A terrain feature-based frequency hopping communication signal failure prediction method

Abstract

The application discloses a frequency hopping communication signal failure prediction method based on terrain features, and relates to the technical field of communication.The application comprises the following steps: acquiring the positions of a sending end and a receiving end and communication path elevation data, and constructing a terrain profile, a line of sight and a shielding feature; segmenting a continuous frequency hopping signal according to hop points, extracting frequency, received power, signal-to-noise ratio, correlation peak, capture time, residual frequency offset and check state, identifying a failure section and generating a failure duration, an event interval and a first recovery hop point label; extracting a mountain trend and a historical receiving state by using a terrain feature branch and a frequency hopping time sequence feature branch respectively, and training a multi-task deep learning model after feature fusion; and outputting a subsequent failure duration, a next failure event interval and a first recovery hop point online, and determining a recovery frequency in combination with a frequency hopping sequence.The application can improve the accuracy and interpretability of frequency hopping link state prediction in a mountainous barrier environment.

Description

Technical Field

[0001] This invention belongs to the field of communication technology, and in particular relates to a method for predicting the failure of frequency hopping communication signals based on terrain features. Background Technology

[0002] Frequency hopping communication refers to a communication method in which the communicating parties periodically or non-periodically switch between multiple hopping frequencies according to a preset frequency hopping sequence. Compared with fixed-frequency communication, frequency hopping communication can reduce the impact of long-term interference or fading of a single frequency on the communication link. However, it also requires the receiving end to complete frequency switching, signal acquisition, synchronization, and data reception within a short dwell time at each hopping point. In flat, open areas with good line-of-sight conditions, the signals at each hopping point can usually maintain a relatively continuous reception state. However, in mountainous areas, canyons, the back side of mountains, and non-line-of-sight areas with tall obstructions, the elevation trend of the mountains and their position relative to the propagation path of the transceiver will change the propagation conditions of the wireless signal, causing some hopping point signals to suffer diffraction loss, multipath fading, and acquisition delay. This can lead to the phenomenon that multiple consecutive hopping points cannot be effectively received, and the signal is restored after passing through the obstructed area.

[0003] In the prior art, Chinese invention patent with authorization announcement number CN114697183B discloses a channel synchronization method based on deep learning, which improves the signal synchronization capability of the communication system by setting up a deep learning channel synchronization network at the receiving end to process the carrier frequency difference and bit timing error in the received signal.

[0004] Chinese invention patent with authorization announcement number CN115276856B discloses a channel selection method based on deep learning, which obtains the energy test statistics of multiple channels and trains a residual network to judge the channel idle probability in order to select the corresponding communication channel.

[0005] Chinese invention patent CN114143145B discloses a channel estimation method based on deep learning, which combines pilot signals, frequency domain channel estimation and neural network interpolation to estimate the channel response of pilot positions and data positions.

[0006] The aforementioned technologies improve the receiving and processing performance of communication systems from the aspects of signal synchronization, channel selection, and channel response estimation, respectively.

[0007] However, the aforementioned technologies primarily focus on the received signal itself, the occupancy status of candidate channels, or the channel response at pilot positions. They do not consider the elevation trend of the mountain between the transmitter and receiver, the theoretical propagation path, and the Fresnel clearance as environmental priors for changes in the frequency-hopping signal state. In mountainous communication, even if the received power or signal-to-noise ratio decreases at a certain moment, it is difficult to determine, based solely on the signal characteristics of a single hop, whether the anomaly is caused by instantaneous noise, random fading, or by the receiver gradually entering the area behind the mountain or blocked by the ridge. Furthermore, it is difficult to determine, based on existing hop signal data, how long the continuous failure state will last or at which subsequent hop the signal will reappear.

[0008] Especially when the receiving radio moves along a mountain road or the outer side of the mountain, the terrain elevation profile, line-of-sight clearance, and mountain obstruction position between the transmitting and receiving radios continuously change as their relative positions change. If a fixed threshold is still used to judge the received power or signal-to-noise ratio, the current jump point anomaly can only be identified after the signal has failed. It is difficult to use the mountain trend and the continuous historical jump point status to predict subsequent failure sections, and it is also impossible to further obtain the duration of signal failure, the interval between adjacent failure events, and the first recovery jump point.

