Edge Computing Scheduling Method and System for Heterogeneous Multi-Source Sensors

Abstract

The present invention relates to the technical field of data scheduling, and particularly to an edge computing scheduling method and system for heterogeneous multi-source sensors. The method includes the following steps: collecting raw data by using heterogeneous sensors built in an intelligent safety helmet; generating data fingerprints for the raw data to obtain an encrypted heterogeneous data set; dividing the heterogeneous data set into task types to generate a standardized task queue; monitoring the node status of an edge computing node to construct a resource dynamic perception matrix; performing resource allocation and security isolation on the standardized task queue according to the resource dynamic perception matrix to generate a resource scheduling log; inputting the resource scheduling log into an AI edge computing security coprocessor for multi-objective optimization to generate a task scheduling strategy; and performing trusted verification on the task scheduling strategy. The present invention improves the credibility and stability of task scheduling through intelligent data encryption, dynamic resource scheduling, secure and trusted verification, and anomaly detection.

Description

Technical Field

[0001] The present invention relates to the technical field of data scheduling, and in particular to an edge computing scheduling method and system for heterogeneous multi-source sensors. Background Art

[0002] Early edge computing was mainly applied to simple sensor data processing. However, with the increase in heterogeneous sensors, how to reasonably schedule data from different sensors and make full use of the advantages of edge computing has become a new challenge. In recent years, researchers have proposed a variety of edge computing scheduling methods, mainly focusing on dynamic scheduling according to the data types, computing requirements, and network conditions of different sensors. The initial scheduling methods were relatively simple and often relied on fixed rules and static configurations. With the in-depth research, more and more dynamic scheduling methods based on artificial intelligence and machine learning have been proposed, which can optimize task allocation in real time and improve the overall efficiency and resource utilization rate of the system. However, currently, traditional task scheduling often ignores multi-objective optimization and the trusted verification of task scheduling strategies, which easily leads to unstable task execution or tampering of scheduling strategies, and further results in low credibility and stability of task scheduling. Summary of the Invention

[0003] Based on this, it is necessary to provide an edge computing scheduling method and system for heterogeneous multi-source sensors to solve at least one of the above technical problems.

[0004] To achieve the above object, an edge computing scheduling method for heterogeneous multi-source sensors, the method includes the following steps:

[0005] Step S1: Collect raw data using heterogeneous sensors built in an intelligent safety helmet; generate data fingerprints for the raw data to obtain an encrypted heterogeneous data set; perform task type division on the heterogeneous data set to generate a standardized task queue;

[0006] Step S2: Monitor the node status of edge computing nodes to construct a resource dynamic perception matrix; perform resource allocation security isolation on the standardized task queue according to the resource dynamic perception matrix to generate a resource scheduling log;

[0007] Step S3: Input the resource scheduling log into an AI edge computing security coprocessor for multi-objective optimization to generate a task scheduling strategy; perform trusted verification on the task scheduling strategy, and deploy the task scheduling strategy to a target edge node according to the result of the trusted verification to generate a final scheduling instruction set;

[0008] Step S4: Collect process data of node scheduling for the final scheduling instruction set to obtain task execution data; detect abnormal execution types in the task execution data, and dynamically adjust the task scheduling policy for scheduling allocation, generate a task execution report and a security audit log, and synchronize them to the emergency command center.

[0009] The present invention provides an efficient and secure task scheduling and execution solution through the combination of intelligent safety helmets and edge computing technology. First, raw data is collected by heterogeneous sensors built into the intelligent safety helmet, and data fingerprints are generated and encrypted to obtain a heterogeneous data set. Then, these data sets are divided into task types to form a standardized task queue. By monitoring the status of edge computing nodes, a dynamic resource perception matrix is constructed, and resources are allocated to the standardized task queue based on the matrix and security isolation is performed to generate a resource scheduling log. Subsequently, the resource scheduling log is input into an AI edge computing security coprocessor for multi-objective optimization to generate a task scheduling policy, and the policy is verified for trustworthiness. Finally, according to the verification result, the task scheduling policy is deployed to the target edge node to generate a final scheduling instruction set. During the task execution process, the system collects data on the node scheduling process to obtain task execution data and detects abnormal execution types. If an abnormality is detected, the system dynamically adjusts the task scheduling policy and generates a task execution report and a security audit log, which are synchronously sent to the emergency command center in real time. Through this process, the task scheduling efficiency can be improved, the security and reliability of the task execution process can be enhanced, resource allocation can be optimized, the stable operation of the system in a complex environment can be ensured, and a quick emergency response can be made when an abnormal situation occurs. Therefore, the present invention improves the credibility and stability of task scheduling through intelligent data encryption, dynamic resource scheduling, security and trustworthiness verification, and anomaly detection.

[0010] Preferably, step S1 includes the following steps:

[0011] Step S11: Collect raw data using heterogeneous sensors built into the intelligent safety helmet;

[0012] Step S12: Perform data preprocessing on the raw data to generate a standard heterogeneous data set, where the preprocessing includes data cleaning, data denoising, missing value filling, and data standardization;

[0013] Step S13: Mark the collection timestamp of the standard heterogeneous data set; extract the device identifier of the standard heterogeneous data set; perform data hash encryption on the standard heterogeneous data set based on the device identifier to generate an encrypted heterogeneous data set;

[0014] Step S14: Divide the encrypted heterogeneous data set into task types to generate heterogeneous data task types;

[0015] Step S15: Sort the heterogeneous data task types through the collected timestamps to generate a standardized task queue.

[0016] Through multi-level data processing and task sorting, the present invention effectively optimizes task management, improves data security, and enhances the stability and scalability of the system. First, the raw data is collected by the heterogeneous sensors built in the intelligent safety helmet, providing rich collection information for subsequent processing. Then, the raw data is preprocessed, including data cleaning, denoising, missing value filling, and standardization, to ensure the quality and usability of the data. Next, by annotating the timestamps of the data, extracting device identifiers, and performing data hashing encryption, the security and integrity of the data are guaranteed, preventing the risks of data tampering and leakage. Subsequently, by classifying the task types and sorting the tasks for the encrypted heterogeneous data set, the execution order of the tasks is effectively organized, ensuring the efficiency of task scheduling and the orderly operation of the system. This series of processing methods not only improves the security, traceability, and reliability of the data, but also ensures the efficiency and stability of the task scheduling process, greatly enhancing the overall performance and adaptability of the intelligent safety helmet system.

[0017] Preferably, step S2 includes the following steps:

[0018] Step S21: Deploy edge computing nodes;

[0019] Step S22: Monitor the node status of the edge computing nodes to generate edge computing node status data, where the node status monitoring includes computing load, storage margin, and network bandwidth status;

[0020] Step S23: Construct a resource dynamic perception matrix based on the edge computing node status data; calculate the task priorities for the standardized task queue, and adjust the task order of the standardized task queue according to the task priorities to generate a dynamic priority queue;

[0021] Step S24: Perform resource allocation security isolation on the dynamic priority queue through a trusted execution environment to generate a resource scheduling log.

[0022] The present invention optimizes the efficiency of resource allocation and task scheduling and enhances the stability and security of the system by deploying edge computing nodes and monitoring their states in real time. First, deploying edge computing nodes provides powerful computing and storage support for task execution. Then, by monitoring the states of the nodes, data such as computing load, storage margin, and network bandwidth status are collected to ensure that the system operates efficiently under various resource usage conditions. Based on these node state data, a resource dynamic perception matrix is constructed, which can reflect the resource status of the nodes in real time, thus providing a reliable basis for task priority calculation. Subsequently, the standardized task queue is adjusted according to the calculated task priorities to generate a dynamic priority queue, ensuring that important tasks are processed first and improving the response speed and efficiency of task execution. Finally, through a trusted execution environment, secure isolation of resource allocation for the task queue is performed to generate a resource scheduling log. This process strengthens the security of the system and ensures the stability of resource allocation and the integrity of data. Overall, this process optimizes resource management and task scheduling through real-time monitoring, dynamic priority adjustment, and secure isolation mechanisms, improving the adaptability, stability, and security of the system in complex environments.

[0023] Preferably, step S24 includes the following steps:

[0024] Step S241: Initialize the trusted execution environment: the secure memory size ranges from 512 MB to 16 GB, the encrypted storage space ranges from 256 MB to 4 TB, the number of CPU trusted execution cores is between 1 and 8 cores, the TEE OS startup time is about 50 to 500 ms, the encrypted algorithm throughput can reach 100 to 5000 Mbps, the remote authentication time is about 10 to 200 ms, the trusted measurement time is 5 to 100 ms, the storage capacity of the key management module is between 128 KB and 32 MB, and the maximum number of concurrent trusted applications supported is 10 to 1000;

[0025] Step S242: Verify the integrity of the trusted execution environment to generate environment verification data; load security policies for the environment verification data to generate policy configuration data; divide resource boundaries for the policy configuration data to generate isolation domain data;

[0026] Step S243: Extract the task attributes of the dynamic priority queue and perform queue scheduling on the dynamic priority queue to generate scheduling sequence data; perform resource requirement analysis on the scheduling sequence data to generate resource mapping data;

[0027] Step S244: Use the isolation domain data to perform secure isolation and allocation of the resource mapping data to generate a resource scheduling log.

[0028] The present invention ensures data security and efficient resource management during task scheduling by introducing a Trusted Execution Environment (TEE), effectively enhancing the security, reliability, and flexibility of the system. First, the trusted execution environment is initialized. By setting appropriate parameters such as secure memory, encrypted storage space, and computing resources, it is ensured that the TEE has sufficient processing power to meet the security and resource requirements of different tasks. The setting of multiple parameters in this process, such as the throughput of the encryption algorithm and the storage capacity of the key management module, ensures that the system can efficiently process large-scale tasks and perform encrypted storage management. Then, by verifying the integrity of the trusted execution environment, environment verification data is generated and security policies are loaded, further strengthening the system's protection ability and avoiding potential security vulnerabilities. Subsequently, by extracting task attributes and queue scheduling from the dynamic priority queue, scheduling sequence data is generated, and resource mapping data is generated in combination with resource requirement analysis to ensure that the priorities of tasks and the required resources are reasonably allocated. Finally, the resource mapping data is securely isolated using isolated domain data to ensure data security during resource allocation, and a resource scheduling log is generated. This process not only enhances the security during task scheduling, preventing potential resource leakage or data tampering, but also optimizes the system's resource utilization efficiency and task execution efficiency through dynamic adjustment and isolation mechanisms. Overall, this process greatly improves the adaptability, stability, and security of the task scheduling system in high-security environments.

