Cloud edge collaborative edge end device operator hot update method, system and device, and medium

Abstract

The invention relates to the technical field of edge computing and artificial intelligence model updating, and provides a cloud edge collaborative edge end equipment operator hot updating method, system and device and a medium. The method comprises the steps that a cloud detects a new version of an operator and generates an incremental update package between the new version and the old version; the edge node pulls the incremental update package and performs multiple security verification on the incremental update package; a double-instance inference engine is deployed in the edge node, operators passing verification are preloaded to a standby engine, and hot switching from an operation engine to the standby engine is achieved through a state synchronization mechanism; monitoring the running state of the new engine after switching, if the running state is abnormal, triggering a rollback mechanism, and switching back to the original engine; and optimizing edge node resources, cleaning old version operators, and reporting an update state and a resource use condition to the cloud. Core mechanisms such as differential updating, double-instance engine hot switching, multiple safety verification and dynamic resource scheduling are fused, and therefore efficient, safe and non-perceptual operator updating is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of edge computing and artificial intelligence model updating, and relates to a cloud-edge collaborative edge device operator hot updating method, system, device and medium. BACKGROUND

[0002] In the current digital and intelligent wave, edge computing technology, as a new computing mode, is gradually penetrating into industrial internet, intelligent transportation, intelligent security and other fields, and has become a key force to promote the transformation and upgrading of these industries.

[0003] Edge computing greatly reduces data transmission delay and improves system response speed and data processing efficiency by processing computing tasks and data storage close to data sources (i.e. edge devices), and is particularly suitable for application scenarios with high requirements for real-time, reliability and continuity.

[0004] With the wide application of edge computing technology, edge devices (such as industrial controllers, roadside AI computing units, intelligent cameras, etc.) are deployed in large numbers and widely integrated with deep learning operators to perform localized inference tasks.

[0005] These operators, as the core components of the model, directly determine the accuracy and efficiency of the inference task.

[0006] However, as business needs change, model performance continues to be optimized, and potential defects are repaired, periodic updating of operators on edge devices becomes a necessary and frequent operation. Existing edge operator updating schemes have many limitations, which seriously restrict the further development and application of edge computing technology. Specifically, these limitations mainly manifest in the following aspects: Large update traffic and high latency: Traditional full update method requires transmission of complete operator files, with file size usually reaching hundreds of MB to several GB. In the case of limited bandwidth between edge nodes and the cloud, this will result in long update time, and even update failure in weak network environment.

[0007] High risk of service interruption: Most existing updating schemes adopt the "stop service updating" mode, i.e. completely interrupting inference service during updating, with interruption time often exceeding 30 seconds. This is unacceptable for scenarios such as industrial quality inspection and real-time traffic monitoring that require high continuity of service. Service interruption can lead to serious consequences such as production line stoppage, traffic congestion, etc., affecting overall business operation efficiency and safety.

[0008] Insufficient security and compatibility guarantees: Existing solutions rely heavily on simple hash verification to ensure the integrity and legitimacy of update packages during the update process, lacking multi-layered security protection mechanisms. This makes update packages vulnerable to malicious tampering attacks, threatening system security. Furthermore, the lack of sufficient consideration of the compatibility between operators and edge device hardware resources, as well as existing inference services, before the update may lead to anomalies in the inference service after the update, such as performance degradation or crashes.

[0009] Inadequate resource scheduling mechanisms: During the update process, existing solutions often fail to dynamically adapt to the real-time resource status of edge nodes (such as memory usage and computing load). This can lead to update tasks and inference services competing for resources, causing performance degradation for both and even system crashes. This conflict is particularly pronounced under resource constraints, severely impacting system stability and reliability.

[0010] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0011] To provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. This summary is not intended as a general commentary, nor is it intended to identify key / important components or describe the scope of protection of these embodiments, but rather as a prelude to the detailed description that follows.

[0012] This disclosure provides a cloud-edge collaborative edge device operator hot update method, system, device, and medium, which integrates core mechanisms such as differential update, dual instance engine hot switching, multiple security checks, and dynamic resource scheduling to achieve efficient, secure, and seamless operator updates.

