A forest fire remote communication unmanned aerial vehicle airborne spread spectrum method and device

Abstract

This invention discloses an airborne spread spectrum method and device for remote communication in forest fires using unmanned aerial vehicles (UAVs), relating to the field of data measurement technology. The method includes: acquiring a composite disturbance quantity containing signal strength, signal-to-noise ratio, and asymmetric jitter; calculating the historical distribution and channel volatility using a dynamic observation window; performing nonlinear weighting based on the dynamic variance of the historical distribution and the hardware response limit to extract the effective dimension, then calculating the coupling attenuation rate and obtaining a convergence threshold value based on the volatility; calculating a dynamic bias quantity using asymmetric features, and then obtaining the gain gradient value from the bias quantity and the disturbance quantity to determine the spread spectrum parameters; finally, dividing the parameters into wake-up micropulses and configuration data packets, triggering a parallel blind detection mechanism via a synchronization control channel to complete the switching of the spread spectrum factor and coding rate. This invention has the advantages of strong anti-interference, high adaptability, and anti-lockdown.

Description

Technical Field

[0001] This invention relates to the field of data measurement technology, specifically to an unmanned aerial vehicle (UAV) airborne spread spectrum method and device for remote communication of forest fires. Background Technology

[0002] In forest fire emergency rescue and fire detection, drones need to penetrate deep into the fire zone, and establishing a stable and reliable airborne communication link is a key prerequisite for ensuring remote command and dispatch. However, existing drone long-range communication spread spectrum solutions have certain limitations when dealing with the complex physical environment of forest fires.

[0003] First, the high-temperature heat column and rising hot air currents in a fire cause thermal turbulence, leading to changes in the refractive index of the electromagnetic wave propagation medium. This results in thermal flicker and asymmetric Doppler frequency shift jitter in communication signals. Existing channel sensing models based on the assumption of stationary symmetry struggle to accurately quantify this signal fading caused by the combined coupling of thermal radiation and vegetation obstruction, easily leading to errors in compensation threshold calculations. Second, existing dynamic spread spectrum adjustment algorithms have shortcomings in data processing. Directly substituting physical parameters with negative polarity can easily cause computational divergence under low signal-to-noise ratio conditions. Furthermore, when faced with transient electromagnetic interference from equipment operating around the fire, the system struggles to effectively distinguish between natural environmental fading and hardware transient interference, potentially triggering inappropriate spreading factor increases and reducing the system's spectral efficiency and communication throughput. Additionally, the synchronization of spread spectrum parameters at both the transmitting and receiving ends of existing systems relies heavily on the protocol-level handshake of the channel itself. When drones encounter dense canopy obstruction or strong electromagnetic interference from the fire, the degraded service channel makes it difficult to reliably transmit control commands requiring parameter switching to the receiving end. If a data packet containing the spread spectrum configuration is corrupted or lost, the receiving end will retain the original parameters because it has not received instructions, resulting in a mismatch between the local despreading sequences at both ends, triggering physical layer unlocking and communication interruption. Summary of the Invention

[0004] To address the technical problems existing in the background art mentioned above, the present invention provides an airborne spread spectrum method and device for remote communication in forest fires using unmanned aerial vehicles (UAVs). This UAV airborne spread spectrum scheme reduces the dependence on two-end protocol handshakes to achieve parameter synchronization, thereby improving communication continuity in complex fire hazard environments.

[0005] A UAV-borne spread spectrum method for remote communication in forest fires includes: acquiring a composite perturbation quantity; the composite perturbation quantity includes signal strength, signal-to-noise ratio, and asymmetric jitter; acquiring historical distribution and channel volatility based on the composite perturbation quantity and a dynamic observation window; acquiring dynamic variance based on the historical distribution, and performing nonlinear weighting based on the dynamic variance and response limit to acquire the effective dimension; acquiring coupling attenuation rate based on the effective dimension, and acquiring convergence threshold based on coupling attenuation rate and channel volatility; acquiring dynamic bias based on convergence threshold and asymmetric characteristics, and acquiring gain gradient based on dynamic bias and composite perturbation quantity; acquiring spread spectrum parameter values ​​based on gain gradient, and dividing the spread spectrum parameter values ​​into wake-up micropulses and configuration data packets; and triggering parallel blind detection via a synchronization control channel based on the wake-up micropulses and configuration data packets to switch the spread spectrum factor value and coding rate parameter.

[0006] Optionally, obtaining the historical distribution and channel volatility based on the composite disturbance and the dynamic observation window includes: defining the dynamic observation window based on the start point of the current communication frame; extracting the physical extreme values ​​based on the dynamic observation window and constructing a non-negative normalized column using linear translation; and obtaining the historical distribution and channel volatility based on the non-negative normalized column.

[0007] Optionally, obtaining the historical distribution column and channel volatility based on the non-negative normalized column includes: obtaining the adjacent sampling difference based on the non-negative normalized column; obtaining the dynamic smoothing basis based on the non-negative normalized column and the preset anti-zero constant; and obtaining the channel volatility based on the adjacent sampling difference, the dynamic smoothing basis, and the preset compensation term.

[0008] Optionally, nonlinear weighting is performed based on the dynamic variance and response limit values ​​to obtain the effective number of dimensions, including: obtaining the feature change rate based on the historical distribution; performing state comparison based on the feature change rate and response limit values; applying a negative penalty value based on the comparison results to eliminate human interference dimensions and obtain the effective number of dimensions.

[0009] Optionally, the coupling attenuation rate is obtained based on the number of effective dimensions, and the convergence threshold is obtained based on the coupling attenuation rate and the channel volatility, including: performing logarithmic operations and linear combinations based on the number of effective dimensions to obtain the coupling attenuation rate; and performing exponential calculation based on the coupling attenuation rate and the channel volatility to obtain the convergence threshold.

[0010] Optionally, the dynamic bias is obtained based on the convergence threshold and asymmetry features, including: extracting positive and negative Doppler based on the heat flow lift force to obtain asymmetry features; and obtaining the dynamic bias based on the asymmetry features, heat flow sensitivity, and static noise floor value.

[0011] Optionally, the gain gradient value is obtained based on the dynamic bias and the composite perturbation, including: obtaining the absolute deviation value based on the non-negative normalized column of the composite perturbation at the current time and its historical mean; and obtaining the gain gradient value based on the absolute deviation value, the convergence threshold value, and the dynamic bias.

[0012] Optionally, the spread spectrum parameter value is obtained based on the gain gradient value, and the spread spectrum parameter value is divided into a wake-up micropulse and a configuration data packet, including: obtaining an enhancement level label based on the gain gradient value and a preset gradient range, and generating the spread spectrum parameter value; performing structural segmentation based on the spread spectrum parameter value, obtaining the wake-up micropulse and the configuration data packet; and fixing and transmitting the wake-up micropulse based on a preset maximum spread spectrum value.

[0013] Optionally, based on the wake-up pulse and configuration data packet, a parallel blind detection mechanism is triggered through the synchronization control channel to complete the switching of the spreading factor value and coding rate parameter, including: triggering the parallel blind detection mechanism based on the captured wake-up pulse and unverified configuration data packet; starting multiple parallel correlators in the next physical frame based on the parallel blind detection mechanism; and extracting peak locking parameters based on the parallel correlators to complete the switching of the spreading factor value and coding rate parameter.

[0014] A UAV-borne spread spectrum device for remote communication in forest fires is also provided. The device is configured to implement a UAV-borne spread spectrum method for remote communication in forest fires. The device includes: a feature sensing block for acquiring composite perturbation quantities and, based on the composite perturbation quantities and a dynamic observation window, acquiring historical distribution patterns and channel volatility; the composite perturbation quantities include signal strength values, signal-to-noise ratio parameters, and asymmetric jitter; a dimensionality reduction block for acquiring dynamic variance values ​​based on historical distribution patterns and performing nonlinear weighting based on the dynamic variance values ​​and response limit values ​​to acquire the effective number of dimensions; acquiring coupling attenuation rates based on the effective number of dimensions and acquiring convergence threshold values ​​based on coupling attenuation rates and channel volatility; a gradient adjustment block for acquiring dynamic bias quantities based on the convergence threshold values ​​and asymmetric features, and acquiring gain gradient values ​​based on the dynamic bias quantities and composite perturbation quantities; and a blind detection execution block for acquiring spread spectrum parameter values ​​based on the gain gradient values, dividing the spread spectrum parameter values ​​into wake-up micropulses and configuration data packets; and triggering parallel blind detection via a synchronization control channel based on the wake-up micropulses and configuration data packets to complete the switching of spread spectrum factor values ​​and coding rate parameters.

[0015] The beneficial effects of this invention are reflected in: In the UAV-based airborne spread spectrum method for remote communication in forest fires, firstly, by performing non-negative normalization preprocessing on the composite disturbance quantity including signal strength, signal-to-noise ratio, and asymmetric jitter, computational divergence that might be caused by the negative polarity of physical parameters is avoided, providing a reliable data foundation for quantifying the coupling fading of thermal turbulence and vegetation. Secondly, a nonlinear penalty weighting mechanism based on the slew rate limit of the operational amplifier is introduced, which can effectively filter out transient man-made electromagnetic pulse interference generated by the operation of equipment around the fire site, reducing the situation where the system triggers unnecessary spread spectrum compensation due to misjudgment, and helping to maintain spectrum utilization and system throughput. At the same time, the integral asymmetry characteristics of positive and negative Doppler frequency shifts are used to calculate the smoothing bias, avoiding computational anomalies under stable channels, so that the gain gradient value can objectively reflect the fading changes of the channel under thermal radiation. Finally, the parameter synchronization method was improved by dividing the configuration command into a wake-up micropulse and a configuration data packet, and sending them through a synchronization control channel isolated by the physical layer. In the event that the configuration data packet cannot be completely verified due to strong electromagnetic interference at the receiving end, the heterogeneous multi-frequency factor parallel blind detection mechanism can be triggered by the captured wake-up micropulse, and the parameter synchronization switch can be completed without the need for handshake confirmation at the transmitting and receiving ends. This reduces the risk of physical layer delocking and improves the continuity and reliability of critical command transmission in the fire area. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0017] Figure 1 This is a schematic diagram illustrating the steps of the UAV-borne spread spectrum method for remote communication of forest fires according to the present invention; Figure 2 This is a schematic diagram of a portion of steps S1 in the UAV-borne spread spectrum method for remote communication of forest fires according to the present invention; Figure 3 This is a schematic diagram of a portion of step S2 in the UAV-borne spread spectrum method for remote communication of forest fires according to the present invention; Figure 4 This is a schematic diagram of a portion of step S3 in the UAV-borne spread spectrum method for remote communication of forest fires according to the present invention; Figure 5 This is a schematic diagram of a portion of step S4 in the UAV-borne spread spectrum method for remote communication of forest fires according to the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0019] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0020] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0021] This invention provides an airborne spread spectrum method for remote communication in forest fires using unmanned aerial vehicles (UAVs), such as... Figure 1 As shown, in one embodiment, the method includes: S1. Obtain the composite disturbance quantity; the composite disturbance quantity includes signal strength value, signal-to-noise ratio parameter and asymmetric jitter; based on the composite disturbance quantity and dynamic observation window, obtain the historical distribution and channel volatility.

