Multi-terminal collaborative personnel training and performance measurement system based on end-side cloud network architecture

Abstract

The invention belongs to the technical field of intelligent data processing and personnel performance evaluation, and discloses a multi-terminal collaborative personnel training and performance measurement system based on an end-side cloud network architecture. The system comprises a terminal layer, an edge computing layer, a cloud and a network layer. The terminal layer collects multi-source data such as operation behaviors, physiological statuses and environments of personnel training; the edge calculation layer pre-processes the data in real time through a lightweight model and extracts key feature sub-information; the cloud fuses and analyzes the feature data, and constructs a performance evaluation model; and the network layer realizes communication of each layer. According to the invention, the problems of insufficient multi-terminal collaboration, limited data processing and single performance measurement in the prior art are solved, efficient collaborative management and accurate evaluation in the training process are realized, and the training quality and effect of personnel are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent data processing and personnel performance evaluation, and in particular to a multi-terminal collaborative personnel training and performance measurement system based on an end-edge cloud network architecture. Background Art

[0002] Existing personnel training and performance measurement technologies face numerous pressing challenges. For one thing, multi-terminal collaboration in training systems is poor. In traditional training systems, different terminal devices often operate independently, preventing real-time data sharing and collaborative processing. For example, in corporate training scenarios, employees use PCs to learn theoretical courses and mobile devices to record practical exercises. However, the lack of an effective data synchronization mechanism between PCs and mobile devices leads to a disconnected training process and an inability to form a complete training data chain, severely impacting training effectiveness and efficiency.

[0003] On the other hand, data processing capabilities are limited. Most existing systems use a single data processing model. Relying solely on cloud-based processing, when faced with large-scale multimodal data (such as data containing video, physiological signals, and operational behavior), network transmission latency is high, making it unable to meet the high real-time requirements of training scenarios. Relying solely on local processing, on the other hand, lacks sufficient computing resources on the device, making it difficult to conduct in-depth analysis of complex data. For example, in some online practical training, there is a significant delay when the student's operation video data is transmitted to the cloud for analysis, resulting in the inability to provide timely feedback to the student, affecting the quality of the training.

[0004] Furthermore, performance measurement methods are limited. Existing performance evaluations mostly consider only a few dimensions, such as task completion results, overlooking important factors such as behavioral norms, physiological state, and teamwork during training. In team training programs, this single-minded approach fails to accurately measure individual contributions to overall team performance, nor does it identify individual strengths and weaknesses during training, hindering the development and optimization of personalized training programs.

[0005] To sum up, how to design a system that can achieve efficient multi-terminal collaboration, has powerful data processing capabilities, and comprehensive and accurate performance measurement functions has become a key technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] In order to overcome the above-mentioned existing defects, the purpose of the present invention is to provide a multi-terminal collaborative personnel training and performance measurement system based on the end-edge cloud network architecture. Through innovative architectural design and technical application, it solves the problems of insufficient multi-terminal collaboration, limited data processing and incomplete performance measurement in the existing technology, realizes efficient collaborative management and accurate performance evaluation of the personnel training process, and improves the quality and effect of personnel training.

[0007] The present invention discloses a multi-terminal collaborative personnel training and performance measurement system based on a device-edge-cloud-network architecture, comprising:

[0008] The terminal layer includes various types of terminal devices, which are used to collect multi-source data during personnel training. The multi-source data includes operational behavior information, physiological status information, and environmental information;

[0009] The edge computing layer is deployed with edge computing nodes, which are used to perform real-time preprocessing and analysis of the operational behavior information, physiological status information, and environmental information collected by the terminal layer to extract the key feature sub-information from each information. The edge computing nodes extract the key feature sub-information from each information based on a lightweight model. The computational complexity of the lightweight model satisfies the formula:

[0010]

[0011] Where C represents the computational complexity, n is the number of computational steps of the lightweight model, Expressed as the coefficient of each i-step calculation, Expressed as a function of the i-th computation step, Represents the input information of this step;

[0012] The cloud, which is equipped with computing and storage capabilities, receives key feature sub-information extracted by the edge computing layer, performs fusion analysis and model training, and then builds a personnel performance evaluation model.

[0013] The network layer enables communication between the terminal layer, edge computing layer and the cloud.

[0014] Preferably, the terminal devices of the terminal layer include mobile terminals, PC terminals, wearable devices and special training equipment, and each terminal device has a unique identification mark for distinguishing and associating various information in the system.

[0015] Preferably, the edge computing nodes of the edge computing layer are deployed in a distributed manner, distributed at the edge of the network close to the terminal devices, and adjacent edge computing nodes are connected by high-speed data network links to form an edge computing network.