[0013] In some embodiments, the method includes: The cloud detects a new version of the operator, generates an incremental update package between the old and new versions, and encrypts and digitally signs the incremental update package. Edge nodes pull the incremental update package at a predetermined time and perform multiple security checks on it, including signature verification, interface compatibility verification, and functional verification. Deploy dual-instance inference engines in edge nodes, including a running engine and a backup engine. Preload verified operators into the backup engine and achieve hot switching from the running engine to the backup engine through a state synchronization mechanism. After switching, monitor the running status of the new engine. If an abnormality is detected, trigger the rollback mechanism and switch back to the original engine. Optimize edge node resources, clean up old version operators, and report update status and resource usage to the cloud.

[0014] Preferably, the incremental update package is generated in the following specific way: The difference generation module compares the model weights and structural differences between the old and new operators to generate an incremental update package containing only the changed parts; the size of the incremental update package is smaller than the full model package.

[0015] Preferably, the multiple security checks include: Use a pre-set public key to digitally sign and verify the update package; Verify that the new operator is compatible with the input / output interfaces of the current service; Trial reasoning was performed using a standard test set in a safety sandbox to verify whether its performance indicators met the standards.

[0016] Preferably, the hot-switching process includes: Preload the new operators into the backup engine and run the engine to continue processing online requests; State synchronization between the two engines can be achieved through shared memory or lock-free queues; New inference requests are routed to the backup engine, and the switchover is completed after the running engine has finished processing the existing requests.

[0017] Preferably, the rollback mechanism includes: Real-time monitoring of the new engine's inference latency and accuracy; If the metric exceeds the preset threshold, it will automatically switch back to the original engine within milliseconds and report the error information to the cloud.

[0018] In some embodiments, the cloud-edge collaborative edge device operator update system includes a cloud layer, an edge node layer, and a terminal device layer; The cloud layer includes: Operator repository, used to store full versions, historical versions and associated metadata of operators; The difference generation module is used to generate incremental update packages based on the differences between the old and new versions of operators. The global scheduling module is used to determine the update strategy based on the resource status and network conditions reported by the edge node layer. The security audit module is used to digitally sign update packages and record tamper-proof update logs; Message queues are used to broadcast update notifications to specified edge node layers; The edge node layer includes: Edge proxies are used to listen for cloud commands, pull update packages, and manage engine switching processes. A dual-instance inference engine, comprising a runtime engine for online services and a backup engine for preloading; The operator verification module is used to perform multiple security checks on the incremental update package; A dynamic scheduler is used to manage node resources and coordinate the updating of tasks and inference services when resources are scarce. A security sandbox is used to provide an isolated execution environment for functional verification. The exception rollback module is used to automatically trigger a rollback when a new engine exception is detected; The terminal device layer includes: A lightweight inference framework for performing local operator inference on the terminal; Trusted execution environment, used to provide hardware-level isolation protection for the update process; The status reporting module is used to send device status to the edge node layer and receive update instructions.

[0019] Preferably, the difference generation module is specifically used for: Using deep learning model optimization tools, the differences between the old and new versions of operators in terms of model parameters and computation graph structure are calculated, and an incremental update package containing only the changed parts is generated.

[0020] Preferably, the dual-instance inference engine achieves state synchronization between the running engine and the backup engine through shared memory, so as to support uninterrupted hot switching of inference services.

[0021] In some embodiments, the cloud-edge collaborative edge device operator hot update apparatus includes a processor and a memory storing program instructions, wherein the processor is configured to execute the cloud-edge collaborative edge device operator hot update method when running the program instructions.

[0022] In some embodiments, the storage medium stores program instructions that, when executed, perform the cloud-edge collaborative edge device operator hot update method.

[0023] This disclosure provides a cloud-edge collaborative edge device operator hot update method, system, apparatus, and medium, which can achieve the following technical effects: This invention achieves low-traffic, seamless, highly secure, and intelligently schedulable operator updates through core mechanisms such as differential updates, hot switching between dual instance engines, multiple security checks, and dynamic resource scheduling. It effectively solves the four core problems in existing solutions and provides strong technical support for the widespread application of edge computing technology.