[0022] S2. Obtain the dynamic variance value based on the historical distribution, and perform nonlinear weighting based on the dynamic variance value and the response limit value to obtain the effective dimension number; obtain the coupling attenuation rate based on the effective dimension number, and obtain the convergence threshold value based on the coupling attenuation rate and the channel volatility.

[0023] S3. Obtain the dynamic bias based on the convergence threshold and asymmetric characteristics, and obtain the gain gradient based on the dynamic bias and the composite perturbation.

[0024] S4. Obtain the spreading parameter value based on the gain gradient value, and divide the spreading parameter value into wake-up micropulse and configuration data packet; based on the wake-up micropulse and configuration data packet, trigger the parallel blind detection mode through the synchronization control channel to complete the switching of the spreading factor value and coding rate parameter.

[0025] In this embodiment, it should be noted that in S1, to address the extreme channel distortion and physical quantity calculation divergence problems faced by UAVs in fire environments, a robust underlying data representation logic is established by performing non-negative normalization and anti-zero division smoothing on the composite disturbance quantity. In the specific execution process, a dynamic observation window spanning 500 milliseconds is first defined based on the start point of the current communication frame, and signal strength values, signal-to-noise ratio parameters, and asymmetric jitter are extracted at a sampling rate of 200 Hz. Since the underlying raw received signal strength usually has a negative polarity, for example, when a UAV flies into a dense smoke area, the signal strength drops from -65 dBm to -95 dBm. If physical parameters with negative signs are directly substituted into the existing volatility calculation formula, it is easy to cause polarity reversal and mathematically significant division-to-zero errors.

[0026] To address this, a linear shift mechanism is employed. Based on the preset lower reception limit of -120dBm and the maximum saturation value of -30dBm, the -65dBm of the 99th sampling point is mapped to (-65-(-120)) / (-30-(-120))=0.611, and the 100th sampling point is mapped to 0.277, thus unifying all physical quantities within the non-negative closed interval of 0 to 1. After completing the non-negative normalization, the channel volatility is further calculated. For the case where the values ​​of sampling points at the edge of deep fading approach zero infinitely, the formula introduces the logic of taking the maximum value between the non-negative normalized value of the previous sampling point and the preset minimum smoothed positive constant (e.g., 0.001) when calculating the smoothed denominator of the difference between adjacent samples. This calculation logic ensures that when the input signal approaches the noise floor level, the denominator always maintains a positive lower bound greater than or equal to 0.001, avoiding computational overflow. This processing method eliminates the interference of negative polarity of physical dimensions and transient thermal noise on data analysis, providing a reliable data foundation for subsequent accurate quantification of the inherent fading intensity of the channel caused by the coupling of fire thermal turbulence and vegetation obstruction, and ensuring the stability of the entire adaptive adjustment algorithm under extreme physical conditions.

[0027] In S2, after acquiring the historical data distribution characteristics, a nonlinear penalty weighting mechanism based on the hardware slew rate limit is introduced to solve the existing technical problem of difficulty in distinguishing between natural fading and transient human interference. In complex fire rescue scenarios, the start-up of fire pumps or the operation of high-frequency communication equipment below can generate short-term electromagnetic pulses. At this time, the historical distribution of the signal-to-noise ratio (SNR) parameter shows a dynamic variance as high as 45, with its first-order characteristic change rate reaching 50V / us. If a conventional adaptive algorithm is used, this drastic fluctuation would be misjudged as environmental deterioration, triggering a deep spread factor increase, thus needlessly consuming communication bandwidth. This solution compares the characteristic change rate with the theoretical limit of the physical slew rate of the airborne operational amplifier (e.g., 15V / us) in real time. Since 50V / us exceeds the physical limit of the hardware's natural response, the fluctuation is determined to originate from human transient interference. Subsequently, a negative penalty value is applied to the SNR parameter dimension and it is removed from the computation queue, reducing the effective dimension number participating in the computation from 3 to 2. Based on the reduced effective dimension number, a logarithmic operation and linear combination are performed to obtain the coupling attenuation rate, which is then used as an exponential term to calculate the convergence threshold. When the effective dimension number is 2, the calculated coupling attenuation rate is 0.610. This is combined with the previously obtained channel volatility of 0.210, and through exponential calculation, a convergence threshold of 0.184 is finally obtained.

[0028] Furthermore, compared to the original volatility, the adaptive contraction of this threshold value reflects a dynamic reduction in tolerance to the complex environment of the fire scene. This calculation logic of weight reduction and threshold compression can tighten the trigger threshold in advance to reserve link margin when the natural forest fire cover intensifies, and can also filter out numerical changes caused by transient electromagnetic interference from equipment, reducing the occurrence of unnecessary parameter compensation due to misjudgment. While maintaining the connectivity of the communication link, it can effectively protect the spectrum utilization and overall data throughput.

[0029] In S3, to objectively reflect the substantial impact of the unique thermal scintillation phenomenon of fire scenes on communication links, a dynamic smoothing bias mechanism was constructed using the integral asymmetry of Doppler frequency shift to obtain accurate gain adjustment gradients. Conventional communication channel models typically assume that the Doppler frequency shift follows a symmetrical stationary distribution. However, in forest fire environments, high-temperature heat columns and updrafts cause asymmetric changes in the electromagnetic wave refractive index. Extracting data from the current physical frame revealed that the positive Doppler integral value extracted due to the updraft force of the heat flow reaches 280Hz, while the negative Doppler integral value is only 40Hz. Utilizing this asymmetry, the ratio of the absolute value of the difference between the positive and negative Doppler integrals to their sum (with an additional 0.1Hz zero-compensation positive number) was calculated, yielding an asymmetry of 0.749. Subsequently, this asymmetry was multiplied by the heat flow sensitivity of 0.15 and a static noise floor of 0.02 to calculate a dynamic bias of 0.132.

[0030] The introduction of this calculation logic stems from the fact that when the communication link is in a relatively stable state, the convergence threshold approaches zero. At this point, even minute thermal noise can cause the denominator of the gain gradient index to become smaller, leading to an overestimation of the calculated result and thus causing miscompensation. Adding a dynamic bias as a buffer term to the denominator of the gain gradient index provides a physical anti-jitter range based on the actual intensity of thermal turbulence. Considering that the current non-negative normalized signal value has dropped to 0.200 and the historical average is 0.550, an absolute deviation of 0.350 is calculated. Dividing this absolute deviation by the sum of the convergence threshold (0.184) and the dynamic bias (0.132) yields a gain gradient value of 1.107. Without this dynamic bias buffer, the gradient value would reach 1.902. This mechanism avoids mathematical calculation errors under stable channel conditions, enabling a more quantitative measurement of the actual degradation risk of the link under thermal radiation.

[0031] In S4, to address the risk of physical layer unlocking due to control command loss during spread spectrum parameter switching, the underlying configuration distribution and receiver synchronization response logic were reconstructed. Based on the previously calculated gain gradient value of 1.107, it was matched with a preset moderate boost range to generate a spread spectrum parameter value that increases the spread spectrum factor from the original SF7 to SF9. To overcome the problem of protocol-level handshake interruption under adverse channel conditions, the spread spectrum parameter value was structurally segmented, extracting a 4-byte wake-up micropulse and a 12-byte configuration data packet. At the transmitting end, the onboard microprocessor invokes a physically isolated independent synchronization control channel. This control channel is hardware-hardware-defined to support the maximum spread spectrum value SF12, which penetrates dense smoke in a fire scene with strong anti-interference capabilities to transmit the wake-up micropulse, followed by the transmission of the configuration data packet.

[0032] In this application scenario, assuming a sudden canopy fire generates transient electromagnetic interference, the receiver successfully captures the high-penetration wake-up pulse, but fails to pass the cyclic redundancy check (CRC) of the subsequent 12-byte configuration data packet. At this point, because the receiver senses the wake-up state but lacks specific configuration parameters, it directly triggers a heterogeneous multi-frequency factor parallel blind detection mechanism at the beginning of the next physical frame. The receiver's baseband processing chip simultaneously activates three parallel correlators configured with local despreading sequences of SF7, SF9, and SF11 for computation. After low-level parallel correlation integration, the peak value of the SF7 channel is identified as only 15, while the peak value of the SF9 channel reaches 158, exceeding the set 80-level locking threshold. The receiver then extracts SF9 as the peak locking parameter, passively completing the switching of modulation and demodulation parameters for the main service channel. This mechanism, transitioning from a two-end protocol handshake to a physical layer multi-channel parallel blind detection, maintains synchronous parameter switching without requiring retransmission from the transmitter or confirmation from the receiver, reducing the risk of communication interruption due to command packet loss in extreme fire environments.

