Low power consumption internet of things terminal scheduling device and method

Abstract

The application discloses a kind of low-power Internet of Things terminal scheduling equipment and method, the method includes: obtaining preset service template and generating to be scheduled task;According to the channel state, admission evaluation is carried out to to-be-scheduled task, and the target channel corresponding to admission task is determined;Pre-start instruction is sent to terminal associated with admission task, and triggers terminal to execute peripheral power-on and data preparation;The readiness state reported by terminal is received, and time slot allocation is executed on target channel based on the actual data volume carried by it;Time slot allocation result is issued to terminal, to make it wake up internal data plane module and send service data.The application is through network control equipment active awakening and two-stage allocation mechanism based on actual report quantity, while effectively compressing the online length of terminal high-power radio frequency component, can reduce the waste of bandwidth or service truncation caused by data volume fluctuation, is conducive to improving system scheduling success rate.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) communication technology, and in particular to a low-power IoT terminal scheduling device and method. Background Technology

[0002] With the large-scale deployment of IoT services in wide-area scenarios such as forest fire fighting, meteorological monitoring, and hydrological and water conservancy, massive distributed sensors and image acquisition nodes need to periodically transmit collected environmental parameters or multimedia data back to data centers via satellite links or cellular networks. In these field scenarios deployed in areas without mains power coverage, how to maximize the lifespan of solar- or battery-powered terminal nodes through effective network scheduling and communication protocol design, while ensuring the efficiency of concurrent transmission of multi-source data, is an important topic of ongoing research in this field.

[0003] Currently, in backhaul scheduling schemes for wide-area IoT terminals, the most common approach is to use a DAMA (Demand-Based Multiple Access) mechanism that allows terminals to autonomously poll for access or pre-allocate fixed time slots. Under these operating modes, the interaction between the network and the terminal often relies on preset static strategies or passive responses to congestion. Typically, the terminal's RF communication module needs to maintain an online listening state for a considerable period to wait for signaling interaction. Alternatively, under a passive allocation mechanism, regardless of whether the actual amount of data collected by the terminal changes due to environmental fluctuations (such as variations in image compression rates caused by changes in light intensity), the network allocates physical time slots according to a uniform rated bandwidth standard.

[0004] In high-load scenarios with massive concurrent terminals, existing scheduling mechanisms face an irreconcilable contradiction between overall system resource utilization and individual terminal energy consumption control.

[0005] On the one hand, passive or static resource allocation methods are difficult to adapt to sudden or fluctuating real physical data loads, which can easily lead to local idleness or worsening of congestion in channel bandwidth. On the other hand, scheduling strategies that lack global timing coordination and environmental awareness capabilities can easily lead to terminals exhausting their limited power too quickly while waiting in vain and wasting radio frequency power.

[0006] Therefore, there is an urgent need to explore a wide-area IoT collaborative scheduling method that can balance transmission performance and extreme power consumption control in high-concurrency environments. Summary of the Invention

[0007] Purpose of the invention: To provide a low-power IoT terminal scheduling device and method to solve the above-mentioned problems existing in the prior art.

[0008] Technical solution: A low-power IoT terminal scheduling method, comprising: Obtain a preset business template and generate a task to be scheduled based on the business template; Based on the channel state, an admission assessment is performed on the task to be scheduled to determine the admission task and the target channel corresponding to the admission task. Send a pre-start command to the terminal associated with the access task to trigger the terminal to perform peripheral power-on and data preparation; After the receiving terminal completes data preparation, it reports the ready status. Based on the actual amount of service data carried in the ready status, it performs time slot allocation for the terminal on the target channel and obtains the time slot allocation result. The time slot allocation results are sent to the terminal so that the terminal can wake up the internal data plane module to send business data based on the time slot allocation results.

[0009] According to one aspect of this application, the pre-start instruction includes a task identifier, a service timeout period, and a pre-start delay period; sending the pre-start instruction to the terminal associated with the admission task to trigger the terminal to perform peripheral power-on and data preparation includes: After receiving the pre-start instruction, the terminal starts the local timed operation based on the pre-start delay time; After the local timed operation is completed, the terminal performs a power-on operation on the peripheral device and enters the data preparation state.

[0010] According to one aspect of this application, after the terminal completes sending the service data, an asynchronous update step of the service template is also included: Get the amount of pre-configured business data in the current record of the business template; A weighted smoothing calculation is performed based on the actual business data volume and the currently recorded pre-configured business data volume to obtain the updated pre-configured business data volume; Replace the corresponding record in the business template with the updated pre-configured business data volume, so that it can be called when generating the scheduled task in the next business scheduling cycle.

[0011] According to one aspect of this application, the terminal internally includes a control plane processor, a communication receiving bus, and a data plane module; When the terminal is in standby mode, the communication receiving bus remains connected to the control plane processor so that the control plane processor can listen for and receive pre-boot instructions. During the process of the control plane processor responding to the pre-start command to power on peripherals and prepare data, the control plane processor triggers the data bus switch action to switch the connection topology of the communication receiving bus from the control plane processor to the data plane module, so that the data plane module after being woken up can exclusively use the communication receiving bus for business data interaction.

[0012] According to one aspect of this application, after the terminal reports the ready state, it also includes self-protection fallback logic on the terminal side: The terminal starts its internal wait timer; If the terminal does not receive the time slot allocation result before the timer triggers a timeout interrupt, the terminal will power down the powered peripherals and exit the active state to enter the sleep mode to wait for subsequent instructions.

[0013] According to one aspect of this application, before calculating the earliest simulation completion time of the task to be scheduled in multiple candidate channels, a preliminary screening step of constructing multiple candidate channels is also included: Obtain the terminal's hardware carrier support capabilities and the beam coverage parameters of the terminal's current location; Based on the hardware carrier support capability and beam coverage parameters, intersection filtering is performed on all channels of the system to determine multiple candidate channels that meet the terminal communication conditions.

[0014] According to one aspect of this application, after classifying the target assessment channel into corresponding congestion levels based on the expected channel load rate, a degradation anti-oscillation mechanism is also included: When the target assessment channel is at a high level of congestion, continuously monitor the expected channel load rate in the next time period; The target evaluation channel will be downgraded to a lower congestion level only if the expected channel load rate is lower than the preset downgrade threshold for several consecutive scheduling cycles.

[0015] According to one aspect of this application, in the event of admission assessment failure or failure of the data plane module to send business data, an abnormal retry mechanism for the scheduled task is triggered, including: Get the current number of failures for the task to be scheduled; The dynamic retry delay interval is obtained by performing exponential backoff calculation based on the current number of failures. Based on the dynamic retry delay interval, the trigger time of the scheduled task is postponed and it is added back to the scheduling queue for the next cycle.

[0016] According to another aspect of this application, a low-power IoT terminal scheduling device is also provided, comprising: Memory, used to store computer programs; A processor that executes computer execution instructions stored in the memory to implement the low-power IoT terminal scheduling method described in any embodiment.

[0017] Beneficial effects: This invention, through a two-stage allocation mechanism of active wake-up by network control and based on actual data volume, can reduce bandwidth waste or service interruption caused by data volume fluctuations while maximizing the compression of the online time of high-power radio frequency components in the terminal, which is conducive to improving the system scheduling success rate. Attached Figure Description

[0018] Figure 1This is a flowchart of a low-power IoT terminal scheduling method according to the present invention.

