Multi-device cooperative control system and method of a soundproof cabin

Abstract

The application discloses a multi-device cooperative control system and method of a silence cabin, and relates to the technical field of silence cabin control. The method collects the running states of multiple devices in the silence cabin in real time, performs sliding window filtering and dimensionless standardization processing on S1 and S2, and constructs the adaptability vector Fit(i,j) of the device i and the task j in combination with historical execution data. The method can quantitatively evaluate the device resource matching degree and the task time delay tolerance before task allocation, thereby avoiding the resource mismatch problem caused by the traditional task allocation mode based on fixed rules or static priority. The mechanism makes the task scheduling based on real-time state data and historical execution feedback, improves the matching accuracy between the task and the device, reduces the probability of single device overload operation, and improves the overall utilization rate of multiple device resources in the silence cabin and the task execution stability.

Description

Technical Field

[0001] This invention relates to the field of soundproof cabin control technology, specifically to a multi-device collaborative control system and method for soundproof cabins. Background Technology

[0002] With the development of information processing systems and multi-tasking intelligent terminals, the demand for multi-device collaboration in silent cabin-like independent environments is gradually increasing. Especially in high-reliability task scenarios such as remote scientific research and isolated office work in sensitive environments, users often need to rely on multiple devices simultaneously to perform tasks such as inference computing, information interaction, and data processing, which places higher demands on the real-time management and task scheduling capabilities of multiple device resources within the silent cabin.

[0003] Currently, most existing soundproof cabin systems employ fixed priority mechanisms or preset task allocation rules to schedule tasks across multiple devices. This lacks real-time awareness of the devices' current operating status, especially in high-concurrency scenarios with overlapping and volatile resources. These systems cannot dynamically assess device resource load or identify task execution delays and failure risks. This "static binding scheduling mode" easily leads to some devices being over-utilized while others remain idle, resulting in low overall resource utilization, unstable task execution quality, and a lack of anomaly prediction and load balancing mechanisms.

[0004] The main reason for the above problems is that the existing in-cabin scheduling scheme has not established a resource assessment model that links task execution feedback, real-time equipment status, and environmental constraints, nor has it formed a complete dynamic adaptation and migration control strategy. When tasks are incorrectly assigned to equipment that is nearing its load limit or is operating at high temperatures, abnormal phenomena such as slow equipment response, frequent task failures, and inability to recover system resources are very likely to occur. In severe cases, it may trigger alarms in the in-cabin temperature control system, cause system crashes due to task backlog, and even lead to a chain of technical risks such as interruption of user operation experience and paralysis of cabin functions. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a multi-device collaborative control system and method for a soundproof cabin, solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution, comprising the following steps: S1. Connect multiple devices in the soundproof cabin via IoT to collect task status execution feedback parameters of multiple devices in the soundproof cabin in real time during task operation, and form a standardized dataset of device status after preprocessing the task status execution feedback parameters. S2, calculate the fit degree based on the standardized dataset, output the fit degree vector Fit(i,j) between device i and task j, and then construct the task allocation matrix and obtain the preliminary allocation result based on the Fit(i,j) vector between device i and task j. S3. Based on the preliminary allocation results, calculate the scheduling adaptability index Eval, and based on the output of the scheduling adaptability index Eval, make a judgment on the rationality of scheduling. S4, based on the scheduling rationality judgment result, triggers the control strategy result hierarchy division, and executes the relevant control strategy according to the control strategy result hierarchy division result.

[0007] Preferably, S1 includes S11; S11. By setting up collection points on multiple devices in the soundproof cabin and interconnecting them with the collection devices in the collection points through the Internet of Things, the task status execution feedback parameters of the multi-state devices in the soundproof cabin during the task operation are collected in real time, and then the collected task status execution feedback parameters are transmitted to the collaborative control server through the Internet of Things. The data collection points include device-level data collection points and task-level data collection points; The task status execution feedback parameters include status parameters and task execution feedback parameters; The state parameters include the CPU availability vector CPUava(i,t) of device i at time t, the memory remaining vector Mem(i,t) of device i at time t, and the interface temperature rise rate vector Thrtmal(i,t) of device i at time t. The task execution feedback parameters include the actual response delay vector Latency(i,j) of device i and task j and the task identification count vector Fail(i,j) of device i and task j. Among them, device-level acquisition points are set at the CPU core, memory management unit, and network interface of each device; status parameters are collected in real time through system call interfaces. The task layer collection points are set in the application service container and process instance of each device, and task execution feedback parameters are collected in real time through the task scheduling system plugin and container monitoring interface.

