model from the model library according to the business type identification; determine the specific parameter configuration of the model according to the business content description; use the selected training data set to train and optimize the basic model; generate an optimized model adapted to the business type; The cloud server packages the generated multiple optimization models into a first model set, and sends a model distribution request to the edge server, wherein the model distribution request includes: metadata information of each optimization model, including model version, applicable business type and resource requirements; model deployment configuration information; model update strategy; The edge server receives the model distribution request and performs the following operations: verifying the integrity and security of the model set; creating a model operating environment according to the model deployment configuration information; storing the first model set in a preset model library; and setting a regular inspection and update mechanism according to the model update strategy.

3. The AI ​​edge computing method according to claim 2, characterized in that: The step of the edge server selecting a corresponding first business processing terminal from the business processing terminals for connection according to the task plans and data processing requirements of the multiple business processing terminals includes: The edge server establishes a terminal management table, which includes: the device identification of each business processing terminal, the network connection status of each business processing terminal, the task plan information of each business processing terminal, and the data processing requirement information of each business processing terminal; wherein the task plan information includes: task type identification task execution time period, task priority and task plan status; the data processing requirement information of each business processing terminal includes: processing type, data scale, real-time requirements and accuracy requirements; The edge server performs terminal classification according to the terminal management table, specifically including: grouping terminals according to task type; grouping terminals according to data processing type; within each group, sorting terminals according to task priority; marking terminals with multi-task processing capabilities; The edge server selects the first service processing terminal according to the following conditions, including: judging the task time overlap and selecting a terminal with matching time resources; judging the matching degree between the terminal's data processing capability and the demand; evaluating the stability of the network connection state; and calculating the task priority comprehensive score; wherein the calculation formula of the task priority comprehensive score is: score = w1×time matching degree + w2×processing capability matching degree + w3×network stability + w4×task priority, where w1, w2, w3, and w4 are preset weight coefficients; The edge server establishes a connection with the selected first business processing terminal, including: sending a connection request to the first business processing terminal; waiting for the terminal to respond and verifying the connection status; establishing a secure communication channel; synchronizing the task plan and data processing configuration; updating the connection status in the terminal management table; The edge server establishes a task execution monitoring mechanism, including: regularly checking the connection status; monitoring the task execution progress; recording data processing performance indicators; and dynamically adjusting the terminal selection strategy according to the monitoring results.

4. The AI ​​edge computing method according to claim 3, characterized in that: The step of generating a second model set by the edge server according to the terminal attribute data, a preset first model adjustment algorithm and the first model set includes: The edge server parses and classifies the terminal attribute data to determine the terminal hardware attributes, terminal software attributes and terminal function attributes; wherein the terminal hardware attributes include: processor architecture type, processor performance parameters, memory capacity and type, and whether there is a dedicated AI accelerator; the terminal software attributes include: operating system type and version, AI framework support and device driver compatibility; the terminal function attributes include: supported sensor types, actuator types and communication interface types; The edge server performs model analysis on each model in the first model set to obtain model structure information, model operation requirements and model application scenarios; wherein the model structure information includes: number and type of network layers, parameter scale and computational complexity; model operation requirements include: minimum computing resource requirements, memory usage requirements and response time requirements; model application scenarios include: applicable task types, input data requirements and output format requirements; The edge server executes a first model adjustment algorithm, specifically including: adjusting the implementation of the model calculation layer according to the terminal processor architecture; optimizing the parameter storage structure according to the terminal memory capacity; adjusting the operator implementation according to whether there is an AI accelerator; selecting the corresponding quantization accuracy from a predefined quantization accuracy level set according to the terminal performance, performing weight quantization, and performing activation value quantization; calculating parameter importance scores, removing low-importance parameters, and retraining the pruned model structure; The edge server performs adaptability verification on the optimized model, including: simulating the terminal hardware environment and configuring the target software environment; testing computing performance, memory usage and response delay, and testing various input scenarios; verifying output accuracy; and checking the exception handling mechanism; The edge server integrates the optimized models that have passed verification into a second model set, including: generating a model description file; packaging the model file; configuring model loading parameters; and setting version control information.

