Internet of Things edge computing method and system oriented to multi-industry elastic deployment

Abstract

The invention discloses an Internet of Things edge computing method and system oriented to multi-industry elastic deployment. When the edge device is accessed, the device resource state matrix is constructed through the device type, the resource state vector and the industry feature tag, and the service constraint is matched. A hardware abstraction layer drives a micro-library set based on an equipment type index basis, and optimal drive is screened through mutual information maximization so as to adapt to hardware and drive. After a service event request is received, a basic operation instance is created, a resource application is initiated to a management program, and the management program synchronizes an industry feature tag to a micro-library arrangement engine; the engine screens an adaptive micro-library combination based on a dynamic micro-library pool in combination with an industry feature tag, and loads the micro-library combination to a basic operation instance to form a function execution instance. And the function execution instance completes the business task according to the service constraint and feeds back a result. The problems of framework rigidity, low hardware adaptation efficiency, resource waste and the like existing in traditional edge device access are effectively solved.

Description

An IoT edge computing method and system for flexible deployment across multiple industries Technical Field

[0001] This invention relates to the field of IoT edge computing technology, and in particular to an IoT edge computing method and system for flexible deployment across multiple industries. Background Technology

[0002] With the rapid development of IoT technology, edge computing, as a key link connecting terminal devices and cloud platforms, has been widely used in various industries such as industrial control, financial transactions, transportation, and smart education due to its advantages such as low latency, high reliability, and local data processing.

[0003] Flexible deployment across multiple industries is one of the core requirements of IoT edge computing. This necessitates that edge computing systems be able to flexibly adapt to hardware environments and efficiently respond to business requests based on the business characteristics, device types, and resource status of different industries. As the core carrier coordinating device management, resource scheduling, and business execution, the performance and adaptability of the edge computing operating system directly determine the application effectiveness of IoT edge computing solutions.

[0004] Although existing edge computing operating systems have certain capabilities in device collaboration and business support, there are still many shortcomings that need to be addressed in real-world application scenarios for flexible deployment across multiple industries: First, the problem of real-time performance loss is prominent: Under dynamic load scenarios, the response latency of existing systems fluctuates too much.

[0005] Secondly, cross-industry adaptation is difficult: the driver modules in existing systems are highly coupled with the hardware platform, and there are significant differences in business protocols and functional requirements across different industries. The lack of an effective adaptation mechanism in existing systems hinders the rapid deployment and flexible expansion of edge computing systems in multiple industry scenarios.

[0006] Therefore, existing technologies still need improvement and development. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to provide an IoT edge computing method and system for flexible deployment across multiple industries, addressing the aforementioned deficiencies of the prior art. This aims to solve the problem that the prior art is unable to guarantee real-time performance and cross-industry flexible adaptation.

[0008] The technical solution adopted by this invention to solve the problem is as follows: Firstly, embodiments of this invention provide an IoT edge computing method for flexible deployment across multiple industries. The method includes: when an edge device is connected, establishing a device resource state matrix through device type, resource state vector, and industry feature tags, and matching service constraints according to the industry feature tags; constructing a hardware abstraction layer, indexing a basic driver micro-library set through device type, and selecting the optimal driver based on maximizing mutual information; receiving a business event request; the business event request carrying corresponding industry feature tags and service constraints; creating a basic running instance based on the business event request and initiating a resource request to a management program; synchronizing the industry feature tags to the micro-library orchestration engine through the management program; pre-constructing a dynamic micro-library pool, and determining a micro-library combination based on the industry feature tags and the dynamic micro-library pool through the micro-library orchestration engine; loading the micro-library combination into the basic running instance to obtain a function execution instance; executing the corresponding business task based on the service constraints through the function execution instance and feeding back the execution result.

[0009] In one implementation, the steps of constructing a hardware abstraction layer, indexing a basic driver micro-library set by device type, and selecting the optimal driver based on maximizing mutual information include: in the hardware abstraction layer, pre-binding a corresponding basic driver micro-library set for each device type; when an edge device is connected, the hardware abstraction layer indexes the pre-bound basic driver micro-library set based on the device type and selects the optimal driver based on maximizing mutual information.

[0010] In one implementation, the step of creating a basic running instance based on the business event request includes: receiving the business event request through the Tenon module and creating a basic running instance based on a lifecycle control algorithm; wherein, when creating the basic running instance, a memory threshold must be met.

[0011] In one implementation, the Tenon module is further configured to: distribute real-time interrupts based on the priority weight of instructions through an interrupt routing model to ensure the priority processing of critical instructions, including financial transaction instructions.

[0012] In one implementation, the steps of constructing a dynamic micro-library pool include: using atomization encapsulation technology to obtain several micro-libraries with different functions, and defining the attributes of the micro-libraries in the form of 5-tuples to achieve plug-and-play functionality of the micro-libraries.

[0013] In one implementation, the micro-library types include: encryption libraries and stream processing libraries. In one implementation, the micro-library attributes include: functional description text, hardware dependency set, interface specification, and upper bound on time consumption.

[0014] In one implementation, the step of determining a micro-library combination based on the industry feature tags and the dynamic micro-library pool through the micro-library orchestration engine includes: converting the industry feature tags into a micro-library probability distribution through the micro-library orchestration engine; the micro-library probability distribution is used to reflect the applicability probability of each micro-library; based on the micro-library probability distribution, micro-libraries are selected from the dynamic micro-library pool, and a micro-library dependency graph without resource conflicts is constructed; and the micro-library combination is determined according to the micro-library dependency graph.

[0015] In one implementation, the step of filtering micro-libraries from the dynamic micro-library pool further includes: verifying the hardware dependency set of the micro-libraries and excluding combinations of micro-libraries with incompatible hardware architectures.

[0016] In one implementation, the step of determining a micro-library combination based on the micro-library dependency graph includes: determining the micro-library combination based on the micro-library dependency graph with latency and power consumption as optimization objectives.

[0017] In one implementation, after the step of selecting micro-libraries from the dynamic micro-library pool, the method further includes: verifying the hardware dependency set of the selected micro-libraries based on the optimal driver, and determining whether the hardware requirements of the optimal driver match the hardware dependency set of the selected micro-libraries, so as to ensure that the hardware computing power can be called normally after the micro-libraries are loaded.

[0018] In one implementation, the step of executing a corresponding business task based on the service constraints through the function execution instance includes: executing the corresponding business task based on the multi-objective scheduling strategy function and the industry feature label switching scheduling algorithm through the function execution instance, and under the condition of satisfying the service constraints.

[0019] In one implementation, the step of switching the scheduling algorithm based on the multi-objective scheduling strategy function and the industry feature label includes: switching the scheduling algorithm to task scheduling based on fixed priority according to the industry feature label of the financial scenario; switching the scheduling algorithm to task scheduling optimization based on the earliest deadline according to the industry feature label of the education scenario; and switching the scheduling algorithm to task scheduling based on a hybrid criticality strategy according to the industry feature label of the transportation scenario.

