Wireless network resource scheduling method and system

Abstract

The invention discloses a wireless network resource scheduling method and a wireless network resource scheduling system, relates to the technical field of wireless network scheduling, and aims to identify an area with abnormal and prominent resource requirements by performing standardization processing on the resource requirements so as to provide a clear quantitative basis for subsequent strategy adjustment. Compared with the existing empirical or fixed threshold scheduling mechanism, the method has dynamic adaptability and cross-regional comparability, and is convenient to realize a partition and hierarchical scheduling strategy. Through construction of scheduling strategy adjustment amount, three key control dimensions in scheduling logic, namely bandwidth channel priority, transmission time slot allocation and cache scheduling sequence, are broken through, so that each control module can perform linkage adjustment based on a unified scheduling urgency degree signal. And common problems of unbalanced resource configuration and the like are avoided. Through a standardized scheduling gradient calculation mechanism, the resource urgency degrees of different regions are quantitatively compared, and the problem of wrong resource investment caused by inflexible threshold setting and incomparable regions in a traditional scheduling system is solved.

Description

A method and system for scheduling wireless network resources Technical Field

[0001] This invention relates to the field of wireless network scheduling technology, specifically to a wireless network resource scheduling method and system. Background Technology

[0002] In wireless communication, wireless resource scheduling is a key factor determining network performance and user experience, encompassing sub-fields such as channel allocation, time slot allocation, buffer scheduling, and access control strategies. Specifically, in multi-user access environments, competition for resources such as bandwidth, time slots, and buffers among different terminal devices intensifies, requiring scheduling systems to dynamically allocate resources based on network conditions to meet quality of service requirements.

[0003] Most existing wireless resource scheduling systems employ real-time feedback mechanisms based on the current network state for scheduling decisions, such as bandwidth allocation and time slot adjustment based on channel quality, number of connected devices, or real-time load. These methods typically rely on instantaneous or short-term state parameters and lack the ability to model the periodic characteristics behind user connection behavior. Therefore, when facing scenarios such as fluctuations in user traffic, sudden changes in regional popularity, and periodic access peaks, they often fail to identify potential resource pressure in advance, resulting in delayed scheduling responses.

[0004] The root cause of these problems lies in the fact that traditional scheduling systems only understand user behavior at the current state response level, ignoring the long-term rhythmic characteristics of user access behavior. For example, a large number of users have high-frequency access behavior during fixed time periods every day or specific time periods every week. If the system cannot model and predict this behavior pattern, it will fail to prepare resources in advance before the peak access of hot users, resulting in a delay in scheduling response. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a wireless network resource scheduling method and system, which solves the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a wireless network resource scheduling system, comprising a user connection behavior extraction module, a rhythm model construction module, a resource demand prediction module, a regional scheduling priority determination module, a dynamic scheduling strategy control module, and a scheduling instruction output and feedback module; the user connection behavior extraction module extracts connection behavior data of user equipment from the network access control system and preprocesses it to obtain a connection behavior dataset WB; the rhythm model construction module extracts the user's unit-time access density p based on the connection behavior dataset WB and constructs a rhythm function A; the resource demand prediction module integrates the rhythm function A and the user spatial distribution to predict the regional resource scheduling demand trend and obtain a regional predicted resource demand function Qr; the regional scheduling priority determination module standardizes the regional predicted resource demand function Qr and calculates the scheduling tilt D; the dynamic scheduling strategy control module dynamically adjusts the resource allocation strategy based on the scheduling tilt D and constructs a scheduling strategy adjustment amount ΔS; the scheduling instruction output and feedback module sends the resource allocation strategy to the wireless resource scheduling controller, obtains the evaluation effect Eadj, and provides feedback.

[0007] Preferably, the user connection behavior extraction module includes a connection behavior cleaning and matching unit and a behavior data normalization and standardization unit. The connection behavior cleaning and matching unit extracts connection behavior data of user equipment from the network access control system, including user connection event LS, session duration TC, and data transmission start and end traffic SV, and fits it into an initial dataset CW. The user connection event LS is obtained through the user's connection status, with 1 indicating a successful connection and 0 indicating a failure. The session duration TC is obtained by the difference between the session end time and the session start time. The data transmission start and end traffic SV is obtained by the difference between the data's end traffic and its start traffic. The initial dataset CW is cleaned to remove data with duplicate access records and instantaneous event noise caused by reconnection. The data normalization and standardization unit processes the initial data... The initial dataset CW is time-aligned and normalized to obtain the connection behavior dataset WB. The time alignment process involves reconstructing the time series by aligning it with a fixed window. Within the fixed time window, a continuous time interval is selected starting from the current time point. Within this time interval, the data of each user in the initial dataset CW at each subdivided time point is read one by one. For each time point read, the data of the initial dataset CW is accumulated into a temporary statistic. After all the data in the entire time window has been accumulated, the accumulated result is divided by the number of time points contained in the time window to obtain the average data value within the time window. The normalization process uses the min-max normalization method to process the initial dataset CW, unifying the data dimensions and scaling the data of each user to the [0,1] interval.

[0008] Preferably, the rhythm model construction module includes a unit for calculating access density per unit time and a unit for fitting periodic rhythm functions. The unit for calculating access density per unit time extracts time series from the connection behavior dataset WB and establishes a model of user access density change per unit time. The entire time period is divided into continuous non-overlapping equal-length windows. In each window, user connection events LS are counted, and the unit-time access density p is calculated. The unit-time access density p is obtained as follows: Based on any time t, a fixed time window length Δt is extended forward. During this time period, the user connection events LS under each time slice are accumulated, and then divided by the window width Δt user connection events LS to obtain the user's unit-time access density p.

