Adaptive topology reconfiguration method for communication links

Abstract

This invention belongs to the field of mobile communication network technology and is used to solve the problem that existing technologies, while utilizing the predictability of node movement, cannot eliminate the inherent reaction gap in reactive handover mechanisms. Specifically, it relates to an adaptive topology reconstruction method for communication links, including: acquiring a node motion state sequence, which includes the node's location information at multiple historical moments; based on the motion state sequence, generating location prediction information for the node within a future time window and a prediction uncertainty metric associated with the location prediction information; determining a target parent node from candidate parent nodes based on the link availability metric, a first type of feature value, a second type of feature value, and dynamically adjusted safety margin parameters, and generating a handover command. This invention, by introducing a node motion state prediction mechanism, advances the trigger point for topology reconstruction from after link quality deterioration to before deterioration, fundamentally eliminating the inherent reaction gap in reactive handover.

Description

Technical Field

[0001] This invention belongs to the field of mobile communication network technology, specifically relating to an adaptive topology reconfiguration method for communication links. Background Technology

[0002] In mobile ad hoc networks, nodes collaborate to build temporary communication topologies to support multi-hop data transmission. Dynamic reconfiguration of network topology is a key mechanism to ensure connectivity and quality of service. Current mainstream topology reconfiguration methods generally adopt a reactive strategy: nodes periodically monitor the link quality of neighboring nodes through beacons. When the quality index of the current link is detected to be lower than a preset threshold, a neighbor scan process is triggered to reselect a parent node and complete the handover.

[0003] The aforementioned reactive mechanism implicitly assumes that link quality degradation is a relatively slow or random process, with a sufficient time window between detection of deterioration and complete link failure, allowing nodes to complete the handover process within this window. However, in practical applications, there exists a special scenario where nodes move at high speeds along predictable trajectories, such as drones flying along predetermined routes or automated guided vehicles operating along fixed tracks. In these scenarios, signal strength attenuation is often not gradual but rather exhibits a step-like decline due to physical obstacles. When a node detects that the signal has fallen below a threshold, it has often entered a communication dead zone, and the beacon used for handover negotiation cannot reach the potential target node, leading to handover failure and communication interruption. This contradiction between the predictability of node movement and the lag in the handover mechanism causes connection interruptions that could otherwise be avoided in deterministic movement scenarios.

[0004] Therefore, how to eliminate the inherent reaction gap in reactive switching mechanisms while taking advantage of the predictability of node movement has become an urgent technical problem to be solved. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive topology reconfiguration method for communication links, which solves the problem that existing technologies, while utilizing the predictability of node movement, cannot eliminate the inherent reaction gap in reactive switching mechanisms.

[0006] The technical problem to be solved by this invention is: how to provide an adaptive topology reconfiguration method for communication links that can eliminate the inherent reaction gap period of reactive switching mechanisms, while taking advantage of the predictability of node movement.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] An adaptive topology reconfiguration method for communication links, applied to mobile network nodes, includes:

[0009] Obtain the motion state sequence of the node, which includes the node's position information at multiple historical moments;

[0010] Based on the motion state sequence, the system generates the node's predicted location information within a future time window, as well as a measurement of the prediction uncertainty associated with the predicted location information.

[0011] Based on location prediction information, prediction uncertainty measurement, and a pre-defined signal propagation model, determine the link availability measurement between the node and at least one candidate parent node at future times.

[0012] Based on the link availability metric, extract the first type of feature value between the node and its current parent node, and the second type of feature value between the node and its candidate parent node. The first type of feature value is used to characterize the expected availability duration of the current link, and the second type of feature value is used to characterize the expected availability duration of the candidate link.

[0013] Based on the first type of feature value, the second type of feature value, and the dynamically adjusted safety margin parameter, the target parent node is determined from the candidate parent nodes, and a switching instruction is generated.

[0014] In response to a switch command, a communication link switch is performed from the current parent node to the target parent node.

[0015] The present invention has the following beneficial effects:

[0016] 1. By introducing a node motion state prediction mechanism, the trigger point for topology reconstruction is moved from after the link quality deteriorates to before the deterioration, fundamentally eliminating the inherent reaction gap period of reactive handover; based on the position prediction information and prediction uncertainty metric generated by Kalman filtering, nodes can predict the link failure time before entering the communication blind zone and initiate handover negotiation in advance, thereby achieving seamless handover of communication links in predictable motion scenarios and significantly reducing the probability of connection interruption caused by physical obstruction;

[0017] 2. By propagating prediction uncertainty to link availability probability, the problem that deterministic prediction cannot reflect prediction confidence is solved. The link availability probability constructed based on expected received signal strength and prediction variance enables handover decisions to automatically weigh the credibility of prediction values: when prediction uncertainty is large, the probability curve tends to flatten, the stable connection period is shortened, the danger approach period is delayed, and nodes automatically adopt conservative strategies to avoid misjudgment due to overconfidence, thereby improving the system's decision robustness in complex maneuvering scenarios.

[0018] 3. By introducing a closed-loop feedback mechanism based on historical switching deviations, dynamic adaptive adjustment of the safety margin parameter is achieved. A proportional-integral controller is used to adjust the deviation between the actual failure time and the predicted critical period, enabling the safety margin to automatically adapt to the time-varying characteristics of prediction accuracy and fluctuations in switching time, achieving a dynamic balance between aggressive and conservative approaches. Compared to a fixed threshold scheme, this mechanism effectively reduces the risk of invalid switching or switching failures due to prediction deviations, improving the long-term stability of topology reconstruction.

[0019] 4. By mining the correlation between the rate of change of acceleration and the steepness of the link availability probability attenuation, implicit perception and parameter adaptation of the channel environment are achieved. Based on environmental state labels, the probability threshold is dynamically adjusted, enabling the system to improve the requirements for stable connections in maneuvering diffusion environments (where signal attenuation is gradual but highly volatile) and provide early warnings in obstructed cutting environments (where signal may drop sharply), thus maintaining consistent handover performance under different channel conditions. This mechanism achieves deep utilization of environmental characteristics without the need for additional sensors, improving the versatility and engineering practicality of the solution in multiple scenarios. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art 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.