[0033] In summary, the UAV-based airborne spread spectrum method for remote communication in forest fires firstly avoids computational divergence caused by the negative polarity of physical parameters by preprocessing the composite disturbance quantity, which includes signal strength, signal-to-noise ratio, and asymmetric jitter, through non-negative normalization. This provides a reliable data foundation for quantifying the coupling fading of thermal turbulence and vegetation. Secondly, the introduction of a nonlinear penalty weighting mechanism based on the slew rate limit of the operational amplifier effectively filters out transient man-made electromagnetic pulse interference generated by the operation of equipment around the fire site, reducing unnecessary spread spectrum compensation triggered by misjudgment and helping to maintain spectrum utilization and throughput. Simultaneously, the calculation of the smoothing bias quantity utilizes the integral asymmetry characteristics of positive and negative Doppler frequency shifts, avoiding computational anomalies under stable channels and ensuring that the gain gradient value objectively reflects the fading changes of the channel under thermal radiation. Finally, the parameter synchronization method was improved by dividing the configuration command into a wake-up micropulse and a configuration data packet, and sending them through a synchronization control channel isolated by the physical layer. In the event that the configuration data packet cannot be completely verified due to strong electromagnetic interference at the receiving end, the heterogeneous multi-frequency factor parallel blind detection mechanism can be triggered by the captured wake-up micropulse, and the parameter synchronization switch can be completed without the need for handshake confirmation at the transmitting and receiving ends. This reduces the risk of physical layer delocking and improves the continuity and reliability of critical command transmission in the fire area.

[0034] like Figure 2 As shown, in one embodiment, the step of obtaining the historical distribution and channel volatility based on the composite disturbance and the dynamic observation window in S1 is broken down into the following sub-steps: S11. Define a dynamic observation window based on the starting point of the current communication frame.

[0035] S12. Extract physical extreme values ​​based on the dynamic observation window and construct a non-negative normalized column using linear translation. Specifically, take the physical extreme values ​​of each dimension as a reference, subtract the lower limit of reception from the current sampling point value, and divide by the difference between the maximum saturation value of the ideal environment and the lower limit of reception to construct a non-negative normalized column in the interval between zero and one.

[0036] S13. Obtain the historical distribution column and channel volatility based on the non-negative normalized column. This step specifically includes: obtaining the adjacent sampling difference based on the non-negative normalized column; obtaining the dynamic smoothing basis based on the non-negative normalized column and the preset anti-zero constant; and obtaining the channel volatility based on the adjacent sampling difference, the dynamic smoothing basis, and the preset compensation term.

[0037] The mathematical expression for channel volatility is as follows:

[0038] in, For the first Channel volatility corresponding to a composite disturbance; For the first The number of effective sampling points for each composite disturbance within the dynamic observation window; For the first The sampling point and the first The adjacent sampling difference of the non-negative normalized column corresponding to each sampling point; According to the first The dynamic smoothing basis is constructed by multiplying the non-negative normalized column corresponding to each sampling point with the preset smoothing ratio threshold; It is a preset compensation term consisting of the ratio of the standard deviation of the non-negative normalized column to the theoretical full-scale span.

[0039] In this embodiment, it should be noted that in S11, during the initial stage of the UAV's reconnaissance mission deep into the core area of ​​the fire, a dynamic observation window for data acquisition is defined. Specifically, the onboard microprocessor uses the start time of the current communication physical frame as the reference anchor point and defines a non-stationary sliding observation window with a span of 500 milliseconds along the historical timeline. The span of the dynamic observation window is determined based on the autocorrelation function analysis of historical channel data. The received signal strength data of the UAV under typical forest fire conditions is collected for 10 consecutive seconds, its autocorrelation function is calculated, and the lag time corresponding to the attenuation to 0.5 is taken. Statistical analysis of historical data from multiple flight missions shows that the lag time distribution is between 400 and 600 milliseconds. The median value is taken, and after considering engineering margins, it is set to 500 milliseconds. This span can capture the channel fading characteristics caused by thermal turbulence and vegetation obstruction in the fire area, while avoiding the dilution of transient changes by an excessively long window. Within this window, low-level raw physical parameters, including received signal strength, signal-to-noise ratio, and microsecond-level asymmetric jitter, are continuously extracted using a high-frequency sampling rate of 200 Hz. The sampling rate is determined based on historical channel fluctuation spectrum analysis. Historical signal data from UAVs at the fire's edge and core areas are collected. Fourier transform analysis of the dominant frequency component of the channel fluctuation reveals that the Doppler jitter frequency caused by thermal turbulence in the fire is mainly concentrated in the 80 Hz to 150 Hz range. Based on the Nyquist sampling theorem, the sampling rate is set to 1.5 times the highest dominant frequency, i.e., 200 Hz. This value is verified through spectrum analysis of historical data from five different fire environments, demonstrating that it can fully preserve the channel's dynamic characteristics without introducing excessive computational burden.

[0040] For example, 100 discrete sampling points can be collected within 500 milliseconds. This dynamic sliding window approach, rather than static global statistics, is primarily to address the spatial heterogeneity of the fire environment. When a drone, cruising at 15 meters per second, rapidly cuts from the sparse forest at the fire's edge into the central area where high temperatures, dense smoke, and thick canopies overlap, the channel state undergoes a dramatic change. Relying on long-term static global statistics would dilute the current sudden fading characteristics with historical data, leading to a delay in the perception of environmental degradation. By using a sliding observation window, the dynamic evolution of thermal turbulence and multipath effects can be closely tracked with near real-time granularity of 500 milliseconds. This step effectively solves the technical problem of delayed environmental state perception in existing communications under the dual effects of high-speed movement and complex terrain, providing a data sample source with high timeliness and a true physical mapping relationship for subsequent channel volatility calculation and adaptive compensation.

[0041] In S12, after extracting the raw data within the observation window, a linear translation mechanism is used to transform the dimensional parameters into dimensionless non-negative normalized series. In actual fire field measurements, the raw physical values ​​of the received signal strength are usually in the negative range. For example, when a drone flies into a dense canopy, the signal strength may drop rapidly from a relatively good -65dBm to -95dBm. If these parameters with negative signs are directly substituted into the subsequent volatility or dispersion calculation formulas, it is very easy to cause polarity reversal during difference or ratio operations, and even trigger division-by-zero anomalies when the denominator approaches the noise floor.

[0042] To mitigate this computational risk, the lower limit of the underlying hardware's receiving sensitivity was pre-set to -120dBm, and the maximum saturation receiving value under ideal lossless conditions was set to -30dBm. The lower limit and maximum saturation value were determined based on historical calibration data from the airborne RF front-end hardware. In sensitivity tests conducted before the drone left the factory, signal strength data was collected 100 times from no signal to saturated output. The lowest signal strength that the receiver could stably demodulate was found to be between -122dBm and -118dBm, with the average of -120dBm taken as the lower limit. The maximum received signal strength measured in an unobstructed, open environment was found to be between -32dBm and -28dBm, with the average of -30dBm taken as the saturation value. This set of values ​​was verified through three independent calibration tests, covering the extreme signal dynamic range that the drone might encounter during actual flight.

[0043] Taking the 99th sampling point as an example, subtracting the lower limit of -120dBm from its -65dBm value yields a relative increment of 55dBm. Dividing this by the ideal span of 90dBm (i.e., -30 minus -120) results in a non-negative normalized value of 0.611. Similarly, the -95dBm value at the 100th sampling point is mapped to 0.277. This linear extremum mapping calculation logic not only unifies the data scale of different physical dimensions such as signal strength and signal-to-noise ratio, making multi-dimensional parameters comparable for algebraic operations within the same mathematical framework, but also eliminates computational anomalies that may be induced by negative data from a mathematical perspective. This step eliminates the interference of physical polarity on channel evaluation and improves the stability of the airborne processor when performing complex floating-point operations in low signal-to-noise ratio edge regions.

[0044] In S13, based on the generated non-negative normalized column, the historical distribution column is further extracted and the channel volatility characterizing the inherent fading intensity of the environment is calculated. In S13, a standardized channel volatility calculation model for quantifying the inherent fading intensity of the environment is constructed, and its specific mathematical expression is as follows: In this expression, Representing the The standardized channel volatility corresponding to each composite disturbance; This represents the total number of valid sampling points within the dynamic observation window; and These represent the feature values ​​of the adjacent next and previous sampling points after non-negative normalization, respectively. It is a minimal smooth positive constant that prevents the denominator from being zero; A ratio threshold for smoothing unstructured transient thermal noise is used to suppress unstructured transient thermal noise. It is the standard deviation of the non-negative normalized sequence within the entire observation window, while This is the theoretical full-scale span constant, used for normalizing the standard deviation. Among them, the minimum smoothed normal coefficient... The value of is determined based on the minimum non-zero value distribution of the non-negative normalization column in historical data. Data from 100 drone flights under extreme fading scenarios such as dense smoke from fires and canopy obstruction were collected. The distribution of the non-negative normalization column under deep signal fading was statistically analyzed, revealing that its minimum value is usually not lower than 0.002. To prevent zero values ​​from appearing in the denominator and to avoid excessive impact on normal calculations, this constant was set to half of the statistical minimum, i.e., 0.001. Verification using historical data from five different fading levels showed that this value ensures operational stability without introducing significant calculation bias. The ratio threshold is... The value is determined based on the distribution analysis of the ratio of adjacent sampling difference to the current sampling value in historical data. Fifty sets of historical data were collected from the UAV in both stable flight and turbulent fire conditions. The ratio of adjacent sampling difference to the current sampling value was calculated. The average value of this ratio was 0.8 in stable conditions and 1.3 in turbulent conditions. The larger of the average values ​​of the two scenarios, 1.5, was taken as the threshold. Cross-validation showed that this value can effectively suppress the abnormal amplification of the ratio caused by transient thermal noise, while preserving the fading characteristics of the real environment.