[0009] Preferably, the periodic rhythm function fitting unit identifies whether the user has repetitive activity patterns within a day, several hours, or a week based on the user's access density p per unit time, and calculates and obtains the rhythm function A; the rhythm function A is obtained in the following way: In the formula, Au(t) represents the rhythm function of user u at time t, W represents the width of the integration window, pu(t) represents the access density of user u per unit time at time t, sin represents the sine function, π represents pi, d represents the integral sign, and Tu represents the dominant period of user u. The dominant period Tu of user u is obtained by performing a fast Fourier transform on the access density p per unit time to identify the dominant frequency component and obtain the dominant period Tu of user u.

[0010] Preferably, the resource demand prediction module includes a user region distribution probability modeling unit and a region resource demand prediction function construction unit. The user region distribution probability modeling unit extracts historical access behavior records of users in different physical regions z, constructs a region distribution probability function reflecting connection tendencies, and obtains the access frequency N. The access frequency N is obtained by: traversing all historical connection records of user u; when it is found that the user's connection behavior at time point ti accesses region z, this event is counted as a valid record; finally, the number of all access events belonging to region z is summed to obtain the user's access frequency N; by analyzing the access frequency N, a user region distribution probability function Xp is constructed. The distribution probability function Xp is obtained by: first, for each user, counting the historical access counts in each region; then, dividing the user's access count in the target region Z by the total access count in all regions to obtain the distribution probability function Xp. The region resource demand prediction function construction unit integrates the rhythm function A of each user and the user region distribution probability function Xp to construct a region predicted resource demand function Qr. The region predicted resource demand function Qr is obtained through the following formula: In the formula, Qr(z,t) represents the regional predicted resource demand function of region z at time t, U represents the set of all users, Au(t) represents the connection activity of user u at time t, and Xpu(z) represents the probability distribution function of user u in region z.

[0011] If a region has multiple users at the peak of their circadian rhythm at a given moment, and these users have a high probability of regional distribution, then the predicted resource pressure in that region at that time point will increase significantly. This prediction function does not rely on real-time detection, but rather infers future resource demand trends through user behavior modeling, thus possessing strong foresight.

[0012] Preferably, the regional scheduling priority determination module includes a regional resource statistical feature extraction unit and a scheduling tilt calculation unit. The regional resource statistical feature extraction unit extracts historical resource demand statistical features based on the regional predicted resource demand function Qr, including the mean resource demand pQr and the standard deviation of resource demand σQr. The mean resource demand pQr is obtained as follows: a complete time period is selected, from the start time to the end time, continuously recording the resource demand of region z at each moment; then, the resource demand values ​​at all moments are summed to obtain a total resource demand; finally, the total is divided by the duration of the entire time period to obtain the mean resource demand pQr. The standard deviation of resource demand σQr is obtained as follows: first, the resource demand at each moment is compared with the regional mean resource demand pQr, and the difference is calculated; then, all differences are squared, and all squared results are summed over the entire time period; the sum is divided by the length of the time period to obtain the average fluctuation intensity of regional resource demand; finally, the square root of this result is taken to obtain the standard deviation of resource demand σQr.

[0013] Preferably, the scheduling tilt calculation unit standardizes the resource demand value at the current time point based on the mean resource demand pQr and the standard deviation of resource demand σQr, constructs the scheduling tilt D, and judges the resource status of the region; the scheduling tilt D is obtained by the following formula: In the formula, D(z,t) represents the scheduling tilt of region z at time t; the resource status of the region is obtained by matching the following formula: when the scheduling tilt of region z at time t is D(z,t) > 1, it indicates that the resources are abnormally tight and need to be scheduled first; when 1 ≥ the scheduling tilt of region z at time t is D(z,t) ≥ -1, it indicates that the current state is consistent with the normal state and no additional intervention is needed; when -1 > the scheduling tilt of region z at time t is D(z,t), it indicates that the resources are surplus and resources should be reclaimed and the priority reduced.

[0014] Preferably, the dynamic scheduling strategy control module includes a scheduling adjustment quantity construction unit and a strategy parameter mapping execution unit. The scheduling adjustment quantity construction unit constructs a scheduling strategy adjustment quantity ΔS associated with the actual resource allocation based on the scheduling skewness D. The formula is as follows: ΔS(z,t) = sd × D(z,t); where ΔS(z,t) represents the scheduling strategy adjustment quantity for region z at time t, and sd represents the strategy sensitivity factor. The strategy parameter mapping execution unit maps the scheduling strategy adjustment quantity ΔS to scheduling strategy parameters, including bandwidth channel priority, transmission time slot allocation, and buffer scheduling weight. The bandwidth channel priority is obtained by using the default scheduling level in the initial state of the region as a basis. The above is obtained by multiplying the scheduling policy adjustment amount ΔS by the scheduling level adjustment coefficient; the transmission time slot allocation is obtained by: first, setting a default base time slot time for each region, which represents the length of each round of transmission time allocated in the region under normal scheduling conditions; then, multiplying the region's current scheduling policy adjustment amount ΔS(z,t) by the sum of the adjustment factor and 1 to obtain a scaling ratio, and then multiplying it by the base time slot time to obtain the transmission time slot allocation; the cache scheduling weight is obtained by adding the scheduling policy adjustment amount ΔS and the cache scheduling adjustment factor to the initial weight used to control the order of cache data block transmission by default.

[0015] Preferably, the scheduling instruction output and feedback module packages the acquired scheduling strategy parameters, including bandwidth channel priority, transmission time slot allocation, and buffer scheduling weight, into scheduling instruction groups according to region, sends them to the wireless resource scheduling controller, and collects the operational indicators after scheduling execution to calculate and obtain the evaluation effect Eadj. The operational indicators include the actual resource allocation value Ra, access latency Tde, and packet loss rate Plo. The actual resource allocation value Ra represents the actual resources allocated at the current time point; the access latency Tde represents the average time elapsed from when a user initiates a connection request to when a connection is successfully established; the packet loss rate Plo represents the proportion of packets lost within a certain time window; the evaluation effect Eadj is obtained using the following formula: In the formula, Eadj(z) represents the evaluation effect of region z, T represents the evaluation effect, and Ra(z,t) represents the actual resource allocation value of region z at time t. The obtained evaluation effect Eadj is compared with the preset evaluation threshold Tead to evaluate the effectiveness of the strategy adjustment and provide feedback. The effectiveness of the strategy adjustment is obtained in the following ways: when the evaluation effect Eadj < the evaluation threshold Tead, it means that the scheduling strategy is accurately matched and effectively executed; when the evaluation effect Eadj ≥ the evaluation threshold Tead, it means that there is a disconnect between prediction and resource allocation, and the parameters need to be recalibrated.