[0016] Preferably, when constructing a personnel performance evaluation model on the cloud, a multi-dimensional evaluation index system is adopted. The multi-dimensional evaluation index system includes at least task completion, operational standardization, behavioral efficiency, physiological fatigue, and teamwork contribution. The weight of each evaluation index is determined by a hierarchical analysis algorithm. The weight calculation formula is:

[0017]

[0018] in, It is expressed as the weight of the kth evaluation indicator, p is the number of people participating in the evaluation, q is the total number of evaluation indicators, Score the importance of the kth indicator for the jth person.

[0019] Preferably, after the cloud receives the key feature sub-information extracted by the edge computing layer, it builds a personnel performance evaluation model based on the federated learning algorithm.

[0020] Preferably, the network layer is based on hybrid transmission of 5G communication and fiber optic network, 5G communication is used for wireless communication between terminal devices and edge computing nodes, and fiber optic network is used for wired communication between edge computing nodes and the cloud and between edge computing nodes.

[0021] Preferably, it also includes a data security module, which is deployed at the terminal layer, edge computing layer and cloud layer to encrypt information at the terminal layer, edge computing layer and cloud layer.

[0022] Preferably, the edge computing nodes of the edge computing layer pre-process the operation behavior information, physiological state information and environmental information, including cleaning, format conversion and normalization processing.

[0023] Compared with the existing technology, the above technical solution has the following beneficial effects:

[0024] 1. Through innovative architecture design and technical applications, this invention addresses the existing issues of insufficient multi-terminal collaboration, limited data processing, and incomplete performance measurement. It enables efficient collaborative management of the personnel training process and accurate performance evaluation, improving the quality and effectiveness of personnel training.

[0025] 2. This invention achieves real-time sharing and precise association of training information through the collaborative operation of multiple types of terminal devices and unique identification design, solving the problem of disconnected data between multiple terminals in traditional systems and significantly improving the consistency and efficiency of the training process.

[0026] 3. This invention utilizes a lightweight model and distributed deployment at the edge computing layer to perform data preprocessing and key feature sub-information close to the terminal device, reducing network transmission pressure and improving the real-time performance of data processing. Combined with the deep analysis capabilities of the cloud, it enables efficient processing and value mining of large-scale multimodal data.

[0027] 4. Based on a multi-dimensional evaluation indicator system and a scientific weight calculation method, it comprehensively evaluates personnel training performance from multiple perspectives, avoiding the limitations of a single evaluation method. The real-time feedback mechanism and personalized training plan development function can adjust training strategies according to personnel performance in a timely manner, achieving precise training and effectively improving the quality and effectiveness of personnel training.

[0028] 5. The setting of the data security module encrypts the information of the terminal layer, edge computing layer and cloud layer throughout the entire process, ensuring the security during the collection, transmission and storage process, and preventing leakage and tampering. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a flow chart of a multi-terminal collaborative personnel training and performance measurement system based on an end-edge cloud network architecture according to the present invention. DETAILED DESCRIPTION

[0030] The advantages of the present invention are further described below with reference to the accompanying drawings and specific embodiments.

[0031] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.

[0032] The terms used in this disclosure are for the purpose of describing specific embodiments only and are not intended to limit the disclosure. As used in this disclosure and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0033] It should be understood that although the terms first, second, third, etc. may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining."

[0034] In the description of the present invention, it should be understood that the terms "longitudinal", "transverse", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention.

[0035] In the description of the present invention, unless otherwise specified and limited, it should be noted that the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a mechanical connection or an electrical connection, or it can be the internal communication between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0036] In the following description, the suffixes such as "module", "component" or "unit" used to represent elements are only used to facilitate the description of the present invention and have no specific meaning. Therefore, "module" and "component" can be used interchangeably.

[0037] See Figure 1 As shown, in this embodiment, a multi-terminal collaborative personnel training and performance measurement system based on an end-edge cloud network architecture will be described in detail.

[0038] Example 1: Skills training for employees of a manufacturing company

[0039] 1. System deployment:

[0040] Edge computing nodes are planned and deployed in distributed fashion across the company's workshops, based on equipment distribution and employee workspaces. Each edge computing node is equipped with high-performance edge computing equipment, such as industrial-grade edge servers with multi-core processors, large memory capacities, and high-speed storage, to ensure efficient processing of data uploaded by surrounding terminal devices. Edge computing nodes are connected via high-speed fiber optic networks, creating a low-latency, high-bandwidth edge computing network that ensures rapid data transmission between nodes.