[0045] This composite superposition calculation logic aims to address the fundamental technical problem of signal jitter characteristics being overwhelmed by background noise in low signal-to-noise ratio edge regions. In conventional statistical processing, variance or standard deviation is typically used to measure signal dispersion. However, drone communication at forest fire sites faces complex physical processes involving alternating multipath effects and signal blockage. Simply calculating the global standard deviation would confuse the slow, gradual decrease in signal strength with severe, localized jitter at high frequencies. Therefore, the first part of the expression uses the absolute value of the difference between adjacent sampling points. To capture high-frequency jitter at the microsecond level. In the denominator The design addresses situations where drones fly into dense tree canopies, causing significant signal attenuation. In this case, its normalized value... It will approach zero infinitely. If we directly use it in the existing formula... As the denominator, the rate of change is calculated; once... Even extremely small, minute differences in the numerators can output an infinitely large ratio, leading to computational overflow and causing data type out-of-bounds errors in the microprocessor. By introducing... The operation forces a positive lower bound to be set for the denominator (e.g., taking...). Multiply this by the ratio threshold. (For example, taking 1.5) is equivalent to building a dynamic and elastic basic reference surface at the bottom layer.

[0046] Looking at the specific data, assuming 100 sampling points were collected within the observation window, the non-negative normalized value of the 99th sampling point is 0.611, while the value of the 100th sampling point drops to 0.277 due to tree shading. The absolute value of the adjacent difference between the two is extracted as 0.334. When calculating the denominator of the dynamic smoothing, since 0.611 is greater than 0.001, we take 0.611 and multiply it by 1.5 to get 0.916. Dividing 0.334 by 0.916 yields the local fluctuation contribution ratio of 0.364. We accumulate and average all 99 local ratios within the window, assuming the average local high-frequency fluctuation is 0.125. At this point, the latter part of the formula... It begins to exert its global compensation effect. Assume the overall standard deviation of these 100 data points. It is 0.085, the full-scale constant. If the value is 1.0, then the global compensation term is 0.085. Adding the local high-frequency fluctuation mean of 0.125 to the global compensation term of 0.085, the final standardized channel volatility is 0.210. This calculation process, which combines the constrained mean of the local high-frequency change rate with the global macroscopic standard deviation, can not only keenly capture the instantaneous multipath distortion caused by the swaying of trees in the fire, but also take into account the overall signal mean subsidence caused by the UAV penetrating deep into the fire. It filters out disordered thermal noise interference caused by the physical hardware receiving sensitivity approaching its limit, providing a benchmark quantitative index that can objectively map the complexity of the physical shielding of the environment. This gives the subsequent parameter compensation mechanism a robust data source input, avoiding overly sensitive adjustment responses when faced with slight environmental changes.

[0047] like Figure 3 As shown, in one implementation, the step of performing nonlinear weight reduction and obtaining the convergence threshold value in S2 is broken down into the following sub-steps: S21. Obtain the characteristic change rate based on the historical distribution.

[0048] S22. Perform state comparison based on characteristic rate of change and response limit value; wherein the response limit value is set by the physical slew rate limit of the airborne operational amplifier.

[0049] S23. Apply a negative penalty value based on the comparison results, remove the human-caused interference dimension, and obtain the effective number of dimensions; specifically: if the dynamic variance value of a certain dimension exceeds the benchmark noise threshold and its characteristic change rate exceeds the response limit value, then it is determined that the fluctuation originates from the human-caused interference dimension, apply a negative penalty value to the dimension and remove it, and obtain the effective number of dimensions after weight reduction.

[0050] S24. Perform logarithmic operations and linear combinations based on the number of effective dimensions to obtain the coupling attenuation rate; specifically: add one to the number of effective dimensions, perform natural logarithmic operations, and divide by the linear combination term after weighting and biasing the number of effective dimensions to obtain the coupling attenuation rate.

[0051] S25. Perform exponential calculation based on the coupling attenuation rate and channel volatility to obtain the convergence threshold value; specifically: multiply the coupling attenuation rate by the negative of the channel volatility as the exponent, raise the power to the natural constant, and then multiply it by the channel volatility to obtain the convergence threshold value.

[0052] In this embodiment, it should be noted that in S21, after the quantification of the basic volatility, the environmental perception enters the feature identification stage, which is responsible for extracting the dynamic variance and feature change rate of each dimension in the historical distribution series. During the continuous advancement of the sliding observation window, the second-order statistical features and first-order derivatives of three dimensions—received signal strength, signal-to-noise ratio, and microsecond-level Doppler asymmetric jitter—are calculated in real time. Taking the signal-to-noise ratio parameter as an example, when a drone enters the core rescue area where high-pressure water cannon firefighting and heavy firefighting machinery operations are underway from the safe airspace outside the fire site, the electromagnetic background noise in the environment will fluctuate drastically.

[0053] At this point, by traversing the non-negative normalized column data of the signal-to-noise ratio within the current 500-millisecond observation window, the dynamic variance value was calculated to be 45, far exceeding the preset steady-state baseline noise threshold of 15. The baseline noise threshold was determined based on the dynamic variance distribution of the channel in a stable, interference-free state in historical data. Twenty historical flight data from the UAV in an open, fire-free, and equipment-interference-free environment were selected, and the dynamic variance of the three dimensions—signal strength, signal-to-noise ratio, and asymmetric jitter—was statistically analyzed. The maximum values ​​were found to be between 12 and 14. Taking the upper limit and adding a 10% margin, this threshold was set to 15. This threshold was verified in five different stable scenarios and can classify more than 99% of natural fluctuations within the normal range, avoiding misjudgments triggered by normal noise.

[0054] Simultaneously, by extracting the voltage change slope of the signal-to-noise ratio between adjacent microsecond-level sampling points, the characteristic change rate of this dimension was calculated to reach 50V / µs. Extracting the dynamic variance value is for macroscopic assessment of whether the channel parameter in this dimension has deviated from a stationary state, while further calculating the first-order characteristic change rate is to capture the transient steepness of the fluctuation. This data extraction logic provides necessary multi-dimensional judgment criteria for subsequently distinguishing between natural environmental fading and sudden man-made interference, moving beyond a single reliance on amplitude-based limit alarms to include a dynamic review dimension based on time change rate, thus constructing a more rigorous preprocessing stage for channel quality assessment data.

[0055] In S22, the high-frequency fluctuation characteristics extracted in S21 are compared with the physical response limits of the underlying hardware to determine the interference source attributes. In nature, channel multipath and shielding fading caused by tree canopies swaying in the wind, drone attitude changes, or rising hot air currents from a fire are physically limited by the speed of macroscopic objects, and the slope of these changes at the electromagnetic wave receiver is usually relatively gentle. However, transient electromagnetic pulses generated by high-power communication base stations and fire pump motor start-stop devices around a fire often have rise times on the order of nanoseconds to microseconds.

[0056] Based on this objective physical law, the physical slew rate limit of the airborne RF front-end operational amplifier (assumed to be 15V / us) was retrieved as a benchmark comparison value. The response limit value was determined based on the datasheet of the airborne RF front-end operational amplifier device and historical measured data. The typical slew rate of the operational amplifier model used was found to be 15V / us in its specification sheet. During the UAV's factory testing, the response waveforms of the RF front-end to step signals were collected 10 times. The measured slew rates ranged from 14.2V / us to 15.1V / us, verifying the accuracy of the device's nominal value. This value, as the hardware physical limit, is used to distinguish between natural fading and man-made transient interference.

[0057] When comparing the characteristic rate of change of the signal-to-noise ratio (SNR) parameter, which is as high as 50V / µs, with the physical slew rate limit of 15V / µs, the logic detector will identify that the characteristic rate of change far exceeds the upper limit of the hardware's linear response capability under natural environmental fading. The logic of using the hardware slew rate limit as the judgment threshold is that any signal fluctuation exceeding this slope does not conform to the continuous law of natural physical environment evolution, and can therefore be directly classified as unnatural, man-made sudden electromagnetic spike interference. This step solves the technical blind spot in existing spread spectrum communication, which tends to attribute all SNR degradation to increased communication distance or obstacle obstruction, and improves the anti-interference discrimination capability in complex electromagnetic rescue environments.

[0058] In step S23, after identifying the interference source attributes, a nonlinear weighting process is performed based on the state comparison results to obtain the effective number of dimensions that truly reflect the complexity of the natural environment. Since the signal-to-noise ratio (SNR) dimension was determined to be affected by transient interference from man-made electromagnetic pulses in the previous step, a negative penalty mechanism is triggered. Instead of normal weight accumulation, a negative penalty value is forcibly applied, completely removing this dimension from the effective parameter queue representing environmental complexity. Therefore, the original three evaluation dimensions—signal strength, SNR, and asymmetric jitter—are reduced to two effective dimensions after weighting and filtering.

[0059] If the transient signal-to-noise ratio drop caused by the water pump motor is directly included in the environmental assessment during a fire, it might be mistakenly assumed that the drone has entered an extremely harsh, densely covered forest area, thus incorrectly triggering an increase in the depth spread factor. This would not only needlessly increase the computational power consumption of the onboard processor but also waste valuable communication bandwidth due to an unnecessary reduction in the coding rate. By eliminating human interference, the focus can be placed on the slowly varying fading characteristics truly caused by the thermodynamics of the fire and the physical obstruction of the trees. This nonlinear weighting and elimination logic effectively blocks the negative transmission of transient electromagnetic spikes to the overall environmental perception coefficient, preventing overreaction of the control strategy and maintaining data throughput levels relatively effectively while ensuring link connectivity.

[0060] In S24, after obtaining the effective dimension number after purification, the coupling attenuation rate used to quantify the overall severity of the current fire situation is calculated through logarithmic operations and linear combination mathematical transformations. Adding 1 to the current effective dimension number 2 yields 3, and then taking its natural logarithm gives a numerator of approximately 1.098. Simultaneously, setting the attenuation coefficient of the linear combination to 0.4 and the bias constant to 1.0, substituting the effective dimension number 2 into the denominator formula yields 1.8. The attenuation coefficient and bias constant were determined based on regression analysis of the effective dimension number and the actual channel fading degree in historical data. Historical data of UAVs in 10 fire environments of different complexities were collected, and the effective dimension number (1 to 3) and the actual measured link fading value were recorded. The linear relationship between the effective dimension number and the fading value was obtained by fitting with the least squares method: fading value = 0.38 × effective dimension number + 0.96. After rounding, the attenuation coefficient was set to 0.4 and the bias constant was set to 1.0. After testing with 5 sets of verification data, this combination can make the correlation coefficient between the output of coupling attenuation rate and the actual link fading reach more than 0.92.