[0009] Therefore, there is a need for a predictive method that can fuse communication path terrain elevation data with continuous frequency hopping signal characteristics, and use deep learning models to establish a mapping relationship between mountain occlusion status, hopping point frequency changes, historical reception status, and signal failure and recovery results. Summary of the Invention

[0010] The purpose of this invention is to provide a method for predicting the failure of frequency-hopping communication signals based on terrain features. By fusing terrain elevation data with the temporal features of frequency-hopping signals and using a multi-task deep learning model for joint prediction, the method can predict the duration of signal failure and the recovery hop point in advance. This solves the problem that existing methods rely only on single-point signal threshold judgment and cannot use terrain priors to predict the failure range and recovery timing.

[0011] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0012] This invention relates to a method for predicting the failure of frequency-hopping communication signals based on terrain features, comprising the following steps:

[0013] The terrain elevation data along the communication path between the transmitting end and the receiving end is obtained, and a terrain occlusion feature characterizing the degree of obstruction of radio wave propagation by the mountain is constructed based on the terrain elevation data.

[0014] The signal reception status parameters of each hop point during frequency hopping communication are obtained to form the timing characteristics of the frequency hopping signal that reflect the changes in historical reception quality.

[0015] Based on the success or failure of acquisition and the data verification results in the timing characteristics of the frequency hopping signal, identify signal failure events, and determine the failure duration of each failure event, the interval between adjacent failure events, and the offset of the first recovery jump point after failure relative to the failure start jump point.

[0016] The terrain occlusion features and the frequency hopping signal timing features are correlated and fused according to time and spatial location to construct a fused feature sequence;

[0017] The fused feature sequence is trained using a multi-task deep learning model, enabling the model to learn the mapping relationship between terrain occlusion, signal temporal changes, and the failure duration, failure event interval, and recovery jump point offset.

[0018] In actual communication, the terrain occlusion features of the current communication path and the timing features of the frequency hopping signal of the most recent historical hop point are input into the trained multi-task deep learning model to obtain the predicted failure duration, the predicted failure event interval, and the predicted recovery hop point offset.

[0019] Based on the position of the current hop point in the frequency hopping sequence and the predicted recovery hop point offset, the hop point for the first recovery after a future failure is determined, and the frequency corresponding to the recovery hop point is obtained from the frequency hopping sequence.

[0020] Furthermore, obtaining the terrain elevation data includes: setting multiple elevation sampling points at preset intervals along the communication path, recording the distance between each sampling point and the transmitting end and its corresponding ground elevation, and arranging them in order of distance to form a terrain elevation sequence.

[0021] Furthermore, constructing the terrain occlusion features further includes:

[0022] Based on the elevation of the top of the transmitting antenna, the elevation of the top of the receiving antenna, and the distance between each sampling point and the transmitting end, the height of the theoretical line-of-sight propagation reference line at each sampling point is determined.

[0023] Calculate the radius of the first Fresnel zone at each sampling point based on the current jump point's operating frequency;

[0024] Using a preset Fresnel clearance coefficient, the theoretical line-of-sight propagation reference line height is subtracted from the sum of the ground elevation of each sampling point and the radius of the first Fresnel zone to obtain the clearance loss of each sampling point;

[0025] When the clearance loss is greater than zero, it is determined that the sampling point has intruded into the preset propagation clearance area.

[0026] Furthermore, the terrain occlusion features include at least: maximum clearance loss, width of continuous paths with clearance loss greater than zero, proportion of occluded paths to all communication paths, maximum upslope, maximum downslope, number of peaks, distance between the maximum occlusion location and the transmitting end, and distance between the maximum occlusion location and the receiving end.

[0027] Furthermore, the timing characteristics of the frequency hopping signal are composed of signal reception status parameters of each hopping point. The signal reception status parameters include at least: hopping point frequency, received power, signal-to-noise ratio, preamble sequence main correlation peak value, ratio of main correlation peak to sidelobe peak, acquisition time, remaining frequency deviation, data verification result, and reception status identifier of the hopping point.

[0028] Furthermore, the identification signal failure event includes:

[0029] When the signal of a certain jump point is successfully captured within the preset capture window and the data verification is passed, the jump point is marked as valid; otherwise, it is marked as invalid.

[0030] The jump point where the reception status changes from valid to invalid is recorded as the invalidation start jump point, and the jump point where the reception status changes from invalid to valid is recorded as the first recovery jump point;

[0031] The failure duration is the sum of the dwell times of all jump points from the failure initiation jump point to the jump point before the first recovery jump point;

[0032] The recovery jump point offset is the difference in sequence number between the first recovery jump point and the failure start jump point;

[0033] The interval between two consecutive failure events is the sum of the dwell times of all jump points between the first recovery jump point of the previous event and the jump point before the start jump point of the next failure event.