[0041] Build an enterprise private cloud platform and deploy multiple high-performance servers to form a server cluster, providing computing and storage capabilities. Redundant servers ensure information security and stable system operation.

[0042] At the network layer, the enterprise campus is fully covered with 5G communication networks, and the base station layout is optimized to ensure signal strength and stability. At the same time, the internal fiber optic network is upgraded to increase network bandwidth, enabling stable and high-speed data transmission between edge computing nodes and the cloud, as well as between each node.

[0043] Employees are equipped with a variety of terminal devices. Customized dedicated operation recording terminals are installed next to the workshop operating equipment. They are durable, easy to operate, and can accurately record operation steps, duration, frequency, and other data. Employees uniformly wear professional-grade smart bracelets that can monitor physiological status data such as heart rate, blood pressure, exercise steps, and exercise intensity in real time. They are also waterproof, dustproof, and have long battery life. Employees are also equipped with high-performance mobile smart terminals to receive training notifications, view detailed learning materials, and take photos or videos of problems and feedback during practical operations. In the office area, employees use high-performance PC terminals to study theoretical courses, conduct online tests, and participate in online training discussions. All terminal devices are given unique and easy-to-identify and manage identifiers. For example, the operation recording terminal number uses the format of "workshop number-equipment number-terminal serial number", and the smart bracelet number is associated with the employee's work number, making it easy for the system to accurately distinguish and associate various types of data.

[0044] 2. Data collection and processing:

[0045] When employees conduct skills training, various terminal devices are activated synchronously to collect multi-source data in real time. For example, when employees operate high-precision machine tools, dedicated operation recording terminals connect to the machine control system via sensors, accurately recording every operation step, including tool changes, parameter adjustments, and machining starts and stops. Operation duration is accurate to the millisecond, and operation frequency is determined through real-time monitoring and statistics of operation instructions. Smart wristbands use multiple sensors, such as photoelectric sensors to monitor heart rate and accelerometers to monitor exercise intensity, to collect data multiple times per second. This data is rapidly transmitted to nearby edge computing nodes via wireless communication technologies such as Bluetooth. Environmental sensors are also deployed throughout the workshop to collect real-time environmental data such as temperature, humidity, noise, and air quality to ensure that environmental factors do not adversely affect employee training and equipment operation. This raw data is transmitted to edge computing nodes via the 5G network with extremely low latency.

[0046] After receiving the raw data, the edge computing node immediately initiates real-time preprocessing. Data cleaning begins, using preset algorithms and thresholds to identify and remove noise and outliers caused by equipment failures, electromagnetic interference, and other factors. For example, for operation log data, if the duration or frequency of an operation clearly exceeds a reasonable range, the system automatically marks it and verifies it. Next, format conversion is performed, unifying the heterogeneous data collected by different terminal devices into a standard JSON format to facilitate subsequent processing and transmission. Data is also normalized. For example, heart rate data is normalized to the [0, 1] interval based on the normal heart rate range, and operation duration data is normalized by comparing it with the standard operation duration.

[0047] Then, the key feature sub-information is extracted using a lightweight model based on deep learning. Assume that the lightweight model contains 3 calculation steps (n=3), and the coefficient of the first calculation step is ,function To characterize the sequence of operation steps, input data For an operation step sequence, such as "start the equipment → adjust parameters → start processing", the feature vector is obtained after encoding. By learning a large amount of historical operation step data, the model can identify the characteristic pattern of the standard operation step sequence and match and extract features of the current operation step sequence. The coefficient of the second calculation step ,function To calculate the normalized value of the operation time, input data The operation time is compared with the standard time to calculate the normalized value to determine whether the operation time is reasonable. The coefficient of the third calculation step is ,function To analyze the changing trend of operation frequency, input data For operation frequency data, the sliding window algorithm is used to analyze the frequency changes. If the operation frequency fluctuates abnormally in a short period of time, it may indicate that there is a problem with the employee's operation or the equipment is operating abnormally. The computational complexity of the calculation model is optimized to ensure accuracy while rationally controlling the resource consumption of edge computing nodes. Key feature sub-information, such as the standardization of operational actions and operational proficiency, is quickly extracted. The processed key feature sub-information is uploaded to the cloud via a fiber optic network in a high-speed and stable manner.

