Method and apparatus for ap switching in pon and onu environment

Abstract

The application is suitable for the technical field of wireless network, and provides an AP switching method and device under PON and ONU environment, which can be applied to Fiber To The Room (FTTR), Fiber To The Room-Business (FTTR-B) and broadband fusion terminal products, the AP switching method under PON and ONU environment comprises the following steps: collecting multi-dimensional link quality parameters of a current access point in real time, and calculating a link quality degradation rate and direction information based on the multi-dimensional link quality parameters; when the link quality degradation rate exceeds a preset trigger threshold, a candidate access point set is obtained; and roaming switching to the target access point is performed. This multi-dimensional analysis method makes the switching process more intelligent, which is helpful to quickly and accurately identify link quality problems and respond in time.

Description

Technical Field

[0001] This invention belongs to the technical field of wireless networks, and particularly relates to an AP switching method and device in PON and ONU environments. Background Technology

[0002] With the rapid development of information technology, users' requirements for network access quality are increasing. Fiber optic access networks (PON), with their advantages of high bandwidth and low latency, have become one of the important technologies for broadband access. In a PON environment, the optical network unit (ONU), acting as a bridge between the user terminal and the optical line terminal (OLT), bears the heavy responsibility of data transmission. In this environment, access point (AP) switching technology is particularly important to improve user experience and ensure network stability and reliability.

[0003] Existing AP handover solutions often rely on single link quality metrics, such as signal strength or packet loss rate, for decision-making, lacking a comprehensive consideration of multi-dimensional link quality parameters. In complex network environments with high user mobility, this approach often fails to respond promptly to changes in link quality, leading to delays or interruptions during handover and negatively impacting the user's internet experience. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide an AP switching method and apparatus in PON and ONU environments to solve the technical problem that traditional methods often cannot respond in a timely manner to changes in link quality in complex network environments and with high user mobility, resulting in delays or interruptions for users during the switching process.

[0005] A first aspect of this invention provides an AP handover method under PON and ONU environments, the AP handover method under PON and ONU environments comprising: Real-time collection of multi-dimensional link quality parameters of the current access point, and calculation of link quality degradation rate and its direction information based on the multi-dimensional link quality parameters; When the link quality degradation rate exceeds a preset trigger threshold, the target scanning sector is determined in the pre-constructed access point topology map according to the direction information, and a directional scan is performed on the candidate access points in the target scanning sector to obtain a set of candidate access points. For each candidate access point in the candidate access point set, multiple optimization objectives are constructed by combining its predicted link quality, load status and historical handover success rate, and the target access point is determined by Pareto optimization. The process involves performing a roaming handover to the target access point, adaptively adjusting the hysteresis parameters based on the roaming context state, and feeding back the adjusted hysteresis parameters to the update process of the trigger threshold. The step of constructing multiple optimization objectives for each candidate access point in the candidate access point set by comprehensively considering its predicted link quality, load status, and historical handover success rate, and determining the target access point through Pareto optimization includes: For each candidate access point in the candidate access point set, its predicted link quality is taken as the first optimization objective, the current load status is taken as the second optimization objective, and the historical handover success rate is taken as the third optimization objective. The first optimization objective, the second optimization objective, and the third optimization objective are normalized respectively to unify the value range of each optimization objective to between 0 and 1; In the set of candidate access points, a dominance relationship is defined as the condition that any candidate access point is not inferior to another candidate access point in all optimization objectives and the difference between them in at least one optimization objective is greater than a threshold. From the set of candidate access points, extract the Pareto optimal solution set that is not dominated by other candidate access points; When the Pareto optimal solution set contains only one candidate access point, it is determined as the target access point; When the Pareto optimal solution set contains multiple candidate access points, the sum of the Euclidean distances from each Pareto optimal solution to all other solutions in the Pareto front is calculated as the inflection point metric, and the candidate access point with the largest inflection point metric is selected as the target access point.

[0006] Furthermore, the step of real-time acquisition of multi-dimensional link quality parameters of the current access point, and calculation of link quality degradation rate and its direction information based on the multi-dimensional link quality parameters, includes: Real-time acquisition of the received signal strength indication, signal-to-noise ratio, data packet retransmission rate and channel utilization of the current access point as multi-dimensional link quality parameters; Each dimension of the multi-dimensional link quality parameters is normalized, and the normalized parameters are calculated by weighted geometric mean according to preset weights to obtain a comprehensive link quality score. The link quality comprehensive scores are arranged in chronological order to construct a link quality vector; The link quality vector is input into a long short-term memory network, and the link quality prediction sequence within a future preset time window is output. Calculate the link quality difference between adjacent time points in the link quality prediction sequence to obtain the link quality change rate sequence; The average of the last preset number of elements in the link quality change rate sequence is taken as the link quality degradation rate. The direction of degradation is determined based on the sign of the link quality degradation rate.

[0007] Furthermore, the step of determining a target scanning sector in a pre-constructed access point topology map based on the direction information when the link quality degradation rate exceeds a preset trigger threshold, and performing a directional scan on candidate access points within the target scanning sector to obtain a set of candidate access points includes: The node corresponding to the current access point and the set of neighboring nodes that have a topological adjacency relationship with the current access point are obtained from the access point topology graph; wherein, the access point topology graph uses the access point as the node and the historical switching frequency and signal coverage overlap between access points as the edge weights. The direction of movement of the current access point in the access point topology map is determined based on the direction information; Centered on the node corresponding to the current access point, with the direction of movement as the axis, and with a preset angle range as the subtended angle, a fan-shaped area is extracted from the access point topology map as the target scanning sector; Obtain the access points corresponding to all neighboring nodes within the target scan sector as a list of access points to be scanned; For each access point in the list of access points to be scanned, a probe request frame is sent and a probe response frame is received on its working channel to collect the real-time link quality parameters of each access point. The access points that receive probe response frames and their corresponding real-time link quality parameters are used to form a set of candidate access points.

[0008] Furthermore, before the step of obtaining the node corresponding to the current access point and the set of neighboring nodes that have a topological adjacency relationship with the current access point from the access point topology graph, the method further includes: Collect historical roaming handover records; each record includes the source access point identifier, the target access point identifier, whether the handover was successful, and the signal strength difference at the time of handover. For any two access point nodes in the access point topology graph that have overlapping signal coverage, count their historical successful handover counts and failed handover counts, and use the Bayesian estimation method to calculate their posterior handover success probability as the weight of the edge between the two nodes. Edges with weights lower than a preset threshold are removed from the access point topology graph, and the remaining edges constitute the adjacency relationship of the access point topology graph. When a new roaming switch is completed, the number of successful and failed switches between the corresponding node pairs is updated based on the switch results, and the edge weights are recalculated.

[0009] Furthermore, the step of constructing multiple optimization objectives for each candidate access point in the candidate access point set by comprehensively considering its predicted link quality, load status, and historical handover success rate, and determining the target access point through Pareto optimization includes: For each candidate access point in the candidate access point set, its predicted link quality is taken as the first optimization objective, the current load status is taken as the second optimization objective, and the historical handover success rate is taken as the third optimization objective. The first optimization objective, the second optimization objective, and the third optimization objective are normalized respectively to unify the value range of each optimization objective to between 0 and 1; In the set of candidate access points, a dominance relationship is defined as the condition that any candidate access point is not inferior to another candidate access point in all optimization objectives and the difference between them in at least one optimization objective is greater than a threshold. From the set of candidate access points, extract the Pareto optimal solution set that is not dominated by other candidate access points; When the Pareto optimal solution set contains only one candidate access point, it is determined as the target access point; When the Pareto optimal solution set contains multiple candidate access points, the sum of the Euclidean distances from each Pareto optimal solution to all other solutions in the Pareto front is calculated as the inflection point metric, and the candidate access point with the largest inflection point metric is selected as the target access point.

[0010] Further, the step of setting the predicted link quality as the first optimization objective, the current load status as the second optimization objective, and the historical handover success rate as the third optimization objective for each candidate access point in the candidate access point set includes: For each candidate access point in the candidate access point set, the most recently calculated comprehensive link quality score is taken as the predicted link quality of that candidate access point. Send a load query request frame to the candidate access point and receive the currently associated terminal number, channel utilization and available bandwidth information returned by it; The ratio of the current number of associated terminals to the maximum number of associated terminals at the access point is used as the load normalization value. The ratio of the available bandwidth information to the total bandwidth of the access point is used as the normalized value of the available bandwidth; The product of the complement of the channel utilization rate and the normalized value of the available bandwidth is used as the capacity normalized value; The weighted average of the load normalization value and the capacity normalization value is taken as the current load status; The edge weights between the current access point and the candidate access point are queried from the access point topology graph, and these edge weights are used as the historical handover success rate.

[0011] Furthermore, the steps of performing a roaming handover to the target access point, adaptively adjusting the hysteresis parameter according to the roaming context state, and feeding back the adjusted hysteresis parameter to the update process of the trigger threshold include: After completing the roaming handover to the target access point, a roaming context state vector is constructed; wherein, the roaming context state vector includes the terminal mobility speed level, the current service quality level, and the number of historical handovers within a preset time window; Based on the roaming context state vector, a first hysteresis parameter and a second hysteresis parameter are calculated using a preset hysteresis adjustment function; wherein, the hysteresis adjustment function takes each component in the roaming context state vector as input and maps each component to the value range of the hysteresis margin and the minimum dwell time using a linear weighted summation method; wherein, the first hysteresis parameter is the hysteresis margin and the second hysteresis parameter is the minimum dwell time; When the terminal's moving speed level is higher, the hysteresis margin is reduced and the minimum dwell time is shortened; When the current service quality level is higher than the preset level or the number of historical switching times is higher than the preset number, the hysteresis margin is increased by the preset margin and the minimum dwell time is extended by the preset duration. Using the hysteresis margin as an offset, the upper limit trigger threshold is obtained by adding the hysteresis margin to the trigger threshold, and the lower limit trigger threshold is obtained by subtracting the hysteresis margin. A roaming scan is triggered when the link quality degradation rate exceeds the upper limit trigger threshold. Roaming scans are stopped when the link quality degradation rate falls below the lower limit trigger threshold. The minimum dwell time is used as the mandatory waiting time. During the mandatory waiting time after a roaming handover is completed, a new roaming handover is prohibited.

[0012] Further, the step of calculating the first hysteresis parameter and the second hysteresis parameter based on the roaming context state vector using a preset hysteresis adjustment function includes: The terminal mobility speed level, the current service quality level, and the historical handover count are respectively mapped to the corresponding speed value, service quality value, and handover count value. The hysteresis margin is equal to the baseline margin value minus the product of the speed value and the first weighting coefficient, plus the product of the service quality value and the second weighting coefficient, plus the product of the number of handovers and the third weighting coefficient. The minimum dwell time is equal to the base time value minus the product of the speed value and the fourth weighting coefficient, plus the product of the service quality value and the fifth weighting coefficient, plus the product of the number of handovers and the sixth weighting coefficient; wherein the first weighting coefficient, the second weighting coefficient, the third weighting coefficient, the fourth weighting coefficient, the fifth weighting coefficient, and the sixth weighting coefficient are all positive numbers.

[0013] Furthermore, the step of constructing the roaming context state vector after completing the roaming handover to the target access point includes: The terminal's mobility mode is determined based on the sign change sequence of the link quality degradation rate before and after roaming handover. When the symbol change sequence shows alternating positive and negative changes within a preset time period, it is determined to be a round-trip movement mode; otherwise, it is determined to be a one-way movement mode. When the round-trip movement mode is determined, the terminal movement speed level is set to a high level; When the unidirectional movement mode is determined, the terminal movement speed level is determined based on the absolute value of the link quality degradation rate; wherein, the terminal movement speed level includes high level, medium level and low level; The system detects the real-time transmission requirements of currently active services and determines the service quality level of the current service based on the service type mapping table. Among them, real-time audio and video services are mapped to a high level, web browsing services are mapped to a medium level, and background data synchronization services are mapped to a low level. The number of historical roaming switches within the preset time window is counted, and the roaming context state vector is formed by combining the mobile mode determination result, the terminal mobile speed level, the current service quality level, and the number of historical roaming switches.