[0016] A wireless network resource scheduling method includes the following steps: Step 1: A user connection behavior extraction module extracts connection behavior data of user equipment from the network access control system and preprocesses it to obtain a connection behavior dataset WB; Step 2: A rhythm model construction module extracts the user's unit-time access density p based on the connection behavior dataset WB and constructs a rhythm function A; Step 3: A resource demand prediction module integrates the rhythm function A and the user spatial distribution to predict the regional resource scheduling demand trend and obtain a regional predicted resource demand function Qr; Step 4: A regional scheduling priority determination module standardizes the regional predicted resource demand function Qr and calculates the scheduling tilt D; Step 5: A dynamic scheduling strategy adjustment module dynamically adjusts the resource allocation strategy based on the scheduling tilt D and constructs a scheduling strategy adjustment amount ΔS; Step 6: A scheduling instruction output and feedback module sends the resource allocation strategy to the wireless resource scheduling controller, obtains the evaluation effect Eadj, and provides feedback.

[0017] This invention provides a wireless network resource scheduling method and system, which has the following beneficial effects: (1) During system operation, by standardizing resource demand, areas with abnormally high resource demand are identified, providing a clear quantitative basis for subsequent strategy adjustments. Compared with existing experience-based or fixed threshold scheduling mechanisms, this method has dynamic adaptability and cross-regional comparability, making it easier to implement partitioned and hierarchical scheduling strategies. By constructing the scheduling strategy adjustment quantity, the three key control dimensions in the scheduling logic are connected: bandwidth channel priority, transmission time slot allocation, and buffer scheduling order, enabling each control module to make linkage adjustments based on a unified scheduling urgency signal. This not only improves the scheduling response speed and consistency, but also avoids common problems such as control conflicts and resource configuration imbalances.

[0018] (2) Three types of basic behavioral data, including user connection events, session duration, and data transmission traffic, are extracted from the existing network access control system through the connection behavior cleaning and matching unit, and uniformly fitted into an initial dataset. This structured ternary behavior modeling approach avoids the one-sidedness of relying solely on a single event to judge connection load in previous systems, and can more comprehensively restore the actual access status of users. The supporting data cleaning mechanism also removes noise such as duplicate access records and transient anomalies, so that the subsequent modeling process is based on clean and stable data, thereby improving the reliability and behavioral interpretability of the scheduling system from the source.

[0019] By setting a fixed time window to align user behavior data over time, and uniformly resampling and averaging data from different times and proportions, the modeling bias caused by inconsistent data collection time granularity is resolved. Simultaneously, a max-min normalization method is used to standardize the dimensions of data from different users, ensuring comparability under the same evaluation metric. This approach provides a stable and consistent input sequence for subsequent rhythm modeling, making it particularly suitable for cross-sectional joint analysis involving multiple users and regions.

[0020] (3) By using a user area distribution probability modeling unit, a user access preference model for each area is constructed based on historical access records. This approach breaks through the static dependence of traditional scheduling systems on the current connected area, and instead uses users' actual behavioral preferences to predict their possible future location areas. In this way, when the scheduling system performs resource planning, it no longer relies solely on real-time location or the number of accesses, but uses behavioral data to achieve pre-judgment of resources for specific areas, significantly enhancing the system's ability to perceive the potential emergence of hotspot areas.

[0021] (4) The adjustment amount through scheduling strategy is mapped to multiple scheduling control dimensions such as bandwidth channel priority, transmission time slot configuration, and cache scheduling weight. The parameter adjustment of each dimension is not only independent, but also based on the same dynamic driving index, which ensures logical coordination and resource configuration consistency between scheduling behaviors. This is significantly better than the situation in traditional systems where each scheduling strategy module runs independently, lacks linkage, and frequently causes strategy conflicts, which helps to achieve more efficient and non-overlapping resource management. Attached Figure Description

[0022] Figure 1 is a block diagram of a wireless network resource scheduling system according to the present invention; Figure 2 is a step diagram of a wireless network resource scheduling method according to the present invention; Figure 3 is a flowchart of obtaining the evaluation effect of the present invention; Figure 4 is a trend diagram of the predicted resource demand function in different regions according to the present invention. Detailed Implementation

[0023] 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. Embodiments

[0024] This invention provides a wireless network resource scheduling system, as shown in Figures 1 to 4. It includes a user connection behavior extraction module, a rhythm model construction module, a resource demand prediction module, a regional scheduling priority determination module, a dynamic scheduling strategy control module, and a scheduling command output and feedback module. The user connection behavior extraction module extracts connection behavior data of user equipment from the network access control system and preprocesses it to obtain a connection behavior dataset WB. The rhythm model construction module extracts the user's unit-time access density p based on the connection behavior dataset WB and constructs a rhythm function A. The resource demand prediction module integrates the rhythm function A and the user spatial distribution to predict the regional-level resource scheduling demand trend and obtain a regional predicted resource demand function Qr. The regional scheduling priority determination module standardizes the regional predicted resource demand function Qr and calculates the scheduling tilt D. The dynamic scheduling strategy control module dynamically adjusts the resource allocation strategy based on the scheduling tilt D and constructs a scheduling strategy adjustment amount ΔS. The scheduling command output and feedback module sends the resource allocation strategy to the wireless resource scheduling controller, obtains the evaluation effect Eadj, and provides feedback.

[0025] In this embodiment, by combining connection density with periodic activity patterns, the scheduling strategy is no longer based solely on the current load, but possesses predictive capabilities for the future, effectively overcoming the lag problem of traditional scheduling that relies on instantaneous state.