[0019] Figure 2 This is a flowchart of the pre-congestion control steps before the admission assessment of the scheduled task in this invention.

[0020] Figure 3 This is a flowchart illustrating the differential update of the initial time slot resources reserved on the target channel according to the present invention.

[0021] Figure 4 This is a flowchart of the asynchronous update steps for the business template in this invention. Detailed Implementation

[0022] Example 1: This example provides a low-power IoT terminal scheduling method, which mainly includes the following steps: like Figure 1 As shown, in step 101, a preset business template is obtained, and a task to be scheduled is generated based on the business template; Step 102: Obtain the current channel state; perform admission evaluation on the task to be scheduled based on the channel state to determine the admission task and the target channel corresponding to the admission task. Channel status can include, but is not limited to, parameters that reflect channel resource occupancy and communication conditions, such as the number of currently occupied time slots, queuing depth, remaining available channel capacity, and channel quality indicators.

[0023] Step 103: Send a pre-start command to the terminal associated with the admission task to trigger the terminal to perform peripheral power-on and data preparation; the peripheral may specifically include data acquisition units such as environmental sensors or high-definition cameras; at the same time, the data plane module is also powered on and started initialization at this stage.

[0024] Step 104: Receive the ready status reported by the receiving terminal after completing data preparation, and perform time slot allocation for the terminal on the target channel based on the actual amount of service data carried in the ready status, to obtain the time slot allocation result; Step 105: The time slot allocation result is sent to the terminal so that the terminal can wake up the internal data plane module to send service data based on the time slot allocation result. Wake-up means that the data plane module changes from the standby state after initialization to the active transmission state.

[0025] The low-power IoT terminal scheduling method relies on a centralized scheduling architecture, which mainly consists of an IoT platform, network management, network control equipment, and multiple terminals.

[0026] Specifically, the service template in step 101 is a set of policies statically configured in the IoT platform or network management system. The service template contains prior operating parameters for specific terminal groups or business scenarios.

[0027] For example, a business template can be configured as follows: There are a total of 500 terminals. The service triggering cycle is once every hour. The pre-configured service data volume for a single service is 1.0 unit (e.g., normalized baseline). The pre-start delay time is 60 seconds, and the service timeout time is 300 seconds. Based on the service template, the network control device automatically instantiates the corresponding terminal and its pre-configured requirements as a task to be scheduled each time the service triggering cycle arrives, and adds it to the global trigger pending queue maintained by the system.

[0028] Furthermore, regarding step 102, the network control device performs a unified evaluation of tasks in the global trigger queue during each time slot planned broadcast cycle. Channel status reflects the current congestion level and remaining capacity of the physical channel. During the admission evaluation, the network control device compares the estimated completion time of the scheduled task with the service timeout time configured in the service template, while also considering the maximum allowed power-on time on the terminal side as a constraint. Only tasks that simultaneously meet the above time constraints under the current channel capacity are marked as admission tasks and bound to a specific physical resource, i.e., the target channel. In this way, the system intercepts tasks that are bound to fail to transmit due to excessive queuing during the admission phase, reducing invalid channel occupation.

[0029] In some optional implementations, for step 103, the pre-start command sent by the network control device to the terminal is not a direct data transmission command, but rather a device control signaling. At this time, the terminal's data plane module (such as a high-power modem) is not yet powered on, and the terminal relies solely on a low-power control plane processor (such as a microcontroller unit, MCU) to listen to the forward control channel. When the terminal receives the command, the control plane processor takes over the control flow and performs the peripheral power-on operation. The peripheral may specifically include data acquisition units such as environmental sensors or high-definition cameras. Subsequently, the peripheral begins to perform substantive data preparation work such as environmental sampling or image capture and compression.

[0030] Specifically, regarding step 104, after the terminal completes data acquisition and caches it, the control plane processor sends a ready status back to the network control device. Since the amount of data acquired by the same peripheral device fluctuates under different conditions in real-world scenarios—for example, the compressed image size of the same camera may differ by several times between day and night—the ready status should carry the precise number of bytes generated in this acquisition, i.e., the actual amount of business data.

[0031] Upon receiving this status, the network control equipment overturns the original estimated time slot requirement calculated based on the service template and instead calculates the exact number of time slots occupied in the target channel superframe structure based on the actual number of bytes, thereby generating the final time slot allocation result. This two-stage mechanism based on "pre-start-actual quantity-reallocation" reduces the space waste caused by fixed bandwidth reservations or service interruption due to sudden large data volumes.

[0032] In step 105, the network control device transmits the generated time slot allocation result, which includes the specific superframe location and time slot number, through the broadcast channel.

[0033] At this point, the terminal's data plane module has completed power supply and internal initialization during the peripheral power-on phase, and has taken over the communication receiving bus through bus switching.

[0034] After the control plane processor parses its own allocation plan, it powers the data plane module and triggers its network search and synchronization, entering the transmission state. Upon receiving a time slot allocation broadcast containing its own identifier, the data plane module transmits its buffered service data back to the network control equipment via satellite or cellular network within the specified time slot window. After transmission is complete and confirmation is received from the network side, the control plane processor regains control of the communication receive bus through a reverse bus switching operation and immediately cuts off power to peripherals and the data plane module, returning to its ultra-low power standby listening mode.

[0035] This design allows the online time of the terminal's high-power RF unit to be compressed to the absolute physical time span required for data transmission, thus extending the battery life of battery-powered IoT nodes in the field.

[0036] This solution operates at the scheduling and control layer of the IoT system, between the physical layer time slot resource management and the upper-layer business applications. It coordinates the access timing of terminals and the allocation of channel resources through the centralized scheduling strategy of this layer.

[0037] Example 2 further illustrates the pre-entry congestion control steps prior to admission assessment.

[0038] In one possible implementation, the following steps are included: like Figure 2 As shown, in step 201, the pre-configured prediction window length and business triggering period are obtained; Traditional reactive congestion control mechanisms typically intervene only after channel congestion occurs. This delayed response can easily lead to timeouts and failures for a large number of time-sensitive IoT tasks. Therefore, this solution leverages the prior attribute of fixed triggering cycles in IoT terminal services, employing a feedforward predictive congestion control strategy. The network control device reads the preset prediction window length from the system configuration. The prediction window length represents the time span over which the network control device needs to anticipate future system load. In practical engineering deployments, the prediction window length should cover the maximum time span from when a task is granted access to the channel and begins queuing until actual physical transmission begins. The preset prediction window length is configured to be 150 cycles. Those skilled in the art will understand that the specific value of this parameter can be adaptively adjusted according to the actual system scale, channel capacity, and service triggering cycle.

[0039] Step 202: Determine the target evaluation channel; based on the prediction window length and the service triggering period, calculate the expected number of tasks arriving at the target evaluation channel in the future time period; Specifically, before performing admission determination in each broadcast cycle, the network control device acquires the service trigger cycles of all terminal devices bound to the target evaluation channel, and counts the total number of tasks that meet the trigger conditions within a future time period starting from the current system time and ending at the time point obtained by superimposing the prediction window length. Since each physical terminal is typically configured with only a single service template in the operation settings of this embodiment, and the duration of the service trigger cycle is much longer than the execution time of a single data backhaul task, the network control device can use the number of terminals directly instead of the number of tasks for a simplified prediction counting operation, reducing the computing power overhead of the system control plane.