[0020] In one implementation, the step of executing the corresponding business task based on the service constraints through the function execution instance further includes: correcting execution deviations through PID feedback.

[0021] In one implementation, the function execution instance is further used to: calculate the context switching time through a real-time performance model, and perform collaborative optimization of register operations, cache processing, and interrupt nesting parallelism based on a context switching time optimization algorithm.

[0022] In one embodiment, the method further includes: when there is cross-industry and cross-device linkage, determining the collaboration mode of multiple devices across industries through a cross-industry collaboration protocol, and using pipeline timing optimization for cross-device communication.

[0023] In one embodiment, the method further includes: when the business task is a machine vision task, cropping the region of interest of the image using a data flow optimization operator.

[0024] Secondly, embodiments of the present invention also provide an IoT edge computing system for flexible deployment across multiple industries. The system includes: a device access module, used to establish a device resource state matrix based on device type, resource state vector, and industry feature tags when an edge device is accessed, and to match service constraints according to the industry feature tags; a hardware abstraction module, used to construct a hardware abstraction layer, index a basic driver micro-library set through device type, and select the optimal driver based on maximizing mutual information; a request receiving module, used to receive business event requests; the business event request carries corresponding industry feature tags and service constraints; an instance creation module, used to create a basic running instance based on the business event request and initiate a resource request to a management program; a management module, used to synchronize the industry feature tags to a micro-library orchestration engine through the management program; a micro-library module, used to pre-build a dynamic micro-library pool, and determine a micro-library combination based on the industry feature tags and the dynamic micro-library pool through the micro-library orchestration engine; a loading module, used to load the micro-library combination into the basic running instance to obtain a function execution instance; and a task execution and feedback module, used to execute corresponding business tasks based on the service constraints through the function execution instance and provide feedback on the execution results.

[0025] Thirdly, embodiments of the present invention also provide a terminal, the terminal including a memory and one or more processors; the memory stores one or more programs; the programs include instructions for executing the IoT edge computing method for flexible deployment across multiple industries as described above; the processor is used to execute the programs.

[0026] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a plurality of instructions adapted to be loaded and executed by a processor to implement the steps of the IoT edge computing method for flexible deployment across multiple industries as described above.

[0027] The beneficial effects of this invention are as follows: In the implementation of this invention, when edge devices access the network, a device resource state matrix is ​​constructed using device type, resource state vector, and industry feature tags, and service constraints are matched. A hardware abstraction layer indexes a basic driver micro-library set based on device type, and maximizes mutual information to filter the optimal driver for hardware and driver compatibility. Upon receiving a business event request, a basic running instance is created and a resource request is initiated to the management program. The management program synchronizes industry feature tags to the micro-library orchestration engine. This engine, based on a dynamic micro-library pool, filters suitable micro-library combinations using industry feature tags and loads them into the basic running instance to form a functional execution instance. The functional execution instance completes business tasks according to service constraints and provides feedback on the results. This effectively solves the problems of rigid architecture, inefficient hardware adaptation, and resource waste inherent in traditional edge device access. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0029] Figure 1 is a basic flowchart of the IoT edge computing method for flexible deployment across multiple industries provided in an embodiment of the present invention.

[0030] Figure 2 is a detailed framework diagram of the IoT edge computing method for flexible deployment across multiple industries provided in an embodiment of the present invention.

[0031] Figure 3 is a schematic diagram of the hierarchical collaboration mechanism provided in the embodiment of the present invention.

[0032] Figure 4 is a schematic diagram of the processing logic of the micro-library orchestration engine provided in an embodiment of the present invention.

[0033] Figure 5 is a schematic diagram of the linkage mechanism between the micro-library and hardware provided in the embodiment of the present invention.

[0034] Figure 6 is a logical schematic diagram of TenonOS dynamic orchestration provided in an embodiment of the present invention.

[0035] Figure 7 is a schematic diagram of the modules of the IoT edge computing system for flexible deployment across multiple industries provided in an embodiment of the present invention.

[0036] Figure 8 is a schematic block diagram of the terminal provided in an embodiment of the present invention. Detailed Implementation

[0037] This invention discloses an IoT edge computing method and system for flexible deployment across multiple industries. To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention.

[0038] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0039] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0040] To address the aforementioned deficiencies in existing technologies, this invention provides an IoT edge computing method for flexible deployment across multiple industries. As shown in Figure 1, the method specifically includes the following steps: Step S100: When an edge device is connected, a device resource state matrix is ​​established using device type, resource state vector, and industry feature tags, and service constraints are matched based on the industry feature tags.

[0041] Specifically, the processing logic when an edge device connects is shown in the following formula: In the formula, Device type (ARM / x86 / RISC-V); Resource state vector (CPU, memory, IO bandwidth); Industry characteristic tags (finance=0x01, education=0x02, etc.).

[0042] This formula defines the core attribute set of a single edge device. Based on this attribute definition, each edge device will collect data when it connects to the network. , , This information is collected and a device resource status matrix is ​​established. This matrix can be viewed as a real-time computing power fingerprint database for heterogeneous hardware platforms, clearly recording the hardware type, current resource status, and industry scenario of each connected device. Simultaneously, industry-specific tags for the devices are also generated during this process. It will be mapped and matched with the preset SLA policy library to bind the corresponding service constraints: for example, when a certain device When dealing with financial scenarios, the system will automatically match low-latency service constraints from the policy library, so that the device can directly meet the core requirements and standards of the industry after it is connected.

[0043] Step S200: Construct a hardware abstraction layer, index the basic driver micro-library set by device type, and select the optimal driver based on maximizing mutual information.

[0044] In one implementation, the steps of constructing a hardware abstraction layer, indexing a base driver micro-library set by device type, and selecting the optimal driver based on maximizing mutual information include: in the hardware abstraction layer, pre-binding a corresponding base driver micro-library set for each device type; when an edge device is connected, the hardware abstraction layer indexes the pre-bound base driver micro-library set based on the device type and selects the optimal driver based on maximizing mutual information.

[0045] Specifically, the processing logic for this step is shown in the following formula: Driver library selection: .

[0046] In the formula, This is a mutual information function to ensure that the driving characteristics match industry requirements.

[0047] First, through the mapping function This function associates device types with the underlying driver micro-library set. It determines the device type (also known as hardware type). Index to the base driver micro-library set pre-bound for this type. ,and This is a collection of multiple driver micro-libraries adapted to this hardware type. First, the range of drivers compatible with the hardware is narrowed down. Next, the driver micro-libraries are selected from the indexed set of basic driver micro-libraries, choosing those that maximize the mutual information function. The mutual information function measures the correlation between the driver micro-library and industry feature labels, ensuring that the selected driver features match the industry requirements of the device. For example, financial devices prioritize loading national cryptographic algorithm libraries because the mutual information value between the national cryptographic algorithm library and the industry feature labels of the financial scenario is higher.