[0026] By integrating user behavior patterns with their historical access tendencies in various regions, a region-oriented resource demand prediction model was established. Unlike existing systems that solely rely on the number of devices or short-term statistics to determine hotspots, this invention can identify potential behavioral hotspot areas and pre-allocate resources before congestion occurs, effectively avoiding network instability caused by resource response delays.

[0027] By standardizing resource demands, regions with exceptionally high resource requirements are identified, providing a clear quantitative basis for subsequent strategy adjustments. Compared to existing experience-based or fixed-threshold scheduling mechanisms, this method offers dynamic adaptability and cross-regional comparability, facilitating the implementation of zoned and hierarchical scheduling strategies. By constructing scheduling strategy adjustment parameters, three key control dimensions in the scheduling logic are integrated: bandwidth channel priority, transmission time slot allocation, and buffer scheduling order, enabling each control module to make coordinated adjustments based on a unified scheduling urgency signal. This not only improves scheduling response speed and consistency but also avoids common problems such as control conflicts and resource allocation imbalances. Example

[0028] This embodiment is an explanation of Embodiment 1. Please refer to Figures 1 and 3. Specifically: the user connection behavior extraction module includes a connection behavior cleaning and matching unit and a behavior data normalization and standardization unit; the connection behavior cleaning and matching unit extracts the connection behavior data of the user equipment from the network access control system, including user connection event LS, session duration TC, and data transmission start and end traffic SV, and fits it into an initial dataset CW; wherein, the user connection event LS is obtained through the user's connection status, with 1 indicating a successful connection and 0 indicating a failure; the session duration TC is obtained through the difference between the session end time and the session start time; the data transmission start and end traffic SV is obtained through the difference between the data's end traffic and the data's start traffic; the initial dataset CW is cleaned to remove data with duplicate access records and instantaneous event noise generated by reconnection; The initial dataset CW is time-aligned and normalized according to the normalization and standardization units to obtain the connection behavior dataset WB. The time alignment process is performed by aligning and reconstructing the time series using a fixed window. Within the fixed time window, a continuous time interval is selected from the current time point as the starting position. Within this time interval, the data of the user in the initial dataset CW at each subdivided time point is read one by one. For each time point read, the data of the initial dataset CW is accumulated into a temporary statistic. When all the data in the entire time window has been accumulated, the accumulated result is divided by the number of time points contained in the time window to obtain the average data value within the time window. The normalization process is performed on the initial dataset CW using the min-max normalization method to unify the data dimensions and scale the data of each user to the [0,1] interval.

[0029] The rhythm model construction module includes a unit for calculating access density per unit time and a unit for fitting periodic rhythm functions. The unit for calculating access density per unit time extracts time series data from the connection behavior dataset WB and establishes a model of user access density changes per unit time. The entire time period is divided into continuous, non-overlapping, equal-length windows. In each window, user connection events LS are counted, and the access density per unit time p is calculated. The access density per unit time p is obtained as follows: Based on any time t, a fixed time window length Δt is extended forward. The user connection events LS under each time slice are accumulated within the time period, and then divided by the window width Δt user connection events LS to obtain the user's access density per unit time p.

[0030] The periodic rhythm function fitting unit identifies whether users have repetitive activity patterns within a day, several hours, or a week based on the user's access density p per unit time, and calculates and obtains the rhythm function A; the rhythm function A is obtained in the following way: In the formula, Au(t) represents the rhythm function of user u at time t, W represents the width of the integration window, pu(t) represents the access density of user u per unit time at time t, sin represents the sine function, π represents pi, d represents the integral sign, and Tu represents the dominant period of user u. The dominant period Tu of user u is obtained by performing a fast Fourier transform on the access density p per unit time to identify the dominant frequency component and obtain the dominant period Tu of user u.

[0031] In this embodiment, a connection behavior cleaning and matching unit extracts three types of basic behavioral data from the existing network access control system: user connection events, session duration, and data transmission traffic, and uniformly fits them into an initial dataset. This structured ternary behavior modeling approach avoids the one-sidedness of relying solely on a single event to judge connection load in previous systems, and can more comprehensively restore the actual user access status. The accompanying data cleaning mechanism also removes noise such as duplicate access records and transient anomalies, ensuring that the subsequent modeling process is based on clean and stable data, thereby improving the reliability and behavioral interpretability of the scheduling system from the source.

[0032] By setting a fixed time window to align user behavior data over time, and uniformly resampling and averaging data from different times and proportions, the modeling bias caused by inconsistent data collection time granularity is resolved. Simultaneously, a max-min normalization method is used to standardize the dimensions of data from different users, ensuring comparability under the same evaluation metric. This approach provides a stable and consistent input sequence for subsequent rhythm modeling, making it particularly suitable for cross-sectional joint analysis involving multiple users and regions.

[0033] The rhythm model construction module quantifies the intensity of user behavior within a certain time period by constructing access density per unit time. It also identifies the dominant connection cycle of users based on a periodic function fitting method, thereby generating a dynamic rhythm function. This function realistically reflects the repetitive patterns of user behavior over time, enabling the system to predict whether users will exhibit high-frequency connection behavior before a certain future moment arrives, thus possessing time-oriented predictive perception capabilities.

[0034] Example 3 is an explanation of Example 2. Please refer to Figures 3 and 4. Specifically, the resource demand prediction module includes a user area distribution probability modeling unit and an area resource demand prediction function construction unit. The user area distribution probability modeling unit extracts historical access behavior records of users in different physical areas z, constructs an area distribution probability function reflecting connection tendency, and obtains the access frequency N. The access frequency N is obtained by traversing all historical connection records of user u. When it is found that the user's connection behavior at time point ti is connected to area z, this event is counted as a valid record. Finally, all records belonging to area z are... The number of access events is accumulated to obtain the user's access frequency N. By analyzing the access frequency N, a distribution probability function Xp for the user's region is constructed. The distribution probability function Xp is obtained as follows: First, for each user, the historical access count in each region is counted. Then, the user's access count in the target region Z is divided by the total access count across all regions to obtain the distribution probability function Xp. The regional resource demand prediction function construction unit integrates the rhythm function A of each user and the distribution probability function Xp of the user's region to construct the regional predicted resource demand function Qr. The regional predicted resource demand function Qr is obtained using the following formula: In the formula, Qr(z,t) represents the regional predicted resource demand function of region z at time t, U represents the set of all users, Au(t) represents the connection activity of user u at time t, and Xpu(z) represents the probability distribution function of user u in region z.