[0034] Furthermore, the association and fusion of the terrain occlusion features and the frequency hopping signal timing features includes:

[0035] When the locations of the transmitting and receiving ends are fixed, the terrain occlusion features of the same communication path are copied and spliced ​​with the signal timing features of multiple consecutive hops.

[0036] When the location of the transmitting or receiving end changes, the corresponding terrain occlusion features are calculated based on the actual transmitting and receiving positions at the time of each hop. These terrain occlusion features are then spliced ​​together with the timing features of the frequency hopping signal at the same time after splicing and aligning them in time and space.

[0037] Furthermore, the multi-task deep learning model includes:

[0038] The terrain feature extraction branch is used to extract mountain trend features from the terrain elevation sequence and the net clearance loss sequence;

[0039] The temporal feature extraction branch uses gated recurrent units, long short-term memory networks, or temporal convolutional networks to extract the reception state change features of continuous historical jump points.

[0040] The feature fusion layer is used to splice or weightedly fuse the mountain trend features with the received state change features;

[0041] The multi-task output layer outputs the predicted failure duration, the predicted failure event interval, and the predicted recovery jump point offset, respectively.

[0042] Furthermore, when training the multi-task deep learning model, a multi-task loss function is used, which is a weighted sum of the regression loss between the predicted failure duration and the actual value, the regression loss between the predicted failure event interval and the actual value, the regression loss between the predicted recovery jump point offset and the actual value, and the model parameter regularization loss.

[0043] Furthermore, in the actual communication process, a sliding time window is used to maintain the features of the most recent fixed number of historical hop points. After each current hop point reception process is completed, the actual reception status of the hop point and the corresponding terrain features are added to the window and the oldest features in the window are discarded. Then, the prediction is re-executed. When the recovery hop point indicated by the predicted recovery hop point offset exceeds the current frequency hopping sequence period, the recovery hop point number is cyclically mapped according to the cycle period of the frequency hopping sequence to determine the corresponding recovery frequency.

[0044] The present invention has the following beneficial effects:

[0045] 1. This invention incorporates terrain elevation trends, line-of-sight clearance, and Fresnel zone intrusion into the frequency-hopping communication signal failure prediction process, moving beyond relying solely on received power or the signal-to-noise ratio of a single hop to determine link status. By fusing relatively stable prior knowledge of mountain obstruction with dynamically changing hop frequencies, acquisition times, correlation peaks, and verification states, the model can distinguish between continuous signal failures caused by persistent mountain obstruction and short-term signal anomalies caused by random noise and transient fading. This improves the accuracy and interpretability of failure status assessment in mountainous, canyon, and non-line-of-sight environments.

[0046] 2. This invention employs a multi-task deep learning model to simultaneously predict the duration of signal failure, the interval between adjacent failure events, and the first recovery hop point, and combines this with a preset frequency hopping sequence to obtain the recovery frequency. Compared to simply outputting a binary classification result indicating whether the signal is valid or invalid, this invention can further provide the expected duration of signal silence, when it may fail again, and at which hop it will recover. This provides directly usable prediction results for setting receive waiting time, data buffering, determining retransmission timing, and providing link status indications, reducing the time and processing resource consumption caused by continuous blind searching and invalid reception during frequency hopping communication in mountainous areas.

[0047] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0048] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments 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.

[0049] Figure 1 This is a flowchart illustrating a frequency hopping communication signal failure prediction method based on terrain features according to the present invention.

[0050] Figure 2 This is a schematic diagram of the multi-task deep learning model structure for fusing terrain elevation features and frequency hopping signal temporal features according to the present invention.

[0051] Figure 3 This is a schematic diagram illustrating the relative positions of the transmitting radio, receiving radio, and the moving path of the receiving end under real mountainous terrain conditions according to the present invention.

[0052] Figure 4 This is a schematic diagram showing the relationship between the terrain elevation profile, theoretical propagation reference line, and Fresnel clearance boundary of the communication path between the transmitting and receiving ends of this invention.

[0053] Figure 5 This is a schematic diagram of the simulation results of the frequency hopping signal receiving power, signal-to-noise ratio, and effective and ineffective states during the movement of the receiver of this invention along the outer side of the mountain.

[0054] Figure 6 This is a schematic diagram illustrating the correspondence between the moving distance of the receiving end and the failure and recovery intervals of the frequency hopping communication signal in this invention.