[0040] Step 203: Based on the expected number of arriving tasks and the total pre-acquired channel capacity, calculate the expected channel load rate of the target evaluation channel, and classify the target evaluation channel into the corresponding congestion level according to the expected channel load rate. The network control equipment calculates the expected channel load rate of the target evaluation channel using the following formula: ρ=(M*N avg ) / (H*R c ); Where ρ is the expected channel load rate, M is the expected number of arriving tasks, and N is the expected number of arriving tasks. avg The average time slot requirement of the terminals bound to this channel, H is the number of cycles included in the prediction window length, and R is the average time slot requirement of the terminals bound to this channel. cThe target channel is evaluated to determine the total number of time slots that can be allocated within each superframe structure. The average time slot requirement is calculated by averaging the data volumes of various pre-configured services. After obtaining the expected channel load rate, the network control device compares it with preset lower and upper thresholds and performs a three-level state classification. The lower threshold is set to 0.6, and the upper threshold is set to 0.85. Specifically, when ρ ≤ the lower threshold, it is classified as level 0, indicating that the current channel capacity is sufficient; when ρ > the lower threshold and ρ ≤ the upper threshold, it is classified as level 1, indicating a channel resource warning; when ρ > the upper threshold, it is classified as level 2, indicating that the channel is severely congested.

[0041] Step 204: Based on the congestion level and the priority carried in the tasks to be scheduled, perform pre-filtering on the tasks to be scheduled to obtain a set of candidate evaluation tasks. After defining the specific congestion level of the target assessment channel, the network control equipment performs pre-access intervention based on the differentiated levels. If the current state is determined to be level 0, the system imposes no restrictions, and all tasks to be scheduled directly enter the candidate assessment task set. If the current state is determined to be level 1, the system initiates a partial rate limiting strategy, allowing only tasks with a priority value ≥ a preset threshold to enter the candidate assessment task set, while low-priority tasks below the preset threshold are directly deferred. The preset threshold is set to 3 (in a priority value range of 1 to 5). If the current state is determined to be level 2, the system enters an extreme rate limiting mode, allowing only the highest priority tasks to enter the candidate assessment task set, and simultaneously reporting channel congestion alarm events to the network management system through the internal operation and maintenance interface. Through this hierarchical filtering mechanism, the system proactively intercepts secondary traffic before physical congestion actually occurs, which is beneficial to ensuring the survival of core high-priority services.

[0042] Step 205: Perform the operation of determining the admission task and target channel from the candidate evaluation task set; After the dual filtering based on predicted load and priority, tasks successfully retained in the candidate evaluation task set will be sent to a more refined multi-channel time constraint calculation process. The network control device will further perform simulation evaluation within this filtered set to ultimately identify the admissible tasks for scheduling and bind them to specific physical channel resources.

[0043] Step 206: When the target assessment channel is at the first congestion level (higher congestion level), continuously monitor the expected channel load rate in the next time period. Step 207: The target evaluation channel is downgraded to the second congestion level (lower congestion level) only if the expected channel load rate is lower than the preset downgrade threshold for multiple consecutive scheduling cycles. The first congestion level has higher priority than the second congestion level. "Multiple consecutive" refers to a preset number of consecutive cycles, such as two or more consecutive cycles.

[0044] To prevent frequent fluctuations in congestion levels near the boundary threshold, this embodiment constructs a smooth degradation anti-oscillation mechanism.

[0045] Specifically, when the target assessment channel is in a high-load state at level 1 or level 2, even if the network control device detects that the expected channel load rate briefly drops below the corresponding lower threshold at a certain independent moment, the rate limiting state will not be immediately lifted. The system mandates that the expected channel load rate must remain below the admission threshold corresponding to the current congestion level for multiple consecutive broadcast cycles (i.e., the upper threshold is used as the degradation judgment benchmark when downgrading from level 2 to level 1, and the lower threshold is used as the degradation judgment benchmark when downgrading from level 1 to level 0) before a formal state degradation operation will be triggered.

[0046] For example, the number of smoothing cycles used for anti-oscillation determination can be set to three consecutive cycles. This smoothing mechanism improves the steady-state maintenance capability of the control plane state machine and can reduce secondary congestion caused by drastic fluctuations in scheduling strategies in large-scale terminal groups.

[0047] Based on the above embodiments, an alternative implementation method is provided. In the above-mentioned pre-congestion control stage, in addition to relying on the statically pre-configured service triggering cycle for task quantity prediction, an adaptive estimation method based on the actual historical arrival rate can also be adopted. The network control device counts the total number of tasks that actually arrive in the previous broadcast cycle in real time, and uses an update coefficient to smoothly iterate it with the historical comprehensive estimate, calculating the current adaptive estimated arrival rate using the following formula: λ new =β*λ real +(1-β)*λ old ; Where, λ new Let β be the updated adaptive estimated arrival rate calculated in this iteration, and λ be the updated control coefficient that is >0 and <1. real λ represents the actual number of tasks that arrived in the previous cycle. old The historical estimated arrival rate is retained from the last updated record. Subsequently, the network control device directly replaces the arrival quantity indicator in the prediction window statistical formula with this adaptive estimated arrival rate for subsequent load factor calculations. Specifically, the network control device uses the adaptive estimated arrival rate λ newMultiplying the number of cycles H contained in the prediction window yields the estimated total number of arriving tasks within the prediction window. This number replaces the expected number of arriving tasks M obtained by counting each task based on the triggering conditions in the original formula, and is used for subsequent load factor calculation.

[0048] This exponential smoothing-based estimation alternative is more sensitive to uncertain scenarios such as holidays or sudden natural disasters that cause sudden changes in actual business volume, further enhancing the environmental adaptability of the congestion control system.

[0049] In the face of the risk of channel overload caused by the concentrated outbreak of high-concurrency tasks, the use of feedforward expected load rate estimation to replace the traditional reactive control can proactively implement graded flow limiting before physical congestion occurs, which can reduce channel collapse and help improve the survival rate of core high-priority services.

[0050] Example 3 further illustrates the channel selection and admission decision process in the admission evaluation stage. In one possible implementation, it includes the following steps: Step 301: Obtain the terminal's hardware carrier support capability and the beam coverage parameters of the terminal's current location; The binding mapping conditions between the scheduled tasks and physical channels are determined by the system's preset service template configuration. For tasks with dynamic scheduling requirements, the network control equipment first performs a multi-channel initial screening mechanism. The network control equipment extracts the hardware carrier support capabilities from the capability set reported by the terminal device during network registration. This capability defines the frequency band range that the terminal's radio frequency module can physically support for communication. Simultaneously, for terminals relying on satellite networks, the network control equipment dynamically deduces and obtains beam coverage parameters based on ephemeris data.

[0051] Step 302: Based on the hardware carrier support capability and beam coverage parameters, perform intersection filtering on all channels of the system to determine multiple candidate channels that meet the terminal communication conditions. After obtaining the basic physical parameters, the network control device performs hierarchical filtering to construct a candidate channel set. Based on the full channel list configured by the system, the network control device calculates the hardware carrier support constraint subset and the beam space coverage constraint subset, and then takes the mathematical intersection of the two.

[0052] Furthermore, the network control equipment will also eliminate physical channels that are currently in a state of severe congestion alarm, and through multi-layer intersection filtering operations, determine multiple candidate channels that are reachable in both electrical characteristics and spatial links.