[0035] In this embodiment, a user region distribution probability modeling unit is used to construct a user access preference model for each region based on historical access records. This approach breaks through the static dependence of traditional scheduling systems on the current connected region, instead using users' actual behavioral preferences to predict their potential future location regions. As a result, when planning resources, the scheduling system no longer relies solely on real-time location or the number of accesses, but instead uses behavioral data to achieve region-oriented pre-judgment of resources, significantly enhancing the system's ability to perceive the potential emergence of hotspot areas.

[0036] In this embodiment, the system not only constructs a spatial probability function based on users' connection preferences in different regions, but also integrates it with the user's own connection rhythm function to form a spatiotemporally driven prediction mechanism. Compared to traditional schemes that only consider time trends or spatial loads, this mechanism can more precisely characterize the specific scheduling demand trends of "a certain type of user may appear in a certain region at a certain time." This strategy greatly improves the accuracy of regional resource allocation, and is especially suitable for network environments with high concurrency access and strong behavioral concentration, effectively reducing resource misallocation and network fluctuations caused by prediction errors.

[0037] By calculating the output of the regional resource demand prediction function, the system provides a clear and continuously computable input for the subsequent priority determination module. The prediction results not only reflect the strength of users' connection intentions over a certain period of time, but also possess the characteristics of being standardized and comparable across regions, providing a basis for prioritizing scheduling across different regions. This mechanism solves the problems of traditional systems struggling to quantify and prioritize in multi-region, multi-user concurrent scenarios, and the resulting ambiguity in decision-making. It facilitates advanced resource management strategies such as precise scheduling, regional control, and dynamic tilting.

[0038] Example 4 is an explanation of Example 3. Please refer to Figures 3 and 4. Specifically, the regional scheduling priority determination module includes a regional resource statistical feature extraction unit and a scheduling tilt calculation unit. The regional resource statistical feature extraction unit extracts historical resource demand statistical features of the region based on the regional predicted resource demand function Qr, including the mean resource demand pQr and the standard deviation of resource demand σQr. The mean resource demand pQr is obtained by selecting a complete time period, from the start time point to the end time point, and continuously recording the resource demand of region z at each moment. Then... The resource demand values ​​at all times are summed up to obtain a total resource demand. Finally, the total is divided by the duration of the entire time period to obtain the mean resource demand pQr. The standard deviation of resource demand σQr is obtained as follows: First, the resource demand at each time point is compared with the mean resource demand pQr of the region, and the difference is calculated. Then, all differences are squared, and the squared results are summed up over the entire time period. The sum is divided by the length of the time period to obtain the average fluctuation intensity of regional resource demand. Finally, the square root of this result is taken to obtain the standard deviation of resource demand σQr.

[0039] The scheduling tilt calculation unit standardizes the resource demand value at the current time point based on the mean resource demand pQr and the standard deviation of resource demand σQr, constructs the scheduling tilt D, and judges the resource status of the region; the scheduling tilt D is obtained by the following formula: In the formula, D(z,t) represents the scheduling tilt of region z at time t; the resource status of the region is obtained by matching the following formula: when the scheduling tilt of region z at time t is D(z,t) > 1, it indicates that the resources are abnormally tight and need to be scheduled first; when 1 ≥ the scheduling tilt of region z at time t is D(z,t) ≥ -1, it indicates that the current state is consistent with the normal state and no additional intervention is needed; when -1 > the scheduling tilt of region z at time t is D(z,t), it indicates that the resources are surplus and resources should be reclaimed and the priority reduced.

[0040] In this embodiment, a regional resource statistical feature extraction unit is used to quantitatively analyze the resource demand of each region within a set period, calculating the mean and standard deviation of the resource demand to reflect the normal resource utilization level and fluctuation range of the region. Compared with the problems of traditional systems using fixed thresholds or uniform strategies across the network, this mechanism can establish a local benchmark with context-aware capabilities based on the historical behavior of each region, significantly improving the adaptability and accuracy of scheduling strategies.

[0041] By comparing the current resource demand with historical averages and standardizing it using standard deviation, the system outputs a dimensionless scheduling skewness index to measure the degree of anomaly in current demand. This index is comparable across different regions, not only solving the misalignment problem in multi-region scheduling priority evaluation but also providing continuous and hierarchical input parameters for subsequent scheduling control modules. This facilitates fine-grained control strategies such as elastic resource allocation, dynamic window adjustment, and on-demand forwarding control.

[0042] Based on the positive and negative range of the scheduling tilt value, the current resource status is divided into three categories: resource shortage, normal range, and resource surplus, and different scheduling response mechanisms are dynamically triggered accordingly. In particular, when resources suddenly increase, the system can identify the abnormal state immediately and upgrade the region to a scheduling priority, avoiding the problems of slow response and lagging resource allocation in the face of sudden traffic surges in traditional mechanisms. In the state of resource surplus, redundant configurations can be actively reclaimed to achieve resource reallocation and overall scheduling efficiency optimization.