[0029] Preferably, the multi-objective optimization of inputting the resource scheduling log into the AI edge computing security coprocessor in step S3 includes:

[0030] Input the resource scheduling log into the AI edge computing security coprocessor for optimization target setting to generate optimization target setting data, where the optimization target setting data includes a latency target, an energy consumption target, and a security risk target;

[0031] Perform latency minimization optimization calculation on the latency target to obtain latency minimization optimization data, where the latency minimization optimization calculation formula is as follows:

[0032]

[0033] In the formula, is the overall latency after optimization, is the computing latency of the th task, is the communication latency of the th task, is the total number of tasks;

[0034] Perform energy consumption equalization optimization calculation on the energy consumption target to obtain energy consumption equalization optimization data, where the energy consumption equalization optimization calculation formula is as follows:

[0035]

[0036] Wherein, is the optimized overall energy consumption, is the computing energy consumption of the th module, is the communication energy consumption of the th module, and are the weight coefficients of the computing energy consumption and communication energy consumption of module ; is the total number of modules;

[0037] Perform security risk suppression optimization calculation on the security risk target to obtain security risk suppression optimization data, and the security risk suppression optimization calculation formula is as follows:

[0038]

[0039] Wherein, is the optimized security risk, is the vulnerability risk of the th task, is the intrusion risk of the th task, and are the weight coefficients of the vulnerability risk and intrusion risk; is the total number of tasks;

[0040] Perform task scheduling on the resource scheduling log according to the delay minimization optimization data, energy consumption balancing optimization data, and security risk suppression optimization data to generate a task scheduling strategy.

[0041] Through the multi-objective optimization mechanism of the AI edge computing security coprocessor, the present invention significantly improves the efficiency, security, and resource utilization of task scheduling, ensuring that the system effectively reduces latency, energy consumption, and security risks while operating efficiently. First, using the optimization objective settings, the AI edge computing security coprocessor converts the resource scheduling logs into multi-dimensional optimization objective data, including latency, energy consumption, and security risks. This process provides a scientific basis for subsequent scheduling decisions through clear optimization objectives. Then, by minimizing the latency, the computing and communication delays during task execution can be reduced, ensuring that the system's response speed and task processing efficiency reach the optimal level. Through the energy consumption balancing optimization, the system can effectively balance the computing and communication energy consumption when processing tasks, reducing unnecessary energy consumption, thereby improving the overall energy efficiency performance, which is particularly important in resource-constrained environments. Finally, through the security risk suppression optimization, the system can effectively control the potential vulnerabilities and intrusion risks of tasks, reduce security hazards, and ensure the security of data and task execution processes. This series of optimization calculations not only improve the stability and security of the system but also ensure the efficiency of resource scheduling. Ultimately, by integrating the optimized data in these three aspects, the system can generate a task scheduling strategy that meets the requirements of latency, energy consumption, and security, thereby achieving the optimal allocation of resources and the efficient execution of tasks on the premise of ensuring system security. Overall, this optimization process significantly improves the system's adaptability in complex environments, reduces latency and energy consumption, and enhances the security guarantee during task execution.

[0042] Preferably, the trustworthy verification of the task scheduling strategy in step S3 includes:

[0043] Extract the source code and configuration files of the task scheduling strategy for static analysis, including verifying the logical correctness of the task scheduling algorithm, policy constraint conditions, and security control measures, and generating static verification data for the task scheduling strategy;

[0044] Use a trustworthy execution environment to dynamically verify the task scheduling strategy, simulate the scheduling process, monitor the behavior and resource usage of the task scheduling strategy in the real operating environment, and generate dynamic verification data for the task scheduling strategy;

[0045] Integrate the static verification data and dynamic verification data of the task scheduling strategy into a trustworthy verification result.

[0046] The present invention combines static analysis and dynamic verification to ensure the logical correctness, execution security, and rationality of resource utilization of the task scheduling strategy before implementation, thereby greatly enhancing the trust and reliability during the task scheduling process. First, by extracting the source code and configuration files of the task scheduling strategy for static analysis, the logical correctness of the task scheduling algorithm can be systematically checked to ensure that the constraint conditions and security control measures of the strategy are not overlooked. This static verification ensures the correctness of the task scheduling strategy in the design phase, avoiding potential vulnerabilities or errors. The static verification data of the task scheduling strategy generated by this analysis provides a basis for subsequent verification. Using a trusted execution environment to perform dynamic verification on the task scheduling strategy, by simulating the scheduling process and monitoring the performance of the strategy in a real operating environment, the behavior, resource usage, and potential risk points of the task scheduling strategy in actual operation can be more realistically reflected. The dynamic verification data generated by this process can further verify the effectiveness and security of the strategy during runtime. Combining the static verification data and the dynamic verification data to generate a trusted verification result, thereby ensuring the security and feasibility of the task scheduling strategy during actual deployment. Through this trusted verification process, potential risks and unexpected behaviors can be identified before task scheduling, reducing the failure rate and potential security hazards during system execution. Overall, the trusted verification process greatly enhances the reliability of the task scheduling strategy, ensuring that the system can operate safely, stably, and effectively in a complex environment.

[0047] Preferably, deploying the task scheduling strategy to the target edge node according to the result of the trusted verification in step S3 includes:

[0048] Deploy the task scheduling strategy to the target edge node for node status detection to generate node status data;

[0049] Perform deployment path planning on the node status data to generate deployment plan data;

[0050] Perform network connection testing on the node status data through the deployment plan data to generate network quality data;

[0051] Establish a secure channel for the network quality data to generate channel configuration data;

[0052] Use the channel configuration data to distribute task instructions to the deployment plan data to generate scheduling instruction data.

[0053] The present invention ensures that the task scheduling strategy can be efficiently and securely deployed to the target edge node by combining the trusted verification results, node status monitoring, network testing, and secure channel construction. First, the task scheduling strategy is deployed to the target edge node according to the trusted verification results, and the node status is detected to generate node status data, which ensures that the task scheduling strategy can be reasonably deployed according to the actual node resource situation. Next, the deployment path of the node status data is planned to generate deployment plan data, ensuring that the task can be executed on a suitable path and avoiding resource conflicts. The network connection of the node status is tested through the deployment plan data to generate network quality data, ensuring a stable network connection during the task deployment process and avoiding task delays caused by poor network quality. Subsequently, based on the network quality data, a secure channel is established to generate channel configuration data, thereby ensuring the security of the data transmission process and preventing data leakage or tampering. Finally, the task instructions of the deployment plan are distributed using the channel configuration data to generate scheduling instruction data, ensuring that the task can be successfully executed on the target node. Overall, this step optimizes the deployment efficiency, security, and reliability of the task scheduling, ensuring that the task is successfully executed on the edge node as expected.

[0054] Preferably, step S4 includes the following steps:

[0055] Step S41: Collect the node scheduling process data of the final scheduling instruction set, including the real-time recording of the task execution time, resource consumption, task status, and node response information, to generate task execution data;

[0056] Step S42: Detect the abnormal execution type of the task execution data to generate abnormal task execution data;

[0057] Step S43: Perform task computing power allocation scheduling on the task scheduling strategy according to the abnormal task execution data, generate a task execution report and a security audit log, and synchronize them to the emergency command center.

[0058] The present invention realizes real-time monitoring and dynamic optimization of the task execution process through step S4 to ensure the accuracy of scheduling and the efficient execution of tasks. First, in step S41, data collection on the node scheduling process of the final scheduling instruction set is carried out, including real-time recording of task execution time, resource consumption, task status, and node response information, thereby generating detailed task execution data. Then, in step S42, detection of abnormal execution types of task execution data is carried out, including task timeout exception, resource overrun exception, task failure exception, and node response delay exception, etc., to ensure accurate identification of potential problems and generate abnormal task execution data. Finally, in step S43, the system performs task computing power allocation scheduling on the task scheduling strategy according to the abnormal task execution data, generates a detailed task execution report and a security audit log, and synchronizes them to the emergency command center for further analysis and decision-making. The implementation of this step can effectively improve the stability of the system, the reliability of task execution, and the response speed of exception handling, providing comprehensive security protection for intelligent task scheduling.

[0059] Preferably, step S42 includes the following steps:

[0060] Step S421: When any of the following situations occurs, it is determined as a task timeout exception and task timeout exception data is obtained: the task execution time deviates from the predetermined maximum execution time by more than ±10%, or the task fails to complete within the specified time;

[0061] Step S422: When any of the following situations occurs, it is determined as a resource overrun exception and resource overrun exception data is obtained: the CPU usage rate exceeds 95%, the memory usage rate exceeds 90%, or the storage usage exceeds 85%;

[0062] Step S423: When the following situations occur simultaneously, it is determined as a task failure exception and task failure exception data is obtained: the number of task failures continuously exceeds 3 times, and the number of retries after each failure does not exceed 2 times;

[0063] Step S424: When the following situations occur simultaneously, it is determined as a node response delay exception and node response delay exception data is obtained: the node response time continuously exceeds 5 seconds, and the duration of the node not responding exceeds 5 minutes;

[0064] Step S425: Integrate the task timeout exception data, resource overrun exception data, task failure exception data, and node response delay exception data to obtain abnormal task execution data.