[0053] Step 303: Obtain multiple candidate channels; for the scheduled task that is not bound to a fixed physical channel, calculate the earliest simulation completion time of the scheduled task in each of the multiple candidate channels. For tasks to be scheduled that are marked as having no fixed binding association in the service template configuration, the network control device dynamically traverses the multiple candidate channels identified above and calculates the earliest simulation completion time under the specific channel conditions for each channel. The specific calculation is implemented through the following formula: T comp =max(T now +T pre +T ready ,T tail )+N req *T frame / R c ; Among them, T comp T represents the earliest simulation completion time on the current candidate channel. now T represents the current system absolute time. pre T is the pre-start delay time. ready T represents the inherent preparation time from power-on to ready reporting by the terminal. tail N represents the simulation tail completion time of the current candidate channel queuing queue. req T represents the number of predicted time slots required for the tasks to be scheduled. frame R is the physical superframe period length. c This represents the total time slot capacity available for allocation per superframe for the current candidate channel. Using this formula, the network control device can transform the heterogeneous queuing delays of each channel into a unified time scale.

[0054] Step 304: Based on the earliest simulation completion time, calculate the expected online duration of the task to be scheduled on each candidate channel; compare the earliest simulation completion time and the corresponding expected online duration with the pre-configured service timeout time and maximum online duration constraint, respectively. After obtaining the simulation results for each candidate channel, the network control device performs a double boundary check. The network control device extracts the system's pre-configured service timeout. Simultaneously, by subtracting the sum of the current time and the pre-start delay from the earliest simulation completion time, the network control device calculates the estimated online duration of the terminal in a non-sleep state and extracts the pre-configured maximum online duration constraint. The network control device then performs a numerical comparison operation between the calculated two time-dimensional indicators and their corresponding preset boundary parameters.

[0055] Step 305: Select candidate channels that simultaneously meet the constraints of service timeout time and maximum online duration, and construct a set of feasible channels; After numerical comparison, the network control device will simultaneously extract candidate channels based on the above two constraint verification conditions and merge them to construct a feasible channel set. All channels included in this set can ensure that the task can successfully complete physical transmission without timeout failure or exhausting the terminal's preset battery limit.

[0056] Step 306: If the feasible channel set is not empty, select the candidate channel with the earliest simulation completion time and the smallest time from the feasible channel set as the target channel, and confirm the task to be scheduled as an admission task; if the feasible channel set is empty, perform delay or rejection processing on the task to be scheduled.

[0057] When the verification results indicate that the feasible channel set contains at least one valid element, the network control device executes the optimal path selection mechanism. The network control device traverses the earliest simulation completion time records of each feasible channel within the set, retrieves the specific channel containing the minimum value through numerical comparison operations, and directly binds it as the target channel.

[0058] The network control device updates the status of the task to be scheduled from queue suspended to access permitted, and simultaneously updates the simulated tail time record of the target channel for use in the queuing prediction calculation of other tasks.

[0059] Step 307: If the set of feasible channels is empty, select the candidate channel with the smallest constraint deviation from multiple candidate channels as the current optimal channel, obtain the earliest simulation completion time and the corresponding expected online duration of the current optimal channel, and execute the following admission decision logic. When the system operates under high concurrency or extremely resource-constrained conditions, causing all candidate channels to fail to meet the aforementioned dual boundary constraints, the feasible channel set becomes empty. At this point, the system triggers an edge degradation processing path. The network control device rescans all non-compliant candidate channels, selects the channel with the smallest sum of absolute deviations in the dual constraint calculation as the current optimal channel, and extracts the earliest simulation completion time and corresponding expected online duration for that channel, before transferring it to a refined three-state auxiliary decision-making module.

[0060] Step 308: Determine whether the expected online duration exceeds the maximum online duration constraint. If it exceeds the maximum online duration constraint, determine that the task to be scheduled will be postponed to the next scheduling cycle. In the auxiliary decision-making module, power consumption limits are given the highest execution priority. The network control device prioritizes assessing the terminal's power availability and directly determines the relationship between the expected online time and the maximum online time constraint. If an excess is found, it indicates that continued queuing will inevitably lead to over-discharge of the terminal's battery. The network control device then forcibly determines to perform a delay operation on this task, causing it to remain in the global queue awaiting system resource utilization.

[0061] Step 309: If the maximum online duration constraint is not exceeded, calculate the timeout margin ratio based on the earliest simulation completion time, the business timeout time, and the task trigger time carried in the task to be scheduled; where the timeout margin ratio represents the relative time margin of the current scheduling scheme in the total timeout time window. Provided that the power consumption constraint test is passed, the network control device introduces statistical fault-tolerant variables to implement refined access control. The network control device calculates the timeout margin ratio using the following formula: δ=(T timeout -T comp ) / (T timeout -T trigger ); Where δ is the overtime margin ratio, T timeout T is the absolute timestamp of the business timeout time. comp T is the earliest simulation completion time. trigger Let be the initial task trigger time of the task to be scheduled. The numerator of this formula represents the remaining safety redundancy length derived from the current queuing queue, while the denominator represents the total lifespan of the task from its inception to its mandatory completion. This calculation mechanism provides a dimensionless, anti-fluctuation scale.

[0062] Step 310: Compare the timeout margin ratio with the pre-configured margin ratio threshold; The network control device extracts the margin ratio threshold issued by the system's operation and maintenance management terminal and compares it with the timeout margin ratio calculated in real time. As an example, this margin ratio threshold can be set to 0.2.

[0063] Step 311: If the timeout margin ratio is greater than or equal to the margin ratio threshold, then the task to be scheduled is confirmed as an admission task, and the current optimal channel is determined as the target channel. When the comparison result satisfies a greater than or equal to relationship, it indicates that the system has sufficient reserved redundancy time. For example, when the preset redundancy ratio threshold is 0.2, even if the terminal side incurs additional time overhead due to uncertainties such as hardware radio frequency search, as long as this additional overhead does not exceed 20% of the total lifecycle, the task can still successfully close the loop. Therefore, the network control device directly issues the access decision.

[0064] Step 312: If the timeout margin ratio is less than the margin ratio threshold but not less than zero, it is determined that the task to be scheduled will be postponed to the next scheduling cycle. While performing the postponement operation, the network control device increments the failure count counter for the task to be scheduled.

[0065] When the comparison result falls within a narrow range between 0 and the margin ratio threshold, it indicates that although the task can theoretically be completed just before the deadline, the system has lost its statistical buffering capability against random fluctuations in the underlying physical links. To avoid a high probability of physical timeout failure during the physical transmission phase, the network control device proactively downgrades the admission decision to a delayed state, keeping the task in the global queue to await reassessment in the next cycle. Simultaneously, the network control device increments the failure count counter for the scheduled task for subsequent retry backoff calculations.

[0066] Step 313: If the timeout margin ratio is less than zero, it is determined that the scheduled task will be rejected.

[0067] When the comparison result is negative, it indicates that based on the current queue backlog, the task's completion time will inevitably cross the service failure boundary. At this time, the network control device immediately performs a forced rejection operation, directly removing the task from the global queue cache and generating a drop event log, completely releasing the deadlocked virtual queuing resources.

[0068] The maximum allowable online duration is incorporated as a hard constraint into the multi-channel earliest completion time optimization process, and a three-state refined decision is made in conjunction with the timeout margin ratio. This not only compresses the power-on time limit of the terminal's high-power RF components to the necessary physical transmission interval, but also provides statistically significant safety redundancy for uncontrollable latency fluctuations such as the underlying network access search.

