Sensing data remote transmission effect optimization method and system based on adaptive filtering

Abstract

The invention discloses a sensing data remote transmission effect optimization method and system based on adaptive filtering, and relates to the technical field of sensing data processing. The method comprises the following steps: acquiring transmission distance sequences at historical recent N moments, and a scene type and a motion mode at the current moment, carrying out standardization processing on transmission distance data, and binding the transmission distance data with the scene type and the motion mode to form a standard feature vector; inputting the transmission distance into a pre-trained transmission distance prediction model, and outputting a basic prediction transmission distance in a future preset time period; combining prediction deviation data of historical similar scenes and standard distance parameter correction of the current scene to obtain a prediction transmission distance; inputting the multi-path time delay extension into a multi-path time delay extension prediction model, and outputting predicted multi-path time delay extension of a future preset time period; and dynamically configuring an adaptive equalization filter order, a transmission code element period and a modulation mode of adaptive filtering so as to realize remote transmission of sensing data. According to the invention, the accuracy and reliability of remote transmission of sensing data are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of sensor data processing technology, specifically to a method and system for optimizing the remote transmission effect of sensor data based on adaptive filtering. Background Technology

[0002] With the widespread application of wireless sensor networks in remote monitoring and control, the high reliability of sensor data transmission has become crucial. In complex environments, signal transmission is susceptible to multipath interference. Adaptive filtering technology can dynamically adjust parameters to compensate for channel distortion, making it an effective anti-interference method.

[0003] However, existing adaptive filtering techniques mostly adjust parameters based on current or historical channel conditions, lacking foresight. Since dynamic changes in the transmission distance of sensor nodes directly increase multipath delay spread, leading to inter-symbol interference and signal distortion, traditional methods struggle to suppress the resulting transmission errors in a timely manner, thus limiting the accuracy and reliability of data transmission. Summary of the Invention

[0004] This application provides a method and system for optimizing the remote transmission effect of sensor data based on adaptive filtering, aiming to solve the technical problem of insufficient accuracy and reliability of data transmission in the prior art.

[0005] In view of the above problems, this application provides a method and system for optimizing the remote transmission effect of sensor data based on adaptive filtering.

[0006] Firstly, this application provides a method for optimizing the remote transmission effect of sensor data based on adaptive filtering, including: Collect the transmission distance sequence of the most recent N time moments, the scene type and motion mode of the current time moment, and after standardizing the transmission distance data, bind it with the scene type and motion mode to form a standard feature vector; The standard feature vector is input into a pre-trained transmission distance prediction model to predict the basic transmission distance for a future preset time period. Based on the prediction deviation data of similar historical scenarios and the standard distance parameters of the current scenario, the basic predicted transmission distance is corrected to obtain the predicted transmission distance. The predicted transmission distance is input into the pre-trained multipath delay spread prediction model, and the predicted multipath delay spread for a future preset time period is predicted and output. Based on the predicted multipath delay spread, the order of the adaptive equalization filter, the transmission symbol period, and the modulation scheme of the adaptive filter are dynamically configured to remotely transmit the sensor data.

[0007] Secondly, this application provides a system for optimizing the remote transmission effect of sensor data based on adaptive filtering, including: The feature vector generation module is used to collect the transmission distance sequence of the most recent N time moments, the scene type and motion mode at the current time, and after standardizing the transmission distance data, it is bound with the scene type and motion mode to form a standard feature vector. The basic distance prediction module is used to input the standard feature vector into the pre-trained transmission distance prediction model and predict the basic transmission distance for a future preset time period. The prediction distance correction module is used to correct the basic predicted transmission distance based on the prediction deviation data of similar historical scenarios and the standard distance parameters of the current scenario, so as to obtain the predicted transmission distance. The multipath delay prediction module is used to input the predicted transmission distance into a pre-trained multipath delay extension prediction model and predict the multipath delay extension for a future preset time period. The parameter configuration transmission module is used to dynamically configure the order of the adaptive equalization filter, the transmission symbol period, and the modulation method of the adaptive filter based on the predicted multipath delay spread, and to remotely transmit the sensor data.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application provides a method and system for optimizing the remote transmission effect of sensor data based on adaptive filtering. By predicting the transmission distance and simultaneously predicting the multipath delay spread, the adaptive filtering parameters are accurately configured in advance, enabling the filter to adapt to changes in the future multipath environment. This effectively compensates for signal distortion and inter-symbol interference caused by the increase in transmission distance, thereby effectively reducing the transmission bit error rate and improving the overall anti-interference capability, accuracy, and reliability of remote sensor data transmission. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 A flowchart illustrating the method for optimizing the remote transmission effect of sensor data based on adaptive filtering, provided in an embodiment of this application. Figure 2 A schematic diagram of the structure of the sensor data remote transmission effect optimization system based on adaptive filtering provided in the embodiments of this application; The components represented by each number in the attached diagram are explained below: Feature vector generation module 11, basic distance prediction module 12, predicted distance correction module 13, multipath delay prediction module 14, parameter configuration transmission module 15. Detailed Implementation