[0061] Finally, dividing the numerator 1.098 by the denominator 1.8 yields a coupling attenuation rate of 0.610. This calculation logic, which places the logarithmic operation in the numerator and the linear combination in the denominator, is used to fit the diminishing marginal effect of the impact of multidimensional environmental characteristics on the communication link in the mathematical model.

[0062] As fire environments become increasingly complex and the number of effective dimensions involved in fluctuations increases, the sensitivity to environmental degradation rises accordingly. However, using a purely linear growth model would lead to excessively large attenuation rates, resulting in overly frequent parameter switching in later stages. The introduction of a logarithmic function flattens the attenuation rate growth curve as the dimension increases, while the linear combination and bias constant in the denominator act as fundamental numerical ballast. This calculation method solves the problems of numerical overflow and weight imbalance that easily occur when fusing multidimensional variables, enabling the output coupled attenuation rate to smoothly and objectively reflect the comprehensive attenuation trend resulting from the superposition of multiple physical factors, including forest fire shading and thermal radiation.

[0063] In S25, the calculated coupling attenuation rate is used to nonlinearly compress the initial channel volatility, ultimately obtaining the critical convergence threshold for link collapse. The mathematical expression for the convergence threshold is as follows: ;in, For channel volatility, This represents the coupling attenuation rate.

[0064] For example, multiplying the coupling attenuation rate of 0.610 by the negative of the initial channel volatility of 0.210 yields an exponential term of -0.128. Raising this to the power of the natural constant e, we get an attenuation factor of 0.879. Then, multiplying this attenuation factor 0.879 again by the initial channel volatility of 0.210, we finally calculate a convergence threshold of 0.184. Compared to the unprocessed initial volatility of 0.210, the threshold value after exponential calculation is reduced to 0.184. The core purpose of this negative exponential calculation logic is to achieve adaptive shrinkage of the tolerance threshold.

[0065] As the fire environment around the drone becomes increasingly complex and the coupling attenuation rate gradually increases, the absolute value of the negative exponent increases accordingly, resulting in a smaller attenuation factor, which further compresses the output convergence threshold. This mechanism means that in harsh forest fire environments, the tolerance for minute signal fluctuations is actively reduced. Signals that would be negligible in open plains can reach the compressed threshold more quickly in the core of the fire, thus triggering the gain compensation mechanism earlier. This step effectively solves the problem of slow response caused by fixed thresholds in dynamically changing extreme environments, providing valuable time margin for adjusting the underlying link and improving the early warning sensitivity of airborne communication when facing signs of deep fading in fire conditions.

[0066] like Figure 4 As shown, in one implementation, the step of obtaining the dynamic bias and gain gradient values ​​in S3 is broken down into the following sub-steps: S31. Extract positive and negative Doppler based on the upward force of heat flow to obtain asymmetric features.

[0067] S32. Obtain the dynamic bias based on the asymmetric characteristics, heat flux sensitivity, and static noise floor value.

[0068] The dynamic bias is mathematically expressed as follows:

[0069] in, This is a dynamic bias value; This is the static noise floor value; This is the positive Doppler integral value extracted within the current physical frame due to the upward force of thermal flow; The negative Doppler integral value is extracted; the integral deviation ratio term constructed from positive and negative Doppler constitutes an asymmetric feature; Heat flux sensitivity; To prevent positive compensation numbers with a denominator of zero, the value of the heat flux sensitivity rate was determined based on the correlation analysis between Doppler asymmetry and actual link bit error rate in historical data. Historical data from UAVs under thermal turbulence conditions in 10 fires of different intensities were collected, recording the changes in Doppler asymmetry and the corresponding link bit error rate. Linear regression fitting showed that for every 0.1 increase in asymmetry, the bit error rate increased by an average of 0.015. To provide an effective buffer for the dynamic bias before the bit error rate deteriorates, the sensitivity rate was set to 0.15. After verification in 5 fire environments, this value allowed the dynamic bias to begin increasing approximately 200 milliseconds before the bit error rate deteriorated, reserving response time for parameter switching.

[0070] S33. Obtain the absolute deviation value based on the non-negative normalized series of the composite disturbance at the current moment and its historical mean; specifically: calculate the absolute value of the difference between the instantaneous value of the non-negative normalized series at the current moment and its historical mean of all sampling points within the dynamic observation window, and use it as the absolute deviation value.

[0071] S34. Obtain the gain gradient value based on the absolute deviation value, the convergence threshold value, and the dynamic bias value.

[0072] The mathematical expression for the gain gradient value is as follows:

[0073] in, For the first The gain gradient value corresponding to each composite disturbance; For the first The absolute deviation value corresponding to each composite disturbance; For the first The convergence threshold value corresponding to each composite perturbation; This is the dynamic bias calculated above.

[0074] In this embodiment, it should be noted that in S31, physical parameters that truly reflect the thermodynamic characteristics of the fire scene, i.e., asymmetric features, are extracted from the underlying Doppler frequency shift data. Existing radio propagation models typically assume that the Doppler frequency shift generated by the moving node follows a symmetrical Gaussian distribution over time integral. However, in the actual environment of forest fires, the high-temperature heat column at the center of the fire line triggers strong updrafts. When the microwave signal emitted by the UAV penetrates this thermally turbulent region, the abrupt change in air refractive index causes the signal phase change to exhibit a directional bias. The frequency domain shift within the current physical frame is continuously monitored by the airborne radio frequency front-end, and time-dimensional integration is performed on both the positive and negative frequency shift deviations.

[0075] For example, within a 20-millisecond communication physical frame, the positive Doppler integral value extracted due to the uplift force of thermal flux reached 280Hz, while the negative Doppler integral value extracted within the same observation window was only 40Hz. Such significantly different integral data is rare in conventional flight environments. By quantifying this physical phenomenon, the two specific asymmetric jitter data points of 280Hz and 40Hz were used as basic inputs to construct an asymmetric feature reflecting the intensity of thermal convection in a fire scene. This step departs from the symmetry assumption in conventional channel estimation, enabling airborne communication to perceive thermal radiation interference and providing physical data support for the subsequent construction of anti-jitter mechanisms.

[0076] In S32, the dynamic bias used to smooth thermal disturbances is calculated. The specific mathematical logic used in this calculation process is as follows: the dynamic bias is equal to the static noise floor value plus the product of the heat flux sensitivity rate and the integral deviation ratio term. The integral deviation ratio term is obtained by dividing the absolute value of the difference between the positive and negative Doppler integral values ​​by the sum of the two and the cumulative sum of the compensation positive number to prevent the denominator from being zero.

[0077] Specifically, a dynamic bias generation model for absorbing thermodynamic disturbances in the fire field was established in S32, and its mathematical expression was set as follows: In this formula, Represents the dynamic bias of the final output; The static noise floor bias constant under conditions without thermal disturbance; and These represent the positive Doppler frequency shift peak integral value and the negative Doppler frequency shift peak integral value extracted by the radio frequency receiver through baseband signal processing within the current time observation window, respectively. A pre-defined heat flux-sensitive weighting factor is used to adjust the response ratio to airflow disturbances; This is a small compensating positive number used to prevent division by zero errors in denominator operations. This computational logic was specifically proposed to address the misleading effect of radio thermal flicker caused by high-temperature hot columns on channel perception. Conventional anti-fading algorithms for UAV communication are mostly based on the premise that the Doppler frequency shift exhibits a symmetrical and stable distribution, assuming that the probability densities of positive and negative frequency shifts are essentially the same. However, in the core area of ​​a real forest fire, high-temperature flames heat the local air, forming a strong convective rising column of hot air reaching high altitudes. When microwave signals penetrate these media where the refractive index changes drastically and inhomogeneously, their phase exhibits a strong asymmetric bias in the frequency domain. To transform this physical characteristic of thermal turbulence into a communication-recognizable variable, the formula constructs... This is the asymmetry proportionality term. The numerator of this term extracts the absolute difference between the positive and negative Doppler integrals, used to measure the degree of imbalance in the shift; the denominator calculates the sum of the two, serving as a numerical normalization, ensuring that the output of this term is always limited to the dimensionless interval between 0 and 1. (Introduction) This is to prevent the denominator from failing when the drone is hovering in a windless environment with zero frequency shift, causing a division by zero error. Among these, the compensation positive number... The value of is determined based on the distribution of the sum of positive and negative Doppler integral values ​​in historical data. Twenty flight history data of the UAV in a hovering windless environment were collected, and the sum of positive and negative Doppler integral values ​​was statistically analyzed. The minimum value is distributed between 0.05Hz and 0.15Hz, and the upper limit of 0.1 is taken as the compensation positive number. This value can prevent the calculation abnormality of zero denominator, and the impact on the asymmetry calculation result is less than 0.3% during normal flight. It has been verified that the calculation error of the subsequent dynamic bias is negligible.

[0078] Taking specific data collection as an example, suppose that the positive Doppler integral value extracted from the updraft of the fire scene within the current physical frame is 280Hz, while the negative Doppler integral value within the same time period is only 40Hz. A standard static noise floor value is set. The heat flux sensitivity is 0.02. The constant is 0.15, preventing zero constant. The static noise floor value is set to 0.1. The static noise floor value is determined based on the statistical analysis of dynamic offset values ​​under historical data in environments without thermal turbulence. Ten historical flight data points of the UAV were selected, taken from scenarios without fire or updraft interference. Under stable channel conditions, the dynamic offset calculation formula was used to inversely extrapolate the results, yielding an offset distribution between 0.018 and 0.022. The median value of 0.02 was taken as the static noise floor value. This value was verified in five different stable scenarios, ensuring that the dynamic offset value is close to this noise floor value under conditions without thermal disturbance, thus avoiding numerical anomalies caused by an excessively small denominator in the gain gradient calculation.