[0069] Example 4 further explains the timing coordination and physical blind spot avoidance mechanism for terminal hardware and software collaboration. In one possible implementation, the following steps are included: Step 401: After receiving the pre-start instruction, the terminal associates the local task context with the task identifier and records the service timeout time as the local timeout benchmark; and starts the local timed operation based on the pre-start delay time. The pre-start command issued by the network control device encapsulates multiple control fields, including task identifier, service timeout, service type, network information associated with the service, and a pre-start delay time for timing alignment. The terminal's control plane processor continuously parses forward link packets in standby mode. When it matches its own identifier information and extracts the complete pre-start command, the terminal's internal state machine transitions from standby to pre-start state. In this state, the control plane processor reads the pre-start delay time parameter carried in the command and assigns it to a local hardware timer to start countdown monitoring. This design allows the network control device to dynamically delay the actual power-on time of the terminal's physical peripherals based on the current channel queue congestion level, further reducing the terminal's power consumption while ensuring service deadlines are met. If the network control device calculates that the terminal needs to be put into operation immediately, it configures the pre-start delay time to zero, and the terminal, upon receiving this, will skip the timer waiting phase and directly proceed to subsequent operations.

[0070] Step 402: After the local timed operation is completed, the terminal performs a power-on operation on the peripheral device and enters the data preparation state.

[0071] When the aforementioned local timer experiences a timing interrupt, the control plane processor drives the state machine to transition to the active ready state. In this state, the control plane processor sends power enable signals to external devices such as sensors and image acquisition units, and simultaneously powers on the data plane module responsible for communication transmission. After the peripherals are powered on, they immediately begin substantive service data acquisition tasks such as physical environment sampling or multimedia data encoding and compression. After all data acquisition and buffering processes are completed without errors, the terminal sends a ready report message back to the network control equipment, and its internal state machine then transitions to the active waiting state, suspending its operation to await the network control equipment's precise channel physical time slot resource scheduling plan.

[0072] Step 403: Record the terminal's status as active on the network control device side; At the same physical moment when the network control device sends a pre-start command to a specific terminal, the device management mapping table within the network control device is simultaneously updated to update the global running flag of that terminal, registering it as active. This control plane action constitutes the first layer of protocol-level protection against the risk of communication deadlock. By performing centralized state tracking on the network control side, the system can accurately grasp at a macro level whether each distributed physical terminal is currently in a high-power-consuming operating mode.

[0073] Step 404: Before the network control device receives a task completion message or a task failure message reported by the terminal, it suspends sending new pre-start instructions to terminals that are recorded as active.

[0074] In real-world high-concurrency IoT scenarios, if a network control device sends multiple scheduling commands to the same physical terminal within a very short time window, and that terminal has already cut off the communication listening path of the control plane processor to execute the first command, subsequent commands will inevitably be lost in the physical link—this is the communication control blind spot. To mitigate this risk, the network control device implements mutual exclusion locking logic based on the aforementioned state tracking. For terminals in an active state, the system background actively intercepts and caches all new scheduling requests for that terminal, prohibiting the sending of new pre-start signaling until a task completion message or task failure message clearly indicating the end of the current closed loop is received, at which point the state lock is released. This eliminates the risk of timing errors caused by commands being sent during the blind spot period at the source.

[0075] Step 405: When the terminal is in standby mode, the communication receiving bus remains connected to the control plane processor so that the control plane processor can listen to and receive the pre-start command. The root cause of the aforementioned physical blind spot lies in the single communication receiver architecture adopted in this embodiment. To reduce terminal standby power consumption, the system incorporates an independent low-power signal demodulation chip within the terminal. When the terminal is in standby mode, the communication receiver bus is physically switched and connected to the low-power control plane processor via a single-pole double-throw switch chip controlled by general-purpose input / output pins. At this time, the high-power data plane main processor is in a power-off state. The control plane processor receives and parses low-speed control frames in the forward control channel through a low-speed serial interface, resulting in extremely low processing load, which is beneficial for improving the microwatt-level extreme standby power consumption performance.

[0076] Step 406: After the control plane processor receives the time slot allocation result, the control plane processor triggers the data bus switch action to switch the connection topology of the communication receiving bus from the control plane processor to the data plane module, so that the data plane module after being woken up can exclusively use the communication receiving bus for service data interaction.

[0077] In the parallel processing pipeline that powers on peripherals, after the data plane module completes its internal initialization and startup, the control plane processor triggers a physical reversal of the single-pole double-throw bus switch chip by pulling up the level signal of the general-purpose input / output pin. This action forcibly disconnects the physical connection topology of the transport stream output by the demodulation chip from the control plane processor and re-anchors it to the input interface of the data plane module. The circuit delay of this physical switch is very short and does not affect the ongoing satellite or cellular network frame synchronization process. After the switch is completed, the data plane module takes over the communication receive bus and exclusively controls the RF demodulated data stream. In this state, the time slot allocation plan broadcast by the network control equipment is directly received and parsed by the data plane module. After the data plane module matches a time slot allocation result containing its own identifier, the terminal state machine transitions from the active waiting state to the active transmission state, and the data plane module performs high-speed service data backhaul within the specified time slot window. When the transmission is completed and a hardware interrupt is triggered, the state machine enters the handover state, the system performs the reverse operation to return the communication receive bus to the control plane processor, and finally returns to the standby state. The underlying driver implementation for the data bus switching action can be adaptively configured according to the specific integration method of the terminal hardware.

[0078] It should be noted that after the bus switch, the mechanism by which the control plane processor receives the time slot allocation result can adopt existing technologies, such as receiving it through an independent low-speed control channel, or forwarding it after parsing by the data plane module. This process is not the focus of this application and will not be described in detail here.

[0079] In some embodiments, the control plane processor may trigger a data bus switch action after receiving the time slot allocation result.

[0080] Alternatively, the control plane processor may receive the time slot allocation result through the independent control channel interface and wake up the data plane module accordingly.

[0081] In addition, during the data preparation phase, the data plane module can also be in a waiting state of power-on initialization without sending data. The relevant configuration parameters can be designed by those skilled in the art according to the workflow.

[0082] Step 407: The terminal starts its internal wait timer; In the reliability design of the terminal hardware architecture, in addition to the protection on the network control side, this embodiment also constructs a second-layer self-protection mechanism on the terminal side to prevent abnormal deadlocks caused by the loss of time slot allocation results issued by the network control device due to wireless channel fading. Specifically, when the terminal state machine resides in the active waiting state, the control plane processor will immediately activate an independently operating watchdog-like waiting timer to begin accumulating the objective time span of waiting for the network control device to respond.

[0083] Step 408: If the terminal has not received the time slot allocation result before the timer triggers a timeout interrupt, the terminal controls the powered peripheral to power off and exits the active state to enter the sleep mode to wait for subsequent instructions.

[0084] The maximum overflow threshold of the waiting timer is pre-configured as a fixed time constant. This time constant can be adaptively configured by those skilled in the art based on the superframe period length of the target channel and the actual network response delay; for example, it can be configured to 60 seconds. If the time constant is exhausted and the control plane processor still has not received a transmission start confirmation signal from the data plane module, it indicates that the network control device may have failed to fulfill its scheduling commitment due to severe queue backlog or channel interference.