[0065] The present invention can effectively monitor and identify problems in the task execution process by determining and collecting data on different abnormal situations that occur during the task execution process, ensuring timely processing and optimizing system performance. First, step S421 can timely identify task timeout anomalies by detecting deviations from the task execution time or failure to complete on time, ensuring that the task is completed within the predetermined time frame. Step S422 monitors resource usage and timely identifies resource overrun anomalies to avoid system performance degradation or collapse due to excessive resource occupancy. Step S423 monitors task failures to ensure that multiple task failures can be identified and corresponding emergency measures can be taken to avoid long-term failure of task execution. Step S424 monitors node response delay anomalies, and can identify long-term node non-response or delay situations, thereby taking measures to avoid system performance bottlenecks or fault expansion. Finally, step S425 integrates various types of abnormal data, generates abnormal task execution data, and helps system managers quickly locate and solve problems that occur during task execution, thereby improving the stability and reliability of the task scheduling system.

[0066] In this specification, an edge computing scheduling system for heterogeneous multi-source sensors is provided, which is used to execute the above-mentioned edge computing scheduling method for heterogeneous multi-source sensors. The edge computing scheduling system for heterogeneous multi-source sensors includes:

[0067] The data acquisition module is used to collect raw data using the heterogeneous sensors built into the smart helmet; generate data fingerprints for the raw data to obtain encrypted heterogeneous data sets; divide the heterogeneous data sets into task types to generate a standardized task queue;

[0068] The task allocation module is used to monitor the node status of edge computing nodes and build a resource dynamic perception matrix; according to the resource dynamic perception matrix, the standardized task queue is allocated and safely isolated, and a resource scheduling log is generated;

[0069] The resource scheduling module is used to input the resource scheduling log into the AI ​​edge computing security coprocessor for multi-objective optimization and generate a task scheduling strategy; perform trusted verification on the task scheduling strategy, and deploy the task scheduling strategy to the target edge node based on the trusted verification results to generate the final scheduling instruction set;

[0070] The anomaly detection module is used to collect node scheduling process data for the final scheduling instruction set to obtain task execution data; detect abnormal execution types in task execution data, and dynamically adjust the task scheduling strategy, generate task execution reports and security audit logs and synchronize them to the emergency command center.

[0071] The beneficial effects of the present invention are as follows: The data acquisition module uses heterogeneous sensors built into the intelligent safety helmet to collect raw data, and generates fingerprints and encrypts the data to ensure the security and integrity of the data. Through task type division and the generation of a standardized task queue, clear guidance can be provided for subsequent task allocation. The task allocation module monitors the status of edge computing nodes, constructs a dynamic resource perception matrix, and based on this matrix, allocates resources and performs security isolation on the task queue, generating a resource scheduling log to provide resource guarantee for task scheduling. The resource scheduling module inputs the resource scheduling log into the AI edge computing security coprocessor for multi-objective optimization, generates a task scheduling strategy, and performs a trusted verification on this strategy. Finally, according to the verification result, the task scheduling strategy is deployed to the target edge node to generate a scheduling instruction set to ensure that the task is executed according to the predetermined plan. The anomaly detection module monitors the task execution process in real time, collects task execution data, timely discovers anomalies in task execution, and dynamically adjusts the task scheduling strategy, generating a task execution report and a security audit log, and synchronizing them to the emergency command center. Through the optimization of multiple links such as data acquisition, task allocation, resource scheduling, and anomaly detection, the system effectively improves the efficiency, reliability, and security of task execution. Therefore, the present invention improves the credibility and stability of task scheduling through intelligent data encryption, dynamic resource scheduling, secure and trusted verification, and anomaly detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 It is a schematic diagram of the step flow of an edge computing scheduling method for heterogeneous multi-source sensors;

[0073] Figure 2 is Figure 1 a detailed implementation step flow diagram of step S2 in

[0074] Figure 3 is Figure 1 a detailed implementation step flow diagram of step S4 in

[0075] The realization, functional features, and advantages of the object of the present invention will be further described in conjunction with the embodiments with reference to the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0076] The technical method of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative work fall within the scope of protection of the present invention.

[0077] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0078] It should be understood that although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0079] To achieve the above object, please refer to Figures 1 to 3 , an edge computing scheduling method for heterogeneous multi-source sensors, the method comprising the following steps:

[0080] Step S1: Collect raw data using the heterogeneous sensors built in the intelligent safety helmet; generate data fingerprints for the raw data to obtain an encrypted heterogeneous data set; divide the heterogeneous data set into task types to generate a standardized task queue;

[0081] Step S2: Monitor the node status of the edge computing node to construct a resource dynamic perception matrix; perform resource allocation and security isolation on the standardized task queue according to the resource dynamic perception matrix to generate a resource scheduling log;

[0082] Step S3: Input the resource scheduling log into the AI edge computing security coprocessor for multi-objective optimization to generate a task scheduling strategy; perform a trust verification on the task scheduling strategy, and deploy the task scheduling strategy to the target edge node according to the result of the trust verification to generate a final scheduling instruction set;

[0083] Step S4: Collect process data of the node scheduling for the final scheduling instruction set to obtain task execution data; detect abnormal execution types in the task execution data, and perform dynamic scheduling allocation adjustment on the task scheduling strategy to generate a task execution report and a security audit log and synchronize them to the emergency command center.

[0084] Through the combination of intelligent safety helmets and edge computing technology, the present invention provides an efficient and secure task scheduling and execution solution. First, raw data is collected by heterogeneous sensors built into the intelligent safety helmet, and data fingerprints are generated. After encryption processing, a heterogeneous data set is obtained. Then, these data sets are divided into task types to form a standardized task queue. By monitoring the status of edge computing nodes, a resource dynamic perception matrix is constructed. According to the matrix, resource allocation and security isolation are performed on the standardized task queue, and a resource scheduling log is generated. Subsequently, the resource scheduling log is input into the AI edge computing security coprocessor for multi-objective optimization to generate a task scheduling strategy, and the strategy is subjected to trusted verification. Finally, according to the verification result, the task scheduling strategy is deployed to the target edge node to generate a final scheduling instruction set. During the task execution process, the system collects data on the node scheduling process to obtain task execution data and detects abnormal execution types. If an abnormality is detected, the system dynamically adjusts the task scheduling strategy and generates a task execution report and a security audit log, which are synchronously sent to the emergency command center in real time. Through this process, the task scheduling efficiency can be improved, the security and reliability of the task execution process can be enhanced, resource allocation can be optimized, the stable operation of the system in a complex environment can be ensured, and a quick emergency response can be made when an abnormal situation occurs. Therefore, through intelligent data encryption, dynamic resource scheduling, secure and trusted verification, and anomaly detection, the present invention improves the credibility and stability of task scheduling.

[0085] In an embodiment of the present invention, with reference to Figure 1 as shown, it is a schematic flowchart of the steps of an edge computing scheduling method for heterogeneous multi-source sensors according to the present invention. In this example, the edge computing scheduling method for heterogeneous multi-source sensors includes the following steps:

[0086] Step S1: Collect raw data using heterogeneous sensors built into the intelligent safety helmet; generate data fingerprints for the raw data to obtain an encrypted heterogeneous data set; divide the heterogeneous data set into task types to generate a standardized task queue;

[0087] In the embodiments of the present invention, multiple sensors are built into the intelligent safety helmet, such as accelerometers, gyroscopes, temperature and humidity sensors, environmental gas detectors (such as CO2 and O2 concentration sensors), heart rate monitors, GPS modules, etc. The sensors continuously collect environmental and physiological data at a fixed frequency (for example, once per second). For example, the temperature and humidity sensors collect the environmental temperature and humidity once per second, and the heart rate monitor collects the wearer's heart rate data every 1 minute. The collected raw data is first stored in the local cache and then uploaded to the edge computing node through the Bluetooth or Wi-Fi module. The raw data is encrypted using a hash algorithm (such as SHA-256) to generate a "fingerprint" of the data. The fingerprint is unique, and after encryption, it can prevent data tampering and ensure data integrity. The encrypted data fingerprint is stored in the local database and uploaded to the data storage system of the edge node. The fingerprint of each data packet is attached with a timestamp and a sensor ID for subsequent verification. The system classifies the collected data based on the data type. For example, physiological data (such as heart rate and temperature) is classified as a health monitoring task, and environmental data (such as temperature and humidity, gas concentration) is classified as an environmental monitoring task. According to factors such as the priority of the task and the processing time requirement, the tasks are sorted by type to generate a standardized task queue. The format of the standardized queue includes information such as task ID, task type, task description, priority, and required resources.

[0088] Step S2: Monitor the node status of the edge computing node to construct a resource dynamic perception matrix; perform resource allocation and security isolation on the standardized task queue according to the resource dynamic perception matrix, and generate a resource scheduling log;

[0089] In the embodiments of the present invention, a resource monitoring module is installed on each edge node to continuously monitor the resource status of the node, such as the processing capacity (CPU usage rate, memory usage rate), storage space, network bandwidth, etc. By periodically collecting the resource status data of the node, a resource dynamic perception matrix is generated. Each item in the matrix represents the resource usage of a node. The matrix update period can be set to once per minute to ensure that the system can perceive resource changes in real time. To improve the accuracy of resource monitoring, the resource data of multiple nodes will pass through a data fusion algorithm to eliminate redundant and incorrect data and form a comprehensive resource perception matrix. According to the information in the resource dynamic perception matrix, the system allocates resources to the standardized task queue. For example, computing resources are preferentially allocated to health monitoring tasks that require high computing power, and storage resources are allocated to environmental monitoring data that requires long-term storage. Virtualization technology (such as Docker containers) is used to isolate each task to ensure that resources between different tasks do not interfere with each other and prevent data leakage or conflicts. After each task scheduling, the system records the resource allocation situation (such as task ID, allocated node, resource type, etc.) to generate a detailed resource scheduling log. This log will be used for subsequent analysis and auditing.