[0048] 3. Performance evaluation and feedback:

[0049] After receiving the key feature sub-information uploaded by the edge computing layer, the cloud uses the constructed multi-dimensional evaluation index system to build a personnel performance evaluation model. The weight of each evaluation index is determined by the hierarchical analysis method. A professional evaluation team consisting of five people (p=5) including experienced training experts, skilled workshop supervisors, and process engineers familiar with the production process are invited to conduct a comprehensive and detailed importance rating of the five evaluation indicators (q=5): task completion, operational standardization, behavioral efficiency, physiological fatigue, and team collaboration contribution. For example, expert 1 scores the importance of task completion as 8 points, that is, The importance of operational standardization is rated as 7 points, i.e. , and so on, according to the calculation formula Calculate the weight of each indicator, assuming the weight of task completion after calculation , the weight of operational norms Then, using a federated learning algorithm, while strictly protecting employee personal data privacy, it integrates multi-source data from different employees and workshops for model training, continuously optimizing the accuracy and adaptability of the performance evaluation model. For example, through the federated learning algorithm, operational data from employees in the same position across different workshops can be jointly analyzed to uncover both differences and commonalities in operational processes and performance across workshops, providing a basis for developing more targeted training programs.

[0050] During employee training, the cloud accurately assesses employees' real-time performance based on a built-in performance evaluation model. If an employee's operating procedures are found to be incorrect, their operating efficiency is low, or they are experiencing high levels of physical fatigue, the evaluation results are immediately fed back to the employee's mobile smart terminal or operation recording terminal through the network layer. For example, if the system detects that an employee has omitted an operating step when operating a machine tool, a prominent prompt box will pop up on the operation recording terminal interface, stating "Incorrect operating steps, please check," along with instructions for the correct operating steps. If an employee's physical fatigue is detected to be high, such as a heart rate that continues to exceed the normal range, the smart bracelet will vibrate to remind them, and the mobile smart terminal will push rest suggestions and relaxation methods. Employees can adjust their operating methods and training rhythm in a timely manner based on this feedback.

[0051] At the same time, the system uses machine learning algorithms to develop personalized training plans for each employee based on the employee's historical training data and performance evaluation results. For employees who are not proficient in operating skills, special training courses for related operations are added, such as arranging more machine tool operation simulation training, and providing detailed operation guidance videos and real-time error correction feedback; for employees who are prone to fatigue, the training intensity and rest time arrangements are adjusted, the continuous working time is shortened from 2 hours to 1.5 hours, the frequency of rest is increased, and suitable fatigue relief training methods are recommended. In the subsequent training process, the system continuously monitors the real-time performance of employees, dynamically adjusts the training plan, and realizes precise training. For example, if the employee's operating proficiency is significantly improved in subsequent training, the system automatically reduces the length of the special training course and adds more challenging comprehensive training content; if the employee's physical fatigue problem is improved, the continuous working time is appropriately extended to improve training efficiency.

[0052] Example 2: School Physical Education Teaching Example

[0053] 1. System deployment:

[0054] Edge computing nodes are installed every 50 meters around the school's playground. Within the gymnasium, nodes are located in areas such as the basketball court and badminton court, based on the venue's functional zoning. Each node is equipped with a dedicated edge computing device featuring an Intel Core i5-12500H processor, 16GB of RAM, and a 512GB solid-state drive, running a customized Linux operating system to ensure efficient and stable data processing. The nodes are connected via the campus's 10G fiber optic network, creating a low-latency, high-bandwidth data transmission channel.

[0055] The school partnered with a professional cloud computing service provider to build a dedicated education cloud platform. This platform consists of a cluster of 20 high-performance servers, each equipped with a dual-socket Intel Xeon Silver 4310 processor, 128GB of memory, and a 2TB solid-state drive, providing robust computing and storage capabilities. The platform utilizes a distributed storage architecture and load balancing technology to meet the data processing needs of the school's 5,000 students undergoing online training simultaneously. Furthermore, the platform employs firewalls, intrusion detection systems, and data encryption modules to ensure the security and privacy of student data.

[0056] At the network level, the school has comprehensively upgraded its campus wireless network, deploying 200 high-performance wireless access points. This provides full 5G-WiFi 6 coverage across sports venues, including the playground and gymnasium. The network boasts peak speeds of up to 3.6 Gbps and average latency below 10ms. Furthermore, the school has partnered with local telecommunications operators to deploy 5G base stations across campus, providing a more stable and fast transmission channel for bandwidth-intensive services such as real-time video data.

[0057] Students are equipped with three types of terminal devices: smart sports bracelets, high-definition sports cameras, and tablet computers. The smart sports bracelet uses a certain brand of professional sports bracelets with a built-in PPG heart rate sensor, a three-axis accelerometer, and a GPS positioning module. It can collect heart rate, exercise steps, exercise distance, calorie consumption, exercise trajectory and other data in real time. The data collection frequency is 1 time per second for heart rate and 5 times per second for exercise steps and distance. The high-definition sports cameras are installed high up in the playground and gymnasium. They use 4K resolution, a 120° wide-angle lens, and a frame rate of 60fps to clearly capture students' exercise details. They also support H.265 video encoding to reduce data transmission volume. The tablets are pre-installed with a customized sports teaching application, through which students can receive teaching materials, participate in online tests, record training experiences, etc. All terminal devices are managed through the school's unified identity authentication system. Each device has a unique identification code and is bound to the student's student ID number to ensure accurate data correspondence.