[0008] Preferably, S1 further includes S12; S12. Receive task status execution feedback parameters in real time in the collaborative control server, and preprocess the task status execution feedback parameters to obtain a standardized dataset of device status. The preprocessing includes sliding window filtering and dimensionless processing. The sliding window filtering uses a fixed-window-length sliding window to calculate the state parameters, perform mean smoothing over a short time interval, and suppress noise and graph compilation interference on the scheduling results. The dimensionless processing involves mapping all vectors in the task execution feedback parameters, after the state parameters have been filtered by a sliding window, to the interval [0,1]. A dimensionless device state matrix is ​​constructed through normalization, where the dimensionless method used is the maximum and minimum value normalization function. The dimensionless state parameters are then integrated with the task execution feedback parameters to obtain a standardized device state dataset.

[0009] Preferably, S2 includes S21; S21. Extract the CPU utilization rate Ecpu(i,j) and the historical actual response time LaT(i,j) of device i and task j from the database of the collaborative control server. Combine the standardized dataset of device status obtained in real time to calculate the fit degree and output the fit degree vector Fit(i,j) of device i and task j. Next, perform adaptation calculations on all devices and tasks to be assigned to form an adaptation vector set, and construct a device and task adaptation matrix based on the adaptation vector set; The fit vector Fit(i,j) between device i and task j is obtained through the following fit calculation; .

[0010] Preferably, S2 further includes S22; S22. The adaptation vector Fit(i,j) between device i and task j is evaluated, and then predictive overload triggered adaptive scheduling control is executed based on the evaluation results. The specific content is as follows: When any device i satisfies one of the following conditions: the fit vector Fit(i,j) between device i and task j is <0.4, the CPU availability vector CPUava(i,t) of device i at time t is <30%, and the historical actual response time LaT(i,j) between device i and task j is >0.8 × the actual response delay vector Latency(i,j) between device i and task j; The system determines that device i has entered a resource critical state and initiates the task migration and scheduling mechanism. The task migration scheduling mechanism writes the corresponding task j into the predicted overload task queue through the task j transfer buffer in the software layer, and reschedules task j when the target device enters the idle execution window. When the operating environment of device i supports a real-time operating system, the execution priority of the current task j on the critical device i is reduced by the task priority replacement interface. When performing task migration, the migration target device is selected from the available devices according to the migration target selection strategy. The migration target selection strategy includes avoiding device i that is operating at high temperature and giving priority to device i that has recently completed the task, thereby completing the dynamic task reallocation.

[0011] Preferably, S3 includes S31; S31. After the task migration scheduling mechanism is completed, calculate the scheduling adaptability index Eval based on the adaptation vector Fit(i,j) of all devices i and tasks j, and analyze the comprehensive scheduling adaptability of the current task allocation scheme at the system level. The scheduling adaptability index Eval is calculated and output using the following algorithm formula; In the formula, N represents the total number of tasks to be assigned, and D represents the set of available devices.

[0012] Preferably, S3 further includes S32; S32. Based on the output of the scheduling adaptability index Eval, determine the rationality of the current task migration scheduling mechanism. The specific determination is as follows: When the scheduling adaptability index Eval ≥ 0.7, it indicates that the current task migration scheduling mechanism is reasonable, and the task allocation is maintained. When the scheduling adaptability index Eval < 0.7, it indicates that the current task migration scheduling mechanism is unreasonable, and the task reallocation strategy should be initiated.

[0013] Preferably, S4 includes S41; S41. After the task reallocation strategy is triggered, the scheduling control logic is structurally divided according to the task scheduling stage, adaptability level, equipment operating status, task execution feedback, and historical failure records. Specifically, this includes: For tasks that are scheduled for the first time in the current scheduling cycle, they are classified as first-level tasks according to the first-level classification criteria and are assigned to the first-level control layer. In the case of non-first scheduling, when the fitness vector Fit(i,j) between device i and task j is less than 0.4, the scheduling path is divided and task j is assigned to the second-level control layer. Under the premise that the fit vector Fit(i,j) between device i and task j is greater than or equal to 0.4, further based on the fact that when the CPU availability vector CPUava(i,t) < 30% and the interface temperature rise rate vector Thrtmal(i,t) of device i at time t is greater than 0.6℃ / s, the corresponding task j is assigned to the third-level control layer. If device i is not currently in a critical state but there are risks of multiple concurrent tasks on the same device, concentrated task density, and structural load, then the corresponding task is classified into the fourth level control layer. If the number of failure records of task j on device i in the scheduling history exceeds a preset number of 6, it will be classified into the fifth level control layer.

[0014] Preferably, S4 further includes S42; S42. After completing the hierarchical division of the control strategy structure, execute the corresponding control strategies for each control level as follows: For the first-level control layer, the standard initial allocation strategy is executed, and task binding is completed directly based on the maximum value of the fit vector Fit(i,j) between device i and task j. For the second-level control level, the task delay binding strategy is implemented. Adaptation tasks with a fit vector Fit(i,j) < 0.4 between device i and task j are temporarily stored in the task waiting area, and devices with higher fit are prioritized. For the third-level control level, a predictive overload control strategy is implemented, and the task is transferred to the equipment whose current temperature and load are in a stable range through a migration mechanism; For the fourth level of control, a resource conflict smoothing strategy is implemented, which controls the task execution frequency by controlling the thread wake-up cycle; For the fifth level of control, a historical failure feedback avoidance strategy is implemented to restrict priority binding again.