[0011] This application provides a method and system for optimizing the remote transmission effect of sensor data based on adaptive filtering, which is used to address the technical problem of insufficient accuracy and reliability of data transmission in the prior art.

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0013] It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to these processes, methods, products, or devices.

[0014] Example 1, as Figure 1 As shown, this application provides a method for optimizing the remote transmission effect of sensor data based on adaptive filtering, the method comprising: S100: Collects the transmission distance sequence of the most recent N time moments, the scene type and motion mode of the current time moment, and after standardizing the transmission distance data, binds it with the scene type and motion mode to form a standard feature vector.

[0015] In this embodiment, the transmission distance sequence of the most recent N moments, the scene type and motion mode at the current moment are collected. After standardizing the transmission distance data, it is bound to the scene type and motion mode to form a standard feature vector. In remote transmission of sensor data, the intensity of multipath interference is directly related to the transmission distance, scene environment, and device motion state. Dynamic changes in transmission distance will change the propagation path of multipath signals, the distribution of obstacles in different scenes will affect the number of multipath components, and the motion mode will determine the stability of distance changes. If only a single data source is used, it is impossible to fully characterize the key factors affecting the multipath environment, and the subsequent prediction model will have insufficient accuracy due to incomplete input features. Therefore, it is necessary to collect three types of data: historical transmission distance, scene type, and motion mode, and unify the data dimensions through standardization to form a standard feature vector that can accurately reflect the current transmission environment, providing a reliable input basis for subsequent transmission distance prediction and multipath delay spread prediction.

[0016] Step S100 in the method provided in this application embodiment includes: Historical transmission distance data is collected at a preset sampling interval, and the transmission distance data of the N most recent moments in history is extracted to form a transmission distance sequence; Environmental features are collected using environmental sensors to identify the current scene type; The velocity change rate is calculated based on the transmission distance sequence, and the motion mode is determined according to the velocity change rate, wherein the motion mode includes stationary, uniform, and variable speed. The transmission distance sequence is normalized to obtain a standard transmission distance sequence; The standard transmission distance sequence, scene type, and motion mode are concatenated in sequence to form a standard feature vector with dimension N+2.

[0017] First, historical transmission distance data is collected at preset sampling intervals, and the transmission distance data from the N most recent moments is extracted to form a transmission distance sequence. Historical transmission distance data between the sensor node and the receiver is collected at preset sampling intervals, and the data from the N most recent moments is extracted from the collected historical data to form the transmission distance sequence. The preset sampling interval refers to a pre-set time interval for periodically collecting transmission distance data, such as 1 second / time or 5 seconds / time, to ensure the regularity of data collection. The transmission distance sequence is an array formed by arranging the transmission distance data from multiple moments in chronological order, reflecting the trend of distance change over time. For example, in a scenario where temperature sensor nodes transmit data to the central control system in an industrial workshop, a preset sampling interval of 2 seconds is set, N=5, meaning data from the 5 most recent moments is extracted. The current time is t. From the historical data collection, the transmission distances of the sensor node at t-8 seconds, t-6 seconds, t-4 seconds, t-2 seconds, and t seconds are extracted, which are 12 meters, 12.1 meters, 12.2 meters, 12.1 meters, and 12 meters respectively, forming the transmission distance sequence: [12, 12.1, 12.2, 12.1, 12].

[0018] Secondly, environmental features are collected through environmental sensors to identify the current scene type. Environmental sensors are deployed at sensing nodes or receiving ends to collect the physical characteristics of the current environment, such as obstacle density, light intensity, and electromagnetic interference intensity. The current scene type is determined using preset scene recognition rules. Environmental sensors are devices that can detect physical or chemical parameters of the environment, such as infrared obstacle sensors, electromagnetic interference detectors, and light sensors. Scene type refers to the category of transmission scene classified according to environmental characteristics, such as industrial workshops, outdoor open fields, indoor offices, and underground parking garages. The distribution of obstacles and the electromagnetic environment differ in different scenes. For example, if an infrared obstacle sensor and an electromagnetic interference detector are deployed at a sensing node in an industrial workshop, with an obstacle density of 3 obstacles per 10 square meters and an electromagnetic interference intensity of 80 dB, the scene type is identified as an industrial workshop based on preset rules: obstacle density ≥ 2 obstacles per 10 square meters and electromagnetic interference intensity ≥ 60 dB.