[0058] 2. Data collection and processing:

[0059] During physical education, various terminal devices work together to collect multi-source data. For example, in basketball class, students wear smart sports bracelets to collect physiological and sports data such as heart rate and number of steps in real time. When students perform 30 minutes of high-intensity basketball training, the heart rate data sequence collected by the bracelet is (Data is collected once per second.) High-definition sports cameras capture students' dribbling, shooting, passing, and other actions from multiple angles. Tablet computers record student behavior data, including the length of time they view instructional videos and their responses to online tests.

[0060] After receiving the raw data, the edge computing node immediately starts the pre-processing process. For the heart rate data collected by the smart sports bracelet, the median filter algorithm is first used to remove the noise data. Assuming that there are abnormal values ​​in the original heart rate data (obviously beyond the normal range), after median filtering, it is corrected to a reasonable value, and then format conversion is performed to convert the heart rate data from the device-specific format to JSON format. Then normalization is performed. It is known that the normal heart rate range for the student of the same age is [60,100]. The actual heart rate value is converted to Normalized to the [0,1] interval, the calculation formula is: .

[0061] Then, the key feature sub-information is extracted using a lightweight model based on deep learning. Assume that the lightweight model contains 3 calculation steps (n=3), and the coefficient of the first calculation step is ,function To calculate the mean of the heart rate data for the past 10 time points, input data , the average heart rate value is calculated ; Coefficients for the second calculation step ,function To calculate the difference between the current heart rate value and the average heart rate value and normalize it, input data , the normalized value is calculated by comparing it with the preset heart rate fluctuation range. Assuming that the preset fluctuation range is [﹣10,10], the normalized value ; Coefficients for the third calculation step ,function To analyze the trend of heart rate difference, input data The heart rate difference data of 5 consecutive time points is analyzed by sliding window algorithm. If the heart rate difference of 3 consecutive time points is increasing, it is judged that the heart rate has an upward trend. The computational complexity of the calculation model is used to reasonably control the resource consumption of edge computing nodes while ensuring the calculation accuracy, and to quickly determine whether the student's exercise intensity is too high.

[0062] For video data captured by high-definition motion cameras, video decoding is first performed to convert the H.265 encoded video into a format that is easy to process. Then, a lightweight model based on a convolutional neural network (CNN) is used to extract motion features. Assuming that the model also contains multiple calculation steps, when calculating the accuracy of the motion, the video image is input into the model and compared with the standard motion template. A similar calculation method is used to calculate the accuracy of the motion. This model efficiently processes data while minimizing computational complexity. For example, when analyzing a shooting motion, the model extracts key features such as hand position and arm angle, compares them with those of a standard shooting motion, calculates the similarity of the motion, and thus assesses the accuracy of the shot. The processed key feature sub-information is uploaded to the cloud, with data upload latency under 100ms.

[0063] 3. Performance evaluation and feedback:

[0064] After receiving key feature sub-information uploaded by the edge computing layer, the cloud constructs a student physical education performance evaluation model using a multi-dimensional evaluation indicator system. The evaluation indicators include four dimensions: physical skill mastery, physical fitness performance, classroom participation, and degree of athletic improvement. An evaluation panel of five physical education teachers and physical education experts was invited to rate the importance of each indicator.

[0065] Expert 1 rated the importance of mastering sports skills as 8 points, i.e. ; The importance of sports physical performance was rated as 7 points, i.e. ; The importance of class participation was rated 6 points, i.e. ; The importance of the degree of exercise progress was rated as 7 points, i.e. Expert 2's scores for each indicator are , , , , and so on, to get the complete scoring matrix .

[0066] According to the formula Calculate the weight of each indicator. First calculate the numerator, for the degree of mastery of sports skills, ; For sports performance, ; Regarding class participation, ; For the extent of exercise progress, . Then calculate the denominator, The weight of the degree of mastery of sports skills is , the weight of sports physical performance , the weight of class participation , the weight of the movement progress .

[0067] During the teaching process, the cloud evaluates students' real-time performance based on a built-in evaluation model. If the system detects a low shooting accuracy score, below 80% of the class average, the evaluation results are immediately fed back to the teacher's teaching terminal and tablet. A prompt window will pop up on the teacher's teaching terminal, displaying the student's name, a video clip of the shooting action, the accuracy score, and a comparison analysis with the standard action. Based on this feedback, the teacher can provide timely on-site guidance to the student, such as correcting the student's shooting gesture and force generation method.