[0043] Example 5 is an explanation of Example 4. Please refer to Figure 1. Specifically, the dynamic scheduling strategy control module includes a scheduling adjustment amount construction unit and a strategy parameter mapping execution unit. The scheduling adjustment amount construction unit constructs a scheduling strategy adjustment amount ΔS associated with the actual resource allocation based on the scheduling tilt D. The formula is as follows: ΔS(z,t) = sd × D(z,t); where ΔS(z,t) represents the scheduling strategy adjustment amount of region z at time t, and sd represents the strategy sensitivity factor. The strategy parameter mapping execution unit maps the scheduling strategy adjustment amount ΔS to scheduling strategy parameters, including bandwidth channel priority, transmission time slot allocation, and buffer scheduling weight. The bandwidth channel priority is obtained by: in the region... The transmission time slot allocation is obtained by multiplying the scheduling policy adjustment amount ΔS by the scheduling level adjustment coefficient on the basis of the default scheduling level in the initial state; the transmission time slot allocation is obtained by first setting a default base time slot time for each region, which represents the length of each round of transmission time allocated in the region under normal scheduling conditions; then, multiplying the current scheduling policy adjustment amount ΔS(z,t) of the region by the sum of the adjustment factor and 1 to obtain a scaling ratio, and then multiplying it by the base time slot time to obtain the transmission time slot allocation; the cache scheduling weight is obtained by adding the scheduling policy adjustment amount ΔS and the cache scheduling adjustment factor to the initial weight used to control the sending order of cache data blocks by default.

[0044] The scheduling instruction output and feedback module acquires scheduling policy parameters, including bandwidth channel priority, transmission time slot allocation, and buffer scheduling weight, and packages them into scheduling instruction groups according to region. These groups are then sent to the wireless resource scheduling controller. The module also collects operational metrics after scheduling execution and calculates the evaluation result Eadj. These operational metrics include the actual resource allocation value Ra, access latency Tde, and packet loss rate Plo. The actual resource allocation value Ra represents the actual resources allocated at the current time point; access latency Tde represents the average time elapsed from initiating a connection request to successfully establishing a connection; and packet loss rate Plo represents the proportion of packets lost within a certain time window. The evaluation result Eadj is obtained using the following formula: In the formula, Eadj(z) represents the evaluation effect of region z, T represents the evaluation effect, and Ra(z,t) represents the actual resource allocation value of region z at time t. The obtained evaluation effect Eadj is compared with the preset evaluation threshold Tead to evaluate the effectiveness of the strategy adjustment and provide feedback. The effectiveness of the strategy adjustment is obtained in the following ways: when the evaluation effect Eadj < the evaluation threshold Tead, it means that the scheduling strategy is accurately matched and effectively executed; when the evaluation effect Eadj ≥ the evaluation threshold Tead, it means that there is a disconnect between prediction and resource allocation, and the parameters need to be recalibrated.

[0045] In this embodiment, scheduling skewness is introduced as a dynamic input variable into the system, and the scheduling policy adjustment amount is generated by multiplying it by a policy sensitivity factor, thereby driving the scheduling behavior to change as needed. This mechanism avoids the traditional one-size-fits-all or statically table-driven policy design of scheduling systems, realizing continuous fine-tuning and rapid response of resource allocation policies. Different regions can have different scheduling response intensities at different times, giving the system a high degree of behavioral adaptability and control flexibility.

[0046] The adjustment of scheduling policies is mapped to multiple scheduling control dimensions, such as bandwidth channel priority, transmission time slot configuration, and cache scheduling weight. The parameter adjustments for each dimension are not only independent but also based on the same dynamic driving indicator, ensuring logical coordination and resource configuration consistency between scheduling actions. This is significantly better than the situation in traditional systems where scheduling policy modules operate independently, lack coordination, and frequently experience policy conflicts, thus contributing to more efficient and non-overlapping resource management.

[0047] The scheduling strategy parameters for each region are packaged into instruction groups and distributed regionally, giving scheduling control spatially directional control. Combined with the ability to continuously adjust regional tilt, differentiated scheduling execution can be achieved by region. This allows for timely resource release in areas experiencing hotspots and proactive resource reduction in low-load areas, achieving precise scheduling logic that allocates only what is needed and avoiding resource allocation redundancy or blind spots. After scheduling execution is complete, the system proactively collects key operational indicators such as actual resource allocation values, average access latency, and data packet loss rate. By comparing these with predicted resource values ​​for each region, evaluation performance indicators are calculated.

[0048] Example 6: A wireless network resource scheduling method, as shown in Figure 2, specifically includes the following steps: Step 1: A user connection behavior extraction module extracts connection behavior data of user equipment from the network access control system and preprocesses it to obtain a connection behavior dataset WB; Step 2: A rhythm model construction module extracts the user's unit-time access density p based on the connection behavior dataset WB and constructs a rhythm function A; Step 3: A resource demand prediction module integrates the rhythm function A and the user spatial distribution to predict the regional resource scheduling demand trend and obtain a regional predicted resource demand function Qr; Step 4: A regional scheduling priority determination module standardizes the regional predicted resource demand function Qr and calculates the scheduling tilt D; Step 5: A dynamic scheduling strategy control module dynamically adjusts the resource allocation strategy based on the scheduling tilt D and constructs a scheduling strategy adjustment amount ΔS; Step 6: A scheduling instruction output and feedback module sends the resource allocation strategy to the wireless resource scheduling controller, obtains the evaluation effect Eadj, and provides feedback.

[0049] In this embodiment, starting from user connection behavior, temporal data features are extracted to construct a rhythm function, fundamentally changing the traditional reactive scheduling model that relies solely on real-time status. By transforming user behavior patterns into predictable data trends and combining them with spatial distribution to predict regional resource demand, the system achieves a shift in scheduling decisions from passive response to proactive planning, significantly enhancing the predictability of scheduling and the ability to pre-allocate resources.

[0050] By employing a standardized scheduling tilt calculation mechanism, the urgency of resources in different regions and at different times can be quantitatively compared, breaking the "resource misallocation" problem caused by rigid threshold settings and incomparable regions in traditional scheduling systems. The dynamic calculation of scheduling priorities enables the system to identify spatiotemporal distribution and adjust local responses, allowing for early intervention before a surge in hotspot areas and automatic reduction of allocation in low-demand areas, thus achieving a resource allocation strategy that differentiates between priorities.