[0019] Furthermore, the velocity change rate is calculated based on the transmission distance sequence, and the motion mode is determined according to the velocity change rate. The motion modes include stationary, uniform, and variable speed. The instantaneous velocity within each time interval is calculated based on the distance difference between adjacent moments in the transmission distance sequence and the sampling interval. Then, the ratio of the difference between adjacent instantaneous velocities to the sampling interval is calculated, i.e., the velocity change rate. The motion mode is classified according to the magnitude of the velocity change rate, including stationary, uniform, and variable speed. The velocity change rate reflects the degree of velocity change over time, and the calculation formula is: Velocity change rate = (Instantaneous velocity at the next moment - Instantaneous velocity at the previous moment) / Sampling interval. The motion mode refers to the category classified according to the movement state of the sensing node: stationary means the velocity is close to 0 and the rate of change is 0; uniform speed means the velocity is stable and the absolute value of the rate of change is ≤ a preset threshold, such as 0.1 m / s. 2 ; Variable speed refers to the absolute value of the rate of change of speed being greater than a preset threshold.

[0020] For example, based on the transmission distance sequence [12, 12.1, 12.2, 12.1, 12] and the sampling interval of 2 seconds, the instantaneous velocity within each time interval is calculated: the instantaneous velocity from t-8 seconds to t-6 seconds = (12.1-12) / 2 = 0.05 m / s; similarly, the instantaneous velocities for the other time intervals are 0.05 m / s, -0.05 m / s, and -0.05 m / s, respectively. Then, the ratio of the difference between adjacent instantaneous velocities to the sampling interval is calculated, yielding a velocity change rate of 0 m / s. 2 -0.05 m / s 2 0 m / s 2 The preset threshold for the rate of change of velocity is 0.1 m / s. 2 If the absolute value of all rates of change is ≤0.1 and the absolute value of the velocity is stable at around 0.05 m / s, the motion mode is determined to be uniform.

[0021] Further, the transmission distance sequence is normalized to obtain a standard transmission distance sequence. A data normalization method is used to transform each data point in the transmission distance sequence to a unified interval of [0,1], eliminating interference from data of different magnitudes on subsequent models. Normalization is a data preprocessing technique that maps the original data to a fixed interval through linear transformation. Standard transmission distance = (original value - sequence minimum value) / (sequence maximum value - sequence minimum value). For example, the transmission distance sequence is [12,12.1,12.2,12.1,12], where the sequence minimum value = 12 and the maximum value = 12.2. For the transmission distance 12, normalization yields a standard transmission distance = (12-12) / (12.2-12) = 0. Similarly, the normalized values ​​for the other transmission distances are 0.5, 1, 0.5, and 0, respectively, resulting in the standard transmission distance sequence: [0,0.5,1,0.5,0].

[0022] Finally, the standard transmission distance sequence, scene type, and motion mode are concatenated in sequence to form a standard feature vector with dimension N+2. The scene type and motion mode are converted into numerical values ​​according to a preset encoding rule, and then concatenated in the order of standard transmission distance sequence, scene type encoding, and motion mode encoding to form a standard feature vector with a total dimension of N+2. The standard feature vector is a numerical array containing multiple dimensions of features, each dimension corresponding to a key input feature, which can be directly input into the subsequent pre-trained model. For example, if N=5, the standard transmission distance sequence is [0,0.5,1,0.5,0]; according to the preset encoding rule, the industrial workshop scene type is converted to the numerical value 1, and uniform speed is converted to the numerical value 1; the three are combined in sequence to form a standard feature vector with dimension 5+2=7: [0,0.5,1,0.5,0,1,1].

[0023] In this embodiment, by simultaneously collecting three types of data—transmission distance sequence, scene type, and motion mode—the key factors affecting multipath environments are comprehensively covered, avoiding subsequent prediction biases due to missing features. Simultaneously, normalization eliminates differences in the magnitude of distance data, and encoding conversion transforms non-numerical features into numerical values ​​recognizable by the model, ensuring that the feature vectors can adapt to subsequent pre-trained models. The resulting standard feature vectors provide reliable support for the accuracy of transmission distance prediction and multipath delay spread prediction, indirectly laying the foundation for subsequent adaptive filtering parameter optimization and improving the anti-interference capability and reliability of remote transmission of sensor data.