[0068] At the same time, the system develops a personalized physical education learning plan for each student based on the student's historical learning data and performance evaluation results. For students with weaker physical skills, such as students with a shooting percentage of less than 30%, the system recommends targeted training courses, including a 30-minute video tutorial of fixed-point shooting exercises every day, and sets daily check-in tasks. For students whose physical fitness needs to be improved, such as students with a longer average heart rate recovery time, a plan of aerobic jogging training three times a week, 40 minutes each time, is formulated, and simple stretching exercises are recommended. In subsequent teaching, the system continues to monitor changes in students' performance and dynamically adjusts the learning plan. For example, if a student's shooting percentage increases to 40% after two weeks of training, the system automatically adjusts the training plan and adds mobile shooting training content to help students continuously improve their physical education learning results.

[0069] Example 3: Military Physical Training Example

[0070] 1. System deployment:

[0071] At military training bases, edge computing nodes are deployed in tactical training areas, physical training areas, and field training areas, based on the training area's functions and topographical distribution. Each node utilizes a ruggedized industrial-grade server with an IP67 rating, capable of stable operation in extreme temperatures ranging from -40°C to 70°C. The server is equipped with an Intel Xeon Platinum 8380 processor, 256GB of RAM, a 4TB solid-state drive, and a real-time operating system, ensuring timely and reliable data processing. Nodes are connected via a dedicated military fiber optic network and wireless ad hoc networks, forming a redundant data transmission network that ensures stable data transmission in complex electromagnetic environments.

[0072] The military has built a highly secure military cloud platform within its internal data centers. This platform, comprised of a cluster of 50 supercomputers, utilizes quantum encryption technology to ensure data security. With a computing power of 100 petaflops per second and petabyte-level storage capacity, it can handle the simultaneous training data processing and analysis needs of tens of thousands of soldiers. Multi-level access control mechanisms and a data desensitization system ensure that only authorized personnel can access and process relevant data.

[0073] At the network layer, a combination of 5G, satellite, and military-specific wireless communication technologies is utilized. Within the training base, the 5G network provides full regional coverage. Using millimeter-wave frequencies, it boasts peak speeds of up to 10Gbps and average latency below 5ms, meeting the demands of real-time HD video transmission and rapid command interaction. For remote areas such as field training, portable satellite communication terminals are deployed, enabling data transmission via military satellites, ensuring soldiers can maintain unimpeded communication with the command center in any environment.

[0074] Soldiers are equipped with three types of terminal devices: smart combat wristbands, high-definition tactical cameras, and portable military terminals. In addition to standard heart rate, blood oxygen, and motion data collection functions, the smart combat wristbands also integrate Beidou positioning modules, air pressure sensors, and tactical communication modules, enabling real-time acquisition of soldiers' geographic location and altitude information, and supporting short-range encrypted communication. The high-definition tactical camera, mounted on the soldier's helmet, features a 12-megapixel, 130° wide-angle lens, a 120fps frame rate, infrared night vision, and image stabilization, enabling clear recording of detailed movements during training and combat. The portable military terminal features a ruggedized design that is waterproof, dustproof, and drop-resistant, and runs a customized military training management system. Soldiers can use it to receive training missions, access military materials, and upload training data. All terminal devices utilize national encryption algorithms for data encryption, and the device identifier is strictly linked to the soldier's military registration information to ensure data security and traceability.

[0075] 2. Data collection and processing:

[0076] During physical training, various terminal devices comprehensively collect multi-source data of soldiers. For example, during field weight-bearing marching training, smart combat bracelets collect real-time data such as the soldier's heart rate, blood oxygen saturation, number of steps, distance traveled, marching speed, and geographic location. Assume that during a 3-hour march, the heart rate data sequence of a soldier is (collected once per second), the blood oxygen saturation data sequence is [ High-definition tactical cameras record soldiers' marching postures, tactical movements, and other video footage. Portable military terminals record data such as soldiers' receipt and completion of training tasks, and the time and content of their access to military materials.

[0077] After receiving the raw data, the edge computing node performs strict data cleaning and preprocessing. For the heart rate and blood oxygen saturation data collected by the smart combat bracelet, the Kalman filter algorithm is used to remove noise and outliers. For example, if the heart rate value at a certain moment is , which is obviously beyond the normal range. After Kalman filtering, it is corrected to a reasonable value. The data format is converted to a data format that meets military standards. Then normalization is performed. It is known that the normal heart rate range of the soldier is [60,100] and the normal range of blood oxygen saturation is [95,100]. The heart rate value is converted to Normalized to [0,1], the calculation formula is ; The blood oxygen saturation value Normalized to [0,1], the calculation formula is .