[0014] Furthermore, after the step of feeding back the adjusted hysteresis parameter to the update process of the trigger threshold, the method further includes: Record the execution result of this roaming handover; wherein, the execution result includes whether the handover was successful, the handover time, and the change in the overall link quality score before and after the handover; When the handover time exceeds a preset time threshold or the change in the overall link quality score before and after the handover is negative, the current roaming handover will be marked as an inefficient handover. The percentage of inefficient switching events within a preset statistical period is recorded. When the proportion of inefficient switching exceeds a preset threshold, the trigger threshold is increased by a preset value; When the proportion of inefficient switching is lower than the preset proportion threshold and the switching success rate meets the preset success rate condition, the trigger threshold is reduced by a preset value. The adjusted trigger threshold will be applied to the next link quality degradation rate assessment process.

[0015] A second aspect of the present invention provides an AP switching device in a PON and ONU environment, comprising: The link quality prediction module is used to collect multi-dimensional link quality parameters of the current access point in real time, and calculate the link quality degradation rate and its direction information based on the multi-dimensional link quality parameters. The directional scanning module is used to determine the target scanning sector in a pre-constructed access point topology map based on the direction information when the link quality degradation rate exceeds a preset trigger threshold, and to perform directional scanning on the candidate access points in the target scanning sector to obtain a set of candidate access points. The target selection module is used to construct multiple optimization targets for each candidate access point in the candidate access point set by comprehensively considering its predicted link quality, load status and historical handover success rate, and determine the target access point through Pareto optimization. The roaming execution module is used to perform roaming handover to the target access point, adaptively adjust the hysteresis parameter according to the roaming context state, and feed back the adjusted hysteresis parameter to the update process of the trigger threshold. The step of constructing multiple optimization objectives for each candidate access point in the candidate access point set by comprehensively considering its predicted link quality, load status, and historical handover success rate, and determining the target access point through Pareto optimization includes: For each candidate access point in the candidate access point set, its predicted link quality is taken as the first optimization objective, the current load status is taken as the second optimization objective, and the historical handover success rate is taken as the third optimization objective. The first optimization objective, the second optimization objective, and the third optimization objective are normalized respectively to unify the value range of each optimization objective to between 0 and 1; In the set of candidate access points, a dominance relationship is defined as the condition that any candidate access point is not inferior to another candidate access point in all optimization objectives and the difference between them in at least one optimization objective is greater than a threshold. From the set of candidate access points, extract the Pareto optimal solution set that is not dominated by other candidate access points; When the Pareto optimal solution set contains only one candidate access point, it is determined as the target access point; When the Pareto optimal solution set contains multiple candidate access points, the sum of the Euclidean distances from each Pareto optimal solution to all other solutions in the Pareto front is calculated as the inflection point metric, and the candidate access point with the largest inflection point metric is selected as the target access point.

[0016] A third aspect of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the AP switching method under PON and ONU environments described in the first aspect above.

[0017] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the AP switching method under PON and ONU environments described in the first aspect.

[0018] The beneficial effects of this invention compared to existing technologies are as follows: By collecting multi-dimensional link quality parameters in real time and calculating the link quality degradation rate and its direction information based on these parameters, handover decisions no longer rely on a single indicator. This multi-dimensional analysis method makes the handover process more intelligent, helping to quickly and accurately identify link quality problems and respond promptly. When the link quality degradation rate exceeds a preset threshold, the method can quickly locate the target scan sector, thereby reducing the search range of candidate APs. This optimization process significantly improves handover efficiency and reduces user latency during handover. By comprehensively analyzing the predicted link quality, load status, and historical handover success rate of candidate access points, multiple optimization objectives are constructed, and Pareto optimization is used to determine the target access point. This method can select higher-quality access points, thereby significantly improving the handover success rate and reducing service interruptions and network instability caused by handover failures. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A schematic flowchart of an AP switching method in a PON and ONU environment provided by the present invention is shown; Figure 2 This diagram illustrates an AP switching device in a PON and ONU environment according to an embodiment of the present invention. Figure 3 A schematic diagram of a terminal device provided in an embodiment of the present invention is shown. Detailed Implementation

[0021] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0022] This invention provides an AP switching method and apparatus in PON and ONU environments to solve the technical problem that traditional methods often fail to respond promptly to changes in link quality in complex network environments and with high user mobility, leading to delays or interruptions for users during the switching process.

[0023] First, this invention provides an AP switching method in both PON and ONU environments. Please refer to [link / reference]. Figure 1 , Figure 1 A schematic flowchart illustrating an AP switching method in a PON and ONU environment provided by this invention is shown. Figure 1 As shown, the AP switching method in the PON and ONU environments may include the following steps: Step 101: Collect multi-dimensional link quality parameters of the current access point in real time, and calculate the link quality degradation rate and its direction information based on the multi-dimensional link quality parameters; This step forms the perceptual foundation of the entire roaming decision-making process. Its core objective is to continuously monitor the link status between the terminal and the current access point from multiple dimensions, and to proactively quantify the changing trends and directions of link quality through a time series prediction model. Traditional roaming triggering mechanisms rely solely on instantaneous threshold comparisons of received signal strength indicators, which have two major drawbacks: first, a single indicator cannot comprehensively reflect the true communication quality of the link, such as scenarios where signal strength is acceptable but channel utilization is too high, leading to a sharp drop in actual throughput; second, instantaneous threshold comparisons lack the ability to predict future link trends, often triggering roaming only after the link quality has severely deteriorated, resulting in increased handover latency or even service interruption. This step, by collecting multi-dimensional link quality parameters and utilizing long short-term memory networks for prediction, achieves a shift from passive response to proactive prediction.

[0024] The multi-dimensional link quality parameters collected by the terminal include four dimensions: ① Received signal strength indicator—reflects the power level of the radio frequency signal received by the terminal, and is a fundamental indicator of the physical layer quality of the link; ② Signal-to-noise ratio (SNR)—reflects the relative strength of the useful signal relative to the background noise, directly related to demodulation performance and data transmission reliability; ③ Packet retransmission rate—reflects the data transmission reliability of the link layer. An increased retransmission rate means increased channel interference or signal attenuation; ④ Channel utilization—reflects the current channel congestion level. High utilization means insufficient available air interface resources, and even with good signal strength, service throughput cannot be guaranteed. These four dimensions comprehensively characterize link quality from the physical layer, data link layer, and resource scheduling layer, providing a sufficient information foundation for subsequent comprehensive scoring and trend prediction.

[0025] Specifically, step 101 includes steps A1 to A7: Step A1: Real-time acquisition of the received signal strength indication, signal-to-noise ratio, data packet retransmission rate, and channel utilization of the current access point as multi-dimensional link quality parameters; The terminal periodically obtains the above four parameters from the currently associated access point through the link quality reporting interface provided by its wireless network card driver. The collection period can be 200 milliseconds, which ensures timely capture of link quality changes without increasing terminal power consumption and air interface overhead due to excessively frequent queries.

[0026] The received signal strength indicator typically ranges from -100dBm to -30dBm, with higher values ​​indicating stronger signals. The signal-to-noise ratio (SNR) typically ranges from 0dB to 60dB, with higher values ​​indicating better signal quality. The packet retransmission rate is expressed as a percentage, ranging from 0 to 1, with higher values ​​indicating more severe packet loss. Channel utilization is also expressed as a percentage, ranging from 0 to 1, with higher values ​​indicating a busier channel. These four parameters are encapsulated into a four-dimensional vector with a timestamp during each data acquisition, providing raw data input for subsequent steps.

[0027] Step A2: Normalize each dimension of the multi-dimensional link quality parameters, and calculate the comprehensive link quality score by weighted geometric mean of the normalized parameters according to preset weights. Since the physical dimensions and value ranges of the four dimensions are different, direct mathematical calculations are meaningless. Therefore, it is necessary to first normalize each dimension to a unified interval of 0 to 1. Normalization methods are divided into two categories based on the characteristics of the indicators: for positive indicators where larger values ​​are better (received signal strength indicator and signal-to-noise ratio), a linear mapping method is used to map the original values ​​to the 0 to 1 interval; for negative indicators where larger values ​​are worse (data packet retransmission rate and channel utilization), an inverse linear mapping is used to ensure that the normalized values ​​also satisfy the semantics of "larger is better".

[0028] Taking the received signal strength indication as an example, the normalization formula is: RSSI_norm=(RSSI-RSSI_min) / (RSSI_max-RSSI_min); Here, RSSI is the currently collected value, and RSSI_min and RSSI_max are the preset minimum and maximum boundary values, respectively. For negative indicators such as packet retransmission rate, the normalization formula is R_norm = 1 - R; where R is the currently collected value. The other two dimensions use a similar normalization strategy. After normalization, a weighted geometric mean is used to calculate the overall link quality score. The advantage of the weighted geometric mean over the weighted arithmetic mean is that it is more sensitive to extreme low values ​​in each dimension—when the normalized value of any dimension approaches zero, the overall score will be significantly lowered, which aligns with the physical intuition of the link quality bottleneck effect. The formula for calculating the overall score is Q = ∏(pi wi ); where pi is the normalization parameter of the i-th dimension, and wi is the preset weight of the i-th dimension, satisfying ∑wi=1. For example, the preset weights of each dimension are shown in Table 1 below: Table 1: Weight Allocation Table for Multidimensional Link Quality Parameters The weights are set based on the degree of impact of each dimension on actual communication quality: signal-to-noise ratio (SNR) directly affects demodulation performance and has the highest weight; received signal strength indication (RSI) and data packet retransmission rate (DRG) reflect the quality of the physical layer and data link layer, respectively, and have the next highest weight; channel utilization reflects the degree of resource contention, and its direct impact on a single terminal is relatively indirect, so it has the lowest weight. These weights are set according to network planning requirements before the device leaves the factory and are stored in the terminal's configuration file.

[0029] Step A3: Arrange the comprehensive link quality scores in chronological order to construct a link quality vector; After each data collection cycle, the terminal calculates a comprehensive link quality score. The comprehensive scores from the most recent M collection cycles are then arranged chronologically to construct a link quality vector of length M: Q_vec = [Q(t-M+1), Q(t-M+2), ..., Q(t)]. Here, M can be 30, corresponding to 30 historical data collection cycles, covering approximately a 6-second historical window within a 200-millisecond collection cycle. The length of this vector must meet the input sequence length requirements of the time series prediction model, but should not be too long to avoid introducing outdated data and reducing the timeliness of the prediction.

[0030] When a new collection cycle arrives, the link quality vector is updated using a sliding window method—the oldest element in the vector is removed, and the latest comprehensive score is appended to the end of the vector, keeping the vector length constant at M.

[0031] Step A4: Input the link quality vector into the Long Short-Term Memory network and output the link quality prediction sequence within the future preset time window; Long Short-Term Memory (LSTM) networks are recurrent neural networks with gating mechanisms, adept at modeling long-term dependencies in time series. This step utilizes an LSM network to model the temporal pattern of the link quality vector, outputting a link quality prediction sequence Q_pred=[Q_hat(t+1), Q_hat(t+2), ..., Q_hat(t+N)] for the next N acquisition cycles. Here, N can be 10, corresponding to a prediction window of approximately 2 seconds, a length sufficient to cover the time required for a complete roaming handover process.

[0032] The core structure of a Long Short-Term Memory (LSTM) network consists of three gating units: the forget gate, the input gate, and the output gate. The forget gate determines how much information from the previous cell state is retained, and its calculation formula is f(t) = σ(Wf·[h(t-1), Q(t)] + bf). The input gate determines how much new information is written into the cell state, and its calculation formula is i(t) = σ(Wi·[h(t-1), Q(t)] + bi). The calculation formula for the candidate cell state is c_tilde(t) = tanh(Wc·[h(t-1), Q(t)] + bc). The cell state update formula is C(t) = f(t)·C(t-1) + i(t)·c_tilde(t). The calculation formulas for the output gate and the hidden state are o(t) = σ(Wo·[h(t-1), Q(t)] + bo) and h(t) = o(t)·tanh(C(t)), respectively. Where σ is the Sigmoid activation function, tanh is the hyperbolic tangent activation function, Wf, Wi, Wc, and Wo are the weight matrices of each gate, and bf, bi, bc, and bo are the bias vectors of each gate. These parameters are obtained through training with a large amount of historical link quality data during the offline training phase.