[0024] S200: Input the standard feature vector into the pre-trained transmission distance prediction model to predict and output the basic predicted transmission distance for a future preset time period.

[0025] In this embodiment, the standard feature vector is input into a pre-trained transmission distance prediction model to predict the basic transmission distance for a preset future time period. An LSTM network is used to construct the transmission distance prediction model. Through supervised training using historical standard feature vectors and corresponding real future distances, the model acquires temporal prediction capabilities. The standard feature vector output from S100 is then input into the trained model to obtain the basic predicted transmission distance for the preset future time period, providing crucial information for subsequent distance correction and multipath delay propagation prediction.

[0026] Step S200 in the method provided in this application embodiment includes: The process of constructing the transmission distance prediction model includes: Multiple historical standard feature vectors are collected from historical data to form a sample feature sequence set; Obtain the actual transmission distance for each historical standard feature vector in the sample feature sequence set corresponding to a future preset time period, and label multiple such actual transmission distances as a sample transmission distance set; Constructing a transmission distance prediction model based on LSTM networks; The transmission distance prediction model is trained and validated in a supervised manner using the sample feature sequence set and the sample transmission distance set, and training is completed after convergence.

[0027] First, multiple historical standard feature vectors are collected from historical data to form a sample feature sequence set. From past sensor data transmission records, standard feature vectors from multiple historical moments are extracted, all generated according to the S100 step, with a dimension of N+2. All historical standard feature vectors are then arranged in chronological order to form the sample feature sequence set. This sample feature sequence set serves as the input sample library for model training. It consists of multiple historical standard feature vectors, each containing information on the transmission distance sequence, scene type, and motion pattern at the corresponding historical moment, providing a data foundation for the model to learn the correlation between input features and future distances. For example, from the records of the industrial workshop over the past week, 300 standard feature vectors at different historical moments are extracted, such as [0,0.5,1,0.5,0,1,1] at time t1, [0.5,1,0.8,0.6,0.3,1,1] at time t2, [1,0.8,0.6,0.4,0.2,1,1] at time t3, and so on. The 300 standard feature vectors are then summarized to form a sample feature sequence set.

[0028] Secondly, the actual transmission distance corresponding to each historical standard feature vector in the sample feature sequence set for a predetermined future time period is obtained, and these actual transmission distances are labeled as a sample transmission distance set. For each historical standard feature vector in the sample feature sequence set, its corresponding actual transmission distance within the predetermined future time period is found, i.e., the true value. These true distances are arranged in vector order to form the sample transmission distance set, which serves as the label for model training. The sample transmission distance set is the output label library for model training. Each label corresponds one-to-one with an input vector in the sample feature sequence set, and is used to determine the deviation between the model's predicted value and the true value in supervised training, guiding the adjustment of model parameters.

[0029] For example, the future preset time period is 10 seconds after the current time, and sampling is performed at 2-second intervals, for a total of 5 time periods. For the standard feature vector [0,0.5,1,0.5,0,1,1] at time t1, the actual transmission distances corresponding to t1+2 seconds, t1+4 seconds, t1+6 seconds, t1+8 seconds, and t1+10 seconds are found to be 12.1 meters, 12.2 meters, 12.1 meters, 12 meters, and 11.9 meters, respectively. The actual distances corresponding to 300 standard feature vectors are obtained in this way, forming a sample transmission distance set, such as {[12.1,12.2,12.1,12,11.9],[12.2,12.1,12,11.9,11.8],……}.

[0030] Furthermore, a transmission distance prediction model is constructed based on an LSTM network. Long Short-Term Memory (LSTM) networks are a special type of recurrent neural network that uses a gating mechanism to address the long-range dependency problem of traditional recurrent neural networks. LSTM networks excel at processing time-series data and can accurately capture the distance change patterns caused by motion continuity, making them suitable for predicting future time-series values. The LSTM network structure is built, adapting to the motion continuity requirement through a "gating mechanism": the input layer dimension matches the standard feature vector dimension N+2=7, receiving features containing motion patterns; the hidden layer uses 1-2 layers of LSTM units, such as 64 units, using a forget gate to retain historical motion state information, such as the uniform speed feature from the previous moment, and using an input gate to filter out abnormal changes that do not conform to continuity, such as spurious features from sudden stops, ensuring the model learns the pattern from motion continuity to distance continuity; the output layer dimension matches the sampling number for a preset future time period, such as 5 moments corresponding to dimension 5, outputting the continuously changing predicted distance.