[0015] A multi-device collaborative control system for a soundproof cabin includes a multi-device acquisition module, an adaptability analysis module, a scheduling adaptability analysis module, and a reallocation module; The multi-device acquisition module connects multiple devices in the silent cabin via the Internet of Things, and collects task status execution feedback parameters of multiple devices in the silent cabin in real time during the task operation process. After preprocessing, the task status execution feedback parameters are used to form a standardized dataset of device status. The fit analysis module calculates the fit based on the standardized dataset, outputs the fit vector Fit(i,j) between device i and task j, and then constructs a task allocation matrix and obtains preliminary allocation results based on the Fit(i,j) vector between device i and task j. The scheduling adaptability analysis module calculates the scheduling adaptability index Eval based on the preliminary allocation results, and outputs the results based on the scheduling adaptability index Eval to judge the rationality of the scheduling. The reallocation module triggers the hierarchical division of control strategy results based on the scheduling rationality judgment result, and executes relevant control strategies according to the hierarchical division result of control strategy results.

[0016] This invention provides a multi-device collaborative control system and method for a soundproof cabin. It offers the following advantages: (1) This method collects the operating status of multiple devices in the silencing chamber in real time through S1 and S2, performs sliding window filtering and dimensionless standardization, and constructs a fit vector Fit(i,j) between device i and task j by combining historical execution data. This method can quantitatively evaluate the matching degree of device resources and the tolerance of task latency before task allocation, thereby avoiding the resource mismatch problem caused by traditional task allocation methods based on fixed rules or static priorities. This mechanism enables task scheduling to be based on both real-time status data and historical execution feedback, which improves the matching accuracy between tasks and devices, reduces the probability of overload operation of a single device, and improves the overall utilization rate of resources of multiple devices in the silencing chamber and the stability of task execution.

[0017] (2) Through predictive overload triggered adaptive scheduling control in S2 and hierarchical division of control strategy structure in S4, this method can proactively identify potential overload risks before the equipment enters a resource critical state, and execute differentiated control strategies according to different risk levels, including task migration, priority adjustment, and delayed binding. This hierarchical control mechanism enables the system to no longer rely on post-event remediation, but to suppress load concentration through proactive scheduling intervention, effectively reducing performance fluctuations caused by abnormal temperature rise, CPU resource exhaustion, or task accumulation, and improving the operational reliability and continuity of the silent cabin system in high-concurrency task environments.

[0018] (3) This method combines a refined task matching mechanism based on the fitness vector Fit(i,j) with a dynamic hierarchical control strategy based on the scheduling adaptability index Eval. This method constructs a closed-loop collaborative control system of "prediction and evaluation - risk identification - hierarchical regulation", enabling the scheduling of multiple devices in the silent cabin to have both accurate matching capability and real-time adaptive capability. This closed-loop mechanism can not only achieve load balancing in the initial stage of the task, but also continuously optimize the task distribution structure during operation, reduce the risk of scheduling imbalance at the system level, and improve the overall robustness and long-term stable operation capability of multi-device collaborative work. Attached Figure Description

[0019] Figure 1 This is a schematic diagram illustrating the steps of a multi-device collaborative control method for a soundproof cabin according to the present invention; Figure 2 This is a schematic diagram of the multi-device collaborative control system for a soundproof cabin according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1: Please refer to Figure 1 This invention provides a multi-device collaborative control method for a soundproof cabin. To achieve the above objectives, this invention employs the following technical solution, comprising the following steps: S1. Connect multiple devices in the soundproof cabin via IoT to collect task status execution feedback parameters of multiple devices in the soundproof cabin in real time during task operation, and form a standardized dataset of device status after preprocessing the task status execution feedback parameters. S2, calculate the fit based on the standardized dataset, output the fit vector Fit(i,j) between device i and task j, and then construct the task allocation matrix and obtain the preliminary allocation result based on the Fit(i,j) vector between device i and task j. S3. Based on the preliminary allocation results, calculate the scheduling adaptability index Eval, and based on the output of the scheduling adaptability index Eval, make a judgment on the rationality of scheduling. S4, based on the scheduling rationality judgment result, triggers the control strategy result hierarchy division, and executes the relevant control strategy according to the control strategy result hierarchy division result.