[0090] Step S3: Input the resource scheduling log into the AI edge computing security coprocessor for multi-objective optimization to generate a task scheduling policy; perform trusted verification on the task scheduling policy, and deploy the task scheduling policy to the target edge node according to the result of the trusted verification to generate a final scheduling instruction set;

[0091] In the embodiment of the present invention, by inputting the task scheduling log into the AI coprocessor, a multi-objective optimization algorithm (such as reinforcement learning, particle swarm optimization, etc.) is used for scheduling optimization. The optimization objectives include resource utilization rate, task completion time, network delay, etc. The AI coprocessor dynamically adjusts the execution order of tasks and node allocation according to the real-time node load and task requirements to complete task scheduling in an optimal manner. The trusted computing module is used to perform security verification on the generated task scheduling policy. This includes verifying whether the task scheduling complies with preset security rules (such as data encryption, access control, etc.) to ensure that data leakage or unauthorized access does not occur during the task scheduling process. The scheduling policy is encrypted and authenticated through digital signatures to ensure that the task scheduling policy has not been tampered with and can be traced back to the source. Once the scheduling policy passes the trusted verification, the system will deploy the task scheduling instruction set to the target edge node. The deployment process will involve the encryption and decryption of the instruction set to ensure that the scheduling instructions are not tampered with during transmission. The final scheduling instruction set will be transmitted to the target edge node through the network to start executing the corresponding tasks.

[0092] Step S4: Collect the process data of node scheduling for the final scheduling instruction set to obtain task execution data; detect the abnormal execution types in the task execution data, and perform dynamic scheduling allocation adjustment on the task scheduling policy to generate a task execution report and a security audit log and synchronize them to the emergency command center.

[0093] In the embodiments of the present invention, when a task is executed on a target edge node, the system will collect task execution data in real time, including execution time, resource consumption, task status, etc. The task execution data of all nodes will be uploaded to the data center in real time for subsequent data analysis and auditing. The system uses an anomaly detection algorithm based on machine learning to monitor the task execution data, detect anomalies in task execution (such as execution timeouts, abnormal resource consumption, etc.), and generate alerts. Once an anomaly is detected, the system will perform dynamic adjustments by rescheduling tasks, adjusting resource allocation, etc. For example, if the CPU usage rate of a certain node is too high, the system will migrate some tasks to other nodes. According to the task execution data, the system will generate a detailed task execution report, and the report content includes resource consumption, execution time, abnormal conditions, etc. during task execution. For the scheduling and execution of each task, the system generates detailed security audit logs, including scheduling decisions, task execution, resource allocation, etc. information. The logs will be synchronized to the emergency command center for real-time monitoring and decision-making. All task execution reports and security audit logs will be synchronized to the emergency command center through a secure channel for managers to monitor and process.

[0094] Preferably, step S1 includes the following steps:

[0095] Step S11: Collect raw data using the heterogeneous sensors built into the intelligent safety helmet;

[0096] Step S12: Perform data preprocessing on the raw data to generate a standard heterogeneous data set, where the preprocessing includes data cleaning, data denoising, missing value filling, and data standardization;

[0097] Step S13: Mark the collection timestamp of the standard heterogeneous data set; extract the device identifier of the standard heterogeneous data set; perform data hash encryption on the standard heterogeneous data set based on the device identifier to generate an encrypted heterogeneous data set;

[0098] Step S14: Perform task type division on the encrypted heterogeneous data set to generate heterogeneous data task types;

[0099] Step S15: Sort the heterogeneous data task types by the collection timestamp to generate a standardized task queue.

[0100] In the embodiments of the present invention, multiple sensors (such as accelerometers, gyroscopes, temperature and humidity sensors, heart rate monitors, GPS modules, etc.) are built into the intelligent safety helmet, and these sensors continuously collect data at preset time intervals (such as every second or every minute). The data collected by the sensors includes but is not limited to: Accelerometer and gyroscope data: used to monitor the wearer's motion state (such as acceleration, angular velocity, etc.). Environmental sensor data: temperature and humidity sensors, gas concentration sensors (such as CO2, O2), etc., used to monitor changes in the surrounding environment. Physiological data: monitoring of physiological states such as heart rate and body temperature. Location information: The GPS module records the geographical location of the wearer in real time. Data cleaning is performed on the collected raw data to remove outliers, invalid data, or inconsistent data. For example, filter out heart rate values that do not meet expectations (such as too low or too high heart rate values), remove data anomalies caused by sensor failures, use statistical methods (such as the median method) to replace outliers or abnormal data, and use filtering algorithms (such as low-pass filters, Kalman filters) to remove noise from the raw data. For example, environmental sensor data is affected by external electromagnetic interference, and filters are used to smooth the data and reduce the impact of noise on data quality. Handle missing values, use interpolation methods (such as linear interpolation, spline interpolation) or predict missing values based on historical data to ensure the continuity and integrity of the data. Standardize the collected data, convert the data collected by different sensors (such as acceleration, temperature, humidity, etc.) to a unified scale, and use standardization algorithms (such as Z-Score standardization or Min-Max standardization) to normalize the data to the range of 0 to 1 to ensure consistency in subsequent processing. Add timestamps to each piece of data in the standard heterogeneous dataset. The timestamps are automatically generated by the system and mark the exact time of data collection. Each data record is accompanied by the collection time to ensure that the data can be analyzed and processed in chronological order. The timestamp format can use UNIX timestamps to ensure the accuracy and consistency of time records. Each intelligent safety helmet is equipped with a unique device identifier (such as a device ID). When data is collected, the system automatically extracts the device identifier and stores it together with the data. Based on the device identifier and data content, use a hash encryption algorithm (such as SHA-256) to encrypt the standard heterogeneous dataset. The purpose of hash encryption is to generate irreversible encrypted data to ensure the integrity and privacy of the data. The hash-encrypted data is stored in an encrypted storage system to ensure that the data cannot be tampered with by unauthorized personnel. According to the different sources and contents of the data, the encrypted heterogeneous dataset is divided into different task types. For example: Health monitoring tasks: involve physiological data such as heart rate and body temperature. Environmental monitoring tasks: involve environmental data such as temperature, humidity, and gas concentration. Motion state tasks: involve motion data such as accelerometer and gyroscope data. According to the different task types, each type of dataset will be separately transferred to different processing units for task processing. Each encrypted dataset is marked according to the task type.The system assigns a unique task ID to each data set and records the task type, which can ensure that the task type is clear and conducive to subsequent processing and scheduling. According to the collection timestamp, the heterogeneous data task types are sorted in chronological order, which helps the system to perform priority sorting according to the timing of the tasks. If the task type involves real-time processing (such as health monitoring tasks), the system will give priority to processing the data with a relatively recent timestamp to ensure that the tasks are processed on time. The sorted tasks are generated into a standardized task queue according to attributes such as priority and task type. Each task in the task queue will contain information such as task ID, task type, timestamp, and device identifier. The task queue will be further processed according to the preset scheduling strategy or assigned to the edge computing nodes for execution.

[0101] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes:

[0102] Step S21: Deploy edge computing nodes;

[0103] Step S22: Monitor the node status of the edge computing nodes to generate edge computing node status data, where the node status monitoring includes computing load, storage margin, and network bandwidth status;

[0104] Step S23: Construct a resource dynamic perception matrix based on the edge computing node status data; calculate the task priorities of the standardized task queue, and adjust the task order of the standardized task queue according to the task priorities to generate a dynamic priority queue;

[0105] Step S24: Perform resource allocation security isolation on the dynamic priority queue through the trusted execution environment to generate a resource scheduling log.

[0106] In an embodiment of the present invention, by deploying edge computing nodes at predetermined edge computing locations (such as construction sites, production sites, or data centers), it is ensured that the nodes can effectively process when receiving data streams from smart helmets. Each edge computing node is equipped with appropriate computing resources (such as CPU, GPU, memory, storage) to ensure that the collected massive data can be processed and stored. The edge computing node is connected to the local network to ensure that the node can receive heterogeneous data sets from the smart helmet in real time and can interact with other edge nodes or central servers. Configure the operating system and software environment of the edge computing node, including installing relevant computing frameworks and task scheduling management tools. Configure the firewalls and security measures of each node to ensure security during data transmission. Set up an access control mechanism for the edge computing node to ensure that only authorized devices or users can access the computing and storage resources of the node. Configure the communication interface between the edge computing node and the smart helmet. Common communication protocols include Wi-Fi, Bluetooth, or 5G, etc., to ensure the reliability and low latency of the data transmission process. Configure the interface between the node and the upper scheduling system or cloud platform for subsequent task scheduling and resource optimization. Deploy resource monitoring tools on edge computing nodes to collect computing load data (such as CPU and GPU usage, memory usage, etc.) in real time. These data help understand whether the computing power of the node is sufficient and whether load balancing is required. By regularly sampling computing load data, monitor whether the node is in an overloaded state, and then decide whether to expand computing resources or redistribute tasks. Monitor the storage usage of the node, including hard disk space, cache storage, database storage, etc. This monitoring ensures that data will not be lost or cause system crashes due to insufficient storage capacity. Generate storage margin data through the storage monitoring system to adjust the task data storage strategy in time. Monitor the network bandwidth of the edge computing node in real time to ensure that the node can communicate with other nodes or cloud platforms unimpeded. The monitoring content includes bandwidth usage, latency, data transmission rate, etc. If the network bandwidth is found to be insufficient, it will affect the timely execution of the task, so the bandwidth usage needs to be dynamically optimized. Aggregate the monitoring data such as computing load, storage margin, network bandwidth, etc. to the central monitoring system or the local storage of the edge node to form the edge computing node status data. The status data includes the computing resource usage, storage resource status, network bandwidth, etc. of each node for subsequent task scheduling and resource allocation. Based on the node status data (such as computing load, storage margin, network bandwidth), a resource dynamic perception matrix is ​​constructed. This matrix can reflect the current resource status of each edge computing node and the load capacity of each node. The dynamic perception matrix will contain the resource status of each node (such as real-time data of load, storage and bandwidth) so that the task scheduling system can optimize the scheduling according to the current resource situation. The priority of each task in the standardized task queue is calculated based on factors such as task type, urgency, and estimated execution time.The priority calculation considers the following factors: Task urgency: High-priority tasks such as real-time health monitoring tasks and environmental monitoring tasks should be processed first. Task complexity: Tasks with high computational complexity (such as large-scale data processing tasks) require more computing resources, so their priorities will be dynamically adjusted according to resource availability. Task dependencies: If there are dependencies between tasks (for example, task A must be executed before task B), the system needs to adjust the task order according to the dependencies. According to the task priority calculation results, the task order in the standardized task queue is adjusted to generate a dynamic priority queue. This queue will contain information such as the task ID, task type, priority, and estimated execution time. The adjusted queue ensures that the most urgent and important tasks can be executed first, while those less urgent tasks will be postponed, avoiding overloading of system resources. Deploy a trusted execution environment (TEE) on each edge computing node, such as technologies like Intel SGX or ARM TrustZone. The TEE is used to ensure the security and privacy of task execution during resource scheduling. Through the TEE, tasks are securely isolated to ensure that resource access between different tasks does not conflict or leak sensitive data. Through the trusted execution environment, tasks in the task queue are allocated resources according to their priorities and resource requirements. Resource allocation includes computing resources (such as CPUs, GPUs), storage resources, network bandwidth, etc. Through the TEE, the task allocation process is isolated to ensure that task data is not interfered with or accessed by other tasks, preventing potential security threats (such as malicious attacks or data leaks). Each step in the resource scheduling process is recorded to generate a detailed resource scheduling log. The scheduling log includes information such as task ID, computing resources allocated to the task, storage resources, network bandwidth, priority, and timestamp.