[0031] For example, in the scenario of temperature sensing nodes transmitting data to the central control system in an industrial workshop, an LSTM model is constructed: the input layer dimension is set to 7, matching the standard feature vector dimension; the hidden layer has 1 layer with 64 LSTM units, and the activation function is tanh; the output layer dimension is set to 5, corresponding to the transmission distance prediction at 5 times within the next 10 seconds, and the activation function is linear; at the same time, a Dropout layer is added to prevent overfitting.

[0032] Next, the transmission distance prediction model is trained and validated in a supervised manner using the sample feature sequence set and the sample transmission distance set, completing the training upon convergence. The sample feature sequence set and the sample transmission distance set are divided into 210 training sets and 90 validation sets in a 7:3 ratio. The mean squared error (MSE) between the predicted and true distances is used as the loss function. The Adam optimizer is employed with a learning rate of 0.001, and the model is trained iteratively. After each training round, the model accuracy is evaluated using the validation set. Training stops when the loss function value no longer decreases for five consecutive rounds, indicating model convergence. The trained transmission distance prediction model is then obtained. Supervised training refers to the model learning the mapping relationship through input samples and corresponding labels, calculating the prediction error using the loss function, and then adjusting the network parameters through backpropagation. Convergence means that the model's prediction error stabilizes at a low level and no longer changes with training iterations; at this point, the model possesses reliable prediction capabilities.

[0033] For example, 210 sets of historical feature vectors and the true future distance are used as the training set, and 90 sets are used as the validation set. The loss is calculated using MSE, and the model parameters are updated in each round using the Adam optimizer. The initial loss value is 0.12. After 20 iterations, the loss value drops to 0.008 and remains stable in this range for 5 consecutive rounds. The model converges, the training is completed and saved, and the trained transmission distance prediction model is obtained.

[0034] Finally, the standard feature vector is input into the pre-trained transmission distance prediction model to predict the basic transmission distance for a preset time period. The current time standard feature vector generated in S100 is input into the trained transmission distance prediction model. Based on the learned correlation patterns, the model calculates and outputs the predicted transmission distance for the preset time period, i.e., the basic predicted transmission distance. For example, if the current standard feature vector [0,0.5,1,0.5,0,1,1] generated in S100 is input into the trained transmission distance prediction model, the model outputs the basic predicted transmission distance for the next 10 seconds as [12.0,12.1,12.0,11.9,11.8].

[0035] In this embodiment, supervised training based on real historical data ensures that the transmission distance prediction model can accurately capture temporal correlation features, avoiding the shortcomings of traditional models in handling time dependencies. The output basic predicted transmission distance provides an initial prediction basis for subsequent distance correction and provides a key input for multipath delay spread prediction, further accurately predicting the intensity of multipath interference, supporting the early optimization of adaptive filtering parameters, and laying a key distance prediction foundation for improving the reliability of sensor data transmission.

[0036] S300: Based on the prediction deviation data of similar historical scenarios and the standard distance parameters of the current scenario, the basic predicted transmission distance is corrected to obtain the predicted transmission distance.

[0037] In this embodiment, the basic predicted transmission distance is corrected based on prediction deviation data from similar historical scenarios and standard distance parameters for the current scenario to obtain the predicted transmission distance. The basic predicted transmission distance output by S200 is based on the overall pattern of historical data and does not fully consider the specific deviations and reasonable distance boundaries of the current scenario. Directly using the basic predicted distance would lead to increased prediction deviations in subsequent multipath delay spread predictions. Therefore, a dual mechanism of historical similar scenario deviation correction and current scenario standard range constraint is needed to optimize the basic prediction result, ensuring that the final predicted transmission distance better matches the actual current scenario and provides high-precision input for subsequent multipath prediction.