[0021] Figure 1 This is a flowchart of the method according to Embodiment 1 of the present invention. Detailed Implementation

[0022] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] In mobile ad hoc networks, topology reconfiguration is the core mechanism for ensuring communication continuity. Existing technologies generally employ reactive handover strategies: nodes periodically monitor real-time link quality via beacons, and when the quality index of the current link is detected to be lower than a preset threshold, a neighbor scan is triggered and a new parent node is selected. This mechanism implicitly assumes at the physical level that the link quality degradation process is slow enough, with a sufficient time window between detection of deterioration and complete link failure, allowing nodes to complete handover negotiation and execution within this window.

[0024] However, in scenarios where nodes move at high speeds along predictable trajectories, the above assumptions no longer hold. Taking a drone flying along a predetermined route as an example, when the drone bypasses mountains or buildings, the signal attenuation caused by blockage is not gradual, but rather drops abruptly from usable to completely interrupted within milliseconds. The inherent "detection-decision-execution" process of reactive mechanisms has a reaction gap of hundreds of milliseconds. When a node detects a signal below a threshold, it has often entered a communication dead zone, and the beacon used for handover negotiation cannot reach the potential target node, leading to handover failure and communication interruption. Existing technologies lack the ability to perceive and utilize the future movement trajectory of nodes, and cannot transform the inherent information of "movement predictability" into a pre-input for handover decisions.

[0025] Existing solutions fail to identify a key contradiction: while node movement is continuous and predictable, switching decisions are discrete and reactive. This contradiction leads to repeated exposure of the system to avoidable connection interruptions in deterministic motion scenarios. Specifically, when UAV swarms perform collaborative inspection tasks, some nodes fly across signal-blocked areas along predetermined routes. Existing technology can only record the disconnection state after the signal disappears, but cannot predict the impending link failure and initiate a switch in advance before the node enters the blockage area. Furthermore, when a node re-establishes a connection after flying out of the blockage area, the system only records one successful topology reconstruction, but fails to trace the communication gaps during the controllable flight period caused by the lack of prediction. As a result, the system's adaptive capabilities are limited to post-event remediation, and it cannot proactively avoid link failures.

[0026] If the aforementioned problems are not addressed, the communication reliability of mobile ad hoc networks in deterministic motion scenarios will continue to be limited by the inherent defects of reactive mechanisms. Nodes repeatedly experience avoidable communication interruptions in predictable obstruction areas, leading to delays or loss of critical control commands. Simultaneously, frequent handover failures and reconnection attempts consume valuable channel resources and node energy, further exacerbating network congestion and topology oscillations. Consequently, the system's perception and response to link quality consistently lags behind actual changes in the physical environment, making it impossible to maintain communication continuity and stability when nodes traverse obstruction areas. Ultimately, this restricts the engineering application of mobile ad hoc networks in highly dynamic scenarios such as drone swarms and intelligent transportation.

[0027] Example 1: As Figure 1 As shown, the adaptive topology reconfiguration method for communication links, applied to mobile network nodes, includes:

[0028] Step S1: Obtain the motion state sequence of the node, which includes the node's position information at multiple historical moments;

[0029] The motion state sequence also includes velocity and / or acceleration information of the node at multiple historical moments; the position prediction information includes the predicted position of the node at multiple discrete moments within a future time window; the prediction uncertainty measure includes the predicted position variance corresponding to each predicted position.

[0030] In this embodiment, step S1 is first executed to obtain the motion state sequence of the nodes. This motion state sequence is the basic data source for all subsequent predictions and decisions, and its accuracy directly affects the performance of the entire topology reconstruction method.

[0031] Specifically, the node is equipped with a combined navigation module of Global Positioning System (GPS) and Inertial Measurement Unit (INS), which can output high-frequency motion state measurements at a fixed sampling period. In this embodiment, the sampling period is... The time resolution was set to 100 milliseconds, a choice that balances the accuracy requirements of motion tracking with the processing power of the embedded system. At each sampling time... (in For sampling sequence number, The integrated navigation module synchronously outputs three types of raw measurement data: the first type is position measurement values, which include the longitude, latitude, and altitude coordinates of the node at the current moment, forming a three-dimensional vector. The second type is velocity measurement, which contains velocity components in three orthogonal directions, forming a three-dimensional vector. This data can be obtained directly from the velocity integration results of the inertial measurement unit, or it can be obtained by position difference at consecutive time points; the third type is acceleration measurement values, which contain acceleration components in three orthogonal directions, forming a three-dimensional vector. This data is directly output by the accelerometer in the inertial measurement unit, reflecting the instantaneous acceleration of the node in the body coordinate system.

[0032] Because sensors may experience abnormal jumps due to electromagnetic interference or momentary obstruction in real-world environments, outliers are unavoidable in the raw measurements. To ensure the purity of the input data, the node performs preprocessing on the raw measurements. The specific preprocessing method is as follows: For each measurement component, the system pre-evaluates the measurement noise standard deviation calibrated in the sensor manual, and the standard deviation of each component is measured. Take 0.5 meters and measure the standard deviation of each component of the speed. Take 0.1 m / s as the acceleration, and measure the standard deviation of each component. A value of 0.05 m / s² is used. The node calculates the deviation between the current measurement and the filtered value from the previous moment. If the absolute value of the deviation of any measurement component exceeds three times the corresponding standard deviation, the component is considered an outlier, and the filtered value from the previous moment is used to replace it. Here, we use... The criterion serves as the basis for judgment, and its statistical significance lies in the fact that, under normal circumstances, the measurement error follows a normal distribution, and the probability of exceeding three times the standard deviation is less than 0.3%. Therefore, it can effectively identify and eliminate abnormal jumps.