[0107] Preferably, step S24 includes the following steps:

[0108] Step S241: Initialize the trusted execution environment: The secure memory size ranges from 512 MB to 16 GB, the encrypted storage space ranges from 256 MB to 4 TB, the number of CPU trusted execution cores is between 1 and 8 cores, the TEE OS startup time is approximately 50 to 500 ms, the encrypted algorithm throughput can reach 100 to 5000 Mbps, the remote authentication time is approximately 10 to 200 ms, the trusted measurement time is 5 to 100 ms, the storage capacity of the key management module ranges from 128 KB to 32 MB, and the maximum number of concurrent trusted applications supported is 10 to 1000;

[0109] Step S242: Verify the integrity of the trusted execution environment to generate environment verification data; load security policies for the environment verification data to generate policy configuration data; divide resource boundaries for the policy configuration data to generate isolation domain data;

[0110] Step S243: Extract the task attributes of the dynamic priority queue, perform queue scheduling on the dynamic priority queue to generate scheduling sequence data; perform resource requirement analysis on the scheduling sequence data to generate resource mapping data;

[0111] Step S244: Use the isolation domain data to perform resource security isolation and allocation on the resource mapping data to generate a resource scheduling log.

[0112] In the embodiments of the present invention, by configuring the size of the secure memory required for the Trusted Execution Environment (TEE), which is typically between 512 MB and 16 GB. This secure memory is used to temporarily store data during task execution and ensure the confidentiality and integrity of the data. Set the size of the encrypted storage space, ranging from 256 MB to 4 TB, for storing sensitive data and encrypted task information. The size of the encrypted storage space should be adjusted according to the storage requirements of specific tasks. Configure the number of CPU trusted execution cores supported by the TEE, usually 1 to 8 cores, which will be used to execute trusted tasks and ensure the security of the tasks. Configure an appropriate number of cores to improve computing performance and support multi-task parallel execution. The startup time of the TEE should be between 50 and 500 ms to ensure that the trusted execution environment can start quickly and be ready to execute tasks at the beginning of task scheduling. The throughput of the encryption algorithm should reach 100 to 5000 Mbps to ensure the efficiency of encryption operations to meet the encryption requirements of large data volume tasks. The remote authentication time is controlled within 10 to 200 ms to ensure that edge nodes and other remote devices can quickly verify and establish trusted communication connections. The trusted measurement time is controlled within 5 to 100 ms to ensure that the credibility measurement of tasks and the environment can be completed in a short time for quick execution of security verification. Configure the storage capacity of the key management module between 128 KB and 32 MB to ensure that the keys used in encryption and decryption operations can be securely stored. The maximum number of concurrent trusted applications supported is 10 to 1000 to ensure that the trusted execution environment can handle multiple parallel tasks and applications simultaneously. By using technologies such as hash values, signatures, and encryption checksums, verify the integrity of the TEE environment to ensure that it has not been tampered with. The verification content includes the hardware configuration, operating system, application programs, etc. of the TEE. The environment verification process should be automatically executed during the startup process to ensure that the TEE environment of each node is secure and consistent. After verification, generate an environment verification data record the information during the verification process, such as verification time, verification result, used algorithms, etc. The environment verification data will be used for subsequent policy loading and task execution to ensure that all operations during task execution are carried out in a trusted environment. Load the corresponding security policies according to the environment verification data to ensure that each task in the TEE environment is under security control. The security policies include access control policies, task execution policies, data protection policies, etc. to ensure that all resource allocation and task scheduling are executed in an environment that meets security requirements. Through the loaded security policies, generate policy configuration data, including the permissions for task execution, restrictions on resource usage, data access control, etc. The policy configuration data will be dynamically adjusted according to the environment verification data to ensure that the policies match the current environment and resource status. Based on the policy configuration data, divide the usage boundaries of resources. For example, divide resources such as CPU, memory, and storage into different secure isolation areas (isolation domain data) according to task types and priorities to prevent interference or leakage of sensitive information between different tasks.Extract the relevant attributes of each task from the dynamic priority queue, including task type, priority, computing resource requirements, storage requirements, execution time, etc. The extraction of task attributes helps to understand the execution characteristics of each task, providing a basis for subsequent resource allocation and scheduling. Based on the priority, resource requirements, and execution time of the tasks, queue scheduling is performed on the tasks. According to the priority order of the tasks, the execution order of the tasks is dynamically adjusted to ensure that high-priority tasks are executed first. During the task scheduling process, factors such as dependencies between tasks and resource conflicts also need to be considered to ensure the reasonable use of resources and the efficient execution of tasks. Generate scheduling sequence data for the scheduled tasks, including the execution order of each task, allocated resources, estimated execution time, etc. The scheduling sequence data will be used as a reference for task scheduling, guiding each edge computing node to execute tasks according to the specified order and resource allocation. Conduct a resource requirement analysis for each task in the scheduling sequence to evaluate the requirements of each task for computing resources, storage resources, network bandwidth, etc. The results of the resource requirement analysis will help the system dynamically adjust the resource allocation strategy to ensure that all tasks can be efficiently executed within the allocated resources. Based on the results of the resource requirement analysis, generate resource mapping data, which details the computing, storage, and network resources required for each task. Use the isolation domain data to perform security isolation on the resource mapping data of the tasks to ensure that each task can only access the resources allocated to it and cannot access the resources of other tasks. The isolation domain data can be configured according to policies to ensure that data and resources between different tasks do not interfere with each other, ensuring security and data privacy during task execution. According to the resource mapping data and isolation domain data, perform resource allocation. Allocate appropriate computing resources, storage space, and network bandwidth for each task to ensure that the tasks can be executed smoothly. During the resource allocation process, the system needs to adjust the resource allocation strategy in real time to cope with the dynamically changing resource requirements and task priorities. Record the detailed information of each step of resource allocation during the resource allocation process to generate a resource scheduling log. The log content includes task ID, allocated resource type, resource quantity, timestamp of resource allocation, etc.

[0113] Particularly importantly, step S244 further includes the following steps:

[0114] Step S2441: Evaluate the security trust level of the isolation domain data to generate isolation domain security trust evaluation data;

[0115] Step S2442: Perform dynamic access control mapping on the resource mapping data to generate access control optimization data;

[0116] Step S2443: Perform multi-dimensional resource conflict detection on the resource mapping data to generate resource conflict prediction data, where the multi-dimensional resource conflict detection includes analyzing the temporal trend of computing resources and predicting resource bottleneck points within the next 30 minutes;

[0117] Step S2444: Integrate the security trust evaluation data, access control optimization data, and resource conflict prediction data of the isolation domain, perform adaptive security isolation and optimized allocation on resources, and generate a resource scheduling log.

[0118] In the embodiment of the present invention, by adopting a method based on the Zero Trust Architecture (ZTA), the trust level of the isolation domain is evaluated. The evaluation parameters include: the integrity of the computing environment (such as the verification result of the TEE environment), the security level of the task (based on the sensitivity of the task data), the historical security event record (such as whether there has been an unauthorized access), and the execution status of the access control policy (whether there is an access that violates the policy). Set the security trust score range (0 - 100), and the calculation formula is as follows: ; where is the TEE environment integrity evaluation score, is the security level of the task (such as high, medium, low), is the historical security event impact factor, is the execution status of the access control policy, , , , are weight parameters. According to the trust score, the isolation domain is classified as follows: High security trust (80 - 100): The task can be directly executed. Medium security trust (50 - 79): Additional security verification (such as multi-factor authentication) is required. Low security trust (0 - 49): Isolate and reject the task execution. Generate a trust evaluation log and store it in a trusted database. Adopt the Role-Based Access Control (RBAC) + Attribute-Based Access Control (ABAC) method: Based on the role of the user to which the task belongs, allocate resource access permissions (such as administrator, ordinary user, visitor). Based on the attributes of the task (such as priority, data sensitivity, computing resource requirements), dynamically adjust the access permissions. Calculate the access permission requirements: Whether the task requires high-priority computing resources; Whether the data of the task requires additional encryption; Whether the execution time of the task affects the overall system scheduling. Adopt the principle of least privilege to limit the task from accessing unnecessary resources and reduce security risks. Record the optimized access control rules: The allowed resource range (CPU, storage, bandwidth), the access control level (execute only, read only, read and write), and the access time window (when the task can access the resources). Adopt time series analysis methods (such as ARIMA, LSTM neural network) to model the usage trend of computing resources. The main metrics of concern are: the CPU utilization trend, the storage usage growth trend, and the network bandwidth consumption change trend. Predict the peak time of computing resources through regression analysis: ; where is the future time The predicted value of the resource bottleneck within For different types of resources (CPU, memory, storage, bandwidth) at time The utilization rate within Is the weight parameter, Is the error term. Set the resource usage threshold: When the CPU utilization rate > 80%, there is a competition for computing resources; when the storage utilization rate > 90%, there is a storage bottleneck; when the network bandwidth utilization rate > 85%, it causes communication latency. Generate resource conflict prediction data, and mark the time points and task IDs where bottlenecks occur. According to the security trust evaluation data of the isolation domain, adjust the resource access permissions: For high-trust tasks, allow direct access to the required resources. For medium-trust tasks, add security verification (such as dynamic password authentication). For low-trust tasks, block access or perform security isolation. According to the access control optimization data, adjust the execution order of tasks: High-priority tasks are preferentially allocated computing resources. Low-priority tasks are executed when the resource pressure is small to avoid resource conflicts. According to the resource conflict prediction data, adjust the resource scheduling strategy for the next 30 minutes: If a resource bottleneck is predicted, postpone the execution of some low-priority tasks. If the computing resources are tight, dynamically adjust the task allocation to avoid resource competition. Record the final resource allocation plan for all tasks and generate a resource scheduling log.