[0051] This method maps the uniformly generated scheduling policy adjustments to multiple control dimensions, including bandwidth channels, transmission time slots, and cache scheduling, thereby achieving synchronous optimization of the scheduling policy across physical links, time scheduling, and cache management. Compared to the traditional system's independent-dimensional policy configuration, which is prone to resource contention or control conflicts, this method can coordinate scheduling actions between modules, improving the consistency and accuracy of overall resource allocation.

[0052] After executing the strategy, the system not only collects resource allocation performance metrics but also calculates the error between the strategy and the demand, comparing it with a threshold to determine if the current strategy is suitable. By feeding the evaluation results back to the system, a scheduling self-correction based on execution performance is formed. The system has self-optimization and dynamic parameter adjustment capabilities, effectively preventing strategy mismatch, prediction drift, or resource waste.

[0053] 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 variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended technical solutions and their equivalents.

Claims

1. A wireless network resource scheduling system, characterized in that: It includes a user connection behavior extraction module, a rhythm model construction module, a resource demand prediction module, a regional scheduling priority determination module, a dynamic scheduling strategy control module, and a scheduling instruction output and feedback module. The user connection behavior extraction module extracts the connection behavior data of user equipment from the network access control system, performs preprocessing, and obtains the connection behavior dataset WB. The rhythm model construction module extracts the user's unit time access density p based on the connection behavior dataset WB and constructs the rhythm function A. The resource demand forecasting module integrates the rhythm function A and the user space distribution to predict the regional resource scheduling demand trend and obtain the regional forecast resource demand function Qr. The regional scheduling priority determination module standardizes the regional predicted resource demand function Qr and calculates the scheduling tilt D; the dynamic scheduling strategy control module dynamically adjusts the resource allocation strategy based on the scheduling tilt D and constructs the scheduling strategy adjustment amount ΔS; the scheduling instruction output and feedback module sends the resource allocation strategy to the wireless resource scheduling controller, obtains the evaluation effect Eadj, and provides feedback.

2. The wireless network resource scheduling system according to claim 1, characterized in that: The user connection behavior extraction module includes a connection behavior cleaning and matching unit and a behavior data normalization and standardization unit. The connection behavior cleaning and matching unit extracts connection behavior data of user equipment from the network access control system, including user connection events (LS), session duration (TC), and data transmission start and end traffic (SV), and fits it into an initial dataset (CW). Specifically, the user connection event (LS) is obtained through the user's connection status, with 1 indicating a successful connection and 0 indicating a failure; the session duration (TC) is obtained by the difference between the session end time and the session start time; and the data transmission start and end traffic (SV) is obtained by the difference between the end traffic and the start traffic. The initial dataset (CW) is cleaned to remove data with duplicate access records and transient event noise caused by reconnections. The data normalization and standardization unit then processes the initial dataset (CW)... W undergoes time alignment and normalization to obtain the connection behavior dataset WB. Time alignment is achieved by reconstructing the time series using a fixed window. Within this fixed time window, a continuous time interval is selected, starting from the current time point. Within this time interval, data from the initial dataset CW is read one by one at each subdivided time point. For each time point read, the data from the initial dataset CW is accumulated into a temporary statistic. After all data within the entire time window has been accumulated, the accumulated result is divided by the number of time points contained in the time window to obtain the average data value within the time window. Normalization is performed on the initial dataset CW using the min-max normalization method to unify the data dimensions and scale each user's data to the [0,1] interval.

3. The wireless network resource scheduling system according to claim 2, characterized in that: The rhythm model construction module includes a unit for calculating access density per unit time and a unit for fitting periodic rhythm functions. The unit for calculating access density per unit time extracts time series data from the connection behavior dataset WB and establishes a model of user access density changes per unit time. The entire time period is divided into continuous, non-overlapping, equal-length windows. In each window, user connection events LS are counted, and the access density per unit time p is calculated. The access density per unit time p is obtained as follows: Based on any time t, a fixed time window length Δt is extended forward. The user connection events LS under each time slice are accumulated within the time period, and then divided by the window width Δt user connection events LS to obtain the user's access density per unit time p.

4. A wireless network resource scheduling system according to claim 3, characterized in that: The periodic rhythm function fitting unit identifies whether users have repetitive activity patterns within a day, several hours, or a week based on the user's access density p per unit time, and calculates and obtains the rhythm function A; the rhythm function A is obtained in the following way: In the formula, Au(t) represents the rhythm function of user u at time t, W represents the width of the integration window, pu(t) represents the access density of user u per unit time at time t, sin represents the sine function, π represents pi, d represents the integral sign, and Tu represents the dominant period of user u. The dominant period Tu of user u is obtained by performing a fast Fourier transform on the access density p per unit time to identify the dominant frequency component and obtain the dominant period Tu of user u.

5. A wireless network resource scheduling system according to claim 4, characterized in that: The resource demand prediction module includes a user region distribution probability modeling unit and a region resource demand prediction function construction unit. The user region distribution probability modeling unit extracts historical access behavior records of users in different physical regions z, constructs a region distribution probability function reflecting connection tendencies, and obtains the access frequency N. The access frequency N is obtained by iterating through all historical connection records of user u; when it is found that the user's connection behavior at time point ti accesses region z, this event is counted as a valid record; finally, the number of all access events belonging to region z is summed to obtain the user's access frequency N. By analyzing the access frequency N, a user region distribution probability function Xp is constructed. The distribution probability function Xp is obtained by first counting the historical access counts of each user in each region; then, dividing the user's access count in the target region Z by the total access count of the user in all regions, the distribution probability function Xp is obtained. The region resource demand prediction function construction unit integrates the rhythm function A of each user and the user region distribution probability function Xp to construct the region predicted resource demand function Qr. The region predicted resource demand function Qr is obtained through the following formula: In the formula, Qr(z,t) represents the regional predicted resource demand function of region z at time t, U represents the set of all users, Au(t) represents the connection activity of user u at time t, and Xpu(z) represents the probability distribution function of user u in region z.