[0048] For example, in practical application scenarios, when an edge device connects, the hardware abstraction layer first determines the device type through a mapping function. The mapping index is mapped to a pre-bound set of basic driver micro-libraries, and then the optimal driver that is both compatible with the hardware and meets the industry requirements is selected from this set by using the rule of maximizing mutual information.

[0049] For example, as shown in Figure 6, the core logic of TenonOS dynamic orchestration is as follows: Define a function named lib_orchestrator, with application_profile as the input parameter. The process consists of two steps: requirement feature extraction (based on LLM semantic parsing): The application configuration file app_profile is processed by llm_processor to generate the corresponding requirement feature vector feature_vector, thereby transforming the application requirements into matchable feature data.

[0050] Hardware compatibility verification: The following steps will be based on the feature vector obtained in the previous step to perform hardware_constraint-related compatibility verification: hardware_constraint=check_hardware(feature_vector).

[0051] The above two steps are the core starting logic of the TenonOS dynamic orchestration algorithm. Relying on large language model parsing and hardware adaptation verification, they provide a foundation for subsequent micro-library orchestration.

[0052] Step S300: Receive a business event request; the business event request carries the corresponding industry characteristic tag and service constraints.

[0053] Specifically, as shown in Figure 3, a business event request is initiated through the application layer, carrying corresponding industry characteristic tags and service constraints. The application layer then sends the business event request to the Tenon module (real-time scheduling and execution module).

[0054] Step S400: Create a basic running instance based on the business event request and initiate a resource request to the management program.

[0055] In one implementation, the step of creating a basic runtime instance based on the business event request includes: receiving the business event request through the Tenon module and creating a basic runtime instance based on a lifecycle control algorithm; wherein, when creating the basic runtime instance, a memory threshold must be met.

[0056] Specifically, the basic runtime instance is the Tenon instance, and the process of creating a Tenon instance is managed by the management program (KaiHong-Mortise management program) through a lifecycle control algorithm. When the Tenon module receives a business event request, it initiates a resource request to the management program, triggering the instance creation judgment process. This process can be implemented based on the lifecycle control algorithm formula: In the formula, This represents the total memory resources in the device resource state vector. =100KB is the preset memory threshold. The hypervisor uses this formula to determine: if the total memory of the current device reaches the 100KB threshold, it returns "true" to allow the creation of a Tenon instance; otherwise, it returns "false" and cannot create an instance. In other words, dynamically creating a Tenon instance requires meeting the memory threshold constraint, and the purpose of initiating a resource request to the hypervisor is to allow the hypervisor to verify whether the device resources meet the conditions for instance creation through this algorithm, ensuring that the instance can run with sufficient hardware resources.

[0057] In one implementation, the Tenon module is further configured to: distribute real-time interrupts based on the priority weight of instructions through an interrupt routing model to ensure the priority processing of critical instructions; the critical instructions include financial transaction instructions.

[0058] Specifically, the Tenon module relies on an interrupt routing model to ensure the priority processing of critical instructions. During Tenon's operation, it receives various real-time interrupt requests, such as interruptions for financial trading instructions. The module then distributes and processes these real-time interrupts using the interrupt routing model, which is based on instruction priority weights. To allocate interrupt queues: priority weights The calculation method is based on the response priority of the current instruction. Dividing by the sum of the response priorities of all instructions, ∑RT, interrupts with higher weights will be preferentially dispatched to the corresponding queue.

[0059] For example, when encountering critical instructions such as financial trading orders, their corresponding response priority is set to a higher value, and the calculated priority weight is... The impact will also be greater, so real-time interruptions corresponding to financial trading instructions will be prioritized for processing to ensure that such critical instructions can be responded to in a timely manner.

[0060] Step S500: Synchronize the industry feature tags to the micro-library orchestration engine through the management program.

[0061] Step S600: Pre-build a dynamic micro-library pool, and determine the micro-library combination based on the industry feature tags and the dynamic micro-library pool through the micro-library orchestration engine.

[0062] In one implementation, the steps for constructing a dynamic micro-library pool include: using atomic encapsulation technology to obtain several micro-libraries with different functions, and defining the micro-library attributes in the form of 5-tuples to achieve plug-and-play functionality for the micro-libraries.

[0063] In one implementation, the micro-library types include: encryption library and stream processing library. In another implementation, the micro-library attributes include: functional description text, hardware dependency set, interface specification, and upper bound of time consumption.

[0064] Specifically, the micro-library orchestration engine requires the use of a dynamic micro-library pool. The key to building a dynamic micro-library pool is employing atomic encapsulation technology to separate modules with different functions (such as encryption libraries). Stream processing library The libraries are encapsulated as independent micro-libraries, and then the micro-library attributes of each micro-library are defined using the following five-tuple of micro-library attributes: In the formula, Represents a single micro-repository; IDM is the unique identifier of the micro-repository. Functional description text; Hardware dependency set; Interface specifications; : Upper bound of time consumption.

[0065] The micro-library attributes cover requirements such as functional description text and hardware dependency sets. This is achieved through a unified interface specification. These micro-libraries are plug-and-play and, when aggregated, form a dynamic micro-library pool. For example, an encryption library. Stream processing library All of them belong to the micro-libraries in the dynamic micro-library pool.

[0066] Once the micro-library orchestration engine obtains industry feature tags, it will select suitable micro-libraries from the dynamic micro-library pool based on these tags. For example, when dealing with industry feature tags for financial scenarios, it will prioritize selecting functional description text. A cryptographic library that matches encryption requirements, while also verifying its hardware dependency set. Is it compatible with the current device type? Upper limit of time consumption? Whether it meets the low-latency service constraints, the final micro-library combination that fits the industry needs is determined, thereby providing functional support for the execution of subsequent business tasks.

[0067] In one implementation, the step of determining a micro-library combination based on the industry feature tags and the dynamic micro-library pool through the micro-library orchestration engine includes: converting the industry feature tags into a micro-library probability distribution through the micro-library orchestration engine; the micro-library probability distribution is used to reflect the applicability probability of each micro-library; based on the micro-library probability distribution, micro-libraries are selected from the dynamic micro-library pool, and a micro-library dependency graph without resource conflicts is constructed; and the micro-library combination is determined according to the micro-library dependency graph.

[0068] In one implementation, the step of filtering micro-libraries from the dynamic micro-library pool further includes: verifying the hardware dependency set of the micro-libraries and excluding combinations of micro-libraries with incompatible hardware architectures.

[0069] In one implementation, the step of determining the micro-library combination based on the micro-library dependency graph includes: determining the micro-library combination based on the micro-library dependency graph with latency and power consumption as optimization objectives.

[0070] Specifically, the Micro-Library orchestration engine first processes industry feature labels through an LLM semantic parsing model, and the processing logic is as follows: In the formula, These are industry-specific tags; the finely tuned BERT model... After processing, the output micro-library probability distribution is obtained. This reflects the applicability probability of each micro-library in the dynamic micro-library pool to the current industry scenario. This formula converts industry feature labels into a quantitative distribution of micro-library applicability.