[0033] After outlier removal, the node combines the three types of measurement data into a unified observation vector, denoted as... This vector is a 9-dimensional column vector, containing three position components, three velocity components, and three acceleration components. Simultaneously, the nodes construct corresponding measurement noise covariance matrices. The matrix is ​​a 9×9 diagonal matrix, with the diagonal elements set as the variances of the three positional components. The variance of the three velocity components The variance of the three acceleration components The measurement noise covariance matrix is ​​constructed based on sensor accuracy indicators. Its physical significance lies in quantifying the reliability of the measurement value itself. In the subsequent Kalman filtering step, this matrix will be used to dynamically balance the confidence of the measurement value and the prediction model.

[0034] It is worth noting that this embodiment uses a combined navigation module instead of a single GPS positioning because in mobile scenarios involving drones or vehicles, simple position measurement suffers from low update frequency and easy loss of satellite signals. In contrast, an inertial measurement unit (IMU) can provide high-frequency velocity and acceleration information, maintaining short-term motion state recursion even during brief GPS lock-off. Integrating these three elements to form a motion state sequence ensures data continuity and provides rich observational information for subsequent Kalman filtering. After the above preprocessing, the data output in step S1 is continuous, clean, and possesses statistical characteristics, laying a reliable data foundation for the subsequent Kalman filtering process in step S2.

[0035] Step S2: Based on the motion state sequence, generate the node's position prediction information within the future time window and the prediction uncertainty measure associated with the position prediction information;

[0036] The specific process of generating the forecast uncertainty measure includes:

[0037] A Kalman filter is used to filter the motion state sequence to obtain the state estimation vector and estimation error covariance matrix at the current moment;

[0038] Based on the state estimation vector and the preset motion model, the predicted state vector at each prediction time is extrapolated.

[0039] Based on the estimated error covariance matrix and the motion model, the prediction error covariance matrix corresponding to each prediction time is obtained through propagation.

[0040] The predicted position is extracted from the predicted state vector, and the submatrix corresponding to the predicted position is extracted from the prediction error covariance matrix as the predicted position variance.

[0041] In this embodiment, step S2 takes the preprocessed observation vector and its noise covariance matrix output from step S1, and uses Kalman filtering to achieve the optimal estimation of the node's motion state. Based on the estimation result, it generates position prediction information within the future time window and the corresponding prediction uncertainty metric. The core of this step is to transform the noisy raw measurement data into a statistically optimal motion state estimate, while quantifying the confidence level of the estimate itself, providing an input with both expected value and uncertainty for subsequent link quality prediction.

[0042] At system startup Since the node has not yet obtained a sequence of historical motion states, the Kalman filter is initialized using the measurements from the first two sampling points. Specifically, the values ​​taken are... This means directly using the first measurement as the initial state. The initial covariance matrix P Set as a diagonal matrix, with diagonal elements taking 10 times the variance of the corresponding measurement noise to reflect the low confidence of the initial estimate. Perform a complete Kalman filter recursion starting from the third sampling point.

[0043] Specifically, a node first needs to establish a state-space model describing its own motion. The state vector is defined as a nine-dimensional column vector, denoted as . The state vector contains position, velocity, and acceleration components in three directions. Velocity and acceleration are included in the state vector because in scenarios with high node mobility, position information alone cannot accurately predict future trajectories. The introduction of velocity and acceleration allows the motion model to describe uniform motion, uniform acceleration, and even more complex motion patterns.

[0044] The state transition matrix F describes the evolution of the system state over time. In this embodiment, the F matrix is ​​constructed based on a uniformly accelerated motion model, which assumes that the acceleration remains constant over short time intervals. This assumption has sufficient engineering accuracy under a sampling period Ts = 100 milliseconds. The F matrix is ​​a 9×9 square matrix, and its specific construction method is as follows: the position components are related to the position, velocity, and acceleration of the previous moment, and the corresponding sub-blocks are 3×3 identity matrices. ,by A 3×3 diagonal matrix with diagonal elements ,by A 3×3 diagonal matrix with diagonal elements The velocity component is related to the velocity and acceleration of the previous moment, and the corresponding sub-block is... and The acceleration component is only related to the acceleration of the previous moment, and the corresponding sub-block is... The remaining sub-blocks are all zero matrices. The physical significance of this construction is that the change in position is contributed by both velocity and acceleration, the change in velocity is contributed by acceleration, and acceleration is considered constant in a short time.

[0045] The process noise covariance matrix Q is used to quantify the uncertainty of the motion model itself, i.e., the deviation between the actual motion and the ideal uniform acceleration model. Since the maneuvering behavior of nodes cannot be accurately predicted, a certain error margin must be reserved for the model through the Q matrix. In this embodiment, the Q matrix is ​​set as a 9×9 diagonal matrix, and its diagonal elements are pre-calibrated according to the target's maximum maneuverability: the position process noise variance is 0.01 square meters, the velocity process noise variance is 0.01 square meters per second squared, and the acceleration process noise variance is 0.01 square meters per second squared. These values ​​are selected based on the statistical analysis of the maneuver amplitude of typical UAVs or vehicles under normal flight / driving conditions, for example, the maximum rate of change of acceleration does not exceed 1 meter per second squared, and the corresponding time cumulative effect is converted into the above variance values.

[0046] The observation matrix H maps the state space to the measurement space. Since the observation vector provided in step S1 contains all components of position, velocity, and acceleration, and has the same dimension as the state vector, the H matrix is ​​directly taken as a 9×9 identity matrix. This means that the state space and the observation space are directly corresponding, and the Kalman filter works by filtering out noisy measurements. Optimally fuse the values ​​with the model predictions.

[0047] Based on the above model configuration, the node at each time step The standard recursive procedure for Kalman filtering is followed. First, state prediction is performed, utilizing the previous time step... Optimal state estimation The prior state estimate at the current time is obtained by extrapolating the state transition matrix: This step is equivalent to calculating the most likely value of the current state at the given moment, assuming the motion model is completely accurate. Simultaneously, the prior error covariance matrix propagates in a similar manner. ,in It is the posterior error covariance matrix of the previous time step, reflecting the uncertainty of the estimate at the previous time step; the addition of Q makes the uncertainty grow naturally over time, reflecting the law that the confidence of the model prediction decays over time.