[0038] Step S300 in the method provided in this application embodiment includes: Calculate the similarity between the standard feature vector and the historical standard feature vector, filter out similar historical standard feature vectors that meet the preset similarity threshold, and take the scene corresponding to the similar historical standard feature vector as the historical similar scene. Obtain multiple historical predicted transmission distances and historical actual transmission distances corresponding to similar historical scenarios, calculate the difference between each historical predicted transmission distance and the corresponding historical actual transmission distance, and average the multiple differences to obtain the average prediction deviation data. Using the scene type and motion mode in the current standard feature vector as retrieval constraints, the standard distance parameter range of the current scene is obtained from the pre-stored scene standard parameter library; Using the standard distance parameter range as a boundary constraint, the basic predicted transmission distance is fused with the average prediction deviation data to obtain the predicted transmission distance.

[0039] First, the similarity between the standard feature vector and historical standard feature vectors is calculated. Similar historical standard feature vectors that meet a preset similarity threshold are selected, and the scenes corresponding to these similar historical standard feature vectors are considered historical similar scenes. A similarity algorithm, such as cosine similarity, is used to calculate the similarity between the current standard feature vector and all historical standard feature vectors in the sample feature sequence set. A preset similarity threshold, such as 0.9, is set, and historical standard feature vectors with a similarity ≥ the threshold are selected. The scenes corresponding to these historical standard feature vectors are considered historical similar scenes. Similarity is an indicator that measures the degree of similarity between two feature vectors. Cosine similarity ranges from -1 to 1, with values ​​closer to 1 indicating greater similarity. The preset similarity threshold is a critical value for determining whether a scene is of the same type; it is determined through experimental verification to ensure a high degree of matching between the selected historical scenes and the current scene features. Historical similar scenes refer to historical scenes that are highly similar to the current scene in terms of distance change trends, scene type, and motion patterns. Their deviation patterns are more valuable for correcting the current scene.

[0040] For example, the cosine similarity formula is used to calculate the similarity between the current feature vector [0, 0.5, 1, 0.5, 0, 1, 1] and 300 historical vectors, such as 0.993, 0.965, 0.871, etc. A similarity threshold of 0.9 is set, and five historical standard feature vectors with a similarity ≥ 0.9 are selected, such as [0.1, 0.4, 0.9, 0.6, 0.1, 1, 1], [0, 0.6, 1.0, 0.4, 0, 1, 1], etc. The historical scenarios corresponding to the selected historical standard feature vectors are all industrial workshops with uniform motion, i.e., similar historical scenarios.

[0041] Secondly, multiple historical predicted transmission distances and historical actual transmission distances corresponding to similar historical scenarios are obtained. The difference between each historical predicted transmission distance and its corresponding historical actual transmission distance is calculated, and the average of these differences is taken to obtain the average prediction deviation data. The historical predicted transmission distance refers to the basic prediction result obtained after inputting the predicted transmission distance model into the historical scenario at that time. The historical actual transmission distance refers to the transmission distance actually collected in the future time period for the historical scenario at that time. From the historical records, the historical predicted transmission distance and historical actual transmission distance corresponding to each selected historical scenario are extracted; the difference between the historical predicted transmission distance and the historical actual transmission distance for each historical scenario is calculated, i.e., the single-scenario deviation; the arithmetic mean of all single-scenario deviations is taken to obtain the average prediction deviation data. The average prediction deviation data reflects the common deviation trend of similar historical scenarios; for example, a negative mean indicates that historical predictions are generally too high, and a positive mean indicates that historical predictions are generally too low, which is used to correct the current basic prediction.

[0042] For example, extract the historical predicted transmission distance and historical actual transmission distance data for 5 similar historical scenarios, and calculate the single-scenario deviation: Scenario 1: Historical predicted transmission distance [12.1, 12.2, 12.1, 12.0, 11.9], historical actual transmission distance data [12.0, 12.1, 12.0, 11.9, 11.8], single-scenario deviation [0.1, 0.1, 0.1, 0.1, 0.1]; Similarly, the single-scenario deviations for scenarios 2-5 are [0.08]. The average prediction deviation data is calculated by taking the average of the five groups of deviations: [0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1, 0.1], [0.09, 0.09, 0.09, 0.09], and [0.11, 0.11, 0.1, 0.1, 0.1]. This indicates that the basic predictions for similar scenarios in the past were generally 0.1 meters higher than the actual values.