[0071] Next, when selecting micro-libraries from the dynamic micro-library pool based on the micro-library probability distribution, the dependency graph construction formula will be used: In the formula, It is a collection of selected micro-libraries; , It is the hardware dependency set of the micro-library, edge The condition requires that the intersection of the hardware dependency sets of the two micro-libraries is not empty. In other words, this embodiment needs to verify the hardware dependency sets of the micro-libraries. If the intersection is empty, it means that the hardware architecture adapted to the micro-library is incompatible and will be excluded, in order to construct a micro-library dependency graph without resource conflicts. .

[0072] Finally, based on this micro-library dependency graph, the final micro-library combination is determined using dynamic combinatorial optimization techniques. The calculation formula used by the dynamic combinatorial optimization techniques is shown below: Delay can be used. and power consumption To optimize the final micro-library combination while meeting the total time consumption requirements. By adhering to service constraints, it can be ensured that the combination of micro-libraries is both adapted to industry needs and compatible with hardware resources and meets service standards.

[0073] In one implementation, after the step of selecting micro-libraries from the dynamic micro-library pool, the method further includes: verifying the hardware dependency set of the selected micro-libraries based on the optimal driver, and determining whether the hardware requirements of the optimal driver match the hardware dependency set of the selected micro-libraries, so as to ensure that the hardware computing power can be called normally after the micro-libraries are loaded.

[0074] Specifically, after micro-library screening, a secondary hardware compatibility verification step can be set up. This step requires combining the optimal driver of the hardware abstraction layer with the micro-library attributes. After the micro-library orchestration engine screens micro-libraries based on their probability distribution and initially constructs a micro-library dependency graph without resource conflicts, it will use the optimal driver for further hardware adaptation verification. The optimal driver is selected by the hardware abstraction layer based on the device type, and its hardware requirements are compatible with the architecture of the current edge device. The hardware dependency set in the five-tuple of micro-library attributes of the micro-libraries in the dynamic micro-library pool represents the hardware architecture requirements for micro-library adaptation. This embodiment will check whether the hardware dependency set of the screened micro-libraries matches the hardware requirements of the optimal driver. If they do not match, the micro-library is excluded, thereby avoiding the situation where the micro-library is compatible with the device architecture but does not match the current optimal driver. This ensures that after the micro-library is loaded into the Tenon module, it can normally call the hardware computing power through the optimal driver, ensuring that the function of the micro-library can be stably executed based on the hardware, and also making the hardware adaptability of the micro-library combination more accurate.

[0075] In one implementation, for the Shenzhen Kaihong business adaptation architecture, the underlying hardware adaptation layer mainly includes two modules: a hardware abstraction layer, whose core capability is to achieve compatibility with domestic chip platforms. Specifically, it adapts to various hardware types and architectures across different scenarios, including ARM financial encryption chips serving financial scenarios, RISC-V educational tablet controllers adapted for educational terminals, and x86 traffic signal controllers for the transportation sector. This layer eliminates the heterogeneity of different hardware, supporting the access of domestically produced ecosystem hardware across multiple fields.

[0076] The KaiHong-Mortise dynamic loading module comes with a driver micro-library of only 15KB in size. This micro-library includes a national cryptographic algorithm library and a SOC power consumption control library. It meets the requirements of security encryption, enables fine-grained control of hardware power consumption, and its small size is suitable for resource-constrained hardware deployment environments.

[0077] Furthermore, the hardware abstraction layer's adaptation design includes setting up a driver micro-library mapping table at this layer, clearly defining the correspondence between hardware and micro-libraries through key-value pairs. For example, the "KH-UKey" hardware corresponds to the "Financial Encryption Library v1.3," and the "RV-EDU" hardware corresponds to the "Educational Screen Sharing Library," thereby achieving precise adaptation between different hardware and dedicated function micro-libraries.

[0078] Step S700: Load the micro-library combination into the basic running instance to obtain a functional execution instance.

[0079] Specifically, the micro-library is loaded into the Tenon module through the micro-library orchestration engine, giving the basic running instance (Tenon instance) the ability to execute the current task, thus obtaining the functional execution instance (Tenon LibOS instance).

[0080] As shown in Figure 4, the creation process of functional execution instances relies on the micro-library orchestration engine. The micro-library orchestration engine uses LLM semantic analysis and hardware feature matching as its core capabilities. The specific processing logic is as follows: The input dimensions of the micro-library orchestration engine include three types of information: first, industry scenario labels (e.g., the "low latency" requirement corresponding to financial scenarios); second, device resource status (e.g., current memory and CPU usage); and third, network topology type (e.g., 5G or LoRa network). The first step is to parse the application's SLA (Service Level Agreement) and extract the latency constraints corresponding to the task. The second step is to select suitable components from the micro-library pool based on the scenario: for financial scenarios, components for national cryptographic encryption and transaction verification are selected, while for educational scenarios, streaming media compression algorithm components are selected. The third step is to generate corresponding functional execution instances based on the selected components, thereby completing the orchestration of micro-library components and the construction of the execution carrier.

[0081] Step S800: Execute the corresponding business task based on the service constraints through the function execution instance, and return the execution result.

[0082] In one implementation, the step of executing the corresponding business task based on the service constraints through the function execution instance includes: executing the corresponding business task based on the multi-objective scheduling strategy function and the industry feature label switching scheduling algorithm through the function execution instance, and under the condition of satisfying the service constraints.

[0083] In one implementation, the steps of switching the scheduling algorithm based on the multi-objective scheduling strategy function and the industry feature label include: switching the scheduling algorithm to task scheduling based on fixed priority according to the industry feature label of the financial scenario; switching the scheduling algorithm to task scheduling optimization based on the earliest deadline according to the industry feature label of the education scenario; and switching the scheduling algorithm to task scheduling based on a hybrid criticality strategy according to the industry feature label of the transportation scenario.

[0084] Specifically, the function execution instance will be based on a multi-objective scheduling strategy function (also known as an industry adaptive scheduler), combined with industry characteristic labels. Switch to the appropriate scheduling algorithm. The expression for the multi-objective scheduling policy function is shown below:

[0085] when For financial scenarios, switch to Fixed Priority (RMA) scheduling; when For educational scenarios, switch to Earliest Deadline (EDF) scheduling; when In traffic scenarios, a hybrid criticality scheduling approach is adopted. This allows for scheduling tailored to industry needs, ensuring that the scheduling logic aligns with the characteristics of the scenario. For example, in financial scenarios, fixed priorities guarantee the priority processing of transaction instructions. After completing the business task through the adapted scheduling algorithm, the function execution instance returns the task execution result to the application layer.

[0086] In one implementation, the step of executing the corresponding business task based on the service constraints through the function execution instance further includes: correcting execution deviations through PID feedback.