[0079] After substituting into the formula, the absolute value of the difference in the numerator is 240, and the sum of the denominators is 320.1. The ratio of the two yields an asymmetry of 0.749. This value objectively characterizes the current link as being subjected to high-intensity thermal convection interference. Subsequently, multiplying 0.749 by the sensitivity factor of 0.15 yields a thermally derived bias of 0.112, which is then added to the static noise floor of 0.02, ultimately outputting a dynamic bias of 0.132. The technical value of converting Doppler asymmetry into a bias value lies in providing a flexible anti-jitter buffer for subsequent gain adjustment calculations. If communication quality is judged solely based on the drop in signal amplitude, it is easy to mistake a brief phase loss due to instantaneous changes in air refractive index for physical obstruction caused by dense trees, thus incorrectly triggering deep parameter reconstruction. By injecting this dynamic bias value, which is positively correlated with environmental thermal instability, the judgment threshold can be actively increased when strong airflow disturbances are detected, enhancing hysteresis characteristics. This computational process enables the receiver to maintain its judgment accuracy in complex and ever-changing high-temperature environments, suppresses false warnings caused by non-physical obstacles, and optimizes the scheduling efficiency of underlying physical resources.

[0080] In step S33, after obtaining the dynamic bias for smoothing environmental thermal disturbances, the instantaneous degradation of the communication link at the current moment is evaluated. Specifically, this involves calculating the absolute deviation between the current non-negative normalized instantaneous value of the composite disturbance and its historical mean. Data from all sampling points within a 500-millisecond dynamic observation window recorded in the onboard memory is retrieved, and the non-negative normalized mean of the signal strength value over this historical period is calculated to be 0.550. Simultaneously, the RF receiving module extracts the instantaneous non-negative normalized value of the latest sampling point, which drops to 0.200. At this moment, the drone may have just flown over a dense coniferous forest canopy, where the thick vegetation causes a sudden increase in microwave signal transmission loss. By calculating the difference between the current instantaneous value of 0.200 and the historical mean of 0.550, and taking its absolute value, an absolute deviation of 0.350 is obtained.

[0081] The core purpose of calculating this deviation value is to capture the degree of abrupt changes in channel conditions. Relying solely on long-term historical average data will fail to detect instantaneous physical attenuation, such as that caused by tree canopy obstruction; while relying solely on instantaneous values ​​lacks a reference benchmark to measure the degree of degradation. By calculating the absolute deviation value, the current link state can be relatively quantitatively compared with the stable state of the previous stage, thereby objectively measuring the magnitude of the current signal drop. This data is the direct driving factor in determining whether to increase the spread spectrum gain, ensuring that compensation actions closely follow changes in the physical environment.

[0082] In S34, following the aforementioned evaluation results, the gain gradient value, which ultimately determines the direction of parameter switching, is calculated by integrating various environmental assessment indicators from the previous stage. The formula used in this step is as follows: the gain gradient value equals the absolute deviation value obtained in the previous step, divided by the sum of the convergence threshold value and the dynamic bias.

[0083] Specifically, S34 uses the gain gradient value to calculate the expression. This completes the transformation mapping from environmental status monitoring to the execution of actions based on communication parameters. In this expression, Indicates the first The spread spectrum gain adjustment gradient index corresponding to each composite disturbance; This represents the instantaneous non-negative normalized feature value extracted by the radio frequency receiving channel at the current physical sampling moment; This represents the statistical average of the disturbance over the entire historical sliding observation window; the denominator contains... It is the critical convergence threshold for link collapse after the increase in effective dimension and nonlinear compression in the preceding steps; This is the dynamic smoothing bias extracted in the previous step. The design intent of this calculation logic is to construct an adaptive risk amplification mechanism that balances the instantaneous link degradation magnitude with the overall environmental carrying capacity. In forest fires, drone communication link interruptions often occur when the instantaneous signal drop exceeds the currently tolerable dynamic boundary. The numerator of the formula... By extracting the absolute value of the deviation between the current instantaneous state and the historical stationary mean, the magnitude of sudden degradation in channel quality is objectively characterized. The denominator of the formula incorporates a dual physical constraint: a convergence threshold. This reflects the static decay base caused by the current terrain and vegetation cover; this value is automatically compressed to a smaller value as the environment becomes more severe; dynamic bias. This represents the anti-shake boundary for transient instability caused by high-temperature airflow.

[0084] Taking specific fire detection data as an example, the historical average signal strength recorded within the past 500 milliseconds observation window was 0.550. However, when the drone penetrated a dense coniferous canopy, the current instantaneous non-negative normalized value dropped sharply to 0.200. The numerator calculated the numerical deviation between the two to be 0.350. On the denominator, the convergence threshold value of 0.184, which was reduced and compressed due to the complex fire environment in the pre-calculation, was retrieved and summed with the extracted dynamic bias of 0.132, resulting in a tolerance boundary sum of 0.316. Dividing 0.350 by 0.316 finally yielded a gain gradient value of 1.107. This operational logic of dividing the sudden deviation by the constrained tolerance boundary allows the sensitivity of parameter compensation to self-adjust with the environment. If, under the same signal drop amplitude of 0.350, there is no strong updraft in the environment, the asymmetric dynamic bias... The gain gradient will drop to near the static noise floor of 0.02, and the sum of the denominators will become 0.204. At this point, the output gain gradient will surge to 1.715, triggering a deeper spread factor boost more quickly. Conversely, under the current 0.132 bias buffer with strong thermal disturbances, the calculated gradient value is suppressed to 1.107, stabilizing in the moderate boost range and avoiding overcompensation for transient thermal flicker. This calculation process effectively solves the misalignment problem that single threshold triggering mechanisms are prone to when facing complex physical environments. By associating the numerator and denominator with different physical channel characteristics, it achieves the convergence of multi-dimensional channel information to one-dimensional decision commands. This mechanism transforms the energy fluctuations of the physical layer into clear digital logic judgment criteria, ensuring the accuracy of spread spectrum configuration issuance actions and maintaining the efficient operation of the underlying communication protocol and stable synchronization between the transmitting and receiving ends without generating redundant control signaling overhead.

[0085] like Figure 5 As shown, in one embodiment, the step of obtaining the spreading parameter value and performing the switching in S4 is broken down into the following sub-steps: S41. Obtain the boost level label based on the gain gradient value and the preset gradient range, and generate the spread spectrum parameter value; S42. Perform structural segmentation based on the spread spectrum parameter values ​​to obtain wake-up micropulses and configuration data packets; S43. Solidify and transmit the wake-up micropulse according to the preset maximum spread spectrum value; that is, through the underlying hardware physical connection, configure the synchronization control channel to transmit the wake-up micropulse at the maximum allowed spread spectrum value. S44. Based on the captured wake-up pulse and the unverified configuration data packet, trigger the parallel blind detection mechanism; this mechanism is triggered when the receiving end captures the wake-up pulse through the synchronization control channel but the configuration data packet fails to pass verification due to interference. S45. According to the parallel blind detection system, multiple parallel correlators are activated in the next physical frame; the baseband chip at the receiving end simultaneously activates multiple parallel correlators corresponding to the local sequences of different enhancement levels. S46. Extract peak-locking parameters from the parallel correlator to switch the spreading factor value and coding rate parameter. By extracting the channel parameter with the largest correlation output peak as the peak-locking parameter, the parameter switching of the main service channel is passively completed.

[0086] In this embodiment, it should be noted that in S41, based on the calculated gain gradient value, S41 is responsible for converting the continuous risk assessment into a discrete control command and generating specific spread spectrum parameter values. The airborne microprocessor's storage unit is pre-divided into multiple continuous and non-overlapping gradient value intervals, each interval corresponding to a defined spread spectrum factor enhancement level. For example, [0, 1.0) is set as the maintain-state interval, [1.0, 2.0) as the first-level spread spectrum factor enhancement level interval, and [2.0, 3.0) as the second-level spread spectrum factor enhancement level interval. The boundary values ​​of the gradient interval were determined based on the correlation analysis between the gain gradient value and the optimal spreading factor in historical data. Historical data of the UAV under 20 different channel quality environments were collected, the gain gradient value at each moment was calculated, and the optimal spreading factor required to make the link bit error rate lower than the target value was recorded. Cluster analysis was used to divide the gradient value into three natural clusters with cluster centers of 0.5, 1.5, and 2.5, respectively. The median value of the adjacent clusters was taken as the boundary, resulting in [0, 1.0) corresponding to maintaining the original state, [1.0, 2.0) corresponding to improving by one level, and [2.0, 3.0) corresponding to improving by two levels. After 10 verification data tests, this interval division made the accuracy of parameter switching reach more than 90%.

[0087] The currently acquired gain gradient value of 1.107 is logically compared with these preset intervals to identify that 1.107 falls within the moderate enhancement interval of [1.0, 2.0), thus matching the first-level spreading factor enhancement level label. Based on this level label, the spreading parameter value for switching the spreading factor of the current main service channel from the original SF7 to SF9 is generated, and the corresponding coding rate is adjusted synchronously.

[0088] The computational logic employing interval matching rather than continuous mapping is because the spreading factor of the physical layer hardware presents as a discrete integer step. If the spreading factor is frequently adjusted minutely with the gradient value, repeated oscillations during parameter switching at the interval boundaries can easily occur, leading to significant signaling overhead and hardware reconfiguration power consumption. By pre-setting a gradient interval with a certain span, risk fluctuations within a certain range can be absorbed into a stable configuration state, improving the stability of the airborne RF hardware during parameter reconstruction and the effectiveness of command issuance.

[0089] In S42, after generating the spreading parameter values ​​for the receiver, the configuration command is structurally segmented at the physical layer, extracting two independent data payloads: a wake-up pulse and a configuration data packet. Normally, adaptive communication packages information such as the target spreading factor, coding rate, and execution timestamp into a single data frame. However, in a forest fire environment, the service channel may already be experiencing a high bit error rate. If a large, long data packet is sent, and some bits are corrupted by an electromagnetic pulse, the entire packet will be discarded by the receiver due to checksum failure. Therefore, the command is decomposed, extracting a 4-byte wake-up pulse, which does not contain specific configuration information and serves only as a physical credential to trigger the underlying logic; and simultaneously extracting a 12-byte configuration data packet to carry the target parameters, including SF9, and the checksum. The lengths of the wake-up micropulse and the configuration data packet were determined based on statistical analysis of the data packet transmission success rate in historical channels. Fifty communication history data of the UAV in a strong electromagnetic interference environment in a fire scene were collected, and the transmission success rate of data packets of different lengths was statistically analyzed. It was found that the success rate of data packets with a length of less than or equal to 4 bytes was 96%, the success rate of data packets with a length of 12 bytes was 78%, and the success rate of data packets with a length of more than 20 bytes dropped to below 50%. Therefore, the wake-up micropulse was fixed to 4 bytes to ensure a high penetration probability, and the configuration data packet was set to 12 bytes to achieve a balance between success rate and information carrying capacity. This length combination was verified in 5 fire scenes and could improve the triggering success rate of the parallel blind detection mechanism to over 92%.