[0022] In this embodiment, the method collects and standardizes the operating status of multiple devices within the silent cabin in real time through S1, enabling the scheduling system to obtain continuous and quantifiable device load characteristic data. This aims to avoid resource misjudgment in dynamic operating environments caused by traditional task allocation methods that rely on static configuration or empirical rules. In practical applications, without unified preprocessing, parameters such as CPU utilization and response latency collected from different devices exhibit differences in scale and fluctuation, easily leading to scheduling decisions deviating from the true load state, resulting in tasks piling up on a single device. By driving the adaptation calculation in S2 with a standardized dataset, task allocation is based on a comprehensive evaluation of the device's real-time carrying capacity and historical execution characteristics. This proactively increases the load distance between devices before tasks enter the execution phase, reducing the probability of local hotspots. Furthermore, S3 uses the scheduling adaptability index Eval to perform system-level verification of the overall allocation structure. Its physical meaning is to evaluate whether the current task layout is close to a balanced and stable state from a global perspective, rather than focusing only on the instantaneous load of a single device. When Eval falls below a threshold, it indicates a potential risk of task concentration. In this case, S4 introduces a tiered control strategy to implement differentiated interventions for tasks with different risk levels. For example, tasks on high-load devices may be migrated, or the binding of poorly compatible tasks may be delayed. This tiered response mechanism can proactively adjust for issues such as device temperature rise, computing power bottlenecks, or intensive concurrent tasks, effectively preventing response delays or task failures caused by sudden overload. Through this implementation process, the system continuously maintains a load-balanced state among multiple devices during operation, making task execution paths more stable and controllable. This improves the real-time response capability and long-term operational reliability of the silent cabin in high-concurrency application scenarios, and reduces performance fluctuations and system anomalies caused by concentrated resource consumption.

[0023] Example 2: Please refer to Figure 1 Specifically: S1 includes S11; S11. By setting up collection points on multiple devices in the soundproof cabin and interconnecting them with the collection devices in the collection points through the Internet of Things, the task status execution feedback parameters of the multi-state devices in the soundproof cabin during the task operation are collected in real time, and then the collected task status execution feedback parameters are transmitted to the collaborative control server through the Internet of Things. Data collection points include device-level data collection points and task-level data collection points; Task status execution feedback parameters include status parameters and task execution feedback parameters; The state parameters include the CPU availability vector CPUava(i,t) of device i at time t, the memory remaining vector Mem(i,t) of device i at time t, and the interface temperature rise rate vector Thrtmal(i,t) of device i at time t. The task execution feedback parameters include the actual task response delay vector Latency(i,j) for device i and task j, and the task identification count vector Fail(i,j) for device i and task j. Among them, device-level acquisition points are set at the CPU core, memory management unit, and network interface of each device; status parameters are collected in real time through system call interfaces. The task layer collection points are set in the application service containers and process instances of each device, and task execution feedback parameters are collected in real time through the task scheduling system plugin and container monitoring interface.

[0024] S1 also includes S12; S12. Receive task status execution feedback parameters in real time in the collaborative control server, and preprocess the task status execution feedback parameters to obtain a standardized dataset of device status. The preprocessing includes sliding window filtering and dimensionless processing. Sliding window filtering uses a sliding window with a fixed window length to calculate the state parameters, perform mean smoothing over a short time interval, and suppress the interference of noise and graph compilation on the scheduling results. Dimensionless processing involves mapping all vectors from the task execution feedback parameters, after sliding window filtering, to the interval [0,1]. A dimensionless device state matrix is ​​constructed through normalization, using a maximum and minimum value normalization function. The dimensionless state parameters are then integrated with the task execution feedback parameters to obtain a standardized device state dataset.

[0025] In this embodiment, method S11 collects feedback parameters such as CPU utilization, remaining memory, interface temperature rise rate, task response latency, and failure count through device and task layer collection points, and uploads the data to the collaborative control server in real time. This is to avoid the scheduling blind spot problem caused by the traditional scheduling system's reliance on static device identifiers or task tags. For example, if device performance is judged solely by device model, it is difficult to reflect the actual operating load of the device in a timely manner during peak task periods, resulting in high-frequency tasks being repeatedly scheduled to resource-constrained devices, forming scheduling hotspots. This collection mechanism enables the scheduling system to have real-time judgment capabilities for time-series loads. S12 sets up sliding window filtering and dimensionless processing, the core purpose of which is to improve the data's resilience and comparability. In real-world scenarios, even if different devices are running the same task, their resource parameters may vary significantly due to differences in architecture, cold start status, and other factors. For example, a device may have high CPU utilization but stable response; without uniform scaling, this device may be misjudged as overloaded. By using max-min normalization, all device status data are uniformly mapped to the 0-1 range, which can then be directly used for horizontal comparison in the adaptation model. This avoids adaptation bias caused by inconsistent indicator scales and effectively improves the sensitivity and robustness of the scheduling model to changes in resource status.