[0087] Specifically, during task execution, the function execution instance feeds back control equations (PID feedback equations) to ensure that service constraints are met. The feedback control equations are shown below: ; This represents the deviation between the actual response time and the SLA response time specified in the service constraints. It is the output of the PID controller, through the proportional term and integral terms Real-time correction of execution deviations. By continuously adjusting the execution process through this equation, it can be ensured that the financial transaction response time is less than the preset time, so that the actual response time meets the service constraints. Finally, the function execution instance executes business tasks through an adapted scheduling algorithm and PID feedback correction method, obtains accurate task execution results, and feeds them back to the application layer.

[0088] For example, when the actual response time of a task... When the CPU utilization exceeds 90%, the dynamic optimization equation is resolved. u(t) in the feedback control equation is used to adjust the scheduling strategy. u(t) is the output of the feedback control, which corrects execution deviations in real time, bringing the actual response time back within the SLA requirements and preventing task timeouts. Releasing computing power, where T represents task latency, E represents device power consumption, and α and β are the corresponding weighting coefficients for latency and power consumption. The dynamic optimization equation aims to minimize the weighted sum of latency and power consumption. Resolving it can optimize resource allocation logic and drive the orchestration engine to update library combinations, thereby releasing redundant computing power and alleviating CPU overload issues.

[0089] In one implementation, the resource scheduler monitors metrics. This can trigger micro-library reorganization. As shown in Figure 5, the mechanism of interaction between micro-library and hardware is illustrated: taking the closed-loop regulation to address CPU overheating as an example, this closed-loop regulation process starts with Mortise resource monitoring, which monitors the hardware resource status in real time; when a CPU overheating alarm is detected, the micro-library scheduler is triggered to intervene. The micro-library scheduler will perform the operation of "unloading the AI ​​inference library" to release computing resources and reduce the CPU load pressure; at the same time, in order to ensure the continuity of core functions, a simplified version of image recognition will be enabled (maintaining basic functions while reducing resource consumption); ultimately, the temperature is restored to the normal threshold, completing the linkage and adaptation between hardware resources and micro-library components.

[0090] In one implementation, the function execution instance is further used to: calculate the context switching time through a real-time performance model, and perform collaborative optimization of register operations, cache processing, and interrupt nesting parallelism based on a context switching time optimization algorithm.

[0091] Specifically, the function execution instance can optimize context switching through the real-time performance model. First, the function execution instance calculates the context switching time based on the real-time performance model, the calculation formula of which is shown below: In the formula, For context switching time; The time spent saving data in the register is the main issue. This refers to the time spent on cache processing; Represents the number of nested interrupt levels; It refers to hardware parallelism; This is the result of rounding up the interrupt nesting level according to the hardware parallelism (corresponding to the parallel processing overhead of interrupt nesting). These are the time consumption coefficients corresponding to the overhead, and these parameters together constitute the calculation dimension of context switching time.

[0092] Based on the calculated context switching time, the function execution instance will use a context switching time optimization algorithm to coordinate the optimization of register operations, cache processing, and interrupt nesting parallelism, so as to achieve the optimization effect of context switching time being less than the preset time. At the same time, this optimization can also support interrupt nesting processing, thereby ensuring that business tasks with high requirements for response time, such as industrial control scenarios, are responded to in a timely manner.

[0093] In one implementation, the method further includes: when there is cross-industry and cross-device collaboration, determining the collaboration mode of multiple devices across industries through a cross-industry collaboration protocol, and using pipeline timing optimization for cross-device communication.

[0094] Specifically, when cross-industry and cross-device collaboration is involved, the industry application processing layer determines the collaboration method for multiple devices through a cross-industry collaboration protocol. The calculation formula corresponding to the cross-industry collaboration protocol is as follows: ; This indicates the collaboration mechanism between equipment in industry A and equipment in industry B. These are the rules for collaborative communication between the two. It is the total time spent on cross-device collaboration, which is the sum of the time spent by each device on its own task processing. Maximum time consumed in cross-device communication links This formula provides a quantitative constraint on the collaboration time of cross-industry devices and determines the processing logic of the collaboration method. Combined with cross-industry collaboration protocols, cross-device communication can adopt pipeline timing optimization (such as cross-industry scenarios where traffic signal equipment triggers financial risk control equipment): by arranging the processing flow of each device and the communication flow between devices in a pipeline manner, device processing and cross-device communication can proceed in parallel as much as possible, thereby optimizing the total time consumption and achieving efficient linkage between cross-industry devices.

[0095] In one implementation, the method further includes: when the business task is a machine vision task, cropping the region of interest of the image using a data flow optimization operator.

[0096] Specifically, machine vision tasks can save computation by cropping out the region of interest (ROI) of an image. For example, when performing machine vision tasks in educational or security scenarios, the ROI can be cropped out first, and then the corresponding task can be performed.

[0097] Figure 2 shows a detailed flowchart of the present invention. This includes a micro-library reorganization feedback loop, representing the dynamic adjustment and cyclical optimization characteristics of the micro-library during operation; the engine section is the key component; and collaborative channels can exist between different industry modules, such as supporting the exchange of risk control data between the financial and transportation sectors, enabling cross-industry module information linkage.

[0098] The application scenarios of this invention include, but are not limited to: 1. Real-time risk control of financial transaction terminals: Usage scenario: During peak transaction periods (500+ transactions per second), bank ATM / POS terminals need to ensure a response latency of ≤8ms.

[0099] Implementation support: (1) Dynamic task decomposition: Transaction instructions are broken down into sub-tasks such as authentication, quota verification, and settlement. (2) Micro-library orchestration and execution: Loading the national cryptographic algorithm library on demand ( ) and atomic transaction library ( ).

[0100] Beneficiary industries: Banking (meets the "Security Specification for Embedded Financial Systems" GB / T 36629).

[0101] For example, the application optimization process for financial industry scenarios is as follows: When a financial terminal initiates a transaction request, it first loads a secure micro-library; then it triggers the KaiHong-Mortise hardware encryption channel (ensuring the secure transmission of transaction data); subsequently, the Tenon instance executes atomic transactions (ensuring the integrity of the transaction); ultimately achieving a response time of <8ms, meeting the national standard requirements in the financial field.

[0102] 2. Smart Education Classroom Multi-Terminal Collaboration: Usage Scenario: Real-time interaction between tablets / projectors / answer devices in a 50-person classroom, with bandwidth fluctuations of 20%-80% under weak network conditions.

[0103] Implementation support: (1) Hardware-insensitive driver: Automatically identifies device type ( (2) Flow control algorithm: dynamic frame reduction strategy .

[0104] Beneficiary industries: Educational informatization (saving 40% in cloud bandwidth costs).

[0105] For example, regarding the application optimization process for two industry scenarios in education, taking the smart classroom scenario as an example: when faced with the video stream of the smart classroom, the stream processing micro-library will be activated first; then, a dynamic frame rate reduction operation will be performed to adjust the video frame rate from 30fps to 15fps; this adjustment can save 40% of bandwidth and achieve smooth adaptation in weak network environments.