[0024] The above general description and the description below are exemplary and illustrative only and are not intended to limit this application. Attached Figure Description

[0025] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations and drawings do not constitute a limitation on the embodiments. Elements having the same reference numerals in the drawings are shown as similar elements. The drawings are not to be scaled. And wherein: Figure 1 Schematic diagram of the method flow of this invention; Figure 2 System architecture diagram of this invention (three levels: cloud, edge node, and terminal device); Figure 3 Flowchart of hot-switching between dual engines at edge nodes; Figure 4 Flowchart of operator update triple check; Figure 5 This is a schematic diagram of the device structure provided in the embodiments of this disclosure. Detailed Implementation

[0026] To provide a more detailed understanding of the features and technical content of the embodiments of this disclosure, the implementation of the embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. The accompanying drawings are for illustrative purposes only and are not intended to limit the embodiments of this disclosure. In the following technical description, for ease of explanation, several details are used to provide a full understanding of the disclosed embodiments. However, one or more embodiments may still be implemented without these details. In other cases, well-known structures and devices may be simplified in their depiction to simplify the drawings.

[0027] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this disclosure described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0028] Example 1 like Figure 1 As shown, a method for hot updating edge device operators in a cloud-edge collaborative manner includes: S1: The cloud detects a new version of the operator, generates an incremental update package between the old and new versions, and encrypts and digitally signs the incremental update package; S2: The edge node pulls the incremental update package at a predetermined time and performs multiple security checks on it, including signature verification, interface compatibility verification and functional verification. S3: Deploy dual-instance inference engines in edge nodes, including a running engine and a backup engine. Preload verified operators into the backup engine and achieve hot switching from the running engine to the backup engine through a state synchronization mechanism. S4: After switching, monitor the running status of the new engine. If an abnormality is detected, trigger the rollback mechanism and switch back to the original engine. S5: Optimize edge node resources, clean up old version operators, and report update status and resource usage to the cloud.

[0029] As a refinement of the above embodiments, step S1, generating the incremental update package, includes: S101: The cloud detects the new version of the operator and compares its parameters and structure with those of the historical version; S102: The differential generation module uses a weighted differential and structural incremental strategy to generate an incremental update package that is much smaller than the full package, and encrypts and digitally signs the package. S103: The global scheduling module decides on the differential update strategy and appropriate update timing based on the status of the target edge node, and sends the update notification through the message queue.

[0030] As a refinement of the above embodiments, in step S2, the edge node pulls the incremental update package at a predetermined time and performs multiple security checks on it, including signature verification, interface compatibility verification, and functional verification, specifically including: S201: The edge agent pulls encrypted incremental update packages via a security protocol at a predetermined time; S202: The operator verification module performs triple verification sequentially, such as... Figure 4 As shown: Signature verification: Verifies that the update package origin is trustworthy and has not been tampered with; Interface verification: Confirm that the input and output of the new operator are compatible with the current service; Functional verification: Trial reasoning is performed using a standard test set in the security sandbox to verify that its performance indicators meet the preset thresholds.

[0031] As a refinement of the above embodiments, such as Figure 3 As shown, step S3, dual-engine preloading and hot-switching, specifically includes: S301: The edge agent preloads new operators in the backup engine, while the running engine continues to process online requests; S302: Achieve state synchronization between the two engines through technologies such as lock-free queues; S303: Trigger hot switch: route newly incoming inference requests to the ready standby engine and wait for the running engine to complete the existing requests; S304: Continuously monitor the inference latency and accuracy of the new engine, and confirm the successful switch once it reaches a stable standard.

[0032] As a refinement of the above embodiment, step S4 monitors the operating status of the new engine after switching. If an abnormality is detected, a rollback mechanism is triggered to switch back to the original engine. Specifically, this includes: S401: After confirming that the new engine is running stably for a period of time, gradually release the resources occupied by the old engine; S402: If an anomaly is detected in the new engine, the anomaly rollback module immediately switches back to the original engine and reports the error to the cloud, triggering the cloud retry mechanism.

[0033] As a refinement of the above embodiments, step S5 optimizes edge node resources, cleans up old version operators, and reports update status and resource usage to the cloud, specifically including: S501: The dynamic scheduler optimizes the local operator cache, retaining hot operators and cleaning up outdated versions; S502: The edge agent aggregates and reports the key results and resource indicators of this update to the cloud, and the cloud audit module records and stores the logs.

[0034] Example 2 like Figure 2 As shown, a cloud-edge collaborative edge device operator update system includes a cloud layer, an edge node layer, and a terminal device layer. The cloud layer includes: Operator repository: Used to store full versions, historical versions and associated metadata of operators, including version number, input and output dimensions, and compatible hardware type; Difference generation module: Based on deep learning model optimization tools, it calculates the differences in parameters and structure between the old and new versions of operators and generates an incremental update package that only contains the changed parts; Global scheduling module: Based on the resource status and network conditions reported by edge nodes, intelligently decides on update strategies, update timing, and task distribution; Security audit module: responsible for digitally signing update packages, recording update logs, and using blockchain technology to ensure the logs are tamper-proof and securely traceable; Message queue: Used to broadcast update notifications to specified edge nodes. The notification content includes version number, update type, and checksum hash value.