[0085] At this point, to strictly prevent the data plane module from continuously absorbing battery power during this ineffective waiting period and forming a zombie task stagnation state, the control plane processor will forcibly take over system control. Its control logic directly skips the normal transmission phase, actively issuing a physical power-off command to shut down all high-power-consuming components, driving the state machine to transition to a timeout sleep state, and reporting a task failure code to the network control device. Subsequently, it resets the hardware to return to the standby listening baseline, improving the survival probability of isolated nodes in harsh communication environments.

[0086] By constructing a dual-layer protection system through active state mutual exclusion locking on the network control side and timeout watchdog sleep mechanism on the terminal side, long-term ineffective radio frequency waste during state machine switching is avoided, thereby improving the survival probability of wide-area isolated nodes in harsh communication environments.

[0087] Example 5 further explains the process of allocating time slots and handling exceptions based on the actual amount of business data carried in the ready state.

[0088] In one possible implementation, the following steps are included: Step 501: Obtain the pre-configured business data volume defined in the business template; The business template provides a static data volume reference during initial setup. This pre-configured business data volume is typically based on historical statistical averages or nominal values ​​set by the business model. In actual IoT deployment environments, the upper-layer business data volume often fluctuates due to differences in the external environment.

[0089] For example, the compressed image data size captured by the same high-definition camera in a clear sky can reach 600KB, while in a foggy sky it may only be 200KB, and in scenarios containing a large amount of high-frequency details, such as a fire scene, it may soar to 900KB. If the allocation relies on a fixed pre-configured service data volume, when the actual collection volume is too large, the data transmission will be forcibly truncated, and when the actual collection volume is too small, it will cause unnecessary waste of bandwidth resources and terminal online time.

[0090] Step 502: Calculate the actual time slot requirement corresponding to the actual amount of business data and the pre-configured time slot requirement corresponding to the pre-configured amount of business data, respectively. After receiving the precise actual amount of service data uploaded by the terminal, the network control device initiates the conversion calculation of resource requirements. The network control device extracts the physical layer attributes of the target channel, including the number of payload bytes carried per timeslot. By dividing the actual service data volume by this number of payload bytes and rounding up, the network control device calculates the actual timeslot requirement. Using the same calculation logic, the network control device calculates the pre-configured timeslot requirement based on the pre-configured service data volume.

[0091] Step 503: Compare the actual time slot demand with the pre-configured time slot demand to determine the target time slot deviation. The network control device calculates the algebraic difference between the actual time slot demand and the pre-configured time slot demand by performing a simple numerical subtraction operation, and defines this difference as the target time slot deviation. This target time slot deviation constitutes the absolute numerical benchmark for subsequent dynamic bandwidth compensation control logic.

[0092] Step 504: Based on the target time slot deviation, perform differential updates on the initial time slot resources reserved on the target channel based on the pre-configured time slot demand to generate time slot allocation results.

[0093] In traditional on-demand multiple access systems, the bandwidth requested by the terminal is the final allocation basis for the network control device, without an intermediate step of reporting the actual data volume after data preparation. This embodiment relies on the special architecture of the network control device actively pre-starting the terminal, utilizing the time difference between the terminal completing data acquisition and the network control device generating a time slot plan to construct a two-stage closed-loop control mechanism: pre-allocation estimation, actual allocation confirmation, and precise time slot allocation. This mechanism enables the network control device to perform secondary adjustments on the reserved physical resources, outputting time slot allocation results that accurately match the actual service load.

[0094] like Figure 3 As shown in step 505, if the actual time slot demand is not greater than the pre-configured time slot demand, then the actual time slot demand is used as the allocation benchmark, and the idle time slot resources corresponding to the target time slot deviation are released. When the comparison results show that the actual time slot demand is less than or equal to the pre-configured time slot demand, it indicates that the amount of business data in the current acquisition environment is lower than or meets expectations. The network control equipment will reduce the original physical resource reservation boundary and generate time slot allocation results using a smaller actual time slot demand.

[0095] The network control equipment removes the excess time slots corresponding to the target time slot deviation from the exclusive pool of the task and returns them to the global available resource pool of the current target channel so that they can be reused by other tasks waiting in the queue.

[0096] Step 506: If the actual time slot demand is greater than the pre-configured time slot demand, then detect the remaining available capacity of the target channel in the current scheduling period. When the comparison results show a sudden surge in large data volume, the network control equipment should assess the current physical channel's capacity limit. The network control equipment should read in real time the number of time slots on the target channel that have not yet been occupied by other admitted tasks within the current superframe period to obtain an accurate value of the remaining available capacity.

[0097] Step 507: When the remaining available capacity can accommodate the target time slot deviation, additional time slot resources corresponding to the target time slot deviation are allocated to the terminal; when the remaining available capacity is insufficient to accommodate the target time slot deviation, a time slot allocation result is generated for the terminal according to the pre-configured time slot demand.

[0098] If the remaining available capacity is greater than or equal to the target time slot offset, it indicates that the current channel is sufficiently idle. The network control device, while maintaining the original pre-configured time slot requirement, performs a resource addition operation, allocating additional time slot quotas to the terminal. The terminal's data plane module operates continuously within this expanded time slot window until all excess service data is transmitted, ensuring a complete single-cycle return of sudden large data blocks.

[0099] Step 508: When the actual time slot demand is greater than the pre-configured time slot demand and the remaining available capacity is insufficient to accommodate the target time slot deviation, generate the time slot allocation result for the terminal for the current scheduling period according to the pre-configured time slot demand. When channel congestion prevents the remaining available capacity from meeting the additional demand, the system implements a partial transmission strategy to prevent overall communication timeout. The network control device still issues a time slot allocation plan according to the initial pre-configured time slot demand. The terminal data plane module sends only the first data block in the queue according to the plan, and retains the unsent data blocks in the local hardware cache sequence. During this period, the terminal control plane processor does not execute the power-down command and remains in an active waiting state.

[0100] Step 509: Obtain the remaining data volume in the actual business data volume that is not covered by the time slot allocation result; After the aforementioned partial transmission phase is completed, the network control device receives a partial completion report message returned by the terminal. This message carries confirmation of the number of bytes successfully sent and the number of remaining bytes in the local cache. The network control device extracts this remaining data and uses it as input parameters for constructing a new task.

[0101] Step 510: Based on the remaining data volume, register and generate a resume transmission subtask on the network control device side, and add the resume transmission subtask to the scheduling queue for time slot allocation in subsequent scheduling cycles.

[0102] The network control device automatically allocates a new task control block in memory to generate a resume subtask for processing legacy data. In the attribute inheritance mechanism, this resume subtask inherits the task identifier of the original business and appends a specific sequence number suffix code to maintain business relevance.

[0103] Specifically, the single-transmission traffic volume of this resume subtask is overwritten with the aforementioned remaining data volume; to prevent residual data from starving indefinitely in the queue, the priority value of this resume subtask is forcibly increased, for example, set to the original priority value plus 1; since the terminal hardware is now operating at full power, the pre-start time of this resume subtask is set to zero; the trigger time is reset to the current system time, and the service timeout time is updated to the original total timeout time minus the consumed time. This resume subtask will be directly inserted at the head of the current channel's pending queue, and will have priority to obtain the time slot plan in the next broadcast cycle. If the resume subtask is not scheduled before the updated service timeout time arrives, the network control device will mark the entire associated task chain as a complete failure and forcibly instruct the terminal to power down.