5. The AI ​​edge computing method according to claim 4, characterized in that: The step of the edge server determining the service processor according to the terminal attribute data, its own service processing capability and resource occupancy rate includes: The edge server establishes a resource evaluation index system, which includes computing resource indicators, network resource indicators and storage resource indicators; wherein the computing resource indicators include: CPU utilization and available core number, GPU / NPU usage, memory utilization and available capacity; network resource indicators include: network bandwidth occupancy, network latency level, connection stability; storage resource indicators include: storage space utilization, I / O performance indicators, cache hit rate; The edge server constructs a processing capability evaluation model, including: Evaluate its own processing capabilities: calculate the current task queue length of the edge server; calculate the average processing delay; predict short-term load trends; Evaluate the first service processing terminal: analyze the computing capability of the first service processing terminal; evaluate the resource margin of the first service processing terminal; calculate the task adaptability; Evaluation of other processing terminals: obtaining real-time status of other terminals; evaluating collaborative processing capabilities; calculating load distribution ratios; The edge server performs multi-dimensional decision calculations, including: Calculate the local processing cost: Cost_local = q1 × computing resource usage + q2 × storage resource usage; Calculate the terminal processing cost: Cost_terminal = q3 × terminal resource usage + q4 × network transmission overhead; Calculate the cost of cooperative processing: Cost_cooperative = q5 × task decomposition overhead + q6 × coordination communication overhead; Computing processing benefits, including: computing processing time benefits; computing resource utilization benefits; computing system stability benefits; Generate decision scores, including: Score = benefit index - cost index; apply constraint adjustment; calculate the final decision score; The edge server establishes a dynamic load balancing mechanism, including: Real-time monitoring of load status, specifically: monitoring resource changes of each processor; recording task processing performance; analyzing load distribution; Perform load adjustment, specifically: set load threshold; calculate load deviation; trigger load migration; Update the processing strategy, specifically: record the processing effect; optimize the decision parameters; adjust the weight coefficient; The edge server determines the final service processor according to the following rules, including: When Score_local is the highest and meets the threshold requirement, the edge server is selected as the processor; the edge server allocates and locks a specific number of processor resources, memory resources, and storage resources from the available computing resource pool according to the resource requirements of the business processing, forms an independent resource space, and initializes the processing environment in the resource space; When Score_terminal is the highest and meets the threshold requirement, the first service processing terminal is selected as the processor; task parameters are configured for the first service processing terminal, and a monitoring mechanism is established; When Score_cooperative is the highest and meets the threshold requirements, the cooperative processing mode is selected, the task boundaries are divided, and the cooperative mechanism is established.

6. The AI ​​edge computing method according to claim 5, characterized in that: The step of, when it is determined that the service processor is the first service processing terminal, the edge server determining the corresponding second model from the second model set according to the service request data, and sending the second model to the first service processing terminal, comprises: The edge server parses the service request data, including: Analyze business type information, specifically: identify the main business type; identify the subtask type; determine the task priority; Analyze data feature information, specifically: analyze data scale; identify data format; determine data quality requirements; Analyze performance requirement information, specifically: obtain real-time requirements; obtain accuracy requirements; obtain resource constraints; The edge server selects candidate models from the second model set, including: Perform feature matching, specifically: matching business type labels; matching data format requirements; matching performance indicator requirements; Evaluate the model's suitability, specifically: calculate feature matching; evaluate performance satisfaction; calculate resource adaptability; Generate a list of candidate models, specifically: sort by matching degree; filter models that do not meet the constraints; record model scores; The edge server performs model combination optimization, including: Analyze task dependencies, specifically: build a task dependency graph; identify critical paths; determine the execution order; Calculate resource constraints, specifically: count available resources; calculate resource requirements; evaluate resource matching; Generate the best combination plan, specifically: calculate the combination benefits; evaluate the feasibility of the combination; select the best plan; The edge server performs a model packaging operation, including: Prepare model files, specifically: organize model parameters; package model structure; generate configuration files; Create deployment instructions, specifically: record dependencies; set deployment sequence; configure run parameters; Perform integrity checks, specifically: verify file integrity; check configuration correctness; generate verification information; The edge server performs model distribution, including: Establish a secure transmission channel, specifically: negotiate encryption method; establish data channel; verify connection security; Perform batch transfers, specifically: set transfer priority; control transfer rate; monitor transfer status; Verify the distribution results, specifically: check the reception integrity; verify the model availability; confirm the deployment success.