[0026] Example 3: Please refer to Figure 1 Specifically: S2 includes S21; S21. Extract the CPU utilization rate Ecpu(i,j) and the historical actual response time LaT(i,j) of device i and task j from the database of the collaborative control server. Combine the standardized dataset of device status obtained in real time to calculate the fit degree and output the fit degree vector Fit(i,j) of device i and task j. Next, the adaptability is calculated for all devices and tasks to be assigned, forming an adaptability vector set. Based on the adaptability vector set, a device-task adaptability matrix is ​​constructed for subsequent task allocation decisions. The fitness vector Fit(i,j) between device i and task j is obtained through the following fitness calculation; ; The formula for calculating the fit is derived from the distance function theory and the principle of resource scheduling fit function in mathematics; In common scheduling systems, distance metrics are used to represent the degree of deviation between the target value and the current state; Where |ab| is the absolute difference function, which is the basic form of Manhattan distance (L1 norm); Then, the value is normalized to the [0,1] interval by linear scaling (i.e., dividing by 2), which conforms to the normalization similarity function pattern in system scheduling; The improvement in this formula is that it incorporates two unrelated resource dimensions (CPU and response latency) into the same evaluation model and uses a weighted average for unified adaptability assessment, thereby simultaneously measuring "resource satisfaction" and "latency tolerance". in, This characterizes the load matching error between the actual CPU required by the device and the currently available CPU; the larger the deviation, the worse the adaptability. The difference between the device's responsiveness to a task and the task's latency tolerance is the greater; the closer they are, the stronger the "latency acceptability".

[0027] S2 also includes S22; S22. The adaptation vector Fit(i,j) between device i and task j is evaluated, and then predictive overload triggered adaptive scheduling control is executed based on the evaluation results. The specific content is as follows: When any device i satisfies one of the following conditions: the fit vector Fit(i,j) between device i and task j is <0.4, the CPU availability vector CPUava(i,t) of device i at time t is <30%, and the historical actual response time LaT(i,j) between device i and task j is >0.8 × the actual response delay vector Latency(i,j) between device i and task j; The system determines that device i has entered a resource critical state and initiates the task migration and scheduling mechanism. The task migration scheduling mechanism writes the corresponding task j into the predicted overload task queue through the task j transfer buffer in the software layer, and reschedules task j when the target device enters the idle execution window. When the operating environment of device i supports a real-time operating system, the execution priority of the current task j on the critical device i is reduced by the task priority replacement interface. When performing task migration, the migration target device is selected from the available devices according to the migration target selection strategy. The migration target selection strategy includes avoiding device i that is operating at high temperature and giving priority to device i that has recently completed the task, thereby completing the dynamic task reallocation.

[0028] In this embodiment, method S21 incorporates historical execution data and real-time standardized status data into the fit calculation to avoid judgment bias caused by relying solely on instantaneous monitoring data for scheduling. In actual operation, some devices may appear idle for a short period, but their historical execution records indicate slow response or sudden increases in resource consumption for specific types of tasks. If historical behavior is ignored and tasks are assigned based solely on the current load, tasks may be assigned to devices that are "instantly idle but unstable in the long term," leading to execution jitter. By calculating the fit based on a distance function, CPU load matching error and response latency difference are simultaneously incorporated into the evaluation model. This allows for a greater comprehensive fit distance between devices before task allocation, making scheduling decisions closer to the actual carrying capacity of the devices and reducing repeated migrations and execution failures caused by incorrect binding. The purpose of setting a predictive overload triggering mechanism in S22 is to intervene in advance before the device enters actual overload, rather than passively remedying it after performance degradation. For example, when the fit is already low and CPU availability continues to decline, continuing to assign tasks to the device can easily lead to load avalanche, causing a sharp increase in response latency. By combining low adaptation thresholds with CPU critical values ​​and historical response deviations, risky devices can be identified in advance, and their load pressure can be relieved through task migration and priority adjustment. This pre-scheduling method is physically equivalent to dynamically smoothing out peak and valley loads for device resources, keeping the overall system operation curve smooth, thereby significantly reducing performance fluctuations caused by sudden overloads and improving the continuous operation stability in a multi-device collaborative environment.

[0029] Example 4: Please refer to Figure 1 Specifically: S3 includes S31; S31. After the task migration scheduling mechanism is completed, calculate the scheduling adaptability index Eval based on the adaptation vector Fit(i,j) of all devices i and tasks j, and analyze the comprehensive scheduling adaptability of the current task allocation scheme at the system level. The scheduling adaptability index Eval is calculated and output using the following algorithm formula; In the formula, N represents the total number of tasks to be assigned, and D represents the set of available devices; , indicating the matching device with the highest adaptability for the j-th task among all devices.

[0030] S3 also includes S32; S32. Based on the output of the scheduling adaptability index Eval, determine the rationality of the current task migration scheduling mechanism. The specific determination is as follows: When the scheduling adaptability index Eval ≥ 0.7, it indicates that the current task migration scheduling mechanism is reasonable, and the task allocation is maintained. When the scheduling adaptability index Eval < 0.7, it indicates that the current task migration scheduling mechanism is unreasonable, and the task reallocation strategy should be initiated.