[0078] Then, the lightweight model deployed on the edge computing node is used to extract key feature sub-information. For example, for motion data, the model is used to analyze the soldier's physical energy consumption and recovery ability. Assume that the lightweight model contains 4 calculation steps (n=4), and the coefficient of the first calculation step is ,function Integrate the movement distance data within a period of time (such as 30 minutes) and input the data The movement distance sequence recorded per second during this period , get the total movement distance ; Coefficients for the second calculation step ,function To calculate the theoretical physical energy consumption value based on the total movement distance D and the soldier's weight m (assuming 70 kg), the empirical formula is used: , input data The total movement distance D and body weight m are used to obtain the theoretical energy consumption value E; the coefficient of the third calculation step is ,function To compare the actual heart rate data with the heart rate range corresponding to the theoretical physical energy consumption, calculate the heart rate deviation value. It is known that the heart rate range corresponding to the theoretical physical energy consumption E is , input data Actual heart rate data , then the heart rate deviation value ; Coefficients for the fourth calculation step ,function To analyze the trend of heart rate deviation, input data The heart rate deviation data of 10 consecutive time points is analyzed by sliding window algorithm. If the heart rate deviation values ​​of 5 consecutive time points are increasing, the predicted physical recovery time will be extended. The computational complexity of the calculation model. While ensuring computational accuracy, the resource consumption of edge computing nodes is reasonably controlled to quickly determine the physical consumption and recovery status of soldiers.

[0079] For video data captured by high-definition tactical cameras, video decoding and noise reduction are first performed. Then, a lightweight model based on deep learning is used to evaluate whether the soldiers' training movements meet military standards and calculate indicators such as movement standardization and accuracy. When calculating movement standardization, a convolutional neural network (CNN) is used to extract features from the video images and compare them with standard military movement templates. The calculation formula is also used. The model's computational complexity is controlled, enabling efficient data processing. For example, when analyzing tactical movements, the model extracts key features such as a soldier's movement path and body posture, compares them with the characteristics of standard tactical movements, and calculates a standardization score for the movement. The processed key feature sub-information is uploaded to the cloud, with data upload latency under 80ms.

[0080] 3. Performance evaluation and feedback:

[0081] After receiving the key feature sub-information uploaded by the edge computing layer, the cloud constructs a multi-dimensional soldier physical training performance evaluation model. The evaluation indicators include four dimensions: physical fitness, military skills mastery, task completion efficiency, and training attitude. Six military training experts and senior instructors were invited to form an evaluation team to rate the importance of each indicator. ; The importance of mastering military skills is rated as 8 points, that is, ; The importance of task completion efficiency is rated as 7 points, that is, ; The importance of training attitude was rated as 8 points, i.e. Expert 2's scores for each indicator are , , , , and so on, to get the complete scoring matrix .

[0082] According to the formula Calculate the weight of each indicator. First calculate the numerator. For physical fitness, ; Regarding the degree of mastery of military skills, ; Task completion efficiency, , regarding training attitude, . Then calculate the denominator, The weight of physical fitness is , the weight of military skill mastery , the weight of task completion rate , the weight of training attitude .

[0083] During training, the cloud evaluates soldiers' real-time performance based on a built-in evaluation model. If the system detects that a soldier's movement standardization score during tactical movement training is below 70% of the average for soldiers in the same batch, the system immediately feeds the evaluation results back to the instructor's command terminal and the soldier's portable military terminal. The instructor's command terminal will display detailed information, including the soldier's name, company affiliation, video clips of the non-standard movement, the standardization score, and a comparative analysis report with standard tactical movement movements. The soldier's portable military terminal will vibrate as a reminder and display brief improvement suggestions, such as "Pay attention to lowering your center of gravity and increasing your movement speed." Instructors can provide timely one-on-one corrective guidance to soldiers based on this feedback, helping them improve their movements through on-site demonstrations and explanations.

[0084] At the same time, the system develops personalized training plans for each soldier based on their historical training data and performance evaluation results. For soldiers with weaker physical fitness, such as those who perform poorly in 3,000-meter run training, the system develops targeted physical training plans, including three additional long-distance runs per week, gradually increasing the distance from 3,500 to 5,000 meters each time. This is combined with core strength training and daily 20-minute exercises such as planks and sit-ups. For soldiers whose military skills need improvement, such as those lacking proficiency in weapon operation, specialized skills training courses are arranged. A virtual simulation training environment is provided, allowing soldiers to repeatedly practice weapon disassembly, assembly, and firing operations in simulated scenarios. Training sessions last at least two hours per day, with professional instructors providing real-time guidance and commentary.