[0033] The fully connected output layer of the network maps the hidden state of the last time step to an N-dimensional prediction vector, i.e., a link quality prediction sequence. This prediction sequence reflects the trend of link quality changes in the near future and forms the basis for subsequent degradation rate calculations.

[0034] As an optional embodiment of this application, link quality prediction can also use a gated recurrent unit network instead of a long short-term memory network. The gated recurrent unit combines the forget gate and the input gate into an update gate, which is simpler in structure and has fewer parameters. When the pattern of the link quality sequence is relatively simple, it can significantly reduce the amount of computation while maintaining similar prediction accuracy, making it more suitable for deployment and operation on embedded terminals with limited computing resources.

[0035] Step A5: Calculate the link quality difference between adjacent time points in the link quality prediction sequence to obtain the link quality change rate sequence; The link quality change rate sequence reflects the time-by-time variation of link quality within the prediction window. For a prediction sequence Q_pred=[Q_hat(t+1), Q_hat(t+2), ..., Q_hat(t+N)], the formula for calculating the change rate sequence is dQ(k)=Q_hat(t+k)-Q_hat(t+k-1), where k ranges from 1 to N. This differencing operation extracts trend information from the prediction sequence, removing interference from the absolute quality level and focusing on the speed and direction of change. When dQ(k) is consistently negative, it indicates that the link quality is continuously decreasing within the prediction window; when dQ(k) alternates between positive and negative values, it indicates that the link quality is unstable.

[0036] Step A6: Take the average of the last preset number of elements of the link quality change rate sequence as the link quality degradation rate; The preset number of terminal elements corresponds to the prediction result closest to the current moment in the prediction sequence, best reflecting the recent trend of link quality changes. Let the number of terminal elements be K, where K can be 5. Then the formula for calculating the link quality degradation rate is D_rate = (1 / K)·∑dQ(k), where k ranges from N-K+1 to N. Taking the average value instead of a single terminal value can smooth out prediction noise to some extent, making the degradation rate indicator more robust. When D_rate is negative, it indicates that the link quality is deteriorating; the larger the absolute value, the faster the degradation. When D_rate is positive or close to zero, it indicates that the link quality is stable or improving.

[0037] Step A7: Determine the degradation direction information based on the sign of the link quality degradation rate; The degradation direction information is a binary flag used to indicate the trend of link quality changes. When D_rate is less than zero, the degradation direction information is marked as degraded, indicating that the link quality is decreasing, the terminal is moving away from the current access point, or the channel environment is deteriorating. When D_rate is greater than or equal to zero, the degradation direction information is marked as stable or improving, indicating that the link quality between the terminal and the current access point remains stable or is improving. This direction information will be used in step 102 to determine the direction of the target scanning sector, enabling roaming scanning to focus on candidate access points in the terminal's movement direction and avoiding the time and power consumption overhead of omnidirectional scanning.

[0038] In the embodiments corresponding to steps A1 to A7, through the cascading processing of the above steps, the terminal realizes a complete sensing link from the acquisition of original link parameters to the prediction of link quality degradation trends. Multidimensional parameter acquisition provides a comprehensive profile of the link status, the weighted geometric mean comprehensive score avoids distortion by a single indicator through the weakest link effect mechanism, the long short-term memory network achieves forward-looking prediction of future link quality by learning temporal patterns, and the change rate sequence and degradation rate calculation extract trend information from the prediction results. The output of the entire sensing link—the link quality degradation rate and its direction information—provides a key decision basis for the subsequent roaming trigger determination and directional scanning in step 102.

[0039] Step 102: When the link quality degradation rate exceeds a preset trigger threshold, the target scanning sector is determined in the pre-constructed access point topology map according to the direction information, and a directional scan is performed on the candidate access points in the target scanning sector to obtain a set of candidate access points; When the absolute value of the link quality degradation rate output in step 101 exceeds the preset trigger threshold, the terminal initiates the roaming scan process. The preset trigger threshold can be -0.05, meaning that roaming is triggered when the average rate of change of the comprehensive link quality score within the future prediction window is less than -0.05. This threshold is set at the factory and can be dynamically adjusted during operation based on the feedback mechanism in step 104.

[0040] The core innovation of this step lies in incorporating the terminal's movement direction information into the scanning range determination process, enabling directional scanning rather than traditional omnidirectional scanning across all channels. During terminal movement, the direction of link quality degradation between the terminal and the current access point contains information about the terminal's movement trend—the degradation direction points away from the current access point, and the candidate access point the terminal is heading towards is most likely to provide better link quality. By extracting a fan-shaped region along the movement direction from a pre-built access point topology map, the scanning range is narrowed from all neighboring nodes to a directionally relevant subset, significantly reducing scanning overhead and power consumption while improving the relevance quality of candidate access points.

[0041] The access point topology map is a crucial data structure in this step. It uses access points as nodes and historical handover frequency and signal coverage overlap between access points as edge weights, depicting the spatial adjacency relationships and handover reliability among access points in the physical deployment environment. This topology map needs to be pre-built before use and dynamically updated after each roaming handover.

[0042] Specifically, step 102 includes steps B1 to B6: Step B1: Obtain the node corresponding to the current access point and the set of neighboring nodes that have a topological adjacency relationship with the current access point from the access point topology graph; wherein, the access point topology graph uses access points as nodes and the historical switching frequency and signal coverage overlap between access points as edge weights. When a terminal accesses the network, it loads a copy of the local access point topology map and locates the node corresponding to the currently associated access point. Starting from that node, it traverses all directly adjacent nodes along the edges of the topology map, forming a neighbor node set. The neighbor node set contains all candidate access points in the physical deployment that have overlapping signal coverage with the current access point, forming the basis of the search space for subsequent directional scanning.

[0043] The construction and maintenance of the access point topology map are crucial for ensuring the accuracy of roaming decisions. The topology map needs to be pre-built when the terminal is first deployed and continuously updated based on roaming handover results during operation. Steps B11 to B14 describe the process of constructing and dynamically maintaining the access point topology map. These steps are completed before step B1 is executed and are triggered to update after each roaming handover.

[0044] Specifically, step B1 includes steps B11 to B14: Step B11: Collect historical roaming handover records; each record includes the source access point identifier, the target access point identifier, whether the handover was successful, and the signal strength difference at the time of handover; After each roaming handover, the terminal records the handover event in its local roaming log. Each record contains four fields: source access point identifier (e.g., BSSID), target access point identifier, handover result (success or failure), and signal strength difference at the time of handover. The signal strength difference is defined as the difference between the received signal strength indications of the target access point and the source access point at the moment of handover; this value reflects the signal quality comparison between the two access points during handover. Historical roaming handover records accumulate as the terminal is used, providing data support for the calculation of topology graph edge weights.

[0045] Step B12: For any two access point nodes with overlapping signal coverage in the access point topology graph, count their historical successful handover counts and failed handover counts, and use the Bayesian estimation method to calculate their posterior handover success probability as the weight of the edge between the two nodes. For any two access point node pairs (A, B) with overlapping signal coverage in the access point topology map, retrieve all handover records from the historical roaming handover records where the source node is A and the destination node is B, and vice versa. Count the number of successful handovers (s) and the number of failed handovers (f). Calculate the posterior handover success probability using a Bayesian estimation method, treating the handover success probability as a random variable following a Beta distribution.

[0046] Within the Bayesian estimation framework, the prior distribution of the success probability of a switchover is denoted as Beta(α0,β0), where α0 and β0 are prior hyperparameters, each taking a value of 1, corresponding to a uniform prior. After observing s successes and f failures, the posterior distribution is Beta(α0+s,β0+f). The expected value of the posterior success probability of a switchover is P = (α0+s) / (α0+s+β0+f).

[0047] The advantage of Bayesian estimation is that when historical data is limited, the prior distribution can avoid extreme values ​​in probability estimation (such as estimating the success rate as 1.0 when only one successful handover is observed), making the estimation results more robust. As historical data accumulates, the posterior distribution gradually becomes data-driven, the influence of the prior gradually weakens, and the estimation results approach the true handover success rate.

[0048] Step B13: Remove edges with weights lower than a preset threshold from the access point topology graph, and retain the edges to form the adjacency relationship of the access point topology graph; Not all access point pairs with overlapping signal coverage are suitable as candidate pairs for roaming handover. When the posterior handover success probability between two access points is too low, it indicates that the reliability of handover to the target access point cannot be guaranteed, and keeping them in the topology graph may lead to incorrect selection. The preset threshold value can be 0.5, meaning edges with a posterior handover success probability lower than 0.5 are deleted. The threshold value can be adjusted according to the reliability requirements of the network deployment—in scenarios with high handover reliability requirements, the threshold can be appropriately increased, while in environments with fewer candidate access points, the threshold can be appropriately decreased to ensure sufficient candidate space.

[0049] After removing low-weight edges, the remaining edges constitute the adjacency relationships of the access point topology graph. These adjacency relationships reflect the actual available roaming switching paths, providing a filtered and reliable candidate set for obtaining neighbor nodes in step B1.

[0050] Step B14: When the new roaming switch is completed, update the number of successful and failed switches between the corresponding node pairs according to the switch results, and recalculate the edge weights; Each time a terminal completes a roaming handover, the handover result is written to the historical roaming handover record, triggering an update of statistics and a recalculation of edge weights between the corresponding node pairs. If the updated edge weight is lower than a preset threshold, the edge is deleted from the topology graph; if a previously deleted edge's weight exceeds the threshold due to the accumulation of new data, it is restored to the topology graph. This dynamic maintenance mechanism allows the access point topology graph to adaptively reflect changes in the network environment—for example, after a new access point is deployed, as terminals gradually accumulate handover data, the edge weights between the new access point and existing access points will gradually converge from prior values ​​to the actual success rate.

[0051] In the embodiments corresponding to steps B11 to B14, a dynamic access point topology graph with Bayesian posterior handover success probability as edge weights is constructed through the cascading processing of the above steps. This topology graph not only depicts the spatial adjacency relationships between access points but also quantifies the reliability of each handover path, providing a high-quality candidate space that has been data-driven and filtered for subsequent neighbor node acquisition in step B1 and historical handover success rate target calculation in step 103. The dynamic update mechanism of the topology graph ensures its timeliness and accuracy in the face of changes in the network environment.

[0052] Step B2: Determine the movement direction of the current access point in the access point topology map based on the direction information; The degradation direction information output in step A7 indicates the trend of link quality changes. This step maps it to the direction of terminal movement in physical space. When the degradation direction is deteriorating, the terminal is moving away from the current access point, and the direction of movement is taken as the azimuth angle indicated by the degradation direction.

[0053] Specifically, the terminal infers the azimuth angle of the movement direction by combining the changing patterns of the multi-dimensional link quality parameters collected in step A1. For example, when the received signal strength indication continues to decrease and the signal-to-noise ratio deteriorates synchronously, the terminal moves outward along the edge of the signal coverage area of ​​the current access point, with the movement direction chosen as the direction of the most significant signal degradation. When the degradation direction is stable or improving, the terminal does not initiate roaming scanning and skips subsequent steps.

[0054] Step B3: Using the node corresponding to the current access point as the center, the direction of movement as the axis, and a preset angle range as the subtended angle, extract a fan-shaped area in the access point topology map as the target scanning sector; In the access point topology diagram, a sector is selected as the target scanning sector, centered on the node corresponding to the current access point, with the movement direction determined in step B2 as the sector axis, and a preset angle range as the subtended angle. The preset angle range can be 120 degrees, that is, extending 60 degrees to each side of the axis centered on the movement direction. This angle range covers the main candidate access points in front of the terminal's movement while excluding irrelevant access points located behind and to the sides of the terminal, effectively reducing the scanning range. The size of the angle range can be adjusted according to the access point deployment density—it can be reduced to 90 degrees in a densely deployed access point environment to further reduce scanning overhead, and expanded to 180 degrees in a sparsely deployed access point environment to ensure a sufficient number of candidates.