[0090] This split-structure logic constructs an asynchronous mechanism at the physical layer of the communication protocol. It addresses the problem that even under severe channel conditions caused by a fire, where longer data packets may struggle to pass through completely, extremely short wake-up pulses, due to their very small duty cycle on the time axis, can penetrate transient interference with a high probability and reach the receiver. This preserves the underlying triggering conditions for maintaining dual-end synchronization even when command transmission is impaired.

[0091] In S43, a special underlying channel mechanism is invoked to ensure priority delivery of the wake-up micropulse for the segmented data payload. The onboard microprocessor no longer relies on the main service data channel to transmit these critical instructions; instead, it directly invokes an independent synchronization control channel isolated at the physical layer hardwire level. According to preset rules, this synchronization control channel is fixed to the maximum allowed spreading value (e.g., SF12) mode to transmit the aforementioned 4-byte wake-up micropulse. In this process, the microprocessor does not depend on the current fading of the main service channel; instead, it forces the RF front-end to sacrifice transmission rate, utilizing the greater processing gain and interference immunity margin provided by SF12 to push the micropulse into the channel.

[0092] The operational logic of employing an independent control channel and fixing the maximum spreading value aims to decouple the service data flow from the control signaling flow at the physical level. In existing frequency band in-band control schemes, once the service channel is interrupted due to vegetation obstruction, the control commands used to salvage the channel are also lost. This step establishes a robust independent transmission path, enabling the establishment of signaling connections even in dense smoke and strong fading. Although SF12 has a low transmission rate, for a micropulse of only 4 bytes, its transmission time is still controlled within milliseconds, ensuring the connectivity of the configuration command pre-triggered signal without affecting response timeliness.

[0093] S44 defines the passive fault-tolerant response logic for the receiver when encountering extreme fire interference. In a real-world scenario, after transmitting the wake-up micropulse, the transmitter immediately transmits a 12-byte configuration data packet. Assume that during this period, a sudden canopy explosion occurs, generating transient electromagnetic shielding that covers the receiving antenna. The receiver's underlying RF circuitry, leveraging the high penetration of SF12, captures the initial wake-up micropulse, but the subsequent configuration data packet experiences numerous bit errors during demodulation and fails cyclic redundancy check. At this point, the configuration data packet is deemed invalid. Under existing protocol frameworks, the receiver, lacking the target configuration parameters, can only remain on the existing SF7 spreading factor, waiting for retransmission from the transmitter. However, this microprocessor, having recorded the arrival event of the wake-up micropulse, clearly senses that the transmitter is about to perform a parameter switch.

[0094] Given this contradictory state of capturing micro-pulses but failing to verify the configuration data packets, the receiving end no longer initiates time-consuming retransmission requests. Instead, it directly triggers a parallel blind detection mechanism at the underlying layer. This mechanism breaks the absolute dependence on command integrity in conventional communication and solves the physical layer deadlock problem caused by handshake failures when channel quality deteriorates. It allows the receiving end to retain the ability to synchronize through underlying probes even in environments lacking specific control parameters, maintaining the continuity of communication link evolution under harsh conditions.

[0095] In S45, once the parallel blind detection mechanism is triggered, the underlying response of the receiver's baseband chip is defined at the microsecond level. When the receiver confirms entry into blind detection mode, its internal hardware clock aligns it to the start of the next physical frame. At this point, the receiver's baseband processing chip is no longer limited to demodulating a single spread spectrum sequence, but simultaneously activates multiple local despreading correlators corresponding to different boost levels. Depending on the application scenario, the receiver's hardware resource pool will open three despreading channels in parallel, loading local pseudo-random despreading sequences corresponding to SF7 (maintain original state), SF9 (first-level boost), and SF11 (second-level boost), respectively. These three correlators will perform parallel correlation integration operations on the same segment of baseband digital sampling stream from the RF front-end within the same reception time window.

[0096] This low-level multi-path parallel computing logic sacrifices the instantaneous computing power of the receiver chip to achieve synchronization continuity in the time domain. It avoids the data frame loss and processing delays that can occur with serial polling. By simultaneously performing parallel verification of potential configuration states at the hardware level, the parameter synchronization process is moved from high-level protocol interaction to low-level physical energy integration comparison, enabling the receiver to decode unknown transmission parameters.

[0097] In S46, the passive reconstruction of service channel parameters is completed by extracting the physical results of parallel computation. After the parallel correlator has run for a set integration period (e.g., 20 milliseconds), the correlation peak energy of each channel is read in real time. According to field data feedback, the correlation peak value corresponding to the SF7 channel is only 15, the peak value corresponding to the SF11 channel is 42, while the correlation channel loaded with the SF9 local sequence has an output peak value of 158, exceeding the preset locking threshold of 80. Based on this clear difference in underlying physical energy, the decision module of the receiver extracts the correlation parameters of the SF9 channel as the peak locking parameters of the current physical frame. Since the peak locking parameters represent the despreading mode with higher energy and stronger matching degree that actually exists in the current channel, the receiver microprocessor then fixes the demodulation register of the main service data channel to SF9 and the corresponding coding rate parameters. Thus, the closed loop of the process from parameter generation, instruction segmentation, asynchronous transmission to blind detection recovery is completed.

[0098] This switching logic, which requires no retransmission from the sending end and no confirmation from the receiving end, avoids the synchronization risks caused by poor uplink quality or two-way handshake delays in remote fire communication. Throughout the process, even with the loss of configuration data payload, it maintains continuous backhaul of airborne detection data and reliable physical connectivity of the communication link under extreme forest fire detection environments.

[0099] Also provided is an airborne spread spectrum device for remote communication of forest fires using an unmanned aerial vehicle (UAV). The device is configured to implement an airborne spread spectrum method for remote communication of forest fires using an UAV. The device includes: The feature sensing block is used to acquire the composite perturbation quantity and, based on the composite perturbation quantity and the dynamic observation window, to acquire the historical distribution and channel volatility; the composite perturbation quantity includes signal strength value, signal-to-noise ratio parameter, and asymmetric jitter; The dimension reduction block is used to obtain the dynamic variance value based on the historical distribution, and to perform nonlinear weight reduction based on the dynamic variance value and the response limit value to obtain the effective dimension number; the coupling attenuation rate is obtained based on the effective dimension number, and the convergence threshold value is obtained based on the coupling attenuation rate and the channel volatility. The gradient adjustment block is used to obtain the dynamic bias based on the convergence threshold and asymmetric features, and to obtain the gain gradient based on the dynamic bias and the composite perturbation. The blind detection execution block is used to obtain the spread spectrum parameter value based on the gain gradient value, divide the spread spectrum parameter value into wake-up micropulse and configuration data packet; based on the wake-up micropulse and configuration data packet, the parallel blind detection system is triggered through the synchronization control channel to complete the switching of the spread spectrum factor value and coding rate parameter.

[0100] To enable those skilled in the art to fully understand and implement the technical solutions described in this specification, the following section, in conjunction with a specific application scenario, provides a detailed deduction and data analysis of the entire implementation principle of this UAV-borne spread spectrum device for remote communication of forest fires.

[0101] When a drone was conducting fire detection missions deep into the core area of ​​a coniferous forest fire at an altitude of 1500 meters and accompanied by gale-force winds of level 8, the airborne communication link faced extreme physical distortion. The device first established a dynamic observation window spanning 500 milliseconds and acquired the composite disturbance of 100 sampling points at a sampling rate of 200 Hz. When extracting signal strength values, the drone entered an area where dense smoke and tree canopy overlapped, causing the original physical extreme value to drop sharply from -65 dBm to -95 dBm. The device then performed linear translation based on a pre-set lower reception limit of -120 dBm and a maximum saturation value of -30 dBm. Taking the 99th sampling point as an example, its non-negative normalized value is mapped to (-65-(-120)) / (-30-(-120))=0.611, while the value of the 100th sampling point is mapped to (-95-(-120)) / 90=0.277, thus constructing a non-negative normalized column in the interval between 0 and 1, eliminating the polarity interference of negative power values ​​in subsequent calculations.

[0102] After acquiring the historical distribution, the device initiates the channel volatility calculation logic. The device extracts the adjacent sampling difference between the 99th and 100th sampling points as |0.277 - 0.611| = 0.334. Simultaneously, the device sets the zero-prevention constant to 0.001 and the smoothing ratio threshold to 1.5. The dynamic smoothing basis corresponding to the 99th sampling point is calculated as max(0.611, 0.001) × 1.5 = 0.916. The volatility contribution ratio at this point is 0.334 / 0.916 = 0.364. The device accumulates and averages the 99 ratios generated from the 100 sampling points, obtaining a base volatility mean of 0.125. At this point, the device extracts the statistical characteristics of the non-negative normalized column, obtaining its standard deviation of 0.085, which is divided by the theoretical full-scale span of 1.0 as a preset compensation term of 0.085. Finally, the channel volatility corresponding to this composite disturbance is calculated to be 0.125 + 0.085 = 0.210. This value objectively reflects the inherent decay intensity caused by the obstruction of tree trunks in the fire area.