[0104] In some optional implementations, the applicability of the aforementioned residual data segmentation and continuation mechanism is subject to specific technical constraints. This strategy relies on the upper-layer service protocol stack adopting a progressive coding format system, such as progressive mode image coding or scalable video coding format conforming to specific standards, or the service itself being a discrete sensor time-series dataset with inherently divisible attributes. Under these conditions, the physically segmented data blocks can still be normally parsed and decoded after being reassembled at the network management control center.

[0105] For service types that do not support segmented transmission, the network control device can perform transmission according to the pre-configured time slot requirements, and the terminal application layer will terminate the transmission after the time slots are exhausted.

[0106] Step 511: Obtain the current number of failures for the task to be scheduled; In severe network conditions characterized by deep fading, IoT terminal systems should possess underlying self-recovery capabilities. The system's backend deploys an anomaly monitoring daemon responsible for capturing the following four types of retry trigger events: a scheduled task fails to obtain access permission in multiple consecutive scheduling cycles' access assessments and has not yet reached the service timeout period; the terminal fails to successfully report a ready message within the pre-startup timeout range; the data plane module throws a radio frequency link interruption anomaly during transmission; and the terminal's active state, maintained by the network control side, exceeds the service-permitted boundary without receiving a completion instruction. "Multiple consecutive scheduling cycles" refers to a preset number of scheduling cycles.

[0107] If a terminal fails to successfully report a ready message within the pre-startup timeout period, it can be understood as the terminal failing to successfully report a ready message within the ready waiting timeout threshold preset by the network control device. The pre-startup timeout can be understood as the difference between the service timeout period and the current time.

[0108] In the case of continuous failure of the first type of admission assessment, the network control device synchronously accumulates the failure count counter during each delayed operation, and performs subsequent exponential backoff calculation based on the counter to gradually lengthen the retry interval and avoid repeated impacts of low-priority tasks on the scheduling queue during high load periods.

[0109] Step 512: Perform exponential backoff calculation based on the current number of failures to obtain the dynamic retry delay interval; The network control device does not immediately resend the task, but instead uses a backoff algorithm to mitigate concurrent conflicts and physical interference. The network control device executes the following exponentially weighted formula: T retry =T base *2 (k-1) ; Among them, T retry To calculate the dynamic retry delay interval of the output, T baseThe basic retry interval time constant pre-configured for the system, T base The time limit is set to 60 seconds. k is a numerical variable representing the current number of failures. Through this exponentially increasing mapping relationship, the cooldown period required for a task to attempt execution before its next attempt will increase exponentially with each failure.

[0110] Step 513: Based on the dynamic retry delay interval, postpone the trigger time of the task to be scheduled and add it back to the scheduling queue of the next cycle; if the current number of failures reaches the pre-configured maximum number of allowed retries, terminate the scheduling of the task to be scheduled and generate a fault alarm.

[0111] After obtaining the dynamic retry delay interval, the network control device adds it to the current system time and writes the sum to the next trigger time field of the task to be scheduled.

[0112] The task to be scheduled is resubmitted to the global trigger queue. During this retry process, the original priority configuration of the task remains frozen, so that its relative ranking weight in the admission evaluation mechanism is not affected.

[0113] The network control device increments the failure count counter by 1. If the accumulated failure count exceeds the pre-configured maximum allowed retries constant (which is set to 5), the network control device will completely terminate the task's scheduling lifecycle, erase the memory record, and send a permanent unreachable fault alarm signal to the external maintenance terminal via the general control interface.

[0114] Example 6 further describes the asynchronous update steps after the terminal completes sending service data. In one possible implementation, the steps include: like Figure 4 As shown, in step 601, obtain the amount of pre-configured business data currently recorded in the business template; After confirming that the terminal has completed sending service data, the network control device triggers a closed-loop optimization process in the background for the service template to which the terminal belongs. This optimization process is executed in a non-real-time path and does not occupy the physical time slot of the current data transmission. The network control device accesses the system's internal policy database, locates the service template associated with the specific terminal, and reads the currently recorded pre-configured service data volume in the template. The pre-configured service data volume represents the nominal data scale value preset for the terminal by the system before the start of this scheduling cycle, based on historical experience or initial calibration. By obtaining this nominal value, the system provides a basic calibration baseline for subsequent adaptive learning algorithms.

[0115] Step 602: Perform a weighted smoothing calculation based on the actual business data volume and the currently recorded pre-configured business data volume to obtain the updated pre-configured business data volume; Specifically, the network control equipment extracts the precise actual business data volume reported by the terminal during this scheduling execution process and uses it as the latest observation sample characterizing the current physical environment's real data generation patterns. To filter out drastic fluctuations caused by occasional environmental noise during a single data acquisition, the network control equipment uses a preset update learning rate parameter to mathematically fuse the latest observation sample with historical baselines. The network control equipment performs weighted smoothing calculations using the following formula: D new =α*D actual +(1-α)*D old ; Among them, D new To calculate the updated pre-configured business data volume of the output, α is the dimensionless update learning rate parameter pre-configured by the system, and D... actual D represents the actual amount of business data acquired in this transaction. old This refers to the amount of pre-configured business data currently recorded above.

[0116] Furthermore, the dimensionless update learning rate parameter is strictly limited to a continuous range greater than 0 and less than 1. For example, this parameter can be set to 0.25. When the parameter value is small, the system operation relies more on long-term historical cumulative averages, exhibiting stronger resistance to impulse interference; when the parameter value is large, the system operation more sensitively tracks the recent data evolution trend. Through this smoothing mechanism based on the principle of exponentially weighted moving averages, the system can reduce systematic prediction biases caused by slow seasonal climate changes or continuous aging of equipment hardware, ensuring that the prior knowledge acquired by the network control equipment continuously closely matches the actual operating trajectory of the physical terminal.

[0117] Step 603: Replace the corresponding record in the business template with the updated pre-configured business data volume for use when generating scheduled tasks in the next business scheduling cycle.

[0118] After obtaining the calculation results, the network control device performs a database overwrite operation. The network control device replaces the old values ​​of the corresponding fields in the original service template with the updated pre-configured service data. When the next service trigger cycle arrives, the network control device will directly read the updated value as the initial input parameter, participating in the pre-congestion control calculation of Example 2 and the earliest completion time estimation of the simulation in Example 3. This asynchronous feedback update link connects discrete single scheduling processes into a closed-loop system with continuous self-learning capabilities in the time dimension. This technique effectively reduces the initial time slot deviation between the initial allocation and actual demand in Example 5, lowers the trigger probability of cross-cycle supplementary allocation or generation of continuation subtasks, and thus globally optimizes the channel utilization efficiency of the multiple access network.

[0119] In some alternative implementations, for service types with drastic periodic fluctuations, such as optical monitoring sensors that generate massive amounts of image data only during specific daytime periods, the aforementioned globally unified weighted smoothing calculation can be extended to an independent multi-state update mechanism based on time slices. The network control device pre-divides multiple independent sets of time intervals in the service template and maintains a pre-configured service data volume state machine for each time interval. After each task is completed, the system extracts the timestamp that triggered the task and maps it to a specific set of time intervals. The system independently performs weighted smoothing calculations only for the data state machine corresponding to the hit time interval. This time-layered alternative optimization strategy reduces mutual smoothing contamination between heterogeneous data sequences across time periods, further improving the accuracy of system resource pre-allocation in scenarios with periodic extreme fluctuations.