[0055] Step B4: Obtain the access points corresponding to all neighboring nodes within the target scan sector as a list of access points to be scanned; From the set of neighboring nodes obtained in step B1, nodes located within the target scan sector are selected. The selection criterion is that the azimuth angle of the neighboring node in the topology map is within the angular range of the target scan sector. The resulting list of access points is the list of access points to be scanned. These access points are located ahead of the terminal's movement direction and are the most likely candidates to provide better link quality. The list of access points to be scanned typically contains 2 to 8 access points, which is much smaller than the total number of neighboring nodes in an omnidirectional scan, significantly reducing scanning overhead.

[0056] Step B5: For each access point in the list of access points to be scanned, send a probe request frame and receive a probe response frame on its working channel to collect the real-time link quality parameters of each access point; The terminal sequentially or in parallel sends probe request frames on the operating channel of each access point in the list of access points to be scanned. Upon receiving a probe request frame, an access point returns a probe response frame, which carries information such as the access point's basic service set identifier, operating channel, supported data transmission rates, and transmission power. Simultaneously with receiving the probe response frame, the terminal measures real-time link quality parameters such as received signal strength indication and signal-to-noise ratio.

[0057] To minimize the impact of scanning on current services, the terminal employs a time-division scanning strategy—the dwell time on the channel at each access point can be set to 30 milliseconds, sending a probe request frame and waiting for a probe response frame. If no response is received within the dwell time, the access point is skipped. During the scanning process, the terminal needs to briefly leave the current channel, which may cause a small amount of data packet buffering. Transmission is resumed from the buffer after the scan is completed, and the impact on upper-layer services is negligible.

[0058] As an optional embodiment of this application, directional scanning can also employ passive scanning instead of active probing—that is, the terminal silently listens for beacon frames on the working channel of each candidate access point without sending probe request frames. Passive scanning does not require consuming air interface resources to send request frames, thus having less impact on current service transmissions, but it requires waiting for the beacon frame broadcast period of the access point (which can be 100 milliseconds), resulting in a longer scanning time. This method is suitable for network environments that are sensitive to air interface occupancy or have short access point beacon periods.

[0059] Step B6: Collect the access points that have received probe response frames and their corresponding real-time link quality parameters to form a candidate access point set; After the scan is complete, all access points that successfully returned probe response frames and their corresponding real-time link quality parameters are summarized into a candidate access point set. Each entry in the candidate access point set includes an access point identifier, operating channel, real-time received signal strength indicator, real-time signal-to-noise ratio, and edge weights from the access point topology graph (i.e., historical handover success probabilities). This set serves as the input for Pareto optimization in step 103 to select the target access point.

[0060] In the embodiments corresponding to steps B1 to B6, the cascading processing of the above steps realizes a complete process from trigger determination to directional scanning and then to candidate set construction. Specifically, the Bayesian edge weighting mechanism of the access point topology graph provides reliable adjacency relationships verified by data; the fan-shaped interception guided by directional information precisely limits the scanning range to the relevant area in the direction of terminal movement; and the time-division active scanning completes the real-time parameter acquisition of candidate access points under controllable service influence. The entire process reduces the time overhead of traditional full-channel omnidirectional scanning from hundreds of milliseconds to tens of milliseconds, gaining valuable time leeway for subsequent target selection in step 103 and roaming execution in step 104.

[0061] Step 103: For each candidate access point in the candidate access point set, construct multiple optimization objectives by combining its predicted link quality, load status and historical handover success rate, and determine the target access point through Pareto optimization; After obtaining the set of candidate access points, the goal of this step is to select an optimal roaming target from among them. Traditional access point selection methods typically use only received signal strength as the sole criterion, choosing the access point with the strongest signal as the handover target. However, the strongest signal does not equate to optimal overall performance—an access point with the strongest signal may have high channel utilization due to too many associated terminals, or it may be at risk of handover failure due to a low historical handover success rate. This step introduces three optimization objectives: predicted link quality, load status, and historical handover success rate. By employing a Pareto multi-objective optimization method, it finds the optimal balance among multiple competing objectives, avoiding the one-sidedness of decision-making based on a single indicator.

[0062] Pareto optimization is a multi-objective optimization framework. Its core idea is that a solution is considered dominant only if it is superior to another solution across all objectives. The final Pareto optimal set is a set of mutually exclusive compromises in the multi-objective space. Compared to traditional methods that weight multiple objectives into a single objective, Pareto optimization does not require pre-setting weight relationships between objectives, avoiding the influence of subjective preferences on the outcome. It is more suitable for roaming decision-making scenarios where there is no clear priority relationship between objectives.

[0063] Specifically, step 103 includes steps C1 to C6: Step C1: For each candidate access point in the candidate access point set, take its predicted link quality as the first optimization objective, the current load status as the second optimization objective, and the historical handover success rate as the third optimization objective. The selection criteria for the three optimization objectives each reflect different dimensions of concern in roaming handover decisions: ① Predicted link quality – reflects the expected level of communication quality after handover to a candidate access point, and is a basic benefit indicator for roaming handover; ② Current load status – reflects the resource busyness of the candidate access point. Selecting an access point with a lighter load helps to obtain higher actual throughput and lower transmission latency; ③ Historical handover success rate – reflects the reliability of handover from the current access point to a candidate access point. Access points with a high historical success rate mean that the handover path has been verified and the risk is low.

[0064] There are interdependent relationships among the three objectives: access points with good signal quality tend to attract more terminal connections, leading to higher loads; access points with light loads may be located in coverage edge areas with generally lower signal quality; and access points with high historical handover success rates may not be able to cover all candidates due to their fixed deployment locations. Therefore, Pareto optimization is needed to find a compromise among the three objectives.

[0065] The specific construction process of the three optimization objectives in step C1 involves multi-dimensional information querying and calculation of candidate access points, which will be elaborated in detail in the following steps C11 to C17.

[0066] Specifically, step C1 includes steps C11 to C17: Step C11: For each candidate access point in the candidate access point set, take the most recently calculated comprehensive link quality score as the predicted link quality of that candidate access point; During the directional scanning process in step B5, the terminal collects the real-time received signal strength indication and signal-to-noise ratio (SNR) for each candidate access point. Based on the scanning results, the terminal recalculates the comprehensive link quality score for each candidate access point using the normalization and weighted geometric mean method from step A2. This score is then used as the predicted link quality for the candidate access point. This score integrates both signal strength and SNR dimensions and uses the weighting configuration from step A2, ensuring consistency and comparability with the current access point link quality assessment.

[0067] Step C12: Send a load query request frame to the candidate access point and receive the currently associated terminal number, channel utilization and available bandwidth information returned by it; The terminal sends a load query request frame to each candidate access point through the load query mechanism in the 802.11 management frame. The candidate access point returns a load information frame containing three fields: the number of currently associated terminals, the current channel utilization, and available bandwidth information. The number of currently associated terminals reflects the number of terminals served by the access point, the channel utilization reflects the busyness of air interface resources, and the available bandwidth information reflects the remaining bandwidth that the access point can allocate to new access terminals under the current load conditions. This information provides the raw data for calculating the load status.

[0068] Step C13: Use the ratio of the current number of associated terminals to the maximum number of associated terminals at the access point as the load normalization value; The load normalization value reflects the terminal association saturation of an access point. The calculation formula is L_norm = N_assoc / N_max; where N_assoc is the current number of associated terminals, and N_max is the maximum number of associated terminals supported by the access point. The smaller this ratio, the more sufficient the access point still has terminal association capacity. The value of N_max comes from the capacity information field reported by the access point during the association phase. When N_assoc approaches N_max, the load normalization value approaches 1, indicating that the access point is about to reach its capacity limit.

[0069] Step C14: Use the ratio of the available bandwidth information to the total bandwidth of the access point as the normalized value of the available bandwidth; The normalized available bandwidth value reflects the remaining bandwidth resources of an access point. The calculation formula is B_avail_norm = B_avail / B_total; where B_avail is the available bandwidth information, and B_total is the total bandwidth of the access point. A larger ratio indicates that the access point can provide more ample bandwidth resources for newly connected terminals. The value of B_total comes from the access point's configuration information and is determined during device deployment.

[0070] Step C15: The product of the complement of the channel utilization rate and the normalized value of the available bandwidth is used as the capacity normalized value; The normalized capacity value combines two factors: channel idleness and bandwidth sufficiency, reflecting the overall available capacity of the access point. The calculation formula is C_norm = (1 - U_ch)·B_avail_norm; where U_ch is the channel utilization rate and 1 - U_ch is the channel idle rate. The product of the channel idle rate and the normalized available bandwidth value takes the larger value when both factors are satisfied, and the smaller value when either factor is insufficient, reflecting the joint constraint of air interface resources and bandwidth resources on available capacity.

[0071] Step C16: Take the weighted average of the load normalization value and the capacity normalization value as the current load state; The current load status combines two dimensions: terminal association saturation and overall available capacity. Since load normalization is a negative indicator (the higher the value, the worse), while capacity normalization is a positive indicator (the higher the value, the better), their directions need to be aligned before weighted averaging. The load status formula is Load = w_L·(1-L_norm) + w_C·C_norm; where w_L and w_C are the weights of the load normalization and capacity normalization dimensions, respectively, satisfying w_L + w_C = 1. w_L can take a value of 0.4, and w_C can take a value of 0.6, meaning the weight of capacity normalization is slightly higher than that of load normalization, as capacity normalization incorporates more information from various dimensions. The load status value ranges from 0 to 1; a higher value indicates a lighter load on the access point and more abundant available resources.

[0072] Step C17: Query the edge weight between the current access point and the candidate access point from the access point topology graph, and use the edge weight as the historical handover success rate; The terminal queries the access point topology graph for the edge weights between the current access point node and the candidate access point node. These edge weights are calculated using the Bayesian estimation method in step B12, i.e., the posterior handover success probability P = (α0 + s) / (α0 + s + β0 + f). This probability value directly reflects the historical reliability of switching from the current access point to the candidate access point, and its value ranges from 0 to 1. For newly deployed access point pairs, due to the lack of historical handover data, the edge weights are determined by the expected value of the prior distribution, i.e., P = α0 / (α0 + β0) = 0.5, representing a neutral expectation.

[0073] In the embodiments corresponding to steps C11 to C17, through the cascaded processing of the above steps, three-dimensional optimization target values ​​are constructed for each candidate access point in the candidate access point set. The predicted link quality comes from the real-time comprehensive score calculation in step A2, the current load status comes from the multi-dimensional information fusion of the load query frame, and the historical handover success rate comes from the Bayesian posterior probability of the access point topology map. All three target values ​​are normalized to the range of 0 to 1, with larger values ​​indicating better performance in that dimension, providing a unified-scale target space for the Pareto optimization in subsequent steps C2 to C6.

[0074] Step C2: Normalize the first optimization objective, the second optimization objective, and the third optimization objective respectively, so that the value range of each optimization objective is unified to between 0 and 1; Since the construction process in steps C11 to C17 has already normalized the three optimization objectives, this step mainly performs consistency checks and boundary pruning. For values ​​that slightly exceed the 0-1 range due to calculation errors, they are pruned to the range boundary. For candidate access points that are missing a certain objective value (such as access points that do not return load information), the median of that objective value in the candidate set is used to fill the gap, avoiding misjudgments of Pareto dominance due to missing data.