[0106] 3. Intelligent Transportation Edge Computing: Use Case: Collaborative optimization of intersection cameras and traffic lights to process traffic flow data that changes within 100ms.

[0107] Implementation support: (1) ROI region clipping: (2) Cross-device collaboration: Tenon instances on the road share memory via KaiHong-Mortise (reduced pixel processing by 63%;) ).

[0108] Beneficiary industries: Vehicle-road cooperation (reducing congestion index by 20%).

[0109] For example, data stream optimization in traffic signal control scenarios achieves efficient processing through the integration of micro-libraries and algorithms: First, the basic processing chain: the raw video stream is first fed into the edge video analysis micro-library, where vehicle location information is obtained through target detection; this information is then passed to the decision-making micro-library, ultimately generating an optimized timing scheme for traffic signals to support intelligent traffic signal control. Simultaneously, the process incorporates algorithm optimization strategies: by dynamically cropping the ROI region (focusing only on the core intersection area), only 20% of the intersection's pixel data is processed, significantly reducing computational load and ultimately achieving a 63% reduction in processing latency, thus improving the efficiency of data stream processing in traffic signal control scenarios.

[0110] 4. Other application scenarios: It can be extended to spaceborne edge computing. For example, Tenon supports fault-tolerant scheduling in space ray environments, or forms an on-orbit autonomous operating system solution (with a single-event upset resistance rate of 99.99%).

[0111] The technical problems solved by this invention include, but are not limited to: 1. Eliminating architectural rigidity during cross-industry deployment (e.g., ATMs cannot reuse educational tablet drivers): Traditional OS statically binds drivers (e.g., KaihongOS requires pre-installed 300+ driver modules), and hardware changes require recompiling the kernel (development cycle ≥ 3 weeks). This invention uses a dynamic micro-library pool: drivers are encapsulated in five-tuples ( The system functions are broken down into pluggable micro-libraries (the smallest unit is 13.6KB); through LLM semantic orchestration: the BERT model is matched with the hardware in real time ( This enables semantic matching of hardware characteristics with industry requirements. Ultimately, it achieves: a 50% increase in resource utilization (dynamic load experiment); and a 70% reduction in hardware adaptation cycle (from 3 weeks to 3 days), achieving elastic resource allocation and real-time hardware assurance in edge environments.

[0112] 2. Ensures hard real-time performance (99.5% task response time <8ms) under mixed workloads; addressing the issues of fixed scheduling strategies leading to large latency fluctuations in financial transactions (violating SLAs) and task backlogs caused by sudden load surges in industrial scenarios. This invention utilizes an industry-adaptive scheduler to ensure that the scheduling logic aligns with the characteristics of the scenario; and employs PID feedback control to quickly correct deviations.

[0113] 3. Ensuring Security and Real-Time Performance: In existing technologies, encryption operations increase latency (national cryptographic authentication latency exceeds 15ms), and the security isolation layer increases inter-process communication (IPC) overhead by 15%. This invention solves these problems through KaiHong-Mortise-Tenon collaboration: it adopts a hardware encryption channel pass-through method (bypassing the kernel), reducing encryption time by 76%; and it achieves fine-grained resource isolation, with memory resources in each isolation zone... ; Build a time-critical micro-library to reserve time slots for real-time tasks ( This ensures the time slot latency of real-time tasks. In this way, while ensuring security, the system's real-time performance is improved and resource consumption is reduced.

[0114] 4. Other issues: (1) Implemented the dynamic reorganization mechanism of micro-libraries: KaiHong-Mortise decomposes the traditional Hypervisor function into composable libraries (such as the lifecycle library with only 356 LoCs); the LLM orchestration engine generates DAG dependency graphs according to scenario requirements (automatically unloads the financial encryption library in the education scenario); solved the problems of micro-library timing consistency (nanosecond-level synchronization) and dependency conflict resolution; (2) Can build a complete end-edge-cloud collaborative system and deploy it on smart mining equipment and traffic signal control nodes; (3) Can provide the TenonBuilder automatic code generation toolchain to improve development efficiency.

[0115] (4) It can provide a LibOS-on-LibOS collaborative architecture (ultra-lightweight management program + real-time OS); an LLM-driven requirements-library mapping model. Hardware-insensitive device driver protocol.

[0116] The core value of the method of this invention is shown in Table 1: Table 1. Comparison results between the present invention and traditional methods

[0117] Based on the above embodiments, the present invention also provides an IoT edge computing system for flexible deployment across multiple industries, as shown in Figure 7. The system includes: a device access module 01, used to establish a device resource state matrix based on device type, resource state vector, and industry feature tags when an edge device accesses the system, and to match service constraints according to the industry feature tags; a hardware abstraction module 02, used to construct a hardware abstraction layer, index the basic driver micro-library set through device type, and select the optimal driver based on maximizing mutual information; and a request receiving module 03, used to receive business event requests; the business event request carries corresponding industry feature tags and service constraints; and instance creation. The module 04 is used to create a basic running instance based on the business event request and initiate a resource request to the management program; the management module 05 is used to synchronize the industry feature tags to the micro-library orchestration engine through the management program; the micro-library module 06 is used to pre-build a dynamic micro-library pool and determine the micro-library combination based on the industry feature tags and the dynamic micro-library pool through the micro-library orchestration engine; the loading module 07 is used to load the micro-library combination into the basic running instance to obtain a function execution instance; the task execution and feedback module 08 is used to execute the corresponding business task based on the service constraints through the function execution instance and provide feedback on the execution result.

[0118] In one implementation, the hardware abstraction module 02 includes: a binding unit, used to pre-bind a corresponding set of basic driver micro-libraries for each device type in the hardware abstraction layer; and a selection unit, used to, when an edge device is connected, index the pre-bound set of basic driver micro-libraries based on the device type through the hardware abstraction layer, and select the optimal driver based on maximizing mutual information.

[0119] In one implementation, the instance creation module 04 includes a Tenon module, used to receive the business event request and create a basic running instance based on a lifecycle control algorithm; wherein, when creating the basic running instance, a memory threshold must be met.

[0120] In one implementation, the Tenon module is further configured to: distribute real-time interrupts based on the priority weight of instructions through an interrupt routing model to ensure the priority processing of critical instructions; the critical instructions include financial transaction instructions.

[0121] In one implementation, the micro-library module 06 includes: a micro-library building unit, which is used to obtain several micro-libraries with different functions using atomic encapsulation technology, and defines the micro-library attributes in the form of a quintuple to achieve plug-and-play functionality of the micro-libraries.

[0122] In one implementation, the micro-library types include: encryption library and stream processing library. In another implementation, the micro-library attributes include: functional description text, hardware dependency set, interface specification, and upper bound of time consumption.