[0048] When the observation vector at the current time is obtained Then, the nodes calculate the Kalman gain matrix. This matrix determines the weights of the predicted and measured values ​​when fusing them. The Kalman gain is calculated as follows: ,in This is the measurement noise covariance matrix constructed in step S1. As can be seen from the formula, when the measurement noise R is small (i.e., the measurement value has high reliability), the Kalman gain increases, and the filter is more inclined to trust the measurement value; conversely, when the prior covariance P⁻ is small (i.e., the model prediction has high reliability), the Kalman gain decreases, and the filter is more inclined to trust the model prediction. This dynamic weighting mechanism is the core advantage of Kalman filtering.

[0049] Next, a state update is performed, where the prior estimate and measurement innovation are weighted and fused to obtain the posterior state estimate at the current time step: Among them, new information items It reflects the deviation between actual measurements and model predictions. The Kalman gain determines what proportion of the deviation is used to correct the prior estimate.

[0050] Finally, update the posterior error covariance matrix: This step reflects the degree to which uncertainty is reduced after fusing the measurements, and the posterior covariance matrix will serve as the input for the recursion at the next time step.

[0051] After the above filtering process, the node obtains the optimal state estimate for the current time. and its estimated error covariance matrix .in The diagonal elements correspond to the variances of the position, velocity, and acceleration estimates, respectively. For example, the variance of the position component is... The elements (1,1), (2,2), and (3,3) in the matrix quantify the degree of certainty about the node's location at the current moment.

[0052] Based on the current state estimate, the node further generates position prediction information for future time windows. This embodiment sets the prediction time window length. Prediction step size: 5 seconds The value is 0.5 seconds, which means that the next 5 seconds are divided into 10 equally spaced prediction times, each prediction time being denoted as 0.5 seconds. Where k = 1, 2, ..., 10. For each prediction time... The node uses the state transition matrix Perform state extrapolation, where The construction method is exactly the same as F, except that the sampling period is changed. Replace with prediction step size Specifically, the future state is predicted as follows: The future error covariance prediction is ,in For process noise over time The accumulation within, take , The unit-time process noise covariance matrix is ​​set in step S2; The setting is based on the target's maximum maneuverability.

[0053] From the predicted state vector Extracting the location component yields the predicted future location value. Meanwhile, from the prediction error covariance matrix... Extract the 3×3 submatrix corresponding to the position, and denote it as the position covariance matrix. The diagonal elements of this matrix represent the variance of the three location component predictions, while the off-diagonal elements reflect the correlation between the components, thus fully preserving the uncertainty information of the predicted locations. For ease of subsequent use, this embodiment further uses the trace of the location covariance matrix as a scalar measure of prediction uncertainty, i.e. This value is equal to the sum of the variances of the positions in the three directions, and intuitively reflects the total uncertainty of the predicted position.

[0054] It is worth noting that prediction uncertainty monotonically increases with the prediction time, which is consistent with the laws of physics: the further into the future the location, the higher the uncertainty naturally becomes. This characteristic will be fully utilized in the subsequent step S4 when calculating the link availability probability, making the long-term link predictions inherently have a larger confidence interval. Through the processing in step S2, the node not only obtains the expected value of its future location, but also obtains complete statistical information characterizing the prediction's reliability, laying a theoretical foundation for subsequent probability-based decision-making.

[0055] Step S3: Based on location prediction information, prediction uncertainty measurement, and preset signal propagation model, determine the link availability measurement between the node and at least one candidate parent node at future times;

[0056] The specific process for determining the link availability metric between a node and at least one candidate parent node at a future time includes:

[0057] For each candidate parent node, the predicted distance between the node and the candidate parent node at each prediction time is calculated based on the predicted position of the node at each prediction time and the position of the candidate parent node.

[0058] Based on the predicted distance and the preset signal propagation model, the expected received signal strength of the node and the candidate parent node at each prediction time is calculated.

[0059] Based on the predicted location variance and the signal propagation model, the predicted variance of the expected received signal strength is calculated.

[0060] Based on the expected received signal strength, prediction variance, and preset link availability threshold, the link availability probability of the node and the candidate parent node at each prediction time is determined, and the link availability probability is used as a link availability metric.

[0061] In this embodiment, step S3, based on the location prediction information and prediction uncertainty metric generated in step S2, and combined with a preset signal propagation model, calculates the link availability probability between the node and each candidate parent node at each future prediction time. The core of this step is to propagate the uncertainty of location prediction to signal strength prediction, thereby obtaining a statistically significant link availability metric, providing a quantitative basis for subsequent handover decisions.

[0062] Before executing step S3, the node needs to pre-configure the relevant parameters of the signal propagation model. This embodiment uses the logarithmic distance path loss model, a classic model in the field of wireless communication describing the attenuation of signal strength with distance. Its mathematical form is that the received signal strength has a linear relationship with the logarithm of the distance. The model includes three key parameters: reference distance... Reference signal strength at the location (Unit: dBm), Path loss exponent n (dimensionless), Standard deviation of shadow fading (Unit: dB). Wherein, reference distance... Usually 1 meter is used. It can be calculated based on the node's transmit power, antenna gain, and operating frequency. For example, under the conditions of 2.4 GHz frequency band, transmit power of 20 dBm, and antenna gain of 0 dBi, At a distance of [meter], P0 is approximately -40 dBm. The path loss exponent n depends on the propagation environment; it is taken as 2.0 in free space and [value] in urban macrocell environments. In this embodiment, n=2.8 is used based on a typical UAV flight scenario. The standard deviation of shadow fading σ_shad reflects the random fluctuation of the signal caused by obstacle occlusion. It can be taken as 4dB in open areas and up to 8dB in urban areas. In this embodiment, it is taken as... Link availability threshold Based on the receiver sensitivity setting, this embodiment uses -95dBm, meaning that the link is considered usable when the received signal strength is higher than -95dBm, and reliable communication cannot be guaranteed if it is lower than this value.