[0075] Step C3: In the set of candidate access points, a dominance relationship is defined as the condition that any candidate access point is not inferior to another candidate access point in all optimization objectives and the difference in at least one optimization objective is greater than a threshold. Pareto dominance is a core concept in multi-objective optimization. For any two access points X and Y in the candidate access point set, the condition for X to dominate Y is defined as follows: X is not inferior to Y in all three optimization objectives, and X is strictly superior to Y in at least one optimization objective with a difference exceeding a preset dominance threshold. The dominance threshold δ can be 0.02. This threshold avoids spurious dominance relationships caused by small numerical differences, making the dominance determination more robust. Formally, X dominates Y if and only if ∀i: fi(X)≥fi(Y) and ∃j: fj(X)-fj(Y)>δ. Here, fi represents the i-th optimization objective function. When the dominance relationship does not hold, X and Y do not dominate each other, and both remain as candidates for Pareto optimality.

[0076] Step C4: Extract the Pareto optimal solution set that is not dominated by other candidate access points from the candidate access point set; The Pareto optimal set is a subset of the candidate access point set that is not dominated by any other candidate access point. The extraction method involves iterating through each access point in the candidate access point set and checking if another access point dominates it. If no dominant access point exists, it is added to the Pareto optimal set. Each access point in the Pareto optimal set represents a unique trade-off among the three optimization objectives—it may be better in predicted link quality, or it may be better in load conditions or historical handover success rate, but there is no single access point that wins across all dimensions.

[0077] Typically, the Pareto optimal solution set contains 1 to 4 candidate access points. When there are many candidate access points and the target values ​​are widely distributed, the Pareto optimal solution set may be large; when a candidate access point is significantly dominant on all targets, the Pareto optimal solution set may contain only one access point.

[0078] Step C5: When the Pareto optimal solution set contains only one candidate access point, determine it as the target access point; When the Pareto optimal solution set contains only one candidate access point, it means that the access point has reached the optimal balance among the three optimization objectives, and there is no competing alternative. Therefore, it is directly identified as the target access point. This is the most ideal decision-making scenario, and no further optimization calculation is required.

[0079] Step C6: When the Pareto optimal solution set contains multiple candidate access points, calculate the sum of the Euclidean distances from each Pareto optimal solution to all other solutions in the Pareto front as the inflection point metric, and select the candidate access point with the largest inflection point metric as the target access point. When the Pareto optimal solution set contains multiple candidate access points, a final target needs to be selected from them. This step uses the inflection point metric method, which finds the solution furthest from all other solutions on the Pareto front. This solution represents a unique choice that is most different from other compromise solutions, and usually corresponds to the candidate access point with the most balanced overall performance.

[0080] For each solution X_i in the Pareto optimal solution set, the inflection point metric is calculated as K(X_i) = ∑||X_i - X_j||_2. Here, ||·||_2 represents the Euclidean distance, summed over all solutions X_j in the Pareto optimal solution set except X_i. The Euclidean distance is calculated in the three-dimensional target space, with three components: normalized predicted link quality, load status, and historical handover success rate. A larger inflection point metric K(X_i) indicates that the solution is further away from other solutions in the target space, meaning its compromise is more unique.

[0081] The candidate access point with the largest inflection point metric is selected as the target access point. The intuitive basis for this selection strategy is that on the Pareto front, the inflection point (i.e., the point with the greatest convexity) usually corresponds to the most balanced compromise after comprehensively considering all objectives, and will not be overly biased towards any single objective.

[0082] As an optional embodiment of this application, when the Pareto optimal solution set contains multiple candidate access points, the determination of the target access point can also use the weighted TOPSIS method instead of the inflection point metric method. The weighted TOPSIS method first calculates the weighted distance ratio from each Pareto optimal solution to the ideal solution and the negative ideal solution. The ideal solution is composed of the maximum value of each objective, and the negative ideal solution is composed of the minimum value of each objective. The formula for calculating the weighted distance ratio is RC_i = D_i. - / (D_i + +D_i - ). Among them, D_i + To solve for the Euclidean distance from X_i to the ideal solution, D_i -To calculate the Euclidean distance from X_i to the negative ideal solution, a larger RC_i indicates that the solution is closer to the ideal solution and further away from the negative ideal solution. The candidate access point with the largest RC_i is selected as the target access point. The weighted TOPSIS method introduces the ideal solution as a reference benchmark, which may provide more intuitive selection results in some scenarios, but it requires setting the weights of each objective.

[0083] In the embodiments corresponding to steps C1 to C6, a target access point selection mechanism based on Pareto multi-objective optimization is realized through the cascading processing of the above steps. The construction of three-dimensional optimization objectives elevates roaming decision-making from a single signal strength comparison to a multi-dimensional evaluation of comprehensive quality, load, and reliability. The extraction of Pareto dominance relationships avoids the influence of subjective weight settings on the decision results, and the inflection point metric selection finds the most balanced compromise solution in the Pareto optimal solution set. The entire selection process is operated on a finite scale of the candidate access point set, with controllable computational complexity, making it suitable for real-time execution under the limited computing resources of the terminal.

[0084] Step 104: Perform a roaming handover to the target access point, adaptively adjust the hysteresis parameter according to the roaming context state, and feed back the adjusted hysteresis parameter to the update process of the trigger threshold; After determining the target access point in step 103, the terminal performs a roaming handover to the target access point. Traditional roaming handover uses fixed hysteresis parameters—namely, fixed signal strength difference margin and minimum dwell time—to control the sensitivity and frequency of roaming triggering. The problem with fixed hysteresis parameters is that for high-speed mobile terminals, a fixed high hysteresis margin may lead to roaming triggering too late, missing the optimal handover opportunity; for low-speed mobile or stationary terminals, a fixed low hysteresis margin may lead to frequent unnecessary roaming handovers, causing a ping-pong effect. The core innovation of this step lies in adaptively adjusting the hysteresis margin and minimum dwell time based on the roaming context state—including the terminal's mobile speed level, current service quality level, and historical handover count—so that the roaming triggering strategy can adapt to different mobile scenarios and service requirements.

[0085] Furthermore, this step also feeds the adjusted hysteresis parameters back into the trigger threshold update process. By recording and analyzing the execution effect of roaming handover, the preset trigger threshold is dynamically adjusted, forming a closed-loop adaptive optimization mechanism. When the roaming handover effect is unsatisfactory (such as excessive handover time or deterioration of link quality after handover), the system automatically increases the trigger threshold to make subsequent roaming more cautious; when the roaming handover effect is good, the system appropriately lowers the trigger threshold to maintain sensitivity to link degradation.

[0086] Specifically, step 104 includes steps D1 to D14: Step D1: After completing the roaming handover to the target access point, construct a roaming context state vector; wherein, the roaming context state vector includes the terminal mobility speed level, the current service quality level, and the number of historical handovers within a preset time window; The roaming context state vector serves as the input basis for adaptive adjustment of hysteresis parameters. Its construction requires comprehensive information from multiple dimensions, including the terminal's mobility mode, service type, and handover history. The roaming context state vector comprises three components: terminal mobility speed level (high, medium, and low), current service quality level (high, medium, and low), and historical handover count (cumulative value within a preset time window). The preset time window can be 60 seconds, a length that reflects the terminal's recent handover frequency without introducing outdated historical data due to an excessively long window.

[0087] The construction of the roaming context state vector involves multiple sub-processes, such as mobility mode determination, speed level estimation, service type identification, and handover history statistics. Steps D11 to D16 elaborate on these processes in detail.

[0088] Specifically, step D1 includes steps D11 to D16: Step D11: Determine the terminal's mobility mode based on the sign change sequence of the link quality degradation rate before and after roaming handover; The terminal's movement pattern is inferred from the sign change sequence of the link quality degradation rate. Before and after each roaming handover, the terminal records the sign (positive or negative) of the link quality degradation rate calculated in step A6. When the terminal moves back and forth between two access points, the sign of the degradation rate alternates between positive and negative—negative when moving away from access point A, positive when moving closer to access point B, and the sign reverses again upon returning. When the terminal moves unidirectionally away from an area, the sign of the degradation rate remains negative. By analyzing the pattern of the sign change sequence, the back-and-forth movement pattern and the unidirectional movement pattern can be distinguished.

[0089] Step D12: When the symbol change sequence shows alternating positive and negative changes within a preset time period, it is determined to be a round-trip movement mode; otherwise, it is determined to be a one-way movement mode. The preset time period can be 120 seconds. Within this time period, if the sign change sequence shows at least two alternations of positive and negative (e.g., negative-positive-negative or positive-negative-positive), it is determined to be a round-trip movement mode. Round-trip movement mode typically occurs when end users move back and forth within a fixed area or shuttle between adjacent rooms; frequent roaming handovers in such scenarios can severely impact service continuity. If the sign change sequence does not show alternations of positive and negative within the preset time period, it is determined to be a unidirectional movement mode, indicating that the terminal moves continuously in one direction.

[0090] Step D13: When the round-trip movement mode is determined, the terminal movement speed level is set to a high level; Round-trip mobility means that the terminal crosses the access point coverage boundary multiple times in a short period of time. Regardless of its actual physical speed, this should be considered a high-mobility scenario from a roaming decision-making perspective. Setting the terminal's mobility speed level to a high level allows for a more aggressive strategy in adjusting subsequent hysteresis parameters—reducing hysteresis margin and minimum dwell time—so that the terminal can complete roaming handovers faster and reduce dwell time at coverage boundaries. The purpose of this setting is to reduce service interruptions in round-trip mobility scenarios by accelerating roaming response. Although this may increase handover frequency, frequent but rapid handovers are a better strategy than the communication quality degradation caused by signal jitter at coverage edges.

[0091] Step D14: When the unidirectional movement mode is determined, the terminal movement speed level is determined based on the absolute value of the link quality degradation rate; wherein, the terminal movement speed level includes high level, medium level and low level; In unidirectional mobility mode, the physical mobility speed of the terminal is indirectly estimated by the absolute value of the link quality degradation rate. A larger absolute value of the degradation rate indicates a faster decline in link quality and a higher terminal mobility speed. Two speed thresholds, V_th1 and V_th2 (V_th1 > V_th2), are defined as follows: when |D_rate| > V_th1, it is classified as high-level; when V_th2 < |D_rate| ≤ V_th1, it is classified as medium-level; and when |D_rate| ≤ V_th2, it is classified as low-level. The value of V_th1 can be 0.15, and the value of V_th2 can be 0.05. These thresholds reflect the rate of change of the overall link quality score at different mobility speeds and can be calibrated at the factory based on typical scenarios.

[0092] Step D15: Detect the real-time transmission requirements of currently active services and determine the service quality level of the current service according to the service type mapping table; among them, real-time audio and video services are mapped to high level, web browsing services are mapped to medium level, and background data synchronization services are mapped to low level. The terminal detects the network traffic characteristics of currently active applications, including traffic patterns, latency sensitivity, and bandwidth requirements, and maps these to quality of service (QoS) levels using a service type mapping table. The service type mapping table is designed based on the tolerance of each service type for roaming interruptions. ① Real-time audio and video services – Highly sensitive to latency and packet loss; roaming interruptions exceeding 100 milliseconds can cause audio and video stuttering or disconnection, mapped to high level. ② Web browsing services – Tolerable for moderate latency, but excessively long switching latency will affect the page loading experience, mapped to medium level. ③ Background data synchronization services – Not sensitive to latency; roaming interruptions of hundreds of milliseconds to several seconds are acceptable, mapped to low level.

[0093] For example, an example of a business type mapping table is shown in Table 2 below: Table 2: Business Type Mapping Table The service type mapping table is stored in the terminal's configuration file. The terminal matches the service type by detecting the application layer protocol characteristics of the currently active application (such as Real-Time Transport Protocol, Hypertext Transfer Protocol, Simple Mail Transfer Protocol, etc.).

[0094] Step D16: Count the number of historical roaming handovers within the preset time window, and combine the mobility mode determination result, the terminal mobility speed level, the current service quality level, and the number of historical roaming handovers to form the roaming context state vector; The terminal counts the number of roaming handovers completed within a preset time window (60 seconds). This value reflects the terminal's recent handover frequency—an excessively high number indicates that the terminal is in a high-speed movement or coverage boundary jitter scenario, requiring increased hysteresis margin to suppress unnecessary handovers. The roaming context state vector is constructed by combining four components: movement mode (round trip or one-way), terminal movement speed level, current service quality level, and historical handover count. This vector will be used as input to the hysteresis adjustment function in step D2 to calculate the adaptive hysteresis margin and minimum dwell time.