[0103] As the drone approached the fire line, the device monitored three characteristic dimensions in the historical distribution. The device detected a dynamic variance as high as 45 in the historical distribution of the signal-to-noise ratio (SNR) parameter, but its characteristic change rate reached 50V / µs. The device immediately compared this with the 15V / µs response limit of the onboard operational amplifier, determining that the sudden change originated from transient man-made electromagnetic interference generated when the fire pump below started. The device applied a negative penalty to the SNR parameter dimension and removed it from the computation queue, resulting in a true effective dimension number of 2 (i.e., signal strength value and asymmetric jitter). The device then added 1 to the effective dimension number 2 and took the natural logarithm, obtaining 1.098, and set the attenuation coefficient of the weighted bias combination to 0.4 and the bias constant to 1.0. The device calculated the coupling attenuation rate as 1.098 / (0.4×2+1.0)=0.610. The device further multiplies the coupling attenuation rate of 0.610 by the negative of the channel volatility of 0.210 to obtain -0.128, raises this to the power of the natural constant e to obtain 0.879, and finally multiplies 0.879 by the channel volatility of 0.210 to obtain a convergence threshold of 0.184. Compared to the original 0.210, the reduction in the threshold value reflects the device's adaptive adjustment of its tolerance to the complex environment of the fire scene.

[0104] After the threshold was set, the drone encountered the thermal convection characteristic of a fire. The strong updraft caused a change in the microwave refractive index, and the device extracted a sharp increase in the positive Doppler integral value to 280Hz within the current physical frame, while the negative Doppler integral value was only 40Hz, exhibiting a strong asymmetry. The device was set to a static noise floor of 0.02, a thermal flux sensitivity of 0.15, and a compensation positive value of 0.1. The device substituted the asymmetry parameters into the calculation: |(280-40) / (280+40+0.1)|=240 / 320.1=0.749. The device multiplied 0.749 by the thermal flux sensitivity of 0.15 to obtain 0.112, and added the static noise floor of 0.02, finally obtaining a dynamic bias of 0.132. This bias injection provides a physical anti-shake buffer for the device in severe thermal turbulence.

[0105] Subsequently, the device performs an overall assessment of the current communication status. The device calculates the historical average signal strength value within the dynamic observation window to be 0.550, while the current instantaneous non-negative normalized value drops to 0.200. The device then calculates the absolute deviation as |0.200 - 0.550| = 0.350. The device divides the absolute deviation of 0.350 by the sum of the convergence threshold of 0.184 and the dynamic bias of 0.132, resulting in a gain gradient of 1.107. Without the buffer of the dynamic bias, this gradient would reach 1.902 and potentially trigger overcompensation. The calculated result of 1.107 allows the device to stably and accurately quantify the true degradation risk of the current link under the combined effects of thermal flicker and canopy shading.

[0106] The device compares the gain gradient value of 1.107 with the medium boost range of [1.0, 2.0] set in the onboard memory to obtain the first-level spread factor boost level label, and generates the spread spectrum parameter value to switch the spread spectrum factor from SF7 to SF9. The device then performs structural segmentation on the spread spectrum parameter value, extracting a 4-byte wake-up micropulse and a 12-byte configuration data packet. The device directly calls the synchronization control channel, which is fixed to the maximum spread spectrum factor SF12, through the physical layer hardwire, penetrates the dense smoke and high temperature of the fire scene to transmit the wake-up micropulse, and transmits the configuration data packet 5 milliseconds later. During this process, a sudden canopy explosion generated full-band electromagnetic shielding, causing the receiver to fail to fully verify the 12-byte configuration data packet.

[0107] Because the receiver's underlying RF circuitry successfully captured the highly penetrating wake-up micropulse under the SF12 spreading factor, the device directly triggered parallel blind detection without obtaining the configuration data packet. At the start of the next physical frame, the receiver's baseband chip simultaneously activated three parallel correlators corresponding to SF7, SF9, and SF11, respectively. After 20 milliseconds of parallel despreading, the correlation peak of the SF7 channel was only 15, the SF11 channel's peak was 42, while the SF9 channel's correlation peak reached 158, far exceeding the set locking threshold of 80. The receiver directly extracted SF9 as the peak locking parameter, passively switching the primary service channel to the new spreading factor value and coding rate parameters. Throughout the entire process, even with the loss of the configuration data payload, continuous transmission of 2kbps critical telemetry data was still ensured, maintaining the smooth operation of the fire detection link.

[0108] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solution of the present invention, and these simple modifications all fall within the protection scope of the present invention.

[0109] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.

[0110] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed by the present invention.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A UAV-borne spread spectrum method for long-range communication in forest fires, characterized in that, The methods include: Obtain the composite disturbance quantity; The composite perturbation includes signal strength value, signal-to-noise ratio parameter, and asymmetric jitter; Based on the composite disturbance and dynamic observation window, the historical distribution and channel volatility are obtained; The dynamic variance value is obtained based on the historical distribution, and nonlinear weighting is performed based on the dynamic variance value and the response limit value to obtain the effective number of dimensions; The coupling attenuation rate is obtained based on the effective number of dimensions, and the convergence threshold is obtained based on the coupling attenuation rate and the channel volatility. The dynamic bias is obtained based on the convergence threshold and asymmetric features, and the gain gradient is obtained based on the dynamic bias and the composite perturbation. The spread spectrum parameter value is obtained based on the gain gradient value, and the spread spectrum parameter value is divided into wake-up micropulse and configuration data packet; Based on the wake-up micropulse and the configuration data packet, a parallel blind detection system is triggered through the synchronization control channel to complete the switching of the spreading factor value and the coding rate parameter.

2. The UAV-borne spread spectrum method for remote communication of forest fires according to claim 1, characterized in that, The step of obtaining the historical distribution and channel volatility based on the composite disturbance and dynamic observation window includes: The dynamic observation window is defined based on the start point of the current communication frame; Physical extrema are extracted based on the dynamic observation window, and a non-negative normalized column is constructed using linear translation. The historical distribution column and the channel volatility are obtained based on the non-negative normalized column.

3. The UAV-borne spread spectrum method for remote communication of forest fires according to claim 2, characterized in that, The step of obtaining the historical distribution column and the channel volatility based on the non-negative normalized column includes: The adjacent sample difference is obtained based on the non-negative normalized column; A dynamic smoothing base is obtained based on the non-negative normalization column and the preset zero-prevention constant; The channel volatility is obtained based on the adjacent sampling difference, the dynamic smoothing base, and the preset compensation term.

4. The UAV-borne spread spectrum method for remote communication of forest fires according to claim 1, characterized in that, The step of performing nonlinear weighting based on the dynamic variance value and the response limit value to obtain the effective number of dimensions includes: The characteristic change rate is obtained based on the historical distribution series; State comparison is performed based on the characteristic rate of change and the response limit value; A negative penalty value is applied based on the comparison results to eliminate human interference dimensions and obtain the effective number of dimensions.

5. The UAV-borne spread spectrum method for remote communication of forest fires according to claim 4, characterized in that, The step of obtaining the coupling attenuation rate based on the effective dimension number and obtaining the convergence threshold based on the coupling attenuation rate and the channel volatility includes: The coupling attenuation rate is obtained by performing logarithmic operations and linear combinations based on the effective dimension number. The convergence threshold is obtained by exponentially calculating the coupling attenuation rate and the channel volatility.

6. The UAV-borne spread spectrum method for remote communication of forest fires according to claim 1, characterized in that, The step of obtaining the dynamic bias based on the convergence threshold and asymmetric features includes: The asymmetric features are obtained by extracting positive and negative Doppler based on the upward force of the heat flow; The dynamic bias is obtained based on the asymmetric characteristics, heat flux sensitivity, and static noise floor value.

7. The UAV-borne spread spectrum method for remote communication in forest fires according to claim 6, characterized in that, The step of obtaining the gain gradient value based on the dynamic bias and the composite perturbation includes: The absolute deviation value is obtained based on the non-negative normalized series of the composite disturbance at the current moment and its historical mean. The gain gradient value is obtained based on the absolute deviation value, the convergence threshold value, and the dynamic bias.

8. The UAV-borne spread spectrum method for remote communication of forest fires according to claim 1, characterized in that, The step of obtaining the spreading parameter value based on the gain gradient value and dividing the spreading parameter value into a wake-up micropulse and a configuration data packet includes: The gain level is obtained based on the gain gradient value and the preset gradient range, and the spread spectrum parameter value is generated. Structural segmentation is performed based on the spread spectrum parameter values ​​to obtain the wake-up micropulse and the configuration data packet; The wake-up micropulse is solidified and emitted according to the preset maximum spread spectrum value.

9. The UAV-borne spread spectrum method for remote communication of forest fires according to claim 8, characterized in that, The step of triggering parallel blind detection via a synchronization control channel based on the wake-up micropulse and the configuration data packet to switch the spreading factor value and coding rate parameter includes: The parallel blind detection mechanism is triggered based on the captured wake-up micro-pulse and the unverified configuration data packet; According to the parallel blind detection mechanism, multiple parallel correlators are activated in the next physical frame; The peak locking parameter is extracted based on the parallel correlator, and the switching between the spreading factor value and the coding rate parameter is completed.

10. An airborne spread spectrum device for remote communication in forest fires, characterized in that, The apparatus is configured to implement an unmanned aerial vehicle (UAV) airborne spread spectrum method for remote communication of forest fires as described in any one of claims 1 to 9, the apparatus comprising: A feature sensing block is used to acquire composite perturbation quantities and, based on the composite perturbation quantities and a dynamic observation window, to acquire historical distribution patterns and channel volatility; the composite perturbation quantities include signal strength values, signal-to-noise ratio parameters, and asymmetric jitter; The dimension reduction block is used to obtain the dynamic variance value based on the historical distribution column, and to perform nonlinear weight reduction based on the dynamic variance value and the response limit value to obtain the effective number of dimensions; to obtain the coupling attenuation rate based on the effective number of dimensions, and to obtain the convergence threshold value based on the coupling attenuation rate and the channel volatility. A gradient adjustment block is used to obtain a dynamic bias amount based on the convergence threshold and asymmetric features, and to obtain a gain gradient value based on the dynamic bias amount and the composite perturbation amount. The blind detection execution block is used to obtain the spreading parameter value based on the gain gradient value, divide the spreading parameter value into a wake-up micropulse and a configuration data packet, and trigger the parallel blind detection mode through the synchronization control channel based on the wake-up micropulse and the configuration data packet to complete the switching of the spreading factor value and the coding rate parameter.

Images