[0063] Each node maintains a neighbor information table, which is updated through periodically broadcast beacon messages. Each neighbor entry contains at least the current location of neighbor node j. If neighboring nodes also have motion state broadcasting capabilities, their motion state estimates and covariances can be stored in the table for synchronous prediction of their future locations. For clarity, this embodiment first considers the case where neighboring nodes are stationary (e.g., ground base stations), and their locations... It is a constant; for moving neighbors, the processing method is the same as for this node, simply substitute their predicted location and covariance into the corresponding calculation.

[0064] The specific calculation process for step S3 is as follows. For each candidate parent node j (including the current parent node and all neighboring nodes), traverse all prediction times k=1, 2, ..., N (N=10) and perform the following operations:

[0065] First, calculate the predicted time. The predicted distance between this node and its neighbor j. The predicted location of this node based on the output of step S2. Current location of neighbors (If the neighbor is stationary), the predicted distance is ;in This represents the Euclidean norm. If a neighbor moves and its predicted location is known, then its predicted location is used. replace .

[0066] Because the predicted location of this node is uncertain, the predicted distance also has variance. This variance can be propagated through error propagation using the node's position covariance matrix. We obtain the result. Define the unit direction vector from the predicted position of this node to the positions of its neighbors as... The variance of the predicted distance is then... The physical meaning of this formula is to project the location covariance along the line of sight to obtain the variance of the distance estimate. If the neighboring property also moves and its location covariance... Given that, the covariances of the two need to be added together, i.e. If the predicted distance Less than the preset minimum distance threshold (In this embodiment, we take 0.01 meters), which indicates that the node and its neighbors almost overlap. At this point, the unit direction vector can be any unit vector, but the calculation of the distance variance requires special handling: due to the position covariance matrix... The projections are equal in any direction, so we can take... That is, the average of the variances of the positions in the three directions. The expected RSSI is... Calculations are performed to avoid the logarithmic function diverging.

[0067] Based on the predicted distance, the expected received signal strength is calculated using the logarithmic distance path loss model: This formula shows that signal strength decreases logarithmically with increasing distance, which is consistent with the basic laws of electromagnetic wave propagation.

[0068] Due to the uncertainty in distance estimation and the presence of shadow fading during propagation, the expected received signal strength is itself a random variable. Its variance consists of two parts: one from the distance estimation variance and the other from the shadow fading variance. The derivative of distance with respect to signal strength can be derived from the path loss model. Therefore, the variance component contributed by distance uncertainty is: After adding the shadow fading variance, the total variance of the RSSI prediction is... At this point, the node has obtained the predicted time. The probability distribution of the received signal strength of neighbor j is approximately a normal distribution with a mean of 1 / j. Standard deviation is The link availability probability is defined as the received signal strength being higher than a preset threshold. The probability, i.e. ,in This is the cumulative distribution function of the standard normal distribution. The value of this function is between 0 and 1. When the expected signal strength is much higher than the threshold and the uncertainty is small, the probability is close to 1; when the expected signal strength is much lower than the threshold, the probability is close to 0; when the expected signal strength is close to the threshold or the uncertainty is large, the probability is in the middle value, reflecting the degree of ambiguity in the prediction.

[0069] Standard normal cumulative distribution function This can be achieved through numerical calculation or table lookup. This embodiment uses a polynomial approximation algorithm for calculation, which has high execution efficiency on embedded platforms.

[0070] Repeat the above calculation for all predicted times k=1…N to obtain the link availability probability sequence within the future time window corresponding to neighbor j. This sequence is the link availability metric in the claims. The same operation is performed on all neighboring nodes to obtain a set of probability curves, which serve as input for step S4.

[0071] It is worth noting that the calculation results in step S3 not only include the expected quality of future links, but also quantify the confidence level of the prediction in probabilistic form. For example, if the predicted location of a neighbor has a large covariance due to drastic maneuvers, its RSSI prediction standard deviation will also be large, resulting in a flatter probability curve, a shorter stable connection period, and a delayed dangerous approach period. This characteristic allows subsequent feature extraction to automatically reflect prediction uncertainty, thereby making more robust decisions.

[0072] Step S4: Based on the link availability metric, extract the first type of feature value between the node and the current parent node and the second type of feature value between the node and the candidate parent node. The first type of feature value is used to characterize the expected availability duration of the current link, and the second type of feature value is used to characterize the expected availability duration of the candidate link.

[0073] The first type of feature value includes the first moment when the link availability probability between the node and the current parent node first falls below the first preset probability threshold; the second type of feature value includes the first duration during which the link availability probability between the node and each candidate parent node continuously exceeds the second preset probability threshold.

[0074] In this embodiment, step S4 is responsible for extracting key feature values ​​for handover decisions from the link availability probability sequence generated in step S3. For the current parent node, a first moment is extracted, i.e., the moment when the link availability probability first falls below a first preset probability threshold; for each candidate parent node, a first duration is extracted, i.e., the duration for which the link availability probability is continuously higher than a second preset probability threshold. These two types of feature values ​​respectively characterize the remaining availability time of the current link and the stable connection capability of the candidate links.

[0075] Before executing step S4, the node pre-sets a first preset probability threshold. Second preset probability threshold In this embodiment, , These are used to define the states of link unavailable and highly reliable, respectively.

[0076] For the current parent node Its discrete probability sequence is ; Nodes search for the first satisfying < index If it does not exist, then in the first moment Pick Otherwise, precise timing can be obtained by linear interpolation between the two points. For each candidate parent node j, starting from the current time, search for nodes that satisfy the following conditions: ≥ The longest continuous interval. Let the starting index of the continuous segment be... End of index Then enter time and departure time Obtained through linear interpolation respectively:

[0077] like =1, = Otherwise utilize -1 and Two-point interpolation: ;

[0078] like , Otherwise utilize and +1 two-point interpolation: ;

[0079] Then the first duration If there are no consecutive intervals that satisfy the condition, then =0.