[0085] During subsequent training, the system continuously monitors changes in soldiers' performance and dynamically adjusts training plans. For example, if a soldier improves their 3,000-meter run time by 30 seconds or their weapon handling proficiency score rises from 60 to 75 after a month of training, the system will automatically adjust the training plan to further increase the intensity of long-distance running training to 6,000 meters. Furthermore, the system will add more complex simulated combat scenarios to weapon handling training, such as nighttime weapon handling training, and set up various emergency fault scenarios, requiring soldiers to quickly troubleshoot and complete shooting missions, in order to continuously improve soldiers' military training levels.

[0086] It should be noted that the embodiments of the present invention have better practicability and do not impose any form of limitation on the present invention. Any technician familiar with the field may use the technical content disclosed above to change or modify it into an equivalent effective embodiment. However, any modification or equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A multi-device collaborative personnel training and performance measurement system based on a device-edge-cloud-network architecture, characterized by: include: The terminal layer includes various types of terminal devices, which are used to collect multi-source data during personnel training, including operational behavior information, physiological status information, and environmental information; The edge computing layer is deployed with edge computing nodes for performing real-time preprocessing and analysis on the operation behavior information, the physiological state information, and the environmental information collected by the terminal layer to extract key feature sub-information from each piece of information. The edge computing nodes extract the key feature sub-information from each piece of information based on a lightweight model, and the computational complexity of the lightweight model satisfies the formula: Where C represents the computational complexity, n is the number of computational steps of the lightweight model, Expressed as the coefficient of each i-step calculation, Expressed as a function of the i-th computation step, Represents the input information of this step; The cloud is deployed with computing and storage capabilities, and is used to receive the key feature sub-information extracted by the edge computing layer, and to build a personnel performance evaluation model after fusion analysis and model training; The network layer is used to implement communication between the terminal layer, the edge computing layer and the cloud.

2. The multi-terminal collaborative personnel training and performance measurement system based on the device-edge-cloud-network architecture according to claim 1 is characterized in that: The terminal devices of the terminal layer include mobile terminals, PC terminals, wearable devices and special training equipment, and each terminal device has a unique identification mark for distinguishing and associating various information in the system.

3. The multi-terminal collaborative personnel training and performance measurement system based on the device-edge-cloud-network architecture according to claim 2 is characterized in that: The edge computing nodes of the edge computing layer are deployed in a distributed manner and are distributed at the edge of the network close to the terminal devices. Adjacent edge computing nodes are connected by high-speed data network links to form an edge computing network.

4. The multi-terminal collaborative personnel training and performance measurement system based on the device-edge-cloud-network architecture according to claim 3 is characterized in that: When the cloud constructs the personnel performance evaluation model, a multi-dimensional evaluation index system is adopted. The multi-dimensional evaluation index system includes at least task completion, operational standardization, behavioral efficiency, physiological fatigue, and teamwork contribution. The weight of each evaluation index is determined by a hierarchical analysis algorithm. The weight calculation formula is: in, It is expressed as the weight of the kth evaluation indicator, p is the number of people participating in the evaluation, q is the total number of evaluation indicators, Score the importance of the kth indicator for the jth person.

5. The multi-terminal collaborative personnel training and performance measurement system based on the device-edge-cloud-network architecture according to claim 4 is characterized in that: After the cloud receives the key feature sub-information extracted by the edge computing layer, it constructs the personnel performance evaluation model based on the federated learning algorithm.

6. The multi-terminal collaborative personnel training and performance measurement system based on the device-edge-cloud-network architecture according to claim 5 is characterized in that: The network layer is based on the hybrid transmission of 5G communication and optical fiber network. The 5G communication is used for wireless communication between the terminal device and the edge computing node, and the optical fiber network is used for wired communication between the edge computing node and the cloud and between each edge computing node.

7. The multi-terminal collaborative personnel training and performance measurement system based on the device-edge-cloud-network architecture according to claim 1 is characterized in that: It also includes a data security module, which is deployed at the terminal layer, the edge computing layer and the cloud, and is used to encrypt information at the terminal layer, the edge computing layer and the cloud.

8. The multi-terminal collaborative personnel training and performance measurement system based on the device-edge-cloud-network architecture according to claim 1 is characterized in that: The edge computing node of the edge computing layer pre-processes the operation behavior information, the physiological state information and the environmental information, including cleaning, format conversion and normalization processing.

Images