[0123] In one implementation, the micro-library module 06 further includes: a probability conversion unit, used to convert the industry feature tags into a micro-library probability distribution through the micro-library orchestration engine; the micro-library probability distribution is used to reflect the applicability probability of each micro-library; a micro-library filtering unit, used to filter micro-libraries from the dynamic micro-library pool based on the micro-library probability distribution, and construct a micro-library dependency graph without resource conflicts; and a combination determination unit, used to determine the micro-library combination according to the micro-library dependency graph.

[0124] In one implementation, the micro-library filtering unit further includes a hardware verification unit, used to verify the hardware dependency set of the micro-libraries and exclude combinations of micro-libraries with incompatible hardware architectures.

[0125] In one implementation, the combination determination unit is specifically used to: determine the combination of micro-libraries based on the micro-library dependency graph, with latency and power consumption as optimization objectives.

[0126] In one implementation, the micro-library module 06 further includes a driver verification unit, used to determine whether the hardware requirements of the optimal driver match the hardware dependency set of the selected micro-library based on the hardware dependency set of the micro-library selected by the optimal driver verification, so as to ensure that the micro-library can normally call hardware computing power after being loaded.

[0127] In one implementation, the task execution and feedback module 08 includes a scheduling unit, which is used to execute the corresponding business task through the function execution instance, based on the multi-objective scheduling strategy function and the industry feature label switching scheduling algorithm, and under the condition of satisfying the service constraints.

[0128] In one implementation, the scheduling unit includes: a financial scheduling unit, used to switch the scheduling algorithm to task scheduling based on fixed priority according to the industry characteristic labels of the financial scenario; an education scheduling unit, used to switch the scheduling algorithm to task scheduling optimization based on the earliest deadline according to the industry characteristic labels of the education scenario; and a transportation scheduling unit, used to switch the scheduling algorithm to task scheduling based on a hybrid criticality strategy according to the industry characteristic labels of the transportation scenario.

[0129] In one implementation, the task execution and feedback module 08 further includes a feedback correction unit for correcting execution deviations through PID feedback.

[0130] In one implementation, the function execution instance is further used to: calculate the context switching time through a real-time performance model, and perform collaborative optimization of register operations, cache processing, and interrupt nesting parallelism based on a context switching time optimization algorithm.

[0131] In one implementation, the system further includes a collaborative optimization module, which, when cross-industry and cross-device linkage occurs, determines the collaboration mode of multiple devices across industries through a cross-industry collaborative protocol and optimizes cross-device communication using pipeline timing.

[0132] In one implementation, the system further includes an image cropping module, used to crop the region of interest of an image using a data flow optimization operator when the business task is a machine vision task.

[0133] Based on the above embodiments, the present invention also provides a terminal, the principle block diagram of which is shown in Figure 8. The terminal includes a processor, a memory, a network interface, and a display screen connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used for communication with external terminals via a network connection. When executed by a processor, this computer program implements an IoT edge computing method for flexible deployment across multiple industries: When an edge device connects, a device resource state matrix is ​​established using device type, resource state vector, and industry feature tags, and service constraints are matched based on the industry feature tags; a hardware abstraction layer is constructed, a basic driver micro-library set is indexed by device type, and the optimal driver is selected based on maximizing mutual information; a business event request is received, carrying the corresponding industry feature tags and service constraints; a basic running instance is created based on the business event request, and a resource request is initiated to the management program; the industry feature tags are synchronized to the micro-library orchestration engine through the management program; a dynamic micro-library pool is pre-built, and the micro-library orchestration engine determines the micro-library combination based on the industry feature tags and the dynamic micro-library pool; the micro-library combination is loaded into the basic running instance to obtain a function execution instance; through the function execution instance, the corresponding business task is executed based on the service constraints, and the execution result is fed back.

[0134] The terminal's display screen can be an LCD screen or an e-ink screen.

[0135] Those skilled in the art will understand that the principle block diagram shown in FIG8 is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the terminal to which the present invention is applied. A specific terminal may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0136] In one implementation, the terminal's memory stores one or more programs, configured to be executed by one or more processors. These programs contain instructions for an IoT edge computing method for flexible deployment across multiple industries: when an edge device connects, a device resource state matrix is ​​established using device type, resource state vector, and industry feature tags, and service constraints are matched based on the industry feature tags; a hardware abstraction layer is constructed, a basic driver micro-library set is indexed by device type, and the optimal driver is selected based on maximizing mutual information; a business event request is received, carrying the corresponding industry feature tags and service constraints; a basic running instance is created based on the business event request, and a resource request is initiated to the management program; the industry feature tags are synchronized to the micro-library orchestration engine through the management program; a dynamic micro-library pool is pre-built, and the micro-library orchestration engine determines micro-library combinations based on the industry feature tags and the dynamic micro-library pool; the micro-library combinations are loaded into the basic running instance to obtain a function execution instance; through the function execution instance, the corresponding business task is executed based on the service constraints, and the execution result is fed back.

[0137] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0138] In summary, this invention discloses an IoT edge computing method and system for flexible deployment across multiple industries. The method includes: when an edge device connects, establishing a device resource state matrix using device type, resource state vector, and industry feature tags, and matching service constraints based on the industry feature tags; constructing a hardware abstraction layer, indexing a basic driver micro-library set using device type, and selecting the optimal driver based on maximizing mutual information; receiving a business event request; the business event request carrying the corresponding industry feature tags and service constraints; creating a basic running instance based on the business event request and initiating a resource request to the management program; synchronizing the industry feature tags to the micro-library orchestration engine through the management program; pre-constructing a dynamic micro-library pool, and determining micro-library combinations based on the industry feature tags and the dynamic micro-library pool through the micro-library orchestration engine; loading the micro-library combinations into the basic running instance to obtain a function execution instance; executing the corresponding business task based on the service constraints through the function execution instance and feeding back the execution result.

[0139] This invention constructs a device resource state matrix based on device type, resource state vector, and industry feature tags during edge device access, and matches service constraints. A hardware abstraction layer indexes a basic driver micro-library set based on device type, and maximizes mutual information to filter for optimal drivers, adapting to both hardware and drivers. Upon receiving a business event request, a basic running instance is created and a resource request is initiated to the management program. The management program synchronizes industry feature tags to the micro-library orchestration engine. This engine, based on a dynamic micro-library pool, filters suitable micro-library combinations based on industry feature tags and loads them into the basic running instance to form a functional execution instance. The functional execution instance completes business tasks according to service constraints and returns results. This effectively solves the problems of rigid architecture, inefficient hardware adaptation, and resource waste inherent in traditional edge device access.

[0140] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. An IoT edge computing method for flexible deployment across multiple industries, characterized in that, The method includes: when an edge device connects, establishing a device resource state matrix based on device type, resource state vector, and industry feature tags, and matching service constraints according to the industry feature tags; constructing a hardware abstraction layer, indexing a basic driver micro-library set through device type, and selecting the optimal driver based on maximizing mutual information; receiving a business event request; the business event request carrying the corresponding industry feature tags and service constraints; creating a basic running instance based on the business event request and initiating a resource request to the management program; synchronizing the industry feature tags to the micro-library orchestration engine through the management program; pre-constructing a dynamic micro-library pool, and determining a micro-library combination based on the industry feature tags and the dynamic micro-library pool through the micro-library orchestration engine; loading the micro-library combination into the basic running instance to obtain a function execution instance; executing the corresponding business task based on the service constraints through the function execution instance and feeding back the execution result.