[0119] Preferably, the multi-objective optimization of inputting the resource scheduling log into the AI edge computing security coprocessor in step S3 includes:

[0120] Input the resource scheduling log into the AI edge computing security coprocessor for optimization target setting, generate optimization target setting data, where the optimization target setting data includes a latency target, an energy consumption target, and a security risk target;

[0121] Perform latency minimization optimization calculation on the latency target to obtain latency minimization optimization data, where the latency minimization optimization calculation formula is as follows:

[0122]

[0123] In the formula, Is the optimized overall latency, Is the th task's computing latency, Is the th task's communication latency, Is the total number of tasks;

[0124] Perform energy consumption equalization optimization calculation on the energy consumption target to obtain energy consumption equalization optimization data, where the energy consumption equalization optimization calculation formula is as follows:

[0125]

[0126] In the formula, is the optimized overall energy consumption, is the th module's computing energy consumption, is the th module's communication energy consumption, and are the weight coefficients of the computing energy consumption and communication energy consumption of module , and is the total number of modules;

[0127] Perform security risk suppression optimization calculation on the security risk target to obtain security risk suppression optimization data. The security risk suppression optimization calculation formula is as follows:

[0128]

[0129] In the formula, is the optimized security risk, is the vulnerability risk of the th task, is the intrusion risk of the th task, and are the weight coefficients of the vulnerability risk and intrusion risk, is the total number of tasks;

[0130] Perform task scheduling on the resource scheduling log according to the delay minimization optimization data, energy consumption balancing optimization data, and security risk suppression optimization data to generate a task scheduling strategy.

[0131] In the embodiments of the present invention, by inputting the resource scheduling log into the AI edge computing security coprocessor, key scheduling information is extracted, such as task execution time, communication overhead, computing resource consumption, security event records, etc. Based on task requirements and system constraints, optimization goals are set, including: reducing task execution time and improving scheduling efficiency; balancing the energy consumption of computing and communication and enhancing the overall energy utilization rate; reducing vulnerability and intrusion risks and improving system security. According to the data set for the optimization goals, an optimization problem is constructed to form a parameter set for inputting optimization calculations. Analyze the task computing time and communication time in the resource scheduling log, and perform the following optimization calculations to obtain the optimal delay: In the formula, is the optimized overall delay, is the computing delay of the th task, is the communication delay of the th task, ​is the total number of tasks; use the multi-objective optimization algorithm (such as dynamic programming, reinforcement learning, or evolutionary algorithm) in the AI co-processor to solve the optimal task allocation scheme to reduce the total latency, generate latency minimization optimization data for subsequent scheduling decisions. Extract the computing energy consumption and communication energy consumption from the resource scheduling log, and perform the following optimization calculations to balance the energy consumption of computing and communication: In the formula, is the optimized overall energy consumption, is the computing energy consumption of the th module, is the communication energy consumption of the and are the weight coefficients of the computing energy consumption and communication energy consumption of module ; is the total number of modules; adjust the task allocation strategy to avoid overloading of certain computing nodes, and at the same time reduce the concentrated distribution of high-energy-consuming tasks, generate energy consumption balancing optimization data to support the scheduling strategy. Based on the log data, calculate the vulnerability risk and intrusion risk of each task, and perform the following optimization calculations to reduce the overall security risk: In the formula, is the optimized security risk, is the vulnerability risk of the th task, is the intrusion risk of the th task, and are the weight coefficients of the vulnerability risk and intrusion risk, is the total number of tasks; adjust the task allocation to reduce the concentrated execution of high-risk tasks. Adopt active defense strategies, such as dynamic access control, anomaly detection mechanisms, etc., generate security risk suppression optimization data for formulating a more secure task scheduling scheme. Based on the latency minimization optimization data, energy consumption balancing optimization data, and security risk suppression optimization data, construct an optimized task scheduling scheme. Use the weighted multi-objective optimization method to balance the importance of different optimization goals. Dynamically adjust the task allocation strategy through an AI model (such as reinforcement learning, neural network, etc.). Allocate tasks to the optimal computing nodes. Adjust the computing and communication resources to achieve energy consumption balance. Adopt security protection measures to reduce vulnerability and intrusion risks. Deploy the task scheduling strategy to the edge computing environment to achieve dynamic task optimization execution.

[0132] Preferably, the trustworthy verification of the task scheduling strategy described in step S3 includes:

[0133] Extract the source code and configuration file of the task scheduling strategy for static analysis, including verifying the logical correctness of the task scheduling algorithm, policy constraint conditions, and security control measures, and generate task scheduling strategy static verification data;

[0134] Dynamically verify the task scheduling strategy using a trusted execution environment, simulate the scheduling process, monitor the behavior and resource usage of the task scheduling strategy in a real operating environment, and generate dynamic verification data for the task scheduling strategy;

[0135] Integrate the static verification data of the task scheduling strategy and the dynamic verification data of the task scheduling strategy into a trusted verification result.

[0136] In the embodiments of the present invention, obtain the core scheduling algorithm code, rule configuration file, security policy definition, etc. from the task scheduling system. Analyze the code structure, extract the scheduling logic, decision rules, and constraint conditions. Use formal verification (such as Hoare logic, model checking) to verify the logical completeness of the scheduling algorithm. Check for problems such as infinite loops, infinite recursions, resource competitions, etc. Ensure that the task scheduling complies with predefined priority rules, latency thresholds, resource allocation policies. Use rule checking tools (such as ESLint, Clang Static Analyzer) to automatically verify whether the scheduling strategy complies with the standards. Analyze whether the task scheduling strategy includes access control, permission management, and defense mechanisms. Use code security audit tools (such as SonarQube, Flawfinder) to detect whether there are security vulnerabilities (such as SQL injection, buffer overflow). Record the code inspection results, logical analysis reports, and security assessment data to form a static verification data set. The trusted execution environment (TEE, Trusted Execution Environment) provides an isolated computing environment to ensure that the task scheduling strategy is executed in a secure environment and avoid external interference. Run the scheduling strategy on a TEE platform such as Intel SGX, AMD SEV, or ARM TrustZone to ensure that the scheduling strategy will not be maliciously tampered with. Track the execution time of tasks to ensure that the latency constraints are met. Monitor the resource occupancy of CPU, memory, network bandwidth, etc. to ensure balanced resource scheduling. Detect whether there are problems such as unauthorized access, abnormal termination, deadlocks, memory leaks, etc. during the task execution process. Identify abnormal patterns through log analysis and anomaly detection algorithms (such as the LSTM anomaly detection model). Monitor whether the access control policy is effective to prevent unauthorized task execution. Detect whether there are unauthorized API calls or file access behaviors. Record the task execution logs, performance monitoring data, and anomaly detection reports to form a dynamic verification data set. Compare the static verification data and the dynamic verification data to analyze whether there are conflicts or inconsistencies. Analyze the historical scheduling data through a machine learning model (such as decision tree, random forest) to optimize the trusted verification result. Set the credibility scoring formula: ; is the logical correctness score, is the constraint compliance score, is the security control compliance score, is the dynamic operation stability score, , , , is the weight coefficient, which can be adjusted according to business requirements.

[0137] Preferably, deploying the task scheduling policy to the target edge node according to the result of the trusted verification in step S3 includes:

[0138] Deploying the task scheduling policy to the target edge node to perform node status detection and generating node status data;

[0139] Performing deployment path planning on the node status data to generate deployment plan data;

[0140] Performing network connection testing on the node status data through the deployment plan data to generate network quality data;

[0141] Establishing a secure channel for the network quality data to generate channel configuration data;

[0142] Distributing task instructions to the deployment plan data by using the channel configuration data to generate scheduling instruction data.