[0031] In this embodiment, the method sets up a calculation step for the scheduling adaptability index Eval in S31. Its fundamental purpose is to integrate the single-task adaptability results, originally scattered across the device layer, into a quantitative index of system-level scheduling rationality, forming a global feedback loop. In practice, multi-task scenarios may exhibit good local device adaptability but uneven overall allocation structure; such deviations cannot be identified by a single adaptability value. Eval aggregates the maximum adaptability matching value for each task, measuring the "overall scheduling fit" of the current scheme and reflecting whether the system is in a resource-coordinated state. Its physical meaning is: if the Eval value is low, it indicates that tasks are mostly assigned to relatively unsuitable devices or that device loads are unbalanced, potentially leading to task congestion, sudden temperature increases, and other hidden dangers. Setting Eval ≥ 0.7 as the threshold for judging the rationality of the scheduling mechanism in S32 is based on the empirical boundary of task success rate and system stability from numerous simulation test results, effectively identifying the scheduling "critical degradation point." The purpose of setting this judgment logic is to transform the evaluation of scheduling effect from subjective "experience judgment" to quantitative "policy switch". Once Eval is lower than the threshold, the reallocation strategy is activated to avoid the problem of "being stuck in poor matching", thereby dynamically optimizing the scheduling path and improving the real-time closed-loop performance of resource matching.

[0032] Example 5: Please refer to Figure 1 Specifically: S4 includes S41; S41. After the task reallocation strategy is triggered, the scheduling control logic is structurally divided according to the task scheduling stage, adaptability level, equipment operating status, task execution feedback, and historical failure records. Specifically, this includes: For tasks that are scheduled for the first time in the current scheduling cycle, they are classified as first-level tasks according to the first-level classification criteria and are assigned to the first-level control layer. In the case of non-first scheduling, when the fitness vector Fit(i,j) between device i and task j is less than 0.4, the scheduling path is divided and task j is assigned to the second-level control layer. Under the premise that the fit vector Fit(i,j) between device i and task j is greater than or equal to 0.4, further based on the fact that when the CPU availability vector CPUava(i,t) < 30% and the interface temperature rise rate vector Thrtmal(i,t) of device i at time t is greater than 0.6℃ / s, the corresponding task j is assigned to the third-level control layer. If device i is not currently in a critical state but there are risks of multiple concurrent tasks on the same device, concentrated task density, and structural load, then the corresponding task is classified into the fourth level control layer. If the number of failure records of task j on device i in the scheduling history exceeds a preset number of 6, it will be classified into the fifth level control layer.

[0033] S4 also includes S42; S42. After completing the hierarchical division of the control strategy structure, execute the corresponding control strategies for each control level as follows: For the first-level control layer, the standard initial allocation strategy is executed, and task binding is completed directly based on the maximum value of the fit vector Fit(i,j) between device i and task j. For the second-level control level, the task delay binding strategy is implemented. Adaptation tasks with a fit vector Fit(i,j) < 0.4 between device i and task j are temporarily stored in the task waiting area, and devices with higher fit are prioritized. For the third-level control level, a predictive overload control strategy is implemented, and the task is transferred to the equipment whose current temperature and load are in a stable range through a migration mechanism; For the fourth level of control, a resource conflict smoothing strategy is implemented. By controlling the thread wake-up cycle, the frequency of task execution is controlled to reduce concurrency impact. For the fifth level of control, a historical failure feedback avoidance strategy is implemented to restrict priority binding again.

[0034] In this embodiment, the method introduces multi-dimensional factors in S41 to classify tasks into structural levels, aiming to solve the scheduling rigidity problem caused by the traditional scheduling scheme's "uniform rules for tasks with different risk levels." Especially when task adaptability is low but not completely mismatched, or when equipment temperature rises rapidly, continuing to use the standard binding strategy can easily induce overheating and performance degradation. By setting a combination of criteria such as "interface temperature rise rate > 0.6℃ / s" and "CPU availability < 30%", the method can accurately identify when a device is about to enter a thermal bottleneck zone. This setting essentially establishes a soft boundary in the "resource utilization - thermal accumulation" space, intervening in scheduling adjustments in advance to avoid performance avalanche caused by "lagging response." In S42, differentiated control logic is set for different levels of strategy, the core of which is to construct a dynamic transition channel from "optimal allocation" to "risk avoidance." For example, to address resource conflicts at the fourth-level control layer, peak shaving and valley filling are achieved by controlling the thread wake-up cycle. This creates execution gaps when task densities highly overlap, preventing sudden concurrency surges that could impact system load. Similarly, the fifth-level control layer's feedback mechanism, which sets a historical failure limit, can filter out duplicate task bindings in scheduling blind spots and penalize high-failure binding paths from historical data, effectively improving the overall success rate. The common goal of these mechanisms is to shift scheduling from a "single-point optimization" model to a "system stability priority" model, achieving three-dimensional coordinated control of temperature, resources, and tasks.