[0043] Furthermore, using the scene type and motion mode in the current standard feature vector as retrieval constraints, the standard distance parameter range for the current scene is obtained from a pre-stored scene standard parameter library. The scene standard parameter library is a static database built based on a large amount of measured data, storing reasonable distance ranges under different scene and motion mode combinations to constrain the reasonableness of predicted values. The standard distance parameter range refers to the physically reasonable interval of transmission distance under the current scene. For example, when moving at a constant speed in an industrial workshop, the distance between the sensor node and the central control unit is usually within 10-13 meters, avoiding predicted values ​​exceeding the actual possible range. A scene standard parameter library is pre-stored, categorizing and storing the corresponding standard distance parameter ranges by scene type and motion mode. Using the scene type and motion mode in the current standard feature vector as retrieval conditions, the corresponding standard distance parameter range is queried and obtained from the scene standard parameter library. For example, by querying the scene standard parameter library, the standard distance parameter range for industrial workshop + constant speed motion is 10.0-13.0 meters, meaning that the transmission distance under the current industrial workshop scene cannot be less than 10 meters or more than 13 meters.

[0044] Furthermore, using the standard distance parameter range as boundary constraints, the basic predicted transmission distance is fused with the average prediction deviation data to obtain the predicted transmission distance. The predicted transmission distance is the final distance prediction result after correction for deviations in similar scenarios and constraints on the current scenario range, making it more accurate and closer to the actual current scenario than the basic prediction. First, the basic predicted transmission distance is corrected for deviation by subtracting the average prediction deviation data from the basic predicted transmission distance. If the average deviation is positive, it indicates that the basic prediction is too high, and the deviation needs to be subtracted; if the average deviation is negative, the deviation is added. Then, it is checked whether the corrected distance is within the standard distance parameter range. If it exceeds the range, it is adjusted to the nearest boundary value. If it is lower than the lower limit, the lower limit is used; if it is higher than the upper limit, the upper limit is used. The final distance obtained is the predicted transmission distance.

[0045] For example, the base predicted transmission distance is [12.0, 12.1, 12.0, 11.9, 11.8], and the average prediction deviation is [0.1, 0.1, 0.1, 0.1, 0.1]. Subtracting the average prediction deviation from the base predicted transmission distance yields the corrected distance [11.9, 12.0, 11.9, 11.8, 11.7]. The corrected distances [11.9, 12.0, 11.9, 11.8, 11.7] are all within the range of 10.0-13.0 meters and require no adjustment. The final predicted transmission distance is: [11.9, 12.0, 11.9, 11.8, 11.7].

[0046] In this embodiment, the common deviations caused by model generalization in the basic prediction are offset by the average deviation correction of similar historical scenarios, making the prediction results more consistent with the actual deviation patterns of the current scenario. By constraining the standard distance range of the current scenario, possible physical anomalies in the basic prediction are eliminated, ensuring the rationality of the prediction results. The final output predicted transmission distance shows improved accuracy compared to the basic prediction of S200, providing a highly reliable distance input for the subsequent accurate prediction of multipath delay spread in S400.

[0047] S400: Input the predicted transmission distance into the pre-trained multipath delay spread prediction model, and predict and output the predicted multipath delay spread for a future preset time period.

[0048] In this embodiment, the predicted transmission distance is input into a pre-trained multipath delay spread prediction model to predict and output the predicted multipath delay spread for a future preset time period. Multipath delay spread refers to the time difference between the multipath signal and the main path signal that arrives at the receiving end via different paths such as reflection and refraction during signal transmission; it can characterize the intensity of multipath interference. A multipath delay spread prediction model based on a nonlinear regression algorithm is constructed. Through supervised training using historical transmission distances and corresponding real multipath delay spreads, the multipath delay spread prediction model learns the nonlinear relationship between transmission distance and multipath delay spread. The predicted transmission distance output by S300 is then input into the trained model to obtain the predicted multipath delay spread for the future preset time period, providing a basis for predicting interference intensity for subsequent dynamic configuration of adaptive filtering parameters.

[0049] Step S400 in the method provided in this application embodiment includes: The construction process of the multipath delay spread prediction model includes: From historical data, we collect multiple historical transmission distances associated with historical scene types and motion patterns to form a sample transmission distance set; For each historical transmission distance in the sample transmission distance set, obtain the real multipath delay spread for a future preset time period, and label multiple real multipath delay spreads as a sample multipath delay spread label set; A multipath delay spread prediction model is constructed based on a nonlinear regression algorithm; The multipath delay spread prediction model is trained and validated in a supervised manner using the sample transmission distance set and the sample multipath delay spread label set, and training is completed after convergence.