[0080] If the two points required for interpolation have equal probabilities, that is... If the ratio is too large, the result cannot be obtained by comparison. In this case, the midpoint of the two points is directly taken as the interpolation result. .

[0081] Through the above calculations, the node obtains the first moment of its current parent node and the first duration of all candidate parent nodes, which serve as input for step S5. This extraction process transforms discrete probabilities into duration indicators with clear physical meaning, providing a quantitative basis for subsequent handover decisions based on dynamic safety margins.

[0082] Step S5: Based on the first type of feature value, the second type of feature value, and the dynamically adjusted safety margin parameter, determine the target parent node from the candidate parent nodes and generate a switching instruction;

[0083] The specific process of determining the target parent node from candidate parent nodes based on the first type of feature value, the second type of feature value, and the dynamically adjusted safety margin parameter includes:

[0084] Obtain historical handover records, which include the deviation between the actual and predicted handover times of at least one historical handover and the historical handover time.

[0085] Based on historical handover times, determine the expected handover time for this operation;

[0086] When the first moment is less than or equal to the sum of the safety margin parameter and the expected time, the target node selection process is triggered.

[0087] During the target node selection process, nodes whose first duration is greater than or equal to the sum of the safety margin parameter and the expected time are selected from the candidate parent nodes to form a candidate set;

[0088] If the candidate set is not empty, the node with the longest first duration is selected from the candidate set as the target parent node; if the candidate set is empty, the node with the longest first time duration is selected from the candidate parent nodes as the target parent node.

[0089] It also includes the step of updating the safety margin parameter:

[0090] After the communication link is switched, record the deviation between the actual time of the switch and the first time, as well as the switching time.

[0091] Store the deviation value and handover time in the historical handover record;

[0092] Based on the deviation value sequence in the historical switching records, a closed-loop control algorithm is used to update the safety margin parameter.

[0093] The safety margin parameters are updated using a closed-loop control algorithm, specifically including:

[0094] The safety margin parameter is updated using a proportional-integral controller, and the update formula is as follows: ;in, For the first Safety margin parameters before the next switch For the updated safety margin parameters, This is the deviation value for the nth switch. This is a preset proportional coefficient. This is the preset integral coefficient.

[0095] After extracting the first moment and the first duration, it also includes:

[0096] Obtain the rate of change of acceleration of the node at the current moment;

[0097] Based on the link availability probability between the node and its current parent node at each prediction time, determine the decay steepness of the link availability probability from the preset first reference probability to the preset second reference probability.

[0098] Based on the comparison results of the rate of change of acceleration and the steepness of decay, an environmental state label is generated;

[0099] Based on the environmental status label, adjust the first preset probability threshold and / or the second preset probability threshold, and use the adjusted threshold to redetermine the first moment and the first duration, so as to determine the target parent node in step S5.

[0100] Based on the comparison between the rate of change of acceleration and the steepness of decay, environmental state labels are generated, including:

[0101] If the rate of change of acceleration is greater than the first preset threshold and the attenuation steepness is less than the second preset threshold, then a first environmental label characterizing the maneuver diffusion environment is generated.

[0102] If the rate of change of acceleration is less than the first preset threshold and the attenuation steepness is greater than the second preset threshold, then a second environmental label representing the occlusion and cutting environment is generated.

[0103] Wherein, the first preset threshold and the second preset threshold are preset constants.

[0104] In this embodiment, step S5 is responsible for determining the target parent node from the candidate parent nodes and generating a handover instruction based on the first moment and first duration extracted and adjusted by environmental perception in step S4, combined with dynamically adjusted safety margin parameters. This step also includes closed-loop updating of the safety margin parameters after the handover is completed, and adaptive adjustment of the probability threshold by the environmental state label. The entire decision-making process takes into account predictive information, historical statistics, and real-time environmental perception to achieve robust proactive topology reconstruction.

[0105] Before executing step S5, the node has already completed the first moment of step S4. and the first duration The initial extraction is then performed. Based on this, the node first generates environmental state labels and adjusts the first preset probability threshold based on the labels. Second preset probability threshold This allows for the re-determination of the first moment and the first duration. Specifically, the node extracts the rate of change of acceleration from its current state of motion. This refers to the rate of change of the magnitude of the acceleration vector, reflecting the severity of the node maneuver. Simultaneously, the attenuation steepness is calculated based on the link availability probability sequence of the current parent node. , defined as the link availability probability from a preset first reference probability Decrease to the preset second reference probability The required average rate of change. In this embodiment, , The formula for calculating the attenuation steepness is: ;in and These are the moments when the probability first drops to 0.9 and 0.1, respectively (obtained through the interpolation method in step S4). If the probability never drops to 0.1, then... Pick .

[0106] rate of change of acceleration With the steepness of decay Compare. If Greater than the first preset threshold and Less than the second preset threshold If so, it is determined that the current environment is a mobile diffusion environment, and a first environment label is generated; if Less than and Greater than If the environment is obstructed or cut off, a second environment tag is generated. In this embodiment, Pick , Pick The specific values ​​can be calibrated according to the actual scenario.

[0107] Based on the generated environment label, the node adaptively adjusts the probability threshold: if the first environment label (mobile diffusion) indicates that the signal attenuation is gradual but the fluctuation is large, the requirement for stable connection should be increased, so the second preset probability threshold is adjusted. Upgraded to Simultaneously, the first preset probability threshold is set. Downgraded to ;in To adjust the step size, this embodiment uses 0.02; if it is a second environmental label (occlusion cutting), it indicates that the signal may drop sharply, and an earlier warning should be issued, therefore... Upgraded to , The threshold remains unchanged; if no label is generated, the threshold remains the same. After adjustment, the node re-executes the interpolation calculation in step S4 to obtain the updated first time step. and the first duration of each candidate parent node .