[0055] Figure 7 This is a schematic diagram illustrating the relationship between the mountain elevation trend, theoretical propagation reference line, Fresnel clearance boundary, maximum clearance loss location, and the location of the transmitting and receiving stations. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Please see Figure 1-7 As shown, this invention is a method for predicting the failure of frequency-hopping communication signals based on terrain features, comprising the following steps:

[0058] The terrain elevation data along the communication path between the transmitting end and the receiving end is obtained, and a terrain occlusion feature characterizing the degree of obstruction of radio wave propagation by the mountain is constructed based on the terrain elevation data.

[0059] The signal reception status parameters of each hop point during frequency hopping communication are obtained to form the timing characteristics of the frequency hopping signal that reflect the changes in historical reception quality.

[0060] Based on the success or failure of acquisition and the data verification results in the timing characteristics of the frequency hopping signal, identify signal failure events, and determine the failure duration of each failure event, the interval between adjacent failure events, and the offset of the first recovery jump point after failure relative to the failure start jump point.

[0061] The terrain occlusion features and the frequency hopping signal timing features are correlated and fused according to time and spatial location to construct a fused feature sequence;

[0062] The fused feature sequence is trained using a multi-task deep learning model, enabling the model to learn the mapping relationship between terrain occlusion, signal temporal changes, and the failure duration, failure event interval, and recovery jump point offset.

[0063] In actual communication, the terrain occlusion features of the current communication path and the timing features of the frequency hopping signal of the most recent historical hop point are input into the trained multi-task deep learning model to obtain the predicted failure duration, the predicted failure event interval, and the predicted recovery hop point offset.

[0064] Based on the position of the current hop point in the frequency hopping sequence and the predicted recovery hop point offset, the hop point for the first recovery after a future failure is determined, and the frequency corresponding to the recovery hop point is obtained from the frequency hopping sequence.

[0065] Furthermore, obtaining the terrain elevation data includes: setting multiple elevation sampling points at preset intervals along the communication path, recording the distance between each sampling point and the transmitting end and its corresponding ground elevation, and arranging them in order of distance to form a terrain elevation sequence.

[0066] Furthermore, constructing the terrain occlusion features further includes:

[0067] Based on the elevation of the top of the transmitting antenna, the elevation of the top of the receiving antenna, and the distance between each sampling point and the transmitting end, the height of the theoretical line-of-sight propagation reference line at each sampling point is determined.

[0068] Calculate the radius of the first Fresnel zone at each sampling point based on the current jump point's operating frequency;

[0069] Using a preset Fresnel clearance coefficient, the theoretical line-of-sight propagation reference line height is subtracted from the sum of the ground elevation of each sampling point and the radius of the first Fresnel zone to obtain the clearance loss of each sampling point;

[0070] When the clearance loss is greater than zero, it is determined that the sampling point has intruded into the preset propagation clearance area.

[0071] Furthermore, the terrain occlusion features include at least: maximum clearance loss, width of continuous paths with clearance loss greater than zero, proportion of occluded paths to all communication paths, maximum upslope, maximum downslope, number of peaks, distance between the maximum occlusion location and the transmitting end, and distance between the maximum occlusion location and the receiving end.

[0072] Furthermore, the timing characteristics of the frequency hopping signal are composed of signal reception status parameters of each hopping point. The signal reception status parameters include at least: hopping point frequency, received power, signal-to-noise ratio, preamble sequence main correlation peak value, ratio of main correlation peak to sidelobe peak, acquisition time, remaining frequency deviation, data verification result, and reception status identifier of the hopping point.

[0073] Furthermore, the identification signal failure event includes:

[0074] When the signal of a certain jump point is successfully captured within the preset capture window and the data verification is passed, the jump point is marked as valid; otherwise, it is marked as invalid.

[0075] The jump point where the reception status changes from valid to invalid is recorded as the invalidation start jump point, and the jump point where the reception status changes from invalid to valid is recorded as the first recovery jump point;

[0076] The failure duration is the sum of the dwell times of all jump points from the failure initiation jump point to the jump point before the first recovery jump point;

[0077] The recovery jump point offset is the difference in sequence number between the first recovery jump point and the failure start jump point;

[0078] The interval between two consecutive failure events is the sum of the dwell times of all jump points between the first recovery jump point of the previous event and the jump point before the start jump point of the next failure event.