7. The AI ​​edge computing method according to claim 6, characterized in that: The step of sending the current environment data, the current operation data and the first control instruction by the edge server to the first business processing terminal to control the first business processing terminal to adjust the second model to obtain the third model according to the current environment data, the current operation data and a preset second model adjustment algorithm includes: The edge server collects and pre-processes environmental data, including: Collect physical environment data such as temperature, humidity, light, vibration, noise, spatial position and movement status; Collect business environment data such as the distribution of peripheral equipment, network environment conditions, and electromagnetic interference conditions; Performing data preprocessing on the physical environment data and the business environment data, specifically: data standardization processing; outlier detection and processing; data time series alignment; The edge server organizes the current operation data, which includes resource usage status, network transmission status, and hardware operation parameters; wherein the resource usage status includes: CPU, GPU usage, memory occupancy, and storage space usage; the network transmission status includes: real-time bandwidth utilization, network delay data, and data packet loss rate; the hardware operation parameters include: processor temperature, power consumption data, and the working status of each component; The edge server generates a first control instruction, which includes a model adjustment strategy, a resource allocation strategy, and an execution control strategy; wherein the model adjustment strategy includes: determining an adjustment target, setting an adjustment range, and defining an adjustment step; the resource allocation strategy includes: allocating computing resources, setting memory limits, and controlling energy consumption levels; the execution control strategy includes: setting execution priority, defining a timeout mechanism, and configuring exception handling; The edge server integrates and sends the data packets, including: Data packaging, specifically: merging environmental data; integrating operational data; including control instructions; Data compression, specifically: selecting a compression algorithm; performing data compression; generating a checksum; Send in batches, specifically: determine the sending order; control the sending rate; monitor the sending status; The edge computing module of the first service processing terminal performs model adjustment, including: Parse received data, specifically: decompress data packets; verify data integrity; classify and store data; Evaluate the necessity of adjustment, specifically: analyze the degree of environmental change; evaluate the degree of performance impact; calculate the benefits of adjustment; Performing model adjustment, specifically: applying the second model adjustment algorithm; fine-tuning parameters; optimizing the model structure; The edge computing module verifies the adjusted third model, including: Perform performance testing, specifically: test computing performance; test response time; test resource usage; Verify functional correctness, specifically: test model accuracy; verify output stability; check for abnormal responses; Generate a verification report, specifically: record the adjustment effect; compile statistics on performance indicators; and provide optimization suggestions.

8. An AI edge computing device, used to execute the AI ​​edge computing method according to any one of claims 1 to 7, characterized in that: include: Data acquisition module and control processing module; wherein, The data acquisition module is configured to: acquire a plurality of first model sets corresponding to the business type of the business processed by the module, wherein the first model sets are generated by the cloud server according to the specific business type and business content; The control processing module is configured to: select a corresponding first business processing terminal from the business processing terminals for connection according to the task plans and data processing requirements of the multiple business processing terminals; The data acquisition module is further configured to: acquire terminal attribute data and service request data of the first service processing terminal; The control processing module is also configured to: Generate a second model set according to the terminal attribute data, a preset first model adjustment algorithm and the first model set; Determine the service processor according to the terminal attribute data, its own service processing capability and resource occupancy rate; When it is determined that the service processor is the first service processing terminal, determining a corresponding second model from the second model set according to the service request data, and sending the second model to the first service processing terminal; The data acquisition module is further configured to: acquire current environment data of the environment in which the first service processing terminal is located and current operation data of the first service processing terminal; The control processing module is also configured to: Sending the current environment data, the current operation data and a first control instruction to the first business processing terminal to control the first business processing terminal to adjust the second model to obtain a third model according to the current environment data, the current operation data and a preset second model adjustment algorithm; The first business processing terminal is controlled to process the business data using the third model, and the first business processing terminal is controlled to perform corresponding operations according to the processing results.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the AI ​​edge computing method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the AI ​​edge computing method as described in any one of claims 1 to 7 are implemented.

Images