[0108] After threshold adaptation is completed, the node enters the decision-making process of the target parent node. The node maintains a historical switch record queue with a fixed length of L=10. Each record contains the deviation between the actual break time and the predicted break time of a historical switch. and the time taken for this switchover , where m is the historical handover index. The actual break point is defined as the moment when the loss of 5 consecutive beacons of the original parent node is first detected after the handover is completed, and the handover time is defined as the time interval from sending the replacement request to receiving the confirmation from the target node.

[0109] Based on historical handover times, nodes determine the expected handover time for this current handover. Take the moving average of the times taken for the three most recent handovers: Where n is the number of times the switch has been completed. If there are fewer than three historical records, the average of the existing records is used; if there are no historical records, the initial empirical value of 0.5 seconds is used.

[0110] The node obtains the current dynamic safety margin parameter. The initial value of this parameter is set to (If there is no historical record, take 2.0 seconds), and then update it through the closed-loop control algorithm.

[0111] The decision trigger condition is: if the first moment after the update satisfy: If so, it is determined that a switch must be performed, triggering the target node selection process.

[0112] During the target node selection process, the node first filters from all candidate parent nodes to find the one that meets the requirements. The nodes form a candidate set S. This condition ensures that the target node has a sufficiently long stable period to accommodate the entire process from the current state to the completion of the switch. If the candidate set S is not empty, the node with the longest first duration is selected from S as the target parent node, i.e. If the candidate set is empty, then select the node with the largest value at the first moment from all candidate parent nodes as the target parent node. ; here If the first moment of the candidate parent node is taken as infinity (or infinity if calculated), then the selection aims to postpone the next switch as much as possible.

[0113] After selecting the target parent node, the node immediately sends a failover request message to the target node through the current link. The message contains the local node ID, the target node ID, and the expected failover time. Upon receiving the request, the target node pre-allocates resources and replies with confirmation. The node then marks the target node as a prospective parent node and generates a switchover command.

[0114] After the switchover is complete (after step S6 is executed), the node updates its safety margin parameters. The actual deviation value of this switchover is recorded. ,in The actual moment of failure and the time taken for this switchover. .Will Store in the history switch record queue and remove the oldest record.

[0115] The safety margin parameter is updated using a proportional-integral controller, and the update formula is as follows: ;in This is the safety margin parameter before the nth switch. For the updated safety margin parameters, This is the deviation value for the nth switch. This is a scaling factor (0.3 in this embodiment). The integral coefficient is 0.1 in this embodiment. This closed-loop control allows the safety margin to be automatically adjusted based on historical deviations: when the actual fracture occurs earlier than the prediction ( When the value is negative, the safety margin increases, and the switching becomes more conservative; conversely, it decreases, and the switching becomes more aggressive.

[0116] Through the above steps, the node achieves dynamic switching decisions based on predictive information, historical statistics, and environmental awareness, and has self-adjustment capabilities, enabling it to adapt to different motion modes and channel environments, ensuring robust and efficient topology reconstruction.

[0117] Step S6: In response to the switching command, perform a communication link switch from the current parent node to the target parent node.

[0118] In this embodiment, step S6 is responsible for responding to the switching command generated in step S5, executing the communication link switch from the current parent node to the target parent node, and recording relevant data after the switch is completed for subsequent parameter updates.

[0119] The node continuously monitors the system clock, and when the expected switching time determined in step S5 is reached... Upon such an event, the handover process is initiated immediately. To avoid missing the optimal handover opportunity due to clock drift or channel abrupt changes, the node simultaneously monitors the real-time received signal strength of the current link. If upon arrival Before, It has fallen below the link availability threshold three times in a row. If the current link is about to break, the node will immediately perform a switchover without waiting. .

[0120] The specific operations for handover execution include: ceasing to send any data frames to the current parent node and updating the parent node identifier in the communication stack to the identifier of the target parent node. Simultaneously, the node sends a formal handover completion acknowledgment frame to the target parent node to activate the resources pre-allocated by the target node. If no response is received from the target node three consecutive times after sending the acknowledgment frame, the handover is deemed to have failed, and the node immediately reverts to the preset emergency handover procedure: scanning all neighboring nodes, selecting the node with the highest real-time received signal strength as the new parent node, and re-establishing the connection.

[0121] After a successful switchover, the node records the switchover time. Its value is the time difference between the moment the succession request is sent in step S5 and the moment the final confirmation response is received from the target node. Simultaneously, the node continuously monitors the connection status with its original parent node to determine the actual moment of connection failure. The actual break point is defined as the moment when the node first detects the loss of five consecutive beacons from its original parent node. If the original parent node's beacons are still receivable after the handover is completed, the handover completion time is taken as the break point. .

[0122] The actual deviation value of this switch and switching time The data is stored in the historical handover record queue for use in the safety margin update in step S5. The historical handover record queue follows a first-in, first-out (FIFO) principle and has a fixed length of L=10 to ensure that statistical information of the most recent ten handovers is always retained.

[0123] Through the above steps, the node has completed a seamless switch from the current parent node to the target parent node, and provided the necessary data support for subsequent adaptive parameter adjustments.

[0124] In summary, this embodiment, by introducing a node motion trajectory prediction mechanism and a prediction uncertainty metric, advances the trigger point for topology reconstruction from after link quality deterioration to before deterioration, fundamentally eliminating the inherent reaction gap in reactive handover. Furthermore, by propagating prediction uncertainty to link availability probability and employing a closed-loop control algorithm to dynamically adjust safety margin parameters, the handover decision adaptively balances the dynamic relationship between prediction accuracy and handover time, avoiding invalid or failed handovers due to prediction bias. Further, by extracting the correlation features between the rate of change of acceleration and the steepness of link availability probability decay, environmental state labels are generated and probability thresholds are adaptively adjusted, enabling the system to maintain robust handover performance in different channel environments such as maneuvering diffusion and occlusion cutting. Thus, this embodiment achieves seamless communication link succession in predictable motion scenarios, significantly reducing the probability of connection interruption due to occlusion. Simultaneously, the parameter adaptive mechanism enhances the system's robustness to complex dynamic environments, providing a reliable technical foundation for the engineering application of mobile ad hoc networks in highly dynamic scenarios.