[0079] Furthermore, the association and fusion of the terrain occlusion features and the frequency hopping signal timing features includes:

[0080] When the locations of the transmitting and receiving ends are fixed, the terrain occlusion features of the same communication path are copied and spliced ​​with the signal timing features of multiple consecutive hops.

[0081] When the location of the transmitting or receiving end changes, the corresponding terrain occlusion features are calculated based on the actual transmitting and receiving positions at the time of each hop. These terrain occlusion features are then spliced ​​together with the timing features of the frequency hopping signal at the same time after splicing and aligning them in time and space.

[0082] Furthermore, the multi-task deep learning model includes:

[0083] The terrain feature extraction branch is used to extract mountain trend features from the terrain elevation sequence and the net clearance loss sequence;

[0084] The temporal feature extraction branch uses gated recurrent units, long short-term memory networks, or temporal convolutional networks to extract the reception state change features of continuous historical jump points.

[0085] The feature fusion layer is used to splice or weightedly fuse the mountain trend features with the received state change features;

[0086] The multi-task output layer outputs the predicted failure duration, the predicted failure event interval, and the predicted recovery jump point offset, respectively.

[0087] Furthermore, when training the multi-task deep learning model, a multi-task loss function is used, which is a weighted sum of the regression loss between the predicted failure duration and the actual value, the regression loss between the predicted failure event interval and the actual value, the regression loss between the predicted recovery jump point offset and the actual value, and the model parameter regularization loss.

[0088] Furthermore, in the actual communication process, a sliding time window is used to maintain the features of the most recent fixed number of historical hop points. After each current hop point reception process is completed, the actual reception status of the hop point and the corresponding terrain features are added to the window and the oldest features in the window are discarded. Then, the prediction is re-executed. When the recovery hop point indicated by the predicted recovery hop point offset exceeds the current frequency hopping sequence period, the recovery hop point number is cyclically mapped according to the cycle period of the frequency hopping sequence to determine the corresponding recovery frequency.

[0089] The specific application of this embodiment is as follows:

[0090] This embodiment uses one transmitting radio and one mobile receiving radio to establish a frequency-hopping communication link. The transmitting radio sends frequency-hopping burst signals according to a preset frequency-hopping sequence, and the receiving radio moves along one side of the mountain. The system acquires the positions, antenna heights, operating frequency ranges, and digital elevation data along the communication path of the two radios, and predicts the duration of the frequency-hopping signal failure and the recovery point after the receiving radio enters the mountain-blocked area based on the elevation trend.

[0091] In this embodiment, the following can be configured:

[0092] The horizontal range of the communication path is 0–20 km;

[0093] The number of terrain elevation sampling points is 401;

[0094] The transmitter is located 1km away;

[0095] The receiving end was moved from 3km to 18km;

[0096] The transmission power is 38dBm;

[0097] The transmitting antenna is 8m high;

[0098] The receiver antenna is 5m high;

[0099] The frequency hopping range is 30–88 MHz;

[0100] The dwell time at the jump point is 0.1s;

[0101] The continuous working time is 12 seconds;

[0102] The noise floor is −112dBm.

[0103] The above values ​​are for illustrative purposes only and do not constitute a limitation on the scope of protection.

[0104] Constructing a topographic elevation profile:

[0105] Multiple elevation sampling points are set along the path between the transmitter and receiver to obtain:

[0106]

[0107] When on-site elevation data cannot be obtained directly, test elevations can be constructed using Gaussian peaks and terrain undulations during the experimental phase.

[0108]

[0109] in:

[0110] Topographical reference elevation;

[0111] The height of the main peak relative to the reference elevation;

[0112] The central location of the main peak;

[0113] Used to indicate the width of a mountain;

[0114] This is the terrain disturbance term used to simulate secondary peaks and slope undulations.

[0115] In the MATLAB interface, you can change the height of the main peak, the center of the mountain, the width of the mountain and the intensity of the undulation by using sliders. You can also import CSV files containing two columns of data: "horizontal position - elevation".

[0116] The purpose of this design is to enable experiments to use actual digital elevation data as well as to quickly construct test scenarios with different occlusion intensities by parametrically configuring the mountain.

[0117] Calculate the theoretical line of sight and propagation clearance loss.

[0118] The elevation of the top of the transmitting antenna is:

[0119]

[0120] The elevation of the top of the receiving antenna is:

[0121]

[0122] in, and These are the locations of the sending and receiving ends, respectively. and These represent the heights of the two antennas relative to the ground.