[0035] (2) Edge node layer, including: Edge Proxy: As a persistent process, it is responsible for listening to cloud update commands, pulling update packages, managing the dual-engine switching process, and reporting the status of this node; Dual-instance inference engine: It includes the running engine of the currently online service and a backup engine for preloading, supports mainstream inference frameworks, and achieves state synchronization through shared memory; Operator verification module: Performs triple security checks, including digital signature-based validity checks, interface definition-based compatibility checks, and functional checks performed in a security sandbox; Dynamic scheduler: Employs cache eviction and load balancing algorithms to manage operator cache and coordinate or distribute update tasks when resources are scarce; Security Sandbox: An isolated execution environment built based on security policies, used to restrict the permissions of update packages to prevent them from causing system-level conflicts or executing malicious code; Anomaly Rollback Module: Monitors the inference performance of the new engine in real time. If an anomaly is detected, it will automatically trigger a rollback operation and report the fault information to the upper layer.

[0036] (3) Terminal equipment layer, including: Lightweight inference framework: used to perform local operator inference on the terminal and process sensitive data; Trusted Execution Environment: Based on hardware security technology, it provides hardware-level isolation protection for the decryption and loading process of update packages; Status reporting module: Periodically sends device status information to edge nodes and receives update instructions from edge nodes.

[0037] Example 3 The following uses an "industrial workshop quality inspection scenario" as an example to specifically illustrate the implementation scheme of this invention: (1) Scenario and requirement definition Application scenario: A parts manufacturing workshop.

[0038] Hardware deployment: 5 edge nodes serve as the computing center for the workshop area. 20 industrial cameras are connected to the edge nodes via a local area network to acquire images of components in real time.

[0039] Software and Services: Deploy deep learning operators based on the ResNet50 model optimized on each edge node for real-time detection of "thread wear" defects in bolts. Single inference latency must be ≤50ms, overall system availability must be ≥99.9%, and update processes must not cause service interruptions.

[0040] (2) Update triggering and cloud decision Update motivation: The cloud monitoring system detected that the current operator version V1.0 had a false detection rate of up to 8% for "thread wear" defects, which could not meet the requirements for process improvement. Therefore, an optimized new operator version V2.0 was released.

[0041] Cloud-based preprocessing: Differential generation. The cloud-based differential generation module compares the model weights and computational graph structure of V1.0 and V2.0, generating an update package containing only parameter changes and structural increments. This incremental update package is 12MB in size (compared to 150MB for the full model package), reducing data volume by 92%. Security hardening: The incremental update package is encrypted using the AES256GCM algorithm, and a digital signature based on HMACSHA256 is generated using the cloud-based private key and appended to the header of the update package. Intelligent scheduling: The global scheduling module polls and finds that the edge nodes in this workshop are in a low-peak business period (average CPU load <20%) between 2:00 AM and 2:30 AM, and that network bandwidth is sufficient. Therefore, it decides to execute the differential update strategy on this batch of nodes at 2:00 AM and accurately send the update notification to the target edge agent through a message queue.

[0042] (3) Edge node update execution without awareness Upon receiving the update notification, the edge agent executes the following precise steps: Step 1: Secure Retrieval and Verification At the scheduled time of 2:00:00, the edge agent pulls an encrypted incremental update package from the cloud via a TLS 1.3 secure channel.

[0043] The triple verification sequence is initiated: Signature verification uses a pre-configured cloud public key to verify the digital signature, confirming the update package's source is trustworthy and has not been tampered with; Interface verification parses the metadata of the V2.0 operator, confirming its input is an RGB image tensor of [3,224,224] and its output is a two-dimensional vector of [normal probability, wear probability], fully compatible with existing service interfaces; Functional verification loads the V2.0 operator in a secure sandbox built on AppArmor and performs DryRun inference using 100 standardized "wear" test images. Test results show that the false positive rate decreased from 8% to 2%, with no memory leaks or runtime errors.