[0035] Example 6: Please refer to Figure 1 and Figure 2 A multi-device collaborative control system for a soundproof cabin includes a multi-device acquisition module, an adaptability analysis module, a scheduling adaptability analysis module, and a reallocation module; The multi-device acquisition module connects multiple devices in the silent cabin via the Internet of Things, and collects the task status execution feedback parameters of multiple devices in the silent cabin in real time during the task operation. After preprocessing, the task status execution feedback parameters are used to form a standardized dataset of device status. The fit analysis module calculates the fit based on a standardized dataset, outputs the fit vector Fit(i,j) between device i and task j, and then constructs a task allocation matrix and obtains preliminary allocation results based on the Fit(i,j) vector between device i and task j. The scheduling adaptability analysis module calculates the scheduling adaptability index Eval based on the preliminary allocation results, and judges the rationality of scheduling based on the output results of the scheduling adaptability index Eval. The reallocation module triggers the hierarchical division of control strategy results based on the scheduling rationality judgment result, and executes the relevant control strategies according to the hierarchical division result of control strategy results.

[0036] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A multi-device collaborative control method for a soundproof cabin, characterized in that: Includes the following steps: S1. Connect multiple devices in the soundproof cabin via IoT to collect task status execution feedback parameters of multiple devices in the soundproof cabin in real time during task operation, and form a standardized dataset of device status after preprocessing the task status execution feedback parameters. S2, calculate the fit degree based on the standardized dataset, output the fit degree vector Fit(i,j) between device i and task j, and then construct the task allocation matrix and obtain the preliminary allocation result based on the Fit(i,j) vector between device i and task j. S3. Based on the preliminary allocation results, calculate the scheduling adaptability index Eval, and based on the output of the scheduling adaptability index Eval, make a judgment on the rationality of scheduling. S4, based on the scheduling rationality judgment result, triggers the control strategy result hierarchy division, and executes the relevant control strategy according to the control strategy result hierarchy division result.

2. The multi-device collaborative control method for a soundproof cabin according to claim 1, characterized in that: S1 includes S11; S11. By setting up collection points on multiple devices in the soundproof cabin and interconnecting them with the collection devices in the collection points through the Internet of Things, the task status execution feedback parameters of the multi-state devices in the soundproof cabin during the task operation are collected in real time, and then the collected task status execution feedback parameters are transmitted to the collaborative control server through the Internet of Things. The data collection points include device-level data collection points and task-level data collection points; The task status execution feedback parameters include status parameters and task execution feedback parameters; The state parameters include the CPU availability vector CPUava(i,t) of device i at time t, the memory remaining vector Mem(i,t) of device i at time t, and the interface temperature rise rate vector Thrtmal(i,t) of device i at time t. The task execution feedback parameters include the actual response delay vector Latency(i,j) of device i and task j and the task identification count vector Fail(i,j) of device i and task j.

3. The multi-device collaborative control method for a soundproof cabin according to claim 2, characterized in that: S1 further includes S12; S12. Receive task status execution feedback parameters in real time in the collaborative control server, and preprocess the task status execution feedback parameters to obtain a standardized dataset of device status. The preprocessing includes sliding window filtering and dimensionless processing. The sliding window filtering uses a fixed-window-length sliding window to calculate the state parameters, perform mean smoothing over a short time interval, and suppress noise and graph compilation interference on the scheduling results. The dimensionless processing involves mapping all vectors in the task execution feedback parameters, after the state parameters have been filtered by a sliding window, to the interval [0,1]. A dimensionless device state matrix is ​​constructed through normalization, where the dimensionless method used is the maximum and minimum value normalization function. The dimensionless state parameters are then integrated with the task execution feedback parameters to obtain a standardized device state dataset.

4. The multi-device collaborative control method for a soundproof cabin according to claim 3, characterized in that: S2 includes S21; S21. Extract the CPU utilization rate Ecpu(i,j) and the historical actual response time LaT(i,j) of device i and task j from the database of the collaborative control server. Combine the standardized dataset of device status obtained in real time to calculate the fit degree and output the fit degree vector Fit(i,j) of device i and task j. Next, perform adaptation calculations on all devices and tasks to be assigned to form an adaptation vector set, and construct a device and task adaptation matrix based on the adaptation vector set; The fit vector Fit(i,j) between device i and task j is obtained through the following fit calculation; 。 5. The multi-device collaborative control method for a soundproof cabin according to claim 4, characterized in that: S2 further includes S22; S22. The adaptation vector Fit(i,j) between device i and task j is evaluated, and then predictive overload triggered adaptive scheduling control is executed based on the evaluation results. The specific content is as follows: When any device i satisfies one of the following conditions: the fit vector Fit(i,j) between device i and task j is <0.4, the CPU availability vector CPUava(i,t) of device i at time t is <30%, and the historical actual response time LaT(i,j) between device i and task j is >0.8 × the actual response delay vector Latency(i,j) between device i and task j; The system determines that device i has entered a resource critical state and initiates the task migration and scheduling mechanism. The task migration scheduling mechanism writes the corresponding task j into the predicted overload task queue through the task j transfer buffer in the software layer, and reschedules task j when the target device enters the idle execution window. When the operating environment of device i supports a real-time operating system, the execution priority of the current task j on the critical device i is reduced by the task priority replacement interface. When performing task migration, the migration target device is selected from the available devices according to the migration target selection strategy. The migration target selection strategy includes avoiding device i that is operating at high temperature and giving priority to device i that has recently completed the task, thereby completing the dynamic task reallocation.