[0123] No. The theoretical line-of-sight height at each sampling point is:

[0124]

[0125] Simply determining whether a mountain exceeds the line of sight is insufficient to fully represent the propagation clearance required for wireless signals; therefore, the radius of the first Fresnel zone must be further calculated.

[0126]

[0127] in:

[0128]

[0129] The speed of electromagnetic wave propagation. This represents the current jump point frequency.

[0130] Using preset clearance coefficient Calculate the net clearance deficit:

[0131]

[0132] In this embodiment, it is acceptable to... ,when When the value is >0, it indicates that the mountain has encroached on the preset propagation clearance area corresponding to the current frequency; The larger the value, the stronger the obstruction by the mountain.

[0133] Extracted from the net clearance loss sequence:

[0134]

[0135] in:

[0136] · This represents the maximum net clearance deficit.

[0137] · The width of the continuous occlusion path;

[0138] · The proportion of the occlusion path;

[0139] · This represents the maximum slope.

[0140] · The number of mountain peaks;

[0141] · The distance between the location of maximum obstruction and the transmitting end;

[0142] · This is the distance between the location of maximum obstruction and the receiving end.

[0143] This design can convert the original high-order program sequence into a model input that reflects "how high the mountain is, how wide the mountain is, which side the mountain peak is located on, and how much the propagation path is blocked".

[0144] Simulate or acquire frequency hopping communication signals:

[0145] The transmitting end follows a preset frequency hopping sequence:

[0146]

[0147] Frequency-hopping burst signals are sent sequentially. The dwell time for each hop is... .

[0148] In practical equipment, the receiver directly records the received power, acquisition time, correlation peak, frequency deviation, and verification results. In MATLAB experiments, free space path loss and terrain-additional loss can be used together to simulate the received power.

[0149] The free space path loss is:

[0150]

[0151] in:

[0152] For the first The distance between the two radio stations when each jump point occurs is expressed in km.

[0153] For the first The jump point frequency is in MHz.

[0154] Set additional terrain loss based on the maximum clearance deficit:

[0155]

[0156] in: For indicator functions;

[0157] Additional losses are incurred when shading occurs;

[0158] This is the incremental loss coefficient corresponding to the unit net air gap.

[0159] The received power is:

[0160]

[0161] in:

[0162] Transmission power;

[0163] This is the perturbation term used to represent the fluctuations in the jump point frequency response and slow fading.

[0164] The signal-to-noise ratio is:

[0165]

[0166] in, To reduce receiver noise.

[0167] Further simulation of the leading correlation peak:

[0168]

[0169] in:

[0170] The reference signal-to-noise ratio is used when the correlation peak is significantly increased;

[0171] The slope coefficient;

[0172] The scaling factor for the attenuation of related peaks due to mountain shading.

[0173] The capture time can be expressed as:

[0174]

[0175] When the mountain blockage increases or the signal-to-noise ratio decreases, the correlation peak decreases, and the receiver needs a longer time to complete the acquisition. When the acquisition time exceeds the dwell time, even if the signal appears briefly, the receiver cannot complete the effective reception at that hop point.

[0176] Determine signal failure and recovery status

[0177] The state of the t-th jump point is defined as follows:

[0178]

[0179] in:

[0180] Preset capture time window;

[0181] The lowest correlation peak threshold;

[0182] This indicates that the data validation has passed;

[0183] 0 indicates that the jump point is in a state of signal failure or silence.

[0184] When the following conditions are met:

[0185]

[0186]

[0187] At that time, the first The jump point was determined as the failure start jump point.

[0188] In the After each jump point, search for the first satisfying condition. The jump point of 1 is determined as the first recovery jump point.

[0189] The failure duration is:

[0190]

[0191] The recovery jump point offset is:

[0192]

[0193] If the previous failure event occurred in the [number]th [year], The first jump point is recovered, and the next failure event starts from the first jump point. Starting from a certain jump point, the failure event interval is:

[0194]

[0195] The purpose of setting the failure duration and failure event interval is to distinguish between two different indicators: "how long the signal remains silent" and "how long it can work stably after the signal is restored".

[0196] Constructing terrain-frequency hopping fusion feature sequences:

[0197] The dynamic signal characteristics of the t-th jump point are:

[0198]

[0199] Recently The time-series features are composed of several jump points:

[0200]

[0201] For the mobile receiver, the terrain occlusion features are calculated based on the receiver location corresponding to each hop point. ,form:

[0202]

[0203] Then concatenate the two types of features:

[0204]

[0205] The resulting model input includes not only "whether the most recent jump points were successfully received", but also "whether the propagation paths corresponding to the most recent jump points gradually enter the mountain-blocked area".