[0095] In the embodiments corresponding to steps D11 to D16, a multi-dimensional roaming context state vector is constructed through the cascading processing of the above steps. This vector integrates the terminal's mobile behavior pattern, mobile speed estimation, service quality requirements, and recent handover history, providing comprehensive decision input for the adaptive adjustment of hysteresis parameters in the subsequent step D2. In particular, the round-trip mobility pattern identification mechanism effectively solves the ping-pong handover problem caused by the terminal jittering back and forth at the coverage boundary.

[0096] Step D2: Based on the roaming context state vector, calculate the first hysteresis parameter and the second hysteresis parameter using a preset hysteresis adjustment function; wherein, the hysteresis adjustment function takes each component in the roaming context state vector as input and maps each component to the value range of the hysteresis margin and the minimum dwell time using a linear weighted summation method; wherein, the first hysteresis parameter is the hysteresis margin and the second hysteresis parameter is the minimum dwell time; The hysteresis adjustment function maps the roaming context state vector to two hysteresis parameters: hysteresis margin and minimum dwell time. Hysteresis margin controls the sensitivity of roaming triggering—a larger margin results in more conservative roaming triggering, requiring more severe link quality degradation to trigger a handover; a smaller margin results in more sensitive roaming triggering, responding more quickly to changes in link quality. Minimum dwell time controls the mandatory waiting time after a roaming handover—a new handover is prohibited within the minimum dwell time after a handover is completed to avoid the ping-pong effect.

[0097] The design principles of the hysteresis adjustment function are as follows: when the terminal's mobility speed is high, reduce the hysteresis margin and shorten the minimum dwell time to enable the terminal to respond quickly to link degradation and switch in a timely manner; when the service quality level is high, increase the hysteresis margin and extend the minimum dwell time to avoid frequent switching from interrupting real-time services; when the historical number of switching is high, appropriately increase the hysteresis margin and extend the minimum dwell time to suppress unnecessary switching.

[0098] The specific calculation of the hysteresis adjustment function involves mapping each component to a numerical value and performing a linear weighted summation, which is elaborated in detail in steps D21 to D23.

[0099] Specifically, step D2 includes steps D21 to D23: Step D21: Map the terminal mobility speed level, the current service quality level, and the historical handover count to the corresponding speed value, service quality value, and handover count value, respectively; The three tier components need to be mapped from discrete tier labels to continuous numerical variables in order to participate in the linear weighted summation calculation. The mapping rule for the speed value V_num is: high tier to 3, medium tier to 2, and low tier to 1. The mapping rule for the quality of service value QoS_num is: high tier to 3, medium tier to 2, and low tier to 1. The handover count value H_num is directly taken as the integer value of the historical handover count. These mapped values ​​are used in the linear weighted summation formulas in subsequent steps D22 and D23.

[0100] Step D22: The hysteresis margin is equal to the baseline margin value minus the product of the speed value and the first weighting coefficient, plus the product of the service quality value and the second weighting coefficient, plus the product of the number of handovers and the third weighting coefficient. The formula for calculating hysteresis margin is H = H_0 - k1·V_num + k2·QoS_num + k3·H_num.

[0101] Where H represents the hysteresis margin, H_0 is the baseline margin value, k1 is the first weighting coefficient, k2 is the second weighting coefficient, and k3 is the third weighting coefficient. The baseline margin value H_0 can be 0.08, k1 can be 0.015, k2 can be 0.01, and k3 can be 0.005. The physical meaning of the above parameters is as follows: the baseline margin value provides the baseline level of the hysteresis margin; the negative coefficient of the speed value reflects the need to reduce the margin when moving at high speed; and the positive coefficients of the service quality value and the number of historical handovers reflect the need to increase the margin when there are high QoS services and high handover frequencies. The final value range of the hysteresis margin is 0.02 to 0.15, and boundary pruning ensures that it does not exceed this range.

[0102] For example, the parameter values ​​of the hysteresis adjustment function are shown in Table 3 below: Table 3: Hysteresis Margin Adjustment Parameters Step D23: The minimum dwell time is equal to the base time value minus the product of the speed value and the fourth weighting coefficient, plus the product of the service quality value and the fifth weighting coefficient, plus the product of the number of handovers and the sixth weighting coefficient; wherein, the first weighting coefficient, the second weighting coefficient, the third weighting coefficient, the fourth weighting coefficient, the fifth weighting coefficient and the sixth weighting coefficient are all positive numbers; The formula for calculating the minimum dwell time is T = T_0 - k4·V_num + k5·QoS_num + k6·H_num.

[0103] Where T is the minimum dwell time (in seconds), T_0 is the baseline time value, k4 is the fourth weighting coefficient, k5 is the fifth weighting coefficient, and k6 is the sixth weighting coefficient. The baseline time value T_0 can be 5 seconds, k4 can be 1.0 second, k5 can be 0.5 seconds, and k6 can be 0.3 seconds. Similar to the calculation of hysteresis margin, the negative coefficient of the speed value reflects the need to shorten the dwell time when moving at high speed, while the positive coefficients of the service quality value and the number of historical handovers reflect the need to extend the dwell time when there are high QoS services and high handover frequencies. The final value of the minimum dwell time ranges from 1 second to 10 seconds, and boundary pruning ensures that it does not exceed this range.

[0104] For example, the minimum dwell time adjustment parameters are shown in Table 4 below: Table 4: Minimum Dwell Time Adjustment Parameters In the embodiments corresponding to steps D21 to D23, the roaming context state vector is transformed into two specific hysteresis parameter values ​​through the cascading processing of the above steps. The linear weighted summation function ensures that the contribution of each component is interpretable and adjustable. The six weight coefficients are calibrated and stored in the configuration file before the device leaves the factory, and can be adjusted according to the optimization requirements of different network deployment scenarios. This adaptive adjustment mechanism enables the roaming triggering strategy to dynamically change according to the real-time mobility status of the terminal and service requirements, overcoming the insufficient adaptability of fixed hysteresis parameters in different scenarios.

[0105] Step D3: When the terminal's moving speed level is higher, reduce the hysteresis margin and shorten the minimum dwell time; This step provides a qualitative explanation of the calculation results in step D2. Since the coefficient for the speed value in the hysteresis margin formula is negative (-k1), the calculated value of the hysteresis margin H is smaller when the terminal's movement speed level is higher (V_num is larger). Similarly, the coefficient for the speed value in the minimum dwell time formula is negative (-k4), and the dwell time T is shorter when the speed level is higher. The purpose of this adjustment strategy is to make roaming triggering more sensitive in high-speed movement scenarios—a smaller hysteresis margin allows the terminal to trigger a scan as soon as the link quality begins to deteriorate, and a shorter minimum dwell time allows the terminal to perform the next handover more quickly to keep up with the movement speed.

[0106] Step D4: When the current service quality level is higher than the preset level or the number of historical handovers is higher than the preset number, increase the hysteresis margin by the preset margin and extend the minimum dwell time by the preset duration; This step also provides a qualitative explanation of the calculation results in step D2. When the service quality level is high (QoS_num=3), the calculated values ​​of hysteresis margin and minimum dwell time increase due to the positive coefficient of QoS_num. When the number of historical handovers is high (H_num is large), the hysteresis margin and minimum dwell time also increase. The preset level can be medium (i.e., QoS_num≥2), and the preset number of handovers can be 3. The purpose of this adjustment strategy is to increase the conservatism of roaming triggering in high-quality service scenarios or high handover frequency scenarios—a larger hysteresis margin avoids frequent unnecessary handovers due to short-term fluctuations in link quality, and a longer minimum dwell time ensures sufficient stabilization time after each handover to maintain service continuity.

[0107] Step D5: Using the hysteresis margin as an offset, add the hysteresis margin to the trigger threshold to obtain the upper limit trigger threshold, and subtract the hysteresis margin to obtain the lower limit trigger threshold; A preset trigger threshold combined with a hysteresis margin forms a dual-threshold strip trigger interval. The formula for calculating the upper trigger threshold is T_high = T_base + H; the formula for calculating the lower trigger threshold is T_low = T_base - H. Where T_base is the preset trigger threshold (e.g., -0.05 set in step 102), and H is the hysteresis margin calculated in step D2. The dual-threshold mechanism introduces a hysteresis effect—the link quality degradation rate needs to exceed the upper trigger threshold to trigger a roaming scan, and once triggered, the scan only stops when the degradation rate falls below the lower trigger threshold. The difference between the upper and lower thresholds, 2H, forms a non-sensitive interval, avoiding frequent start-stop scanning when the degradation rate fluctuates near the threshold.

[0108] Step D6: Trigger a roaming scan when the link quality degradation rate exceeds the upper limit trigger threshold; When the absolute value of the link quality degradation rate D_rate calculated in step A6 exceeds the upper limit trigger threshold T_high, the terminal re-enters the roaming scan process in step 102. The upper limit trigger threshold takes into account hysteresis margin, so that scanning is not triggered when the link quality has just begun to degrade but the degree is still within the margin range. The scanning process is only started when the degree of degradation really requires a switch.

[0109] Step D7: Stop roaming scan when the link quality degradation rate is lower than the lower limit trigger threshold; When the absolute value of the link quality degradation rate falls below the lower threshold T_low, the terminal stops roaming scanning. This mechanism ensures that the termination of the scan also has a hysteresis effect—once the scan is triggered, even if the link quality recovers somewhat, the scan will continue as long as the degradation rate is still above the lower threshold, avoiding the overhead of prematurely terminating the scan due to a brief recovery in link quality, which could lead to the need to restart the scan later.

[0110] Step D8: Using the minimum dwell time as the mandatory waiting time, prohibit triggering a new roaming handover within the mandatory waiting time after completing a roaming handover; The minimum dwell time, acting as a mandatory waiting period, is timed after the terminal completes a roaming handover. During this time, even if the link quality degradation rate exceeds the upper trigger threshold, the terminal will not trigger a new roaming handover. This mechanism effectively prevents the ping-pong effect—the possibility of an unnecessary handover triggered by transient fluctuations in the signal environment immediately after the terminal switches to the target access point. The mandatory waiting period ensures that the terminal has sufficient time to stabilize at the new access point and obtain an accurate link quality assessment before making a decision.

[0111] Step D9: Record the execution result of this roaming handover; wherein, the execution result includes whether the handover was successful, the handover time, and the change in the overall link quality score before and after the handover; After each roaming handover, the terminal records the execution result in its local log. The execution result includes three metrics: ① Handover success – whether the association and authentication process was successfully completed; ② Handover time – the time interval from triggering the roaming scan to completing the target access point association; ③ Change in the overall link quality score before and after the handover – the difference between the overall score of the target access point calculated in step A2 and the overall score of the source access point. These three metrics quantify the effectiveness of the roaming handover from different dimensions, providing data for subsequent inefficient handover assessments and trigger threshold adjustments.

[0112] Step D10: When the handover time exceeds the preset time threshold or the change in the comprehensive link quality score before and after the handover is negative, mark this roaming handover as an inefficient handover; The preset handover time threshold can be set to 500 milliseconds. This threshold reflects the upper limit of handover interruption perceptible to the user—a handover time exceeding 500 milliseconds may lead to real-time service interruption. When the handover time exceeds the preset threshold, it indicates that the handover process is too long, possibly due to improper target access point selection or abnormal authentication process. When the change in the overall link quality score before and after the handover is negative, it indicates that the link quality has actually decreased after the handover, and the actual quality of the target access point is worse than expected. Both of these situations indicate that the roaming handover has failed to achieve the expected results and are marked as inefficient handover.