[0044] Step 2: Dual-engine hot-switching Preloading: After successful verification, the edge agent starts the backup inference engine EngineB, loads the verified V2.0 operators into GPU memory, and initializes the TensorRT context. This process fully utilizes the serialization capabilities of the inference framework. During this period, the currently running engine EngineA continues to process the industrial camera images flowing in online, without the service being aware of it.

[0045] State synchronization: Through lock-free queues and shared memory, the current inference context (such as session state and cache) of EngineA is quickly synchronized to EngineB.

[0046] Traffic migration: At 2:00:05, the edge proxy routes all new inference requests to the ready EngineB. For the remaining requests (approximately 10 images) being processed by EngineA, a completion window of up to 40 milliseconds is granted. Thereafter, all traffic is taken over by EngineB.

[0047] Step 3: Stability Monitoring and Resource Recovery Observation period: After the hot switch is completed, the system enters a 30-minute enhanced observation period. The anomaly rollback module continuously monitors EngineB's inference latency (required to be <50ms) and batch false positive rate.

[0048] Successfully confirmed: During the observation period, EngineB processed more than 10,000 inference requests, and all performance metrics remained stable within the thresholds.

[0049] Resource reclamation: At 2:30:00, the edge agent gradually releases the CPU / GPU memory and computing power resources occupied by EngineA to avoid system jitter caused by releasing resources too quickly.

[0050] (4) Exception handling mechanism In this embodiment, the update process proceeds normally. However, to demonstrate the system's integrity, it is assumed that the false positive rate of Engine B suddenly increases to 5% during the observation period. In this case, the abnormal rollback module will automatically trigger a rollback operation within 100 milliseconds, switching all traffic back to the stable Engine A. Simultaneously, the edge proxy reports detailed error logs (including timestamps, error codes, and abnormal sample characteristics) to the cloud, triggering the cloud's retry analysis mechanism.

[0051] Example 4 Combination Figure 5 As shown, this disclosure provides a cloud-edge collaborative edge device operator hot update apparatus 300, including a processor 304 and a memory 301. Optionally, the apparatus may further include a communication interface 302 and a bus 303. The processor 304, communication interface 302, and memory 301 can communicate with each other via the bus 303. The communication interface 302 can be used for information transmission. The processor 304 can call logical instructions in the memory 301 to execute the cloud-edge collaborative edge device operator hot update method of the above embodiment.

[0052] Furthermore, the logic instructions in the aforementioned memory 301 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0053] The memory 301, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this disclosure. The processor 304 executes functional applications and data processing by running the program instructions / modules stored in the memory 301, thereby implementing the edge device operator hot update method for cloud-edge collaboration in the above embodiments.

[0054] The memory 301 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 301 may include high-speed random access memory and may also include non-volatile memory.

[0055] This disclosure provides a computer-readable storage medium storing computer-executable instructions configured to execute the aforementioned cloud-edge collaborative edge device operator hot update method.

[0056] The aforementioned computer-readable storage medium may be a transient computer-readable storage medium or a non-transitory computer-readable storage medium.

[0057] The technical solutions of this disclosure can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes one or more instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in this disclosure. The aforementioned storage medium can be a non-transitory storage medium, including: a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and other media capable of storing program code. It can also be a transient storage medium.

[0058] The foregoing description and accompanying drawings fully illustrate embodiments of this disclosure to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, procedural, and other changes. The embodiments represent only possible variations. Individual components and functions are optional unless explicitly required, and the order of operation may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. Moreover, the terminology used in this application is for describing embodiments only and is not intended to limit the claims. As used in the description of embodiments and claims, the singular forms “a,” “an,” and “the” are intended to equally include the plural forms unless the context clearly indicates otherwise. Similarly, the term “and / or” as used in this application means including one or more of the associated listed items and all possible combinations thereof. Additionally, when used in this application, the term "comprise" and its variations "comprises" and / or "comprising" refer to the presence of stated features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof. Without further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the process, method, or apparatus that includes said element. In this document, each embodiment may focus on the differences from other embodiments, and similar or identical parts between embodiments can be referred to mutually. For methods, products, etc., disclosed in the embodiments, if they correspond to the method section disclosed in the embodiments, the relevant parts can be referred to the description of the method section.

[0059] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this disclosure. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0060] The methods and products (including but not limited to devices and equipment) disclosed in the embodiments herein can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units may be merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the coupling or direct coupling or communication connection shown or discussed between each other may be through some interfaces, and the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected to implement this embodiment according to actual needs. In addition, the functional units in the embodiments of this disclosure 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.