[0206] Training deep learning models:

[0207] This embodiment can use a GRU network to extract the temporal variation pattern of the fused feature sequence. The input sequence passes through the following steps in sequence:

[0208] Sequence input layer;

[0209] GRU layer;

[0210] Fully connected layer;

[0211] ReLU activation layer;

[0212] Dropout layer;

[0213] Three-output regression layer.

[0214] The model output is:

[0215]

[0216] in:

[0217] To predict the duration of failure;

[0218] To predict the interval of the next failure event;

[0219] To predict the recovery jump point offset.

[0220] The training loss is:

[0221]

[0222] in, Indicates model parameters.

[0223] In this MATLAB experiment, using the specific parameters of Cuihua Mountain in Xi'an as an example, training samples can be generated by randomly changing the following parameters:

[0224] Mountain height;

[0225] Location of the mountain peak;

[0226] Mountain width;

[0227] Intensity of terrain undulation;

[0228] Location of the sending and receiving ends;

[0229] The direction of the radio station's movement;

[0230] Transmission power;

[0231] Low noise level;

[0232] Jump point frequency;

[0233] Jump point dwell time.

[0234] This design enables the model to learn not just a single mountain or a set of frequencies, but to learn the common patterns of change between mountain terrain, propagation paths, and consecutive jump point states.

[0235] Online prediction and recovery frequency determination:

[0236] During actual communication, the system continuously maintains a sliding window containing the K most recent hops.

[0237] After completing the reception processing for each hop:

[0238] Extract the signal features of the current jump point;

[0239] Recalculate terrain obstruction features based on the current radio station location;

[0240] Add the current feature to the sliding window;

[0241] The earliest deletion jump point characteristic;

[0242] Input the trained model;

[0243] Obtain the predicted failure duration, the predicted failure event interval, and the predicted recovery jump point offset.

[0244] Assume the current jump point number is The predicted recovery jump point offset is The initial recovery jump point is:

[0245]

[0246] The corresponding recovery frequency is:

[0247]

[0248] when When the frequency hopping cycle length is exceeded, a circular index is used:

[0249]

[0250] in, This represents the number of hops within one frequency hopping cycle.

[0251] All embodiments of the present invention are within the scope of protection of this patent.

[0252] 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.

[0253] 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 the specific implementations described. 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. A method for predicting the failure of frequency-hopping communication signals based on terrain features, characterized in that, The prediction method includes the following steps: The terrain elevation data along the communication path between the transmitting end and the receiving end is obtained, and a terrain occlusion feature characterizing the degree of obstruction of radio wave propagation by the mountain is constructed based on the terrain elevation data. The signal reception status parameters of each hop point during frequency hopping communication are obtained to form the timing characteristics of the frequency hopping signal that reflect the changes in historical reception quality. Based on the success or failure of acquisition and the data verification results in the timing characteristics of the frequency hopping signal, identify signal failure events, and determine the failure duration of each failure event, the interval between adjacent failure events, and the offset of the first recovery jump point after failure relative to the failure start jump point. The terrain occlusion features and the frequency hopping signal timing features are correlated and fused according to time and spatial location to construct a fused feature sequence; The fused feature sequence is trained using a multi-task deep learning model, enabling the model to learn the mapping relationship between terrain occlusion, signal temporal changes, and the failure duration, failure event interval, and recovery jump point offset. In actual communication, the terrain occlusion features of the current communication path and the timing features of the frequency hopping signal of the most recent historical hop point are input into the trained multi-task deep learning model to obtain the predicted failure duration, the predicted failure event interval, and the predicted recovery hop point offset. Based on the position of the current hop point in the frequency hopping sequence and the predicted recovery hop point offset, the hop point for the first recovery after a future failure is determined, and the frequency corresponding to the recovery hop point is obtained from the frequency hopping sequence.

2. The method for predicting frequency hopping communication signal failure based on terrain features according to claim 1, characterized in that, Acquiring the terrain elevation data includes: setting multiple elevation sampling points at preset intervals along the communication path, recording the distance between each sampling point and the transmitting end and its corresponding ground elevation, and arranging them in order of distance to form a terrain elevation sequence.