[0113] Step D11: Calculate the percentage of inefficient switching events within a preset statistical period; The preset statistical period can be 10 minutes, which covers multiple handover events of the terminal in typical mobile scenarios, ensuring the statistical results are representative. The formula for calculating the proportion of inefficient handovers is R_low = N_low / N_total. Here, N_low is the number of inefficient handovers within the statistical period, and N_total is the total number of handovers within the statistical period. When N_total is zero (i.e., no handovers occurred within the statistical period), R_low takes a zero value, and no threshold adjustment is triggered.

[0114] Step D12: When the proportion of inefficient switching exceeds a preset ratio threshold, the trigger threshold is increased by a preset value; The preset threshold value can be 0.3. This means that when the proportion of inefficient handovers exceeds 30%, the current trigger threshold is considered too sensitive, causing the terminal to trigger handover prematurely when the link quality is still acceptable, resulting in the selection of an undesirable target access point. In this case, the trigger threshold is increased by a preset value, which can be 0.02, meaning T_base is adjusted to T_base + 0.02. The adjusted trigger threshold requires a higher link quality degradation rate to trigger roaming, ensuring that the terminal only performs handover when the link degradation is more severe, thus improving the quality of the target access point during handover.

[0115] Step D13: When the proportion of inefficient switching is lower than the preset proportion threshold and the switching success rate meets the preset success rate condition, the trigger threshold is reduced by a preset value. When the proportion of inefficient handovers is below a preset threshold and the handover success rate meets a preset success rate condition, it indicates that the current trigger threshold is too conservative, and the terminal is delaying handover even when link quality is significantly degraded. The preset success rate condition can be a handover success rate higher than 95%. In this case, the trigger threshold is lowered by a preset value, which can be 0.01, i.e., T_base is adjusted to T_base-0.01. The adjusted trigger threshold makes the terminal more sensitive to link quality degradation, triggering roaming scans and handovers earlier, reducing passive handovers when link quality is already severely degraded.

[0116] Step D14: Apply the adjusted trigger threshold to the next link quality degradation rate determination process; The adjusted trigger threshold is written into the terminal's configuration parameters and takes effect in the next execution cycle of steps 101 to 102. Because the adjustment range of the trigger threshold is small (±0.01 to ±0.02 each time), a single adjustment will not cause drastic changes in roaming behavior. Instead, through cumulative adjustments over multiple statistical cycles, the trigger threshold gradually converges to a level suitable for the current network environment and terminal usage patterns. The lower limit of the trigger threshold is pruned to -0.15, and the upper limit is pruned to -0.01 to ensure that the trigger threshold always remains within a reasonable range.

[0117] In the embodiments corresponding to steps D1 to D14, a complete mechanism is achieved through the cascading processing of the above steps, from roaming context awareness to adaptive adjustment of hysteresis parameters and then to closed-loop optimization of trigger thresholds. The roaming context state vector integrates multi-dimensional information such as mobility mode, speed level, service QoS, and handover history. The linearly weighted hysteresis adjustment function maps the context state to two interpretable parameters: hysteresis margin and minimum dwell time. The dual-threshold strip triggering and minimum dwell time mechanism effectively suppress the ping-pong effect. Inefficient handover statistics and dynamic adjustment of trigger thresholds constitute a closed-loop feedback optimization path. The entire mechanism enables the roaming triggering strategy to dynamically change according to the real-time mobile scenario and service requirements of the terminal, and to gradually optimize to the optimal operating point in long-term operation.

[0118] To verify the technical effectiveness of this embodiment, a comparative experiment was conducted in an enterprise campus network environment containing 50 wireless access points and 200 mobile terminals. The experiment lasted for 72 hours, and the roaming performance indicators of the three methods were statistically analyzed: ① Traditional RSSI threshold method – roaming is triggered only by a fixed threshold based on the received signal strength indication, without prediction and adaptive mechanisms; ② Fixed hysteresis method – roaming is triggered by a fixed hysteresis parameter based on the comprehensive link quality score, without adaptive adjustment; ③ The adaptive predictive roaming method of this embodiment. Evaluation indicators included average handover latency, ping-pong handover rate, handover success rate, and number of service interruptions. The experimental results are shown in Table 5 below.

[0119] For example, the roaming performance comparison results are shown in Table 5: Table 5: Performance Comparison of Different Roaming Methods The experimental results show that, compared with the traditional RSSI threshold method, the adaptive predictive roaming method in this embodiment reduces the average handover latency by 65.5%, the ping-pong handover rate from 18.5% to 2.1%, the handover success rate by 7.1 percentage points, and the number of service interruptions by 89.5%. Compared with the fixed hysteresis method, all indicators are also significantly improved. This verifies the comprehensive effect of multi-dimensional link quality prediction, directional sector scanning, Pareto multi-objective optimization selection, and adaptive adjustment of hysteresis parameters, and proves the effectiveness and superiority of this method in complex wireless network environments such as enterprise campuses.

[0120] In the embodiments corresponding to steps 101 to 103, by collecting multi-dimensional link quality parameters in real time and calculating the link quality degradation rate and its direction information based on these parameters, the handover decision no longer relies on a single indicator. This multi-dimensional analysis method makes the handover process more intelligent, helping to quickly and accurately identify link quality problems and respond promptly. When the link quality degradation rate exceeds a preset threshold, the method can quickly locate the target scan sector, thereby reducing the search range of candidate APs. This optimization process significantly improves the efficiency of handover and reduces user latency during the handover process. By comprehensively analyzing the predicted link quality, load status, and historical handover success rate of candidate access points, multiple optimization objectives are constructed, and Pareto optimization is used to determine the target access point. This method can select higher-quality access points, thereby significantly improving the handover success rate and reducing service interruptions and network instability caused by handover failures.

[0121] like Figure 2 This invention provides an AP switching device in PON and ONU environments. Please refer to [link / reference]. Figure 2 , Figure 2 A schematic diagram of an AP switching device in a PON and ONU environment provided by the present invention is shown, as follows: Figure 2The AP switching device shown in the PON and ONU environments includes: The link quality prediction module 21 is used to collect multi-dimensional link quality parameters of the current access point in real time, and calculate the link quality degradation rate and its direction information based on the multi-dimensional link quality parameters. The directional scanning module 22 is used to determine the target scanning sector in a pre-constructed access point topology map based on the direction information when the link quality degradation rate exceeds a preset trigger threshold, and to perform directional scanning on the candidate access points in the target scanning sector to obtain a set of candidate access points; The target selection module 23 is used to construct multiple optimization targets for each candidate access point in the candidate access point set by comprehensively considering its predicted link quality, load status and historical handover success rate, and determine the target access point through Pareto optimization. The roaming execution module 24 is used to perform a roaming handover to the target access point, adaptively adjust the hysteresis parameter according to the roaming context state, and feed back the adjusted hysteresis parameter to the update process of the trigger threshold.

[0122] This invention provides an AP handover device in PON and ONU environments. By collecting multi-dimensional link quality parameters in real time and calculating the link quality degradation rate and its direction information based on these parameters, handover decisions no longer rely on a single indicator. This multi-dimensional analysis method makes the handover process more intelligent, helping to quickly and accurately identify link quality problems and respond promptly. When the link quality degradation rate exceeds a preset threshold, the method can quickly locate the target scanning sector, thereby reducing the search range of candidate APs. This optimization process significantly improves handover efficiency and reduces user latency during handover. By comprehensively analyzing the predicted link quality, load status, and historical handover success rate of candidate access points, multiple optimization objectives are constructed, and Pareto optimization is used to determine the target access point. This method can select higher-quality access points, thereby significantly improving the handover success rate and reducing service interruptions and network instability caused by handover failures.

[0123] Figure 3 This is a schematic diagram of a terminal device provided in an embodiment of the present invention. Figure 3 As shown, a terminal device 3 in this embodiment includes: a processor 30, a memory 31, and a computer program 32 stored in the memory 31 and executable on the processor 30, such as an AP switching program in a PON and ONU environment. When the processor 30 executes the computer program 32, it implements the steps in the various AP switching method embodiments in a PON and ONU environment described above, for example... Figure 1 Steps 101 to 103 are shown. Alternatively, when the processor 30 executes the computer program 32, it implements the functions of each unit in the above-described device embodiments, for example... Figure 2 The function of the unit shown.

[0124] For example, the computer program 32 can be divided into one or more units, which are stored in the memory 31 and executed by the processor 30 to complete the present invention. The one or more units can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program 32 in the terminal device 3. For example, the specific functions of each unit of the computer program 32 can be divided as follows: The link quality prediction module is used to collect multi-dimensional link quality parameters of the current access point in real time, and calculate the link quality degradation rate and its direction information based on the multi-dimensional link quality parameters. The directional scanning module is used to determine the target scanning sector in a pre-constructed access point topology map based on the direction information when the link quality degradation rate exceeds a preset trigger threshold, and to perform directional scanning on the candidate access points in the target scanning sector to obtain a set of candidate access points. The target selection module is used to construct multiple optimization targets for each candidate access point in the candidate access point set by comprehensively considering its predicted link quality, load status and historical handover success rate, and determine the target access point through Pareto optimization. The roaming execution module is used to perform roaming handover to the target access point, adaptively adjust the hysteresis parameters according to the roaming context state, and feed back the adjusted hysteresis parameters to the update process of the trigger threshold.

[0125] The terminal device includes, but is not limited to, a processor 30 and a memory 31. Those skilled in the art will understand that... Figure 3 This is merely an example of a terminal device 3 and does not constitute a limitation on a terminal device 3. It may include more or fewer components than shown, or combine certain components, or different components. For example, the terminal device may also include input / output devices, network access devices, buses, etc.

[0126] The processor 30 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0127] The memory 31 can be an internal storage unit of the terminal device 3, such as a hard disk or memory of the terminal device 3. The memory 31 can also be an external storage device of the terminal device 3, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device 3. Furthermore, the memory 31 can include both internal and external storage units of the terminal device 3. The memory 31 is used to store the computer program and other programs and data required by the roaming control device. The memory 31 can also be used to temporarily store data that has been output or will be output.

[0128] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0129] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of the present invention. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0130] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0131] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0132] This invention provides a computer program product that, when run on a mobile terminal, enables the mobile terminal to implement the steps described in the above-described method embodiments.

[0133] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0134] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0135] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0136] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0137] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units.

[0138] It should be understood that, when used in this specification and the appended claims, terms include indicating the presence of the described feature, integral, step, operation, element and / or component, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0139] It should also be understood that the terms used in this specification and the appended claims refer to any combination of one or more of the associated listed items and all possible combinations, and include such combinations.

[0140] As used in this specification and the appended claims, the term "if" can be interpreted, depending on the context, as when, once, or in response to determination or in response to detection. Similarly, the phrase "if determined" or "if detected [the described condition or event]" can be interpreted, depending on the context, as once determined or in response to determination or detection [the described condition or event] or in response to detection [the described condition or event].

[0141] Furthermore, in the description of this invention and the appended claims, the terms first, second, third, etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0142] References to one or more embodiments described in this specification mean that a particular feature, structure, or characteristic described in connection with that embodiment is included in one or more embodiments of the invention. Therefore, phrases appearing in different parts of this specification as referring to one embodiment, some embodiments, some other embodiments, and others do not necessarily refer to the same embodiment, but rather mean one or more, but not all, embodiments, unless otherwise specifically emphasized. The terms include, comprise, have, and variations thereof mean including but not limited to, unless otherwise specifically emphasized.