[0143] In the embodiments of the present invention, by collecting the status data of the CPU, memory, storage, bandwidth, running tasks, etc. of the target edge node, the current task load and resource utilization rate are obtained, and a node resource information table is formed. The collected node resource information is analyzed to detect abnormal conditions such as overloading, faults, and packet loss. Threshold detection and machine learning anomaly detection methods are used to generate a node health status report. According to the health status report, node status data is generated, including availability score, running status, load condition, etc. The task scheduling policy after trusted verification is parsed to extract scheduling constraint conditions (computing requirements, bandwidth requirements, latency requirements, etc.). Combining the node status data, a candidate node set is determined. The shortest path algorithm based on graph theory (such as Dijkstra) is used to calculate the optimal deployment path. Combining the real-time network topology, ensure that the path selection meets the QoS (Quality of Service) requirements. Deployment scheme data is formed, including target node selection, task migration path, load balancing strategy, etc. According to the deployment scheme data, connectivity tests (Ping, TCP handshake, etc.) are performed on the candidate edge nodes. Network parameters such as packet loss rate and RTT (Round Trip Time) are monitored to ensure node reachability. Tools such as iPerf are used to evaluate the bandwidth performance to ensure stable transmission of task data. The throughput rate between nodes is calculated, and a suitable network path is selected based on the task requirements. Combining the connectivity test and bandwidth test results, network quality data is generated, including metrics such as latency, jitter, and packet loss rate. An end-to-end encrypted channel is established using TLS (Transport Layer Security) or IPSec, and multi-factor identity authentication is performed using the zero trust architecture (ZTA, Zero Trust Architecture). Secure channels are established through means such as VPN, SSH Tunnel, and TLS to ensure data transmission security, and a data encryption policy is configured, and an encryption algorithm such as AES-256 or a higher level is selected. The configuration information of the secure channel is recorded, including encryption protocol, authentication method, key exchange information, etc. According to the channel configuration data, the task scheduling instruction is encrypted and encapsulated to ensure the integrity and anti-tampering of the instruction. Blockchain technology is used to perform signature verification on the task instruction to improve credibility. Message queue technologies such as MQTT and Kafka are used to ensure reliable transmission of the instruction to the target node. Low-latency instruction issuance is achieved through WebSocket or gRPC. The task instruction distribution result is recorded to confirm whether each target edge node correctly receives and executes the task. Scheduling instruction data is formed, including task allocation logs, execution status, abnormal conditions, etc.

[0144] As an example of the present invention, refer to Figure 3 As shown, in this example, step S4 includes:

[0145] Step S41: Collect the process data of node scheduling for the final scheduling instruction set, including real-time recording of task execution time, resource consumption, task status, and node response information, and generate task execution data;

[0146] Step S42: Detect the abnormal execution type of the task execution data and generate abnormal task execution data;

[0147] Step S43: Perform task computing power allocation scheduling on the task scheduling strategy according to the abnormal task execution data, generate a task execution report and a security audit log, and synchronize them to the emergency command center.

[0148] In the embodiments of the present invention, the start time, end time, running time, and task response latency of a task are recorded. A distributed logging system (such as ELK, Prometheus) is used to continuously monitor the task execution time. The deviation range of the execution time is calculated in combination with the node load situation. The consumption of resources such as CPU, memory, disk I / O, and network bandwidth is monitored. A performance monitoring tool (such as cAdvisor, Grafana) is used to collect the real-time running status of nodes. A resource consumption model is formed to evaluate the scheduling efficiency. The task execution status (success, failure, timeout, abnormal termination, etc.) is recorded. The heartbeat signal, system log, and load situation of the node are collected to ensure the normal execution of the task. The WebSocket or MQTT protocol is used to implement a low-latency data reporting mechanism. Combining the task execution time, resource consumption, task status, and node response data, task execution data is generated. Anomaly detection is performed on the task execution data, including: timeout anomaly (the execution time exceeds the preset threshold), resource overflow anomaly (CPU, memory, bandwidth exceed the available range), task failure anomaly (returning an error code or the task crashing), communication anomaly (the node does not respond for a long time or the packet loss rate is too high). LSTM (Long Short-Term Memory Network) or Isolation Forest is used for anomaly detection to distinguish between normal and abnormal execution situations. The model is trained to learn historical task execution data and dynamically adjust the detection threshold. The information of abnormal tasks is recorded, including task ID, anomaly type, anomaly reason, scope of influence, etc., and abnormal task execution data is generated. Based on the abnormal task execution data, the task scheduling strategy is dynamically adjusted, including: load balancing optimization (migrating high-load tasks to low-load nodes). Computing power allocation optimization (increasing the computing power priority of abnormal tasks to avoid resource bottlenecks). Real-time task retry (re-assigning failed tasks to standby nodes). The overall situation of task execution is recorded, including: total number of tasks, success rate, failure rate, average task execution time, resource utilization rate, type distribution of abnormal tasks, and optimization effect. The detailed log of the task scheduling process is recorded to ensure traceability, including: task execution changes (scheduling adjustment records), anomaly response records (corrective measures and recovery processes), security event records (whether it involves attacks or illegal operations), and blockchain technology is used to ensure that the audit log cannot be tampered with. The task execution report and the security audit log are synchronized to the emergency command center through the Kafka message queue or the RESTful API. The AES-256 encryption is used to transmit data to ensure security. The command center can perform anomaly early warning, decision optimization, and historical data backtracking.

[0149] Particularly importantly, the task computing power allocation scheduling of the task scheduling strategy according to the abnormal task execution data further includes:

[0150] Dynamically perceive the edge node resources for the task scheduling policy based on the abnormal task execution data, and generate the edge computing resource status perception data;

[0151] Perform portrait modeling on the edge computing resource status perception data to generate a computing resource portrait;

[0152] Perform resource fusing on the computing resource portrait through the federated reinforcement learning framework to generate a trusted policy fusing mechanism;

[0153] Use the trusted policy fusing mechanism to conduct resource allocation auditing on the task scheduling policy, and generate a task execution report and a security audit log.

[0154] In the embodiments of the present invention, the resources of edge nodes are dynamically perceived according to various abnormal information extracted from abnormal task execution data (such as task timeout, resource overrun, task failure, etc.). This process generates edge computing resource status perception data by monitoring the resource usage of edge nodes, including real-time data such as CPU, memory, storage, and network bandwidth. These data will reflect the load conditions, resource consumption levels, and operating states of current edge nodes, helping the system to evaluate the resource capabilities and load conditions of each node in real time. Based on the edge computing resource status perception data, the system performs computational resource profiling modeling. This process constructs the computational resource profile of each edge node through machine learning methods (such as clustering analysis, principal component analysis, etc.) using various collected resource consumption data, task execution status, and historical performance data. The computational resource profile includes characteristics such as the processing capacity, resource consumption pattern, and load balancing ability of the node, and is classified according to different working states of the node (such as high load, low load, idle state, etc.). After constructing the computational resource profile, the system will perform resource fusing on the computational resource profile through a federated reinforcement learning framework. The federated reinforcement learning framework allows multiple edge nodes to share information while maintaining data privacy, and optimizes the resource scheduling strategy through collaborative learning. Under this framework, the system evaluates the fusing threshold of resources by simulating different task scheduling schemes and resource allocation strategies, that is, how to quickly and effectively adjust resource allocation in case of overload or resource bottleneck to prevent system crashes or task failures. The generation of the resource fusing mechanism is based on the node load conditions, task priorities, and resource requirements of tasks, ensuring that the system can dynamically adjust under high load or abnormal conditions and avoid single-point overload problems. The generated trusted policy fusing mechanism is used to audit the task scheduling strategy. The trusted policy fusing mechanism automatically adjusts the task scheduling strategy according to the computational resource profile and the resource fusing mechanism under various resource abnormal conditions to ensure reasonable allocation of resources among nodes. Through the audit of the task scheduling strategy, the system can evaluate the effectiveness of the current scheduling strategy and make adjustments for abnormal task execution data. For example, if the resources of some nodes exceed the standard, the system will migrate the relevant tasks to nodes with lower load, or split the computational tasks into smaller subtasks for parallel processing. Finally, the adjustment of the task scheduling strategy will generate a task execution report and a security audit log. The task execution report will detail the changes in various indicators during the task execution process, including task success rate, execution duration, resource consumption, task failure situation, etc.; the security audit log includes detailed records of all resource allocations, scheduling decisions, and change operations during the task scheduling process to ensure the transparency and traceability of operations.

[0155] Preferably, step S42 includes the following steps:

[0156] Step S421: When any of the following situations occurs, it is determined as a task timeout exception, and task timeout exception data is obtained: the task execution time deviates from the predetermined maximum execution time by more than ±10%, or the task fails to be completed within the specified time;

[0157] Step S422: When any of the following situations occurs, it is determined as a resource overrun exception, and resource overrun exception data is obtained: the CPU usage rate exceeds 95%, the memory usage rate exceeds 90%, or the storage usage exceeds 85%;

[0158] Step S423: When the following situations occur simultaneously, it is determined as a task failure exception, and task failure exception data is obtained: the number of task failures continuously exceeds 3 times, and the number of retries after each failure does not exceed 2 times;

[0159] Step S424: When the following situations occur simultaneously, it is determined as a node response delay exception, and node response delay exception data is obtained: the node response time continuously exceeds 5 seconds, and the duration of the node not responding exceeds 5 minutes;

[0160] Step S425: Integrate the task timeout exception data, resource overrun exception data, task failure exception data, and node response delay exception data to obtain abnormal task execution data.

[0161] In the embodiments of the present invention, when the task execution time deviates from the predetermined maximum execution time by more than ±10%, or the task fails to be completed within the specified time, it is determined that the task has a timeout exception. At this time, the system will monitor the execution time of the task, calculate the difference between it and the predetermined maximum execution time. If the difference exceeds ±10%, it is regarded as a timeout; or if the task is not completed within the specified time range, a task timeout exception determination will also be triggered. The task timeout exception data will include the task ID, task start time, predetermined maximum execution time, actual execution time, and the reason for the timeout. When the CPU usage rate exceeds 95%, the memory usage rate exceeds 90%, or the storage usage exceeds 85% during the task execution, it is determined that there is a resource overrun exception. This step monitors the consumption of node resources in real time. Once it is found that the resource usage rate exceeds the set threshold, the system will immediately mark it as a resource overrun exception and generate corresponding resource overrun exception data. These data will include the task ID, CPU usage rate, memory usage rate, storage usage rate, and the specific threshold for resource overrun. When the number of consecutive task failures exceeds 3 times, and the number of retries after each failure does not exceed 2 times, it is determined that there is a task failure exception. This step monitors the execution status of the task and records the situation of task failures. If the task fails continuously more than three times during the execution process, and only no more than two retries are performed after each failure, then this task will be marked as a failure exception. The task failure exception data will include detailed information such as the task ID, number of failures, number of retries, and reason for failure. When the response time of the node continuously exceeds 5 seconds, and the duration of the node not responding exceeds 5 minutes, it is determined that there is a node response delay exception. By monitoring the response time of the node, if the response time exceeds 5 seconds and the node has not responded in the past 5 minutes, it is regarded as a node response delay exception. The node response delay exception data will record information such as the node ID, consecutive response time, duration of non-response, and node status. In the above steps, all the generated exception data will be integrated. The task timeout exception data, resource overrun exception data, task failure exception data, and node response delay exception data will be summarized into a complete set of exception task execution data.