3. The method for predicting frequency hopping communication signal failure based on terrain features according to claim 2, characterized in that, Constructing the terrain occlusion features further includes: Based on the elevation of the top of the transmitting antenna, the elevation of the top of the receiving antenna, and the distance between each sampling point and the transmitting end, the height of the theoretical line-of-sight propagation reference line at each sampling point is determined. Calculate the radius of the first Fresnel zone at each sampling point based on the current jump point's operating frequency; Using a preset Fresnel clearance coefficient, the theoretical line-of-sight propagation reference line height is subtracted from the sum of the ground elevation of each sampling point and the radius of the first Fresnel zone to obtain the clearance loss of each sampling point; When the clearance loss is greater than zero, it is determined that the sampling point has intruded into the preset propagation clearance area.

4. The method for predicting frequency hopping communication signal failure based on terrain features according to claim 3, characterized in that, The terrain occlusion features include at least: maximum clearance loss, width of continuous paths with clearance loss greater than zero, proportion of occluded paths to all communication paths, maximum upslope, maximum downslope, number of peaks, distance between the maximum occlusion location and the transmitting end, and distance between the maximum occlusion location and the receiving end.

5. The method for predicting frequency hopping communication signal failure based on terrain features according to claim 4, characterized in that, The timing characteristics of the frequency hopping signal are composed of the signal reception status parameters of each hopping point. The signal reception status parameters include at least: hopping point frequency, received power, signal-to-noise ratio, preamble sequence main correlation peak value, ratio of main correlation peak to sidelobe peak, acquisition time, remaining frequency deviation, data verification result, and reception status identifier of the hopping point.

6. The method for predicting frequency hopping communication signal failure based on terrain features according to claim 5, characterized in that, The identification signal failure event includes: When the signal of a certain jump point is successfully captured within the preset capture window and the data verification is passed, the jump point is marked as valid; otherwise, it is marked as invalid. The jump point where the reception status changes from valid to invalid is recorded as the invalidation start jump point, and the jump point where the reception status changes from invalid to valid is recorded as the first recovery jump point; The failure duration is the sum of the dwell times of all jump points from the failure initiation jump point to the jump point before the first recovery jump point; The recovery jump point offset is the difference in sequence number between the first recovery jump point and the failure start jump point; The interval between two consecutive failure events is the sum of the dwell times of all jump points between the first recovery jump point of the previous event and the jump point before the start jump point of the next failure event.

7. The method for predicting frequency hopping communication signal failure based on terrain features according to claim 6, characterized in that, The association and fusion of the terrain occlusion features and the frequency hopping signal timing features includes: When the locations of the transmitting and receiving ends are fixed, the terrain occlusion features of the same communication path are copied and spliced ​​with the signal timing features of multiple consecutive hops. When the location of the transmitting or receiving end changes, the corresponding terrain occlusion features are calculated based on the actual transmitting and receiving positions at the time of each hop. These terrain occlusion features are then spliced ​​together with the timing features of the frequency hopping signal at the same time after splicing and aligning them in time and space.

8. The method for predicting frequency hopping communication signal failure based on terrain features according to claim 7, characterized in that, The multi-task deep learning model includes: The terrain feature extraction branch is used to extract mountain trend features from the terrain elevation sequence and the net clearance loss sequence; The temporal feature extraction branch uses gated recurrent units, long short-term memory networks, or temporal convolutional networks to extract the reception state change features of continuous historical jump points. The feature fusion layer is used to splice or weightedly fuse the mountain trend features with the received state change features; The multi-task output layer outputs the predicted failure duration, the predicted failure event interval, and the predicted recovery jump point offset, respectively.

9. The method for predicting frequency hopping communication signal failure based on terrain features according to claim 8, characterized in that, When training the multi-task deep learning model, a multi-task loss function is used, which is a weighted sum of the regression loss between the predicted failure duration and the actual value, the regression loss between the predicted failure event interval and the actual value, the regression loss between the predicted recovery jump point offset and the actual value, and the model parameter regularization loss.

10. The method for predicting frequency hopping communication signal failure based on terrain features according to claim 9, characterized in that, In the actual communication process, a sliding time window is used to maintain the features of the most recent fixed number of historical hop points. After each current hop point is received, the actual reception status of the hop point and the corresponding terrain features are added to the window and the oldest features in the window are discarded. Then the prediction is re-executed. When the recovery hop point indicated by the predicted recovery hop point offset exceeds the current frequency hopping sequence period, the recovery hop point number is cyclically mapped according to the cycle period of the frequency hopping sequence to determine the corresponding recovery frequency.

Images