6. The multi-device collaborative control method for a soundproof cabin according to claim 1, characterized in that: S3 includes S31; S31. After the task migration scheduling mechanism is completed, calculate the scheduling adaptability index Eval based on the adaptation vector Fit(i,j) of all devices i and tasks j, and analyze the comprehensive scheduling adaptability of the current task allocation scheme at the system level. The scheduling adaptability index Eval is calculated and output using the following algorithm formula; In the formula, N represents the total number of tasks to be assigned, and D represents the set of available devices.

7. The multi-device collaborative control method for a soundproof cabin according to claim 6, characterized in that: S3 further includes S32; S32. Based on the output of the scheduling adaptability index Eval, determine the rationality of the current task migration scheduling mechanism. The specific determination is as follows: When the scheduling adaptability index Eval ≥ 0.7, it indicates that the current task migration scheduling mechanism is reasonable, and the task allocation is maintained. When the scheduling adaptability index Eval < 0.7, it indicates that the current task migration scheduling mechanism is unreasonable, and the task reallocation strategy should be initiated.

8. The multi-device collaborative control method for a soundproof cabin according to claim 7, characterized in that: S4 includes S41; S41. After the task reallocation strategy is triggered, the scheduling control logic is structurally divided according to the task scheduling stage, adaptability level, equipment operating status, task execution feedback, and historical failure records. Specifically, this includes: For tasks that are scheduled for the first time in the current scheduling cycle, they are classified as first-level tasks according to the first-level classification criteria and are assigned to the first-level control layer. In the case of non-first scheduling, when the fitness vector Fit(i,j) between device i and task j is less than 0.4, the scheduling path is divided and task j is assigned to the second-level control layer. Under the premise that the fit vector Fit(i,j) between device i and task j is greater than or equal to 0.4, further based on the fact that when the CPU availability vector CPUava(i,t) < 30% and the interface temperature rise rate vector Thrtmal(i,t) of device i at time t is greater than 0.6℃ / s, the corresponding task j is assigned to the third-level control layer. If device i is not currently in a critical state but there are risks of multiple concurrent tasks on the same device, concentrated task density, and structural load, then the corresponding task is classified into the fourth level control layer. If the number of failure records of task j on device i in the scheduling history exceeds a preset number of 6, it will be classified into the fifth level control layer.

9. A multi-device collaborative control method for a soundproof cabin according to claim 8, characterized in that: S4 further includes S42; S42. After completing the hierarchical division of the control strategy structure, execute the corresponding control strategies for each control level as follows: For the first-level control layer, the standard initial allocation strategy is executed, and task binding is completed directly based on the maximum value of the fit vector Fit(i,j) between device i and task j. For the second-level control level, the task delay binding strategy is implemented. Adaptation tasks with a fit vector Fit(i,j) < 0.4 between device i and task j are temporarily stored in the task waiting area, and devices with higher fit are prioritized. For the third-level control level, a predictive overload control strategy is implemented, and the task is transferred to the equipment whose current temperature and load are in a stable range through a migration mechanism; For the fourth level of control, a resource conflict smoothing strategy is implemented, which controls the task execution frequency by controlling the thread wake-up cycle; For the fifth level of control, a historical failure feedback avoidance strategy is implemented to restrict priority binding again.

10. A multi-device collaborative control system for a soundproof cabin, applied to the multi-device collaborative control method for a soundproof cabin as described in any one of claims 1-9, characterized in that: It includes a multi-device acquisition module, an adaptation analysis module, a scheduling adaptability analysis module, and a reallocation module; The multi-device acquisition module connects multiple devices in the silent cabin via the Internet of Things, and collects task status execution feedback parameters of multiple devices in the silent cabin in real time during the task operation process. After preprocessing, the task status execution feedback parameters are used to form a standardized dataset of device status. The fit analysis module calculates the fit based on the standardized dataset, outputs the fit vector Fit(i,j) between device i and task j, and then constructs a task allocation matrix and obtains preliminary allocation results based on the Fit(i,j) vector between device i and task j. The scheduling adaptability analysis module calculates the scheduling adaptability index Eval based on the preliminary allocation results, and outputs the results based on the scheduling adaptability index Eval to judge the rationality of the scheduling. The reallocation module triggers the hierarchical division of control strategy results based on the scheduling rationality judgment result, and executes relevant control strategies according to the hierarchical division result of control strategy results.

Images