2. The IoT edge computing method for flexible deployment across multiple industries according to claim 1, characterized in that, The steps of constructing a hardware abstraction layer, indexing the basic driver micro-library set by device type, and selecting the optimal driver based on maximizing mutual information include: in the hardware abstraction layer, pre-binding the corresponding basic driver micro-library set for each device type; when an edge device is connected, the hardware abstraction layer indexes the pre-bound basic driver micro-library set based on the device type and selects the optimal driver based on maximizing mutual information.

3. The IoT edge computing method for flexible deployment across multiple industries according to claim 1, characterized in that, The steps for creating a basic running instance based on the business event request include: receiving the business event request through the Tenon module and creating a basic running instance based on the lifecycle control algorithm; wherein, when creating the basic running instance, a memory threshold must be met.

4. The IoT edge computing method for flexible deployment across multiple industries according to claim 3, characterized in that, The Tenon module is also used to: distribute real-time interrupts based on the priority weight of instructions through an interrupt routing model to ensure the priority processing of critical instructions, including financial transaction instructions.

5. The IoT edge computing method for flexible deployment across multiple industries according to claim 1, characterized in that, The steps to build a dynamic micro-library pool include: using atomic encapsulation technology to obtain several micro-libraries with different functions, and defining the attributes of the micro-libraries in the form of 5-tuples to achieve plug-and-play functionality of the micro-libraries.

6. The IoT edge computing method for flexible deployment across multiple industries according to claim 5, characterized in that, Micro-library types include: encryption libraries and stream processing libraries.

7. The IoT edge computing method for flexible deployment across multiple industries according to claim 5, characterized in that, Micro-library attributes include: functional description text, hardware dependency set, interface specification, and upper bound of time consumption.

8. The IoT edge computing method for flexible deployment across multiple industries according to claim 1, characterized in that, The steps of determining a micro-library combination based on the industry feature tags and the dynamic micro-library pool using the micro-library orchestration engine include: converting the industry feature tags into a micro-library probability distribution using the micro-library orchestration engine; the micro-library probability distribution being used to reflect the applicability probability of each micro-library; selecting micro-libraries from the dynamic micro-library pool based on the micro-library probability distribution and constructing a micro-library dependency graph without resource conflicts; and determining the micro-library combination based on the micro-library dependency graph.

9. The IoT edge computing method for flexible deployment across multiple industries according to claim 8, characterized in that, The step of filtering micro-libraries from the dynamic micro-library pool also includes: verifying the hardware dependency set of the micro-libraries and excluding combinations of micro-libraries with incompatible hardware architectures.

10. The IoT edge computing method for flexible deployment across multiple industries according to claim 8, characterized in that, The step of determining the micro-library combination based on the micro-library dependency graph includes: determining the micro-library combination based on the micro-library dependency graph with latency and power consumption as optimization objectives.

11. The IoT edge computing method for flexible deployment across multiple industries according to claim 8, characterized in that, The step of selecting micro-libraries from the dynamic micro-library pool also includes: verifying the hardware dependency set of the selected micro-libraries based on the optimal driver, and determining whether the hardware requirements of the optimal driver match the hardware dependency set of the selected micro-libraries, so as to ensure that the hardware computing power can be called normally after the micro-libraries are loaded.

12. The IoT edge computing method for flexible deployment across multiple industries according to claim 1, characterized in that, The steps of executing the corresponding business task based on the service constraints through the function execution instance include: executing the corresponding business task based on the multi-objective scheduling strategy function and the industry feature label switching scheduling algorithm, and satisfying the service constraints.

13. The IoT edge computing method for flexible deployment across multiple industries according to claim 12, characterized in that, The steps of the scheduling algorithm based on the multi-objective scheduling strategy function and the industry feature label switching algorithm include: according to the industry feature label of the financial scenario, switching the scheduling algorithm to task scheduling based on fixed priority; according to the industry feature label of the education scenario, switching the scheduling algorithm to task scheduling optimization based on the earliest deadline; and according to the industry feature label of the transportation scenario, switching the scheduling algorithm to task scheduling based on a hybrid criticality strategy.

14. The IoT edge computing method for flexible deployment across multiple industries according to claim 1, characterized in that, The steps for executing the corresponding business task based on the service constraints through the function execution instance further include: correcting execution deviations through PID feedback.

15. The IoT edge computing method for flexible deployment across multiple industries according to claim 1, characterized in that, The function execution instance is also used to: calculate the context switching time through a real-time performance model, and perform collaborative optimization of register operations, cache processing, and interrupt nesting parallelism based on the context switching time optimization algorithm.

16. The IoT edge computing method for flexible deployment across multiple industries according to claim 1, characterized in that, The method further includes: when there is cross-industry and cross-device linkage, determining the collaboration mode of multiple devices across industries through a cross-industry collaboration protocol, and using pipeline timing optimization for cross-device communication.

17. The IoT edge computing method for flexible deployment across multiple industries according to claim 1, characterized in that, The method further includes: when the business task is a machine vision task, cropping the region of interest of the image through a data flow optimization operator.

18. An IoT edge computing system for flexible deployment across multiple industries, characterized in that, The system includes: a device access module, used to establish a device resource state matrix by means of device type, resource state vector and industry feature tags when an edge device is accessed, and to match service constraints according to the industry feature tags; a hardware abstraction module, used to build a hardware abstraction layer, index the basic driver micro-library set by device type, and select the optimal driver based on maximizing mutual information; a request receiving module, used to receive business event requests; the business event requests carry corresponding industry feature tags and service constraints; an instance creation module, used to create a basic running instance based on the business event request and initiate a resource request to the management program; a management module, used to synchronize the industry feature tags to the micro-library orchestration engine through the management program; a micro-library module, used to pre-build a dynamic micro-library pool, and to determine the micro-library combination based on the industry feature tags and the dynamic micro-library pool through the micro-library orchestration engine; a loading module, used to load the micro-library combination into the basic running instance to obtain a function execution instance; and a task execution and feedback module, used to execute the corresponding business task based on the service constraints through the function execution instance and to provide feedback on the execution result.

19. A terminal, characterized in that, The terminal includes a memory and one or more processors; the memory stores one or more programs; the programs contain instructions for executing the IoT edge computing method for flexible deployment across multiple industries as described in any one of claims 1 to 17; the processors are used to execute the programs.

20. A computer-readable storage medium storing a plurality of instructions, characterized in that, The instructions are applicable to being loaded and executed by a processor to implement the steps of the IoT edge computing method for flexible deployment across multiple industries as described in any one of claims 1 to 17.

Images