6. A wireless network resource scheduling system according to claim 5, characterized in that: The regional scheduling priority determination module includes a regional resource statistical feature extraction unit and a scheduling tilt calculation unit; the regional resource statistical feature extraction unit extracts the historical resource demand statistical features of the region based on the regional predicted resource demand function Qr, including the resource demand mean pQr and the resource demand standard deviation σQr; The mean resource demand pQr is obtained as follows: Select a complete time period, from the start time to the end time, and continuously record the resource demand of region z at each moment; then, sum up the resource demand values ​​at all moments to obtain a total resource demand; finally, divide the total by the duration of the entire time period to obtain the mean resource demand pQr. The standard deviation of resource demand σQr is obtained as follows: First, compare the resource demand at each moment with the mean resource demand pQr of the region and calculate the difference; then, square all the differences, and finally sum up all the squared results over the entire time period. Divide the sum by the length of the time period to obtain the average fluctuation intensity of regional resource demand; finally, take the square root of this result to obtain the standard deviation of resource demand σQr.

7. A wireless network resource scheduling system according to claim 6, characterized in that: The scheduling tilt calculation unit standardizes the resource demand value at the current time point based on the mean resource demand pQr and the standard deviation of resource demand σQr, constructs the scheduling tilt D, and judges the resource status of the region; the scheduling tilt D is obtained by the following formula: In the formula, D(z,t) represents the scheduling tilt of region z at time t; the resource status of the region is obtained by matching the following formula: when the scheduling tilt of region z at time t is D(z,t) > 1, it indicates that the resources are abnormally tight and need to be scheduled first; when 1 ≥ the scheduling tilt of region z at time t is D(z,t) ≥ -1, it indicates that the current state is consistent with the normal state and no additional intervention is needed; when -1 > the scheduling tilt of region z at time t is D(z,t), it indicates that the resources are surplus and resources should be reclaimed and the priority reduced.

8. A wireless network resource scheduling system according to claim 7, characterized in that: The dynamic scheduling strategy control module includes a scheduling adjustment quantity construction unit and a strategy parameter mapping execution unit. The scheduling adjustment quantity construction unit constructs a scheduling strategy adjustment quantity ΔS associated with the actual resource allocation based on the scheduling tilt D. The formula is as follows: ΔS(z,t) = sd × D(z,t); where ΔS(z,t) represents the scheduling strategy adjustment quantity for region z at time t, and sd represents the strategy sensitivity factor. The strategy parameter mapping execution unit maps the scheduling strategy adjustment quantity ΔS to scheduling strategy parameters, including bandwidth channel priority, transmission time slot allocation, and buffer scheduling weight. The bandwidth channel priority is obtained by using the default scheduling level in the initial state of the region. The transmission time slot allocation is obtained by multiplying the scheduling policy adjustment amount ΔS by the scheduling level adjustment coefficient. The method for obtaining the transmission time slot allocation is as follows: First, a default base time slot time is set for each region. This value represents the length of each round of transmission time allocated to the region under normal scheduling conditions. Then, the region's current scheduling policy adjustment amount ΔS(z,t) is multiplied by the sum of the adjustment factor and 1 to obtain a scaling ratio, which is then multiplied by the base time slot time to obtain the transmission time slot allocation. The cache scheduling weight is obtained by adding the scheduling policy adjustment amount ΔS and the cache scheduling adjustment factor to the initial weight used to control the order of cache data block transmission by default.

9. A wireless network resource scheduling system according to claim 8, characterized in that: The scheduling instruction output and feedback module acquires scheduling policy parameters, including bandwidth channel priority, transmission time slot allocation, and buffer scheduling weight, and packages them into scheduling instruction groups according to region. These groups are then sent to the wireless resource scheduling controller. The module also collects operational metrics after scheduling execution and calculates the evaluation result Eadj. These operational metrics include the actual resource allocation value Ra, access latency Tde, and packet loss rate Plo. The actual resource allocation value Ra represents the actual resources allocated at the current time point; access latency Tde represents the average time elapsed from when a user initiates a connection request to when a connection is successfully established; and the packet loss rate Plo represents the proportion of packets lost within a certain time window. The evaluation result Eadj is obtained using the following formula: In the formula, Eadj(z) represents the evaluation effect of region z, T represents the evaluation effect, and Ra(z,t) represents the actual resource allocation value of region z at time t. The obtained evaluation effect Eadj is compared with the preset evaluation threshold Tead to evaluate the effectiveness of the strategy adjustment and provide feedback. The effectiveness of the strategy adjustment is obtained in the following ways: when the evaluation effect Eadj < the evaluation threshold Tead, it means that the scheduling strategy is accurately matched and effectively executed; when the evaluation effect Eadj ≥ the evaluation threshold Tead, it means that there is a disconnect between prediction and resource allocation, and the parameters need to be recalibrated.

10. A wireless network resource scheduling method, applied to a wireless network resource scheduling system according to any one of claims 1 to 9, characterized in that: The process includes the following steps: Step 1: The user connection behavior extraction module extracts connection behavior data of user equipment from the network access control system, performs preprocessing, and obtains a connection behavior dataset WB; Step 2: The rhythm model construction module extracts the user's unit-time access density p based on the connection behavior dataset WB and constructs a rhythm function A; Step 3: The resource demand prediction module integrates the rhythm function A and the user spatial distribution to predict the regional resource scheduling demand trend and obtains the regional predicted resource demand function Qr; Step 4: The regional scheduling priority determination module standardizes the regional predicted resource demand function Qr and calculates the scheduling tilt D; Step 5: The dynamic scheduling strategy adjustment module dynamically adjusts the resource allocation strategy based on the scheduling tilt D and constructs the scheduling strategy adjustment amount ΔS; Step 6: The scheduling instruction output and feedback module sends the resource allocation strategy to the wireless resource scheduling controller, obtains the evaluation effect Eadj, and provides feedback.

Images