[0162] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to cover all changes falling within the meaning and scope of the equivalent elements of the application documents within the present invention.

[0163] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. An edge computing scheduling method for heterogeneous multi-source sensors, characterized in that: The following steps are involved: Step S1: Collecting raw data using heterogeneous sensors built into the smart helmet; Generating data fingerprints for the original data to obtain an encrypted heterogeneous data set; dividing the heterogeneous data set into task types to generate a standardized task queue; Step S2: Monitor the node status of edge computing nodes to build a resource dynamic perception matrix; perform resource allocation security isolation on the standardized task queue according to the resource dynamic perception matrix, and generate a resource scheduling log; Step S3: Input the resource scheduling log into the AI ​​edge computing security coprocessor for multi-objective optimization to generate a task scheduling strategy; The task scheduling strategy is trusted and verified, and the task scheduling strategy is deployed to the target edge node according to the trusted verification result to generate the final scheduling instruction set; wherein, the resource scheduling log is input into the AI ​​edge computing security coprocessor for multi-objective optimization, including: Input the resource scheduling log into the AI ​​edge computing security coprocessor to set the optimization target and generate the optimization target setting data, where the optimization target setting data includes the latency target, energy consumption target and security risk target; The delay minimization optimization calculation is performed on the delay target to obtain the delay minimization optimization data, where the delay minimization optimization calculation formula is as follows: In the formula, is the optimized overall delay, For the The computational latency of each task, For the The communication delay of each task, is the total number of tasks; Energy consumption balance optimization calculation is performed on the energy consumption target to obtain energy consumption balance optimization data, where the energy consumption balance optimization calculation formula is as follows: In the formula, is the overall energy consumption after optimization, For the The computing energy consumption of each module is For the The communication energy consumption of each module is and It is a module The weight coefficient of computing energy consumption and communication energy consumption, is the total number of modules; The security risk suppression optimization calculation is performed on the security risk target to obtain the security risk suppression optimization data, wherein the security risk suppression optimization calculation formula is as follows: In the formula, For optimized security risks, For the The vulnerability risk of each task, For the The invasion risk of each task, and is the weight coefficient of vulnerability risk and intrusion risk, is the total number of tasks; According to the latency minimization optimization data, energy consumption balancing optimization data, and security risk suppression optimization data, the resource scheduling log is used to schedule tasks and generate a task scheduling strategy. Step S4: Collect node scheduling process data for the final scheduling instruction set to obtain task execution data; detect abnormal execution types in the task execution data, and dynamically adjust the task scheduling strategy, generate a task execution report and a security audit log and synchronize them to the emergency command center.

2. The edge computing scheduling method for heterogeneous multi-source sensors according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: collecting raw data using the heterogeneous sensors built into the smart helmet; Step S12: preprocessing the raw data to generate a standard heterogeneous data set, wherein the preprocessing includes data cleaning, data denoising, missing value filling and data standardization; Step S13: marking the acquisition timestamp of the standard heterogeneous data set; extracting the device identification of the standard heterogeneous data set; performing data hash encryption on the standard heterogeneous data set based on the device identification to generate an encrypted heterogeneous data set; Step S14: dividing the encrypted heterogeneous data set into task types to generate heterogeneous data task types; Step S15: sorting heterogeneous data task types by collecting timestamps to generate a standardized task queue.

3. The edge computing scheduling method for heterogeneous multi-source sensors according to claim 1 is characterized in that: Step S2 includes the following steps: Step S21: deploy edge computing nodes; Step S22: Perform node status monitoring on the edge computing node to generate edge computing node status data, wherein the node status monitoring includes computing load, storage margin, and network bandwidth status; Step S23: construct a resource dynamic perception matrix based on the edge computing node status data; calculate the task priority of the standardized task queue, and adjust the task order of the standardized task queue according to the task priority to generate a dynamic priority queue; Step S24: Perform resource allocation security isolation on the dynamic priority queue through the trusted execution environment and generate a resource scheduling log.

4. The edge computing scheduling method for heterogeneous multi-source sensors according to claim 3 is characterized in that: Step S24 includes the following steps: Step S241: Initialize the trusted execution environment: the secure memory size is 512MB to 16GB, the encrypted storage space ranges from 256MB to 4TB, the number of CPU trusted execution cores is between 1 and 8 cores, the TEE OS startup time is 50 to 500ms, the encryption algorithm throughput can reach 100 to 5000Mbps, the remote authentication time is 10 to 200ms, the trusted measurement time is 5 to 100ms, the key management module storage capacity is between 128KB and 32MB, and the maximum number of concurrent trusted applications can support 10 to 1000; Step S242: verifying the integrity of the trusted execution environment and generating environment verification data; performing security policy loading on the environment verification data and generating policy configuration data; performing resource boundary division on the policy configuration data and generating isolation domain data; Step S243: extracting the task attributes of the dynamic priority queue, and performing queue scheduling on the dynamic priority queue to generate scheduling sequence data; performing resource demand analysis on the scheduling sequence data to generate resource mapping data; Step S244: Use the isolation domain data to perform resource security isolation and allocation on the resource mapping data, and generate a resource scheduling log.

5. The edge computing scheduling method for heterogeneous multi-source sensors according to claim 1 is characterized in that: The trusted verification of the task scheduling strategy in step S3 includes: Extract the source code and configuration files of the task scheduling strategy for static analysis, including verifying the logical correctness of the task scheduling algorithm, policy constraints and security control measures, and generating static verification data for the task scheduling strategy; Use the trusted execution environment to dynamically verify the task scheduling strategy, simulate the scheduling process, monitor the behavior and resource usage of the task scheduling strategy in the real operating environment, and generate dynamic verification data for the task scheduling strategy; Integrate the static verification data of task scheduling strategy and the dynamic verification data of task scheduling strategy into a trusted verification result.

6. The edge computing scheduling method for heterogeneous multi-source sensors according to claim 1 is characterized in that: Deploying the task scheduling strategy to the target edge node according to the result of the trusted verification in step S3 includes: According to the results of trusted verification, the task scheduling strategy is deployed to the target edge node to detect the node status and generate node status data; Plan the deployment path for the node status data and generate deployment plan data; Perform network connection test on node status data through deployment plan data to generate network quality data; Establish a secure channel for network quality data and generate channel configuration data; The channel configuration data is used to distribute task instructions to the deployment plan data and generate scheduling instruction data.

7. The edge computing scheduling method for heterogeneous multi-source sensors according to claim 1 is characterized in that: Step S4 includes the following steps: Step S41: collecting node scheduling process data for the final scheduling instruction set, including real-time recording of task execution time, resource consumption, task status and node response information, and generating task execution data; Step S42: performing abnormal execution type detection on the task execution data to generate abnormal task execution data; Step S43: According to the abnormal task execution data, the task scheduling strategy is used to allocate task computing power, generate a task execution report and a security audit log, and synchronize them to the emergency command center.

8. The edge computing scheduling method for heterogeneous multi-source sensors according to claim 7 is characterized in that: Step S42 includes the following steps: Step S421: When any of the following situations occurs, it is determined as a task timeout exception and task timeout exception data is obtained: the task execution time deviates from the predetermined maximum execution time by more than ±10%, or the task fails to be completed within the specified time; Step S422: when any of the following situations occurs, it is determined as a resource over-limit exception, and resource over-limit exception data is obtained: CPU usage exceeds 95%, memory usage exceeds 90%, or storage usage exceeds 85%; Step S423: When the following conditions occur at the same time, it is determined as a task failure exception, and task failure exception data is obtained: the number of task failures exceeds 3 times in a row, and the number of retries after each failure does not exceed 2 times; Step S424: When the following conditions occur at the same time, it is determined that the node response delay is abnormal, and the node response delay abnormal data is obtained: the node response time exceeds 5 seconds continuously, and the duration of the node non-response exceeds 5 minutes; Step S425: Integrate task timeout exception data, resource excess exception data, task failure exception data and node response delay exception data to obtain abnormal task execution data.

9. An edge computing scheduling system for heterogeneous multi-source sensors, characterized in that: Used to execute the edge computing scheduling method for heterogeneous multi-source sensors as claimed in claim 1, the edge computing scheduling system for heterogeneous multi-source sensors includes: The data acquisition module is used to collect raw data using the heterogeneous sensors built into the smart helmet; generate data fingerprints for the raw data to obtain encrypted heterogeneous data sets; divide the heterogeneous data sets into task types to generate a standardized task queue; The task allocation module is used to monitor the node status of edge computing nodes and build a resource dynamic perception matrix; according to the resource dynamic perception matrix, the standardized task queue is allocated and safely isolated, and a resource scheduling log is generated; The resource scheduling module is used to input the resource scheduling log into the AI ​​edge computing security coprocessor for multi-objective optimization and generate a task scheduling strategy; perform trusted verification on the task scheduling strategy, and deploy the task scheduling strategy to the target edge node based on the trusted verification results to generate the final scheduling instruction set; The anomaly detection module is used to collect node scheduling process data for the final scheduling instruction set to obtain task execution data; detect abnormal execution types in task execution data, and dynamically adjust the task scheduling strategy, generate task execution reports and security audit logs and synchronize them to the emergency command center.

Images