[0061] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

Claims

1. A method for hot updating edge device operators in a cloud-edge collaborative manner, characterized in that, Includes the following steps: The cloud detects a new version of the operator, generates an incremental update package between the old and new versions, and encrypts and digitally signs the incremental update package. Edge nodes pull the incremental update package at a predetermined time and perform multiple security checks on it, including signature verification, interface compatibility verification, and functional verification. Deploy dual-instance inference engines in edge nodes, including a running engine and a backup engine. Preload verified operators into the backup engine and achieve hot switching from the running engine to the backup engine through a state synchronization mechanism. After switching, monitor the running status of the new engine. If an abnormality is detected, trigger the rollback mechanism and switch back to the original engine. Optimize edge node resources, clean up old version operators, and report update status and resource usage to the cloud.

2. The edge device operator hot update method for cloud-edge collaboration according to claim 1, characterized in that, The specific method for generating the incremental update package is as follows: The difference generation module compares the model weights and structural differences between the old and new operators to generate an incremental update package containing only the changed parts; the size of the incremental update package is smaller than the full model package.

3. The edge device operator hot update method for cloud-edge collaboration according to claim 1, characterized in that, The multiple security checks include: Use a pre-set public key to digitally sign and verify the update package; Verify that the new operator is compatible with the input / output interfaces of the current service; Trial reasoning was performed using a standard test set in a safety sandbox to verify whether its performance indicators met the standards.

4. The edge device operator hot update method for cloud-edge collaboration according to claim 1, characterized in that, The hot-switching process includes: Preload the new operators into the backup engine and run the engine to continue processing online requests; State synchronization between the two engines can be achieved through shared memory or lock-free queues; New inference requests are routed to the backup engine, and the switchover is completed after the running engine has finished processing the existing requests.

5. The edge device operator hot update method for cloud-edge collaboration according to claim 1, characterized in that, The rollback mechanism includes: Real-time monitoring of the new engine's inference latency and accuracy; If the metric exceeds the preset threshold, it will automatically switch back to the original engine within milliseconds and report the error information to the cloud.

6. A cloud-edge collaborative edge device operator update system for implementing the method of any one of claims 1-5, characterized in that, It includes the cloud layer, the edge node layer, and the terminal device layer; The cloud layer includes: Operator repository, used to store full versions, historical versions and associated metadata of operators; The difference generation module is used to generate incremental update packages based on the differences between the old and new versions of operators. The global scheduling module is used to determine the update strategy based on the resource status and network conditions reported by the edge node layer. The security audit module is used to digitally sign update packages and record tamper-proof update logs; Message queues are used to broadcast update notifications to specified edge node layers; The edge node layer includes: Edge proxies are used to listen for cloud commands, pull update packages, and manage engine switching processes. A dual-instance inference engine, comprising a runtime engine for online services and a backup engine for preloading; The operator verification module is used to perform multiple security checks on the incremental update package; A dynamic scheduler is used to manage node resources and coordinate the updating of tasks and inference services when resources are scarce. A security sandbox is used to provide an isolated execution environment for functional verification. The exception rollback module is used to automatically trigger a rollback when a new engine exception is detected; The terminal device layer includes: A lightweight inference framework for performing local operator inference on the terminal; Trusted execution environment, used to provide hardware-level isolation protection for the update process; The status reporting module is used to send device status to the edge node layer and receive update instructions.

7. The cloud-edge collaborative edge device operator update system according to claim 6, characterized in that, The difference generation module is specifically used for: Using deep learning model optimization tools, the differences between the old and new versions of operators in terms of model parameters and computation graph structure are calculated, and an incremental update package containing only the changed parts is generated.

8. The cloud-edge collaborative edge device operator update system according to claim 7, characterized in that, The dual-instance inference engine achieves state synchronization between the running engine and the backup engine through shared memory, thereby supporting uninterrupted hot switching of inference services.

9. A cloud-edge collaborative edge device operator hot update device, comprising a processor and a memory storing program instructions, characterized in that, The processor is configured to execute the edge device operator hot update method as described in any one of claims 1 to 5 when running the program instructions.

10. A storage medium storing program instructions, characterized in that, When the program instructions are executed, they perform the edge device operator hot update method for cloud-edge collaboration as described in any one of claims 1 to 5.

Images