[0120] By utilizing a two-stage differential update mechanism based on the actual amount of data to be reported when ready, and combined with smooth self-learning prediction across cycles, the bandwidth fragmentation caused by fixed time slot allocation is reduced, which is conducive to improving the adaptation of channel resources to the actual physical payload.

[0121] Compared to the terminal's autonomous polling access mode, the single online duration of the terminal's RF module in this solution is compressed to the physical time span required for actual data transmission, eliminating the need for prolonged idle listening. Compared to the traditional DAMA mode with pre-configured fixed time slot allocation, this solution performs secondary adjustments based on the actual data volume. When the collected data volume is lower than expected, excess time slots can be released for other terminals to reuse; when the data volume is higher than expected, resuming subtasks can be dynamically added or generated, reducing the probability of bandwidth idleness or service interruption caused by fixed allocation.

[0122] Example 7: This example provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the low-power IoT terminal scheduling method of any one of Examples 1 to 6.

[0123] It should be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.

Claims

1. A low-power IoT terminal scheduling method, characterized in that, include: Obtain a preset business template and generate a task to be scheduled based on the business template; Obtain the current channel status, perform admission evaluation on the task to be scheduled based on the channel status, and determine the admission task and the target channel corresponding to the admission task; Send a pre-start command to the terminal associated with the access task to trigger the terminal to perform peripheral power-on and data preparation; After the receiving terminal completes data preparation, it reports the ready status. Based on the actual amount of service data carried in the ready status, it performs time slot allocation for the terminal on the target channel and obtains the time slot allocation result. The time slot allocation results are sent to the terminal so that the terminal can wake up the internal data plane module to send business data based on the time slot allocation results.

2. The method according to claim 1, characterized in that, Before performing admission assessment on the task to be scheduled based on the channel state, a preliminary congestion control step is also included: Obtain the pre-configured prediction window length and business triggering period; Determine the target evaluation channel, and calculate the expected number of tasks arriving at the target evaluation channel in the future time period based on the prediction window length and the service triggering period; Based on the expected number of arriving tasks and the total pre-acquired channel capacity, the expected channel load rate of the target evaluation channel is calculated, and the target evaluation channel is divided into corresponding congestion levels according to the expected channel load rate. Based on the congestion level and the priority carried in the tasks to be scheduled, the tasks to be scheduled are pre-filtered to obtain a set of candidate evaluation tasks. Perform operations to determine the admission task and target channel from the candidate evaluation task set.

3. The method according to claim 1, characterized in that, Based on the channel state, an admission assessment is performed on the task to be scheduled to determine the admission task and the target channel corresponding to the admission task, including: Multiple candidate channels are obtained. For the scheduled task that is not bound to a fixed physical channel, the earliest simulation completion time of the scheduled task is calculated in each of the multiple candidate channels. Based on the earliest completion time of the simulation, calculate the expected online duration of the task to be scheduled on each of the candidate channels; The earliest simulation completion time and the corresponding estimated online duration are compared with the pre-configured service timeout time and maximum online duration constraints, respectively. Candidate channels that simultaneously meet the constraints of service timeout time and maximum online duration are selected and constructed into a set of feasible channels; If the feasible channel set is not empty, the candidate channel with the earliest simulation completion time and the smallest time is selected from the feasible channel set as the target channel, and the task to be scheduled is confirmed as the admission task.

4. The method according to claim 3, characterized in that, If the feasible channel set is empty, select the candidate channel with the smallest constraint deviation from multiple candidate channels as the current optimal channel, obtain the earliest simulation completion time and the corresponding expected online duration of the current optimal channel, and execute the following admission decision logic: Determine whether the expected online duration exceeds the maximum online duration constraint. If it does, determine whether the task to be scheduled will be postponed to the next scheduling cycle. If the maximum online duration constraint is not exceeded, the timeout margin ratio is calculated based on the earliest simulation completion time, the business timeout time, and the task triggering time carried in the task to be scheduled. Compare the timeout margin ratio with the pre-configured margin ratio threshold; If the timeout margin ratio is greater than or equal to the margin ratio threshold, the task to be scheduled will be confirmed as an admission task. If the timeout margin ratio is less than the margin ratio threshold but not less than zero, then the task to be scheduled will be postponed to the next scheduling cycle. If the timeout margin ratio is less than zero, the task to be scheduled is deemed to be rejected.

5. The method according to claim 1, characterized in that, After sending the pre-launch instruction to the terminal associated with the admission task, the process also includes: The terminal's status is recorded as active on the network control device side; Before the network control device receives a task completion message or a task failure message reported by the terminal, it suspends sending new pre-start instructions to terminals that are recorded as active.

6. The method according to claim 1, characterized in that, Based on the actual amount of service data carried in the ready state, time slot allocation is performed for the terminal on the target channel to obtain the time slot allocation result, including: Retrieve the pre-configured business data volume defined in the business template; Calculate the actual time slot requirement corresponding to the actual amount of business data and the pre-configured time slot requirement corresponding to the pre-configured amount of business data, respectively. The actual time slot demand is compared with the pre-configured time slot demand to determine the target time slot deviation. Based on the target time slot deviation, the initial time slot resources reserved on the target channel based on the pre-configured time slot demand are differentially updated to generate the time slot allocation result.

7. The method according to claim 6, characterized in that, Based on the target time slot offset, the initial time slot resources reserved on the target channel are differentially updated, including: If the actual time slot demand is not greater than the pre-configured time slot demand, the actual time slot demand will be used as the allocation benchmark, and the idle time slot resources corresponding to the target time slot deviation will be released. If the actual time slot demand is greater than the pre-configured time slot demand, then the remaining available capacity of the target channel in the current scheduling period is detected. When the remaining available capacity accommodates the target time slot deviation, additional time slot resources corresponding to the target time slot deviation are allocated to the terminal. When the remaining available capacity is insufficient to accommodate the target time slot deviation, an allocation instruction is generated for the terminal according to the pre-configured time slot requirement.

8. The method according to claim 7, characterized in that, The method also includes: If the actual time slot demand is greater than the pre-configured time slot demand, and the remaining available capacity is insufficient to accommodate the target time slot deviation, the time slot allocation result for the current scheduling period is generated for the terminal according to the pre-configured time slot demand. Obtain the amount of remaining data in the actual business data volume that is not covered by the time slot allocation results; Based on the remaining data volume, a resume transmission subtask is registered and generated on the network control device side, and the resume transmission subtask is added to the queue to be scheduled for allocation of time slots in subsequent scheduling cycles.

9. The method according to claim 1, characterized in that, After the terminal completes sending the service data, the process also includes an asynchronous update step for the service template: Get the amount of pre-configured business data in the current record of the business template; A weighted smoothing calculation is performed based on the actual business data volume and the currently recorded pre-configured business data volume to obtain the updated pre-configured business data volume; Replace the corresponding record in the business template with the updated pre-configured business data volume, so that it can be called when generating the scheduled task in the next business scheduling cycle.

10. A low-power Internet of Things (IoT) terminal scheduling device, characterized in that, include: Memory, used to store computer programs; A processor for executing a computer program to implement the steps of the low-power Internet of Things (IoT) terminal scheduling method as described in any one of claims 1 to 9.

Images