[0143] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for AP switching in a PON and ONU environment, characterized in that, The AP switching method under PON and ONU environments includes: Real-time collection of multi-dimensional link quality parameters of the current access point, and calculation of link quality degradation rate and its direction information based on the multi-dimensional link quality parameters; When the link quality degradation rate exceeds a preset trigger threshold, the target scanning sector is determined in the pre-constructed access point topology map according to the direction information, and a directional scan is performed on the candidate access points in the target scanning sector to obtain a set of candidate access points. For each candidate access point in the candidate access point set, multiple optimization objectives are constructed by combining its predicted link quality, load status and historical handover success rate, and the target access point is determined by Pareto optimization. The process involves performing a roaming handover to the target access point, adaptively adjusting the hysteresis parameters based on the roaming context state, and feeding back the adjusted hysteresis parameters to the update process of the trigger threshold. The step of constructing multiple optimization objectives for each candidate access point in the candidate access point set by comprehensively considering its predicted link quality, load status, and historical handover success rate, and determining the target access point through Pareto optimization includes: For each candidate access point in the candidate access point set, its predicted link quality is taken as the first optimization objective, the current load status is taken as the second optimization objective, and the historical handover success rate is taken as the third optimization objective. The first optimization objective, the second optimization objective, and the third optimization objective are normalized respectively to unify the value range of each optimization objective to between 0 and 1; In the set of candidate access points, a dominance relationship is defined as the condition that any candidate access point is not inferior to another candidate access point in all optimization objectives and the difference between them in at least one optimization objective is greater than a threshold. From the set of candidate access points, extract the Pareto optimal solution set that is not dominated by other candidate access points; When the Pareto optimal solution set contains only one candidate access point, it is determined as the target access point; When the Pareto optimal solution set contains multiple candidate access points, the sum of the Euclidean distances from each Pareto optimal solution to all other solutions in the Pareto front is calculated as the inflection point metric, and the candidate access point with the largest inflection point metric is selected as the target access point.

2. The AP switching method under PON and ONU environments as described in claim 1, characterized in that, The steps of real-time acquisition of multi-dimensional link quality parameters of the current access point and calculation of link quality degradation rate and its direction information based on the multi-dimensional link quality parameters include: Real-time acquisition of the received signal strength indication, signal-to-noise ratio, data packet retransmission rate and channel utilization of the current access point as multi-dimensional link quality parameters; Each dimension of the multi-dimensional link quality parameters is normalized, and the normalized parameters are calculated by weighted geometric mean according to preset weights to obtain a comprehensive link quality score. The link quality comprehensive scores are arranged in chronological order to construct a link quality vector; The link quality vector is input into a long short-term memory network, and the link quality prediction sequence within a future preset time window is output. Calculate the link quality difference between adjacent time points in the link quality prediction sequence to obtain the link quality change rate sequence; The average of the last preset number of elements in the link quality change rate sequence is taken as the link quality degradation rate. The direction of degradation is determined based on the sign of the link quality degradation rate.

3. The AP switching method under PON and ONU environments as described in claim 1, characterized in that, The step of determining a target scanning sector in a pre-constructed access point topology map based on the direction information when the link quality degradation rate exceeds a preset trigger threshold, and performing a directional scan on candidate access points within the target scanning sector to obtain a set of candidate access points includes: The node corresponding to the current access point and the set of neighboring nodes that have a topological adjacency relationship with the current access point are obtained from the access point topology graph; wherein, the access point topology graph uses the access point as the node and the historical switching frequency and signal coverage overlap between access points as the edge weights. The direction of movement of the current access point in the access point topology map is determined based on the direction information; Centered on the node corresponding to the current access point, with the direction of movement as the axis, and with a preset angle range as the subtended angle, a fan-shaped area is extracted from the access point topology map as the target scanning sector; Obtain the access points corresponding to all neighboring nodes within the target scan sector as a list of access points to be scanned; For each access point in the list of access points to be scanned, a probe request frame is sent and a probe response frame is received on its working channel to collect the real-time link quality parameters of each access point. The access points that receive probe response frames and their corresponding real-time link quality parameters are used to form a set of candidate access points.

4. The AP switching method under PON and ONU environments as described in claim 3, characterized in that, Before the step of obtaining the node corresponding to the current access point and the set of neighboring nodes that have a topological adjacency relationship with the current access point from the access point topology graph, the method further includes: Collect historical roaming handover records; each record includes the source access point identifier, the target access point identifier, whether the handover was successful, and the signal strength difference at the time of handover. For any two access point nodes in the access point topology graph that have overlapping signal coverage, count their historical successful handover counts and failed handover counts, and use the Bayesian estimation method to calculate their posterior handover success probability as the weight of the edge between the two nodes. Edges with weights lower than a preset threshold are removed from the access point topology graph, and the remaining edges constitute the adjacency relationship of the access point topology graph. When a new roaming switch is completed, the number of successful and failed switches between the corresponding node pairs is updated based on the switch results, and the edge weights are recalculated.

5. The AP switching method under PON and ONU environments as described in claim 1, characterized in that, The step of taking the predicted link quality as the first optimization objective, the current load status as the second optimization objective, and the historical handover success rate as the third optimization objective for each candidate access point in the candidate access point set includes: For each candidate access point in the candidate access point set, the most recently calculated comprehensive link quality score is taken as the predicted link quality of that candidate access point. Send a load query request frame to the candidate access point and receive the currently associated terminal number, channel utilization and available bandwidth information returned by it; The ratio of the current number of associated terminals to the maximum number of associated terminals at the access point is used as the load normalization value. The ratio of the available bandwidth information to the total bandwidth of the access point is used as the normalized value of the available bandwidth; The product of the complement of the channel utilization rate and the normalized value of the available bandwidth is used as the capacity normalized value; The weighted average of the load normalization value and the capacity normalization value is taken as the current load status; The edge weights between the current access point and the candidate access point are queried from the access point topology graph, and these edge weights are used as the historical handover success rate.

6. The AP switching method under PON and ONU environments as described in claim 1, characterized in that, The steps of performing a roaming handover to the target access point, adaptively adjusting the hysteresis parameter according to the roaming context state, and feeding back the adjusted hysteresis parameter to the update process of the trigger threshold include: After completing the roaming handover to the target access point, a roaming context state vector is constructed; wherein, the roaming context state vector includes the terminal mobility speed level, the current service quality level, and the number of historical handovers within a preset time window; Based on the roaming context state vector, a first hysteresis parameter and a second hysteresis parameter are calculated using a preset hysteresis adjustment function; wherein, the hysteresis adjustment function takes each component in the roaming context state vector as input and uses a linear weighted summation method to map each component to the value range of hysteresis margin and minimum dwell time; wherein, the first hysteresis parameter is the hysteresis margin and the second hysteresis parameter is the minimum dwell time; When the terminal's moving speed level is higher, the hysteresis margin is reduced and the minimum dwell time is shortened; When the current service quality level is higher than the preset level or the number of historical switching times is higher than the preset number, the hysteresis margin is increased by the preset margin and the minimum dwell time is extended by the preset duration. Using the hysteresis margin as an offset, the upper limit trigger threshold is obtained by adding the hysteresis margin to the trigger threshold, and the lower limit trigger threshold is obtained by subtracting the hysteresis margin. A roaming scan is triggered when the link quality degradation rate exceeds the upper limit trigger threshold. Roaming scans are stopped when the link quality degradation rate falls below the lower limit trigger threshold. The minimum dwell time is used as the mandatory waiting time. During the mandatory waiting time after a roaming handover is completed, a new roaming handover is prohibited.

7. The AP switching method under PON and ONU environments as described in claim 6, characterized in that, The step of calculating the first hysteresis parameter and the second hysteresis parameter based on the roaming context state vector using a preset hysteresis adjustment function includes: The terminal mobility speed level, the current service quality level, and the historical handover count are respectively mapped to the corresponding speed value, service quality value, and handover count value. The hysteresis margin is equal to the baseline margin value minus the product of the speed value and the first weighting coefficient, plus the product of the service quality value and the second weighting coefficient, plus the product of the number of handovers and the third weighting coefficient. The minimum dwell time is equal to the base time value minus the product of the speed value and the fourth weighting coefficient, plus the product of the service quality value and the fifth weighting coefficient, plus the product of the number of handovers and the sixth weighting coefficient; wherein the first weighting coefficient, the second weighting coefficient, the third weighting coefficient, the fourth weighting coefficient, the fifth weighting coefficient, and the sixth weighting coefficient are all positive numbers.

8. The AP switching method under PON and ONU environments as described in claim 6, characterized in that, The step of constructing the roaming context state vector after completing the roaming handover to the target access point includes: The terminal's mobility mode is determined based on the sign change sequence of the link quality degradation rate before and after roaming handover. When the symbol change sequence shows alternating positive and negative changes within a preset time period, it is determined to be a round-trip movement mode; otherwise, it is determined to be a one-way movement mode. When the round-trip movement mode is determined, the terminal movement speed level is set to a high level; When the unidirectional movement mode is determined, the terminal movement speed level is determined based on the absolute value of the link quality degradation rate; wherein, the terminal movement speed level includes high level, medium level and low level; The system detects the real-time transmission requirements of currently active services and determines the service quality level of the current service based on the service type mapping table. Among them, real-time audio and video services are mapped to a high level, web browsing services are mapped to a medium level, and background data synchronization services are mapped to a low level. The number of historical roaming switches within the preset time window is counted, and the roaming context state vector is formed by combining the mobile mode determination result, the terminal mobile speed level, the current service quality level, and the number of historical roaming switches.

9. The AP switching method under PON and ONU environments as described in claim 1, characterized in that, Following the step of feeding back the adjusted hysteresis parameter to the update process of the trigger threshold, the method further includes: Record the execution result of this roaming handover; wherein, the execution result includes whether the handover was successful, the handover time, and the change in the overall link quality score before and after the handover; When the handover time exceeds a preset time threshold or the change in the overall link quality score before and after the handover is negative, the current roaming handover will be marked as an inefficient handover. The percentage of inefficient switching events within a preset statistical period is recorded. When the proportion of inefficient switching exceeds a preset threshold, the trigger threshold is increased by a preset value; When the proportion of inefficient switching is lower than the preset proportion threshold and the switching success rate meets the preset success rate condition, the trigger threshold is reduced by a preset value. The adjusted trigger threshold will be applied to the next link quality degradation rate assessment process.

10. An AP switching device in a PON and ONU environment, characterized in that, The AP switching device in the PON and ONU environments includes: The link quality prediction module is used to collect multi-dimensional link quality parameters of the current access point in real time, and calculate the link quality degradation rate and its direction information based on the multi-dimensional link quality parameters. The directional scanning module is used to determine the target scanning sector in a pre-constructed access point topology map based on the direction information when the link quality degradation rate exceeds a preset trigger threshold, and to perform directional scanning on the candidate access points in the target scanning sector to obtain a set of candidate access points. The target selection module is used to construct multiple optimization targets for each candidate access point in the candidate access point set by comprehensively considering its predicted link quality, load status and historical handover success rate, and determine the target access point through Pareto optimization. The roaming execution module is used to perform roaming handover to the target access point, adaptively adjust the hysteresis parameter according to the roaming context state, and feed back the adjusted hysteresis parameter to the update process of the trigger threshold. The step of constructing multiple optimization objectives for each candidate access point in the candidate access point set by comprehensively considering its predicted link quality, load status, and historical handover success rate, and determining the target access point through Pareto optimization includes: For each candidate access point in the candidate access point set, its predicted link quality is taken as the first optimization objective, the current load status is taken as the second optimization objective, and the historical handover success rate is taken as the third optimization objective. The first optimization objective, the second optimization objective, and the third optimization objective are normalized respectively to unify the value range of each optimization objective to between 0 and 1; In the set of candidate access points, a dominance relationship is defined as the condition that any candidate access point is not inferior to another candidate access point in all optimization objectives and the difference between them in at least one optimization objective is greater than a threshold. From the set of candidate access points, extract the Pareto optimal solution set that is not dominated by other candidate access points; When the Pareto optimal solution set contains only one candidate access point, it is determined as the target access point; When the Pareto optimal solution set contains multiple candidate access points, the sum of the Euclidean distances from each Pareto optimal solution to all other solutions in the Pareto front is calculated as the inflection point metric, and the candidate access point with the largest inflection point metric is selected as the target access point.

11. A terminal device, characterized in that, The terminal device includes: a memory, a processor, and an AP switching program in PON and ONU environments stored in the memory and executable on the processor. The AP switching program in PON and ONU environments is configured to implement the steps in the AP switching method in PON and ONU environments as described in any one of claims 1 to 9.

12. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps in the AP switching method under PON and ONU environments as described in any one of claims 1 to 9.

Images