[0125] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0126] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0127] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An adaptive topology reconfiguration method for communication links, applied to mobile network nodes, characterized in that, include: Obtain the motion state sequence of the node, the motion state sequence including the position information of the node at multiple historical moments; Based on the motion state sequence, position prediction information of the node within a future time window and a prediction uncertainty measure associated with the position prediction information are generated; Based on the location prediction information, the prediction uncertainty measure, and the preset signal propagation model, the link availability measure between the node and at least one candidate parent node at future times is determined. Based on the link availability metric, extract the first type of feature value between the node and the current parent node and the second type of feature value between the node and the candidate parent node, wherein the first type of feature value is used to characterize the expected availability duration of the current link and the second type of feature value is used to characterize the expected availability duration of the candidate link. Based on the first type of feature value, the second type of feature value, and the dynamically adjusted safety margin parameter, the target parent node is determined from the candidate parent nodes, and a switching instruction is generated; In response to the switching instruction, a communication link switch is performed from the current parent node to the target parent node.

2. The adaptive topology reconfiguration method for communication links according to claim 1, characterized in that, The motion state sequence also includes velocity information and / or acceleration information of the node at multiple historical moments; the position prediction information includes the predicted position of the node at multiple discrete moments within the future time window; the prediction uncertainty measure includes the predicted position variance corresponding to each predicted position.

3. The adaptive topology reconfiguration method for communication links according to claim 2, characterized in that, The specific process for generating the prediction uncertainty measure includes: The motion state sequence is filtered using a Kalman filter to obtain the current state estimation vector and the estimation error covariance matrix. Based on the state estimation vector and the preset motion model, the predicted state vector at each prediction time is extrapolated. Based on the estimated error covariance matrix and the motion model, the prediction error covariance matrix corresponding to each prediction time is obtained through propagation. The predicted position is extracted from the predicted state vector, and a submatrix corresponding to the predicted position is extracted from the prediction error covariance matrix as the predicted position variance.

4. The adaptive topology reconfiguration method for communication links according to claim 2, characterized in that, The specific process for determining the link availability metric between the node and at least one candidate parent node at a future time includes: For each candidate parent node, the predicted distance between the node and the candidate parent node at each prediction time is calculated based on the predicted position of the node at each prediction time and the position of the candidate parent node. Based on the predicted distance and the preset signal propagation model, the expected received signal strength of the node and the candidate parent node at each prediction time is calculated; Based on the predicted location variance and the signal propagation model, calculate the predicted variance of the expected received signal strength; Based on the expected received signal strength, the prediction variance, and the preset link availability threshold, the link availability probability between the node and the candidate parent node at each prediction time is determined, and the link availability probability is used as the link availability metric.

5. The adaptive topology reconfiguration method for communication links according to claim 4, characterized in that, The first type of feature value includes the first moment when the link availability probability between the node and the current parent node first falls below a first preset probability threshold; the second type of feature value includes the first duration during which the link availability probability between the node and each candidate parent node continuously exceeds a second preset probability threshold.

6. The adaptive topology reconfiguration method for communication links according to claim 5, characterized in that, The specific process of determining the target parent node from the candidate parent nodes based on the first type of feature values, the second type of feature values, and the dynamically adjusted safety margin parameter includes: Obtain historical switching records, which include the deviation between the actual and predicted switching times of at least one historical switching event and the historical switching time. Based on the historical handover time, the expected handover time is determined; When the first moment is less than or equal to the sum of the safety margin parameter and the expected time, the target node selection process is triggered; During the target node selection process, nodes whose first duration is greater than or equal to the sum of the safety margin parameter and the expected time consumption are selected from the candidate parent nodes to form a candidate set; If the candidate set is not empty, then the node with the longest first duration is selected from the candidate set as the target parent node; if the candidate set is empty, then the node with the longest first time interval is selected from the candidate parent nodes as the target parent node.

7. The adaptive topology reconfiguration method for communication links according to claim 6, characterized in that, It also includes the step of updating the safety margin parameter: After the communication link switch is performed, the deviation between the actual break time of this switch and the first time is recorded, as well as the switching time of this switch. The deviation value and the switching time are stored in the historical switching record; Based on the deviation value sequence in the historical switching records, the safety margin parameter is updated using a closed-loop control algorithm.

8. The adaptive topology reconfiguration method for communication links according to claim 7, characterized in that, The step of updating the safety margin parameter using a closed-loop control algorithm specifically includes: The safety margin parameter is updated using a proportional-integral controller, and the update formula is as follows: ;in, For the first Safety margin parameters before the next switch For the updated safety margin parameters, This is the deviation value for the nth switch. This is a preset proportional coefficient. The preset integral coefficient.

9. The adaptive topology reconfiguration method for a communication link according to claim 5, characterized in that, After extracting the first moment and the first duration, the process further includes: Obtain the rate of change of acceleration of the node at the current moment; Based on the link availability probability between the node and the current parent node at each prediction time, determine the decay steepness of the link availability probability from a preset first reference probability to a preset second reference probability; Based on the comparison between the acceleration change rate and the attenuation steepness, an environmental state label is generated; Based on the environmental state label, the first preset probability threshold and / or the second preset probability threshold are adjusted, and the first moment and the first duration are re-determined using the adjusted thresholds, so as to determine the target parent node in step S5.

10. The adaptive topology reconfiguration method for a communication link according to claim 9, characterized in that, The generation of environmental state labels based on the comparison between the rate of change of acceleration and the steepness of decay includes: If the rate of change of acceleration is greater than a first preset threshold and the attenuation steepness is less than a second preset threshold, then a first environmental label characterizing the maneuver diffusion environment is generated. If the rate of change of acceleration is less than the first preset threshold and the attenuation steepness is greater than the second preset threshold, then a second environmental label characterizing the occlusion and cutting environment is generated. Wherein, the first preset threshold and the second preset threshold are preset constants.

Images