[0050] First, from historical data, multiple historical transmission distances associated with historical scene types and motion patterns are collected to form a sample transmission distance set. This sample transmission distance set serves as the input sample library for training the multipath delay spread prediction model. It contains specific numerical values ​​of historical transmission distances and implicitly includes corresponding scene and motion pattern features, providing a data foundation for the multipath delay spread prediction model to learn the nonlinear correlation between distance and multipath delay spread. From past sensor data transmission records, multiple historical transmission distances associated with historical scene types and motion patterns are extracted. These distances must cover different distance values, and the scene and motion pattern must be consistent with the current scene, such as both being an industrial workshop with constant speed. The historical transmission distances are then organized in chronological order to form the sample transmission distance set. From the historical records of the industrial workshop + uniform motion scene, 300 transmission distances at different times were extracted, such as 11.7 meters, 11.8 meters, 11.9 meters, 12.0 meters, 12.1 meters, etc., covering the possible range of the current predicted distance, forming a sample transmission distance set containing the transmission distances at 300 different times: [11.7, 11.8, 11.9, 12.0, 12.1, ..., 11.9].

[0051] Secondly, for each historical transmission distance in the sample transmission distance set, the true multipath delay spread for a future preset time period is obtained, and multiple such true multipath delay spreads are labeled as a sample multipath delay spread label set. Multipath delay spread refers to the maximum time difference between the multipath signal and the main path signal arriving at the receiver; a larger value indicates more multipath components and a higher risk of inter-symbol interference. The sample multipath delay spread label set is the output label library for training the multipath delay spread prediction model. Each label corresponds one-to-one with a distance value in the sample transmission distance set, guiding the multipath delay spread prediction model to learn the mapping relationship between input distance and output delay spread. For each historical transmission distance in the sample transmission distance set, the true multipath delay spread value within its corresponding future preset time period is collected using dedicated measurement equipment. The true multipath delay spread values ​​are arranged in distance order to form the sample multipath delay spread label set, which serves as the output label for training the multipath delay spread prediction model. The future preset time period is 10 seconds after the current time. For example, for a sample transmission distance of 11.9 meters, the actual multipath delay spread corresponding to the next 5 moments is measured by a vector network analyzer as [195ns, 198ns, 196ns, 193ns, 190ns]. The actual multipath delay spreads corresponding to 300 sample transmission distances are obtained in this way, and the 300 actual multipath delay spreads are combined into a sample multipath delay spread label set: [[185, 188, 186, 183, 180], [190, 193, 191, 188, 185], ..., [195, 198, 196, 193, 190]].

[0052] Furthermore, a multipath delay spread prediction model is constructed based on a nonlinear regression algorithm. By integrating multiple weak learners, the nonlinear regression algorithm can fit complex curve relationships between variables. Compared to linear regression, it is more suitable for handling the nonlinear correlation between transmission distance and multipath delay spread, which is influenced by multiple factors, thus avoiding prediction bias caused by simple linear assumptions. A regression algorithm suitable for handling nonlinear relationships is selected to construct the multipath delay spread prediction model. The input layer dimension of the multipath delay spread prediction model matches the dimension of the predicted transmission distance; for example, 5 time points correspond to input dimension 5. Through combined learning of multiple decision trees, the nonlinear correlation between transmission distance and multipath delay spread is captured. For example, when the distance increases by 0.1 meters, the delay spread may increase by 3 ns or 5 ns, depending on the obstacle distribution at that distance.

[0053] For example, a multipath delay spread prediction model is constructed using the Gradient Boosting Tree (GBDT) algorithm. The input layer receives the transmission distance at 5 time points. 100 decision trees are set, with the depth of each tree limited to 5 to avoid overfitting. By learning the nonlinear relationship between transmission distance and multipath delay spread in the samples, a mapping model from distance to delay spread is established.

[0054] Next, the multipath delay spread prediction model is trained and validated in a supervised manner using the sample transmission distance set and the sample multipath delay spread label set, completing the training upon convergence. The sample transmission distance set and the sample label set are divided into training and validation sets in a 7:3 ratio. The model is trained iteratively using the mean absolute error (MAE) between the predicted multipath delay spread and the true value as the loss function. After each training round, the accuracy is evaluated using the validation set. When the MAE remains stable at a low level for 5 consecutive rounds, the model converges, training stops, and the data is saved. The mean absolute error (MAE) is an indicator of prediction accuracy; a smaller value indicates a more accurate prediction. Convergence indicates that the model has fully learned the nonlinear characteristics in the samples and can stably output reliable prediction results.

[0055] For example, the 300 sample transmission distance sets and sample label sets are divided into 210 training sets and 90 validation sets in a 7:3 ratio. The training sets are used for model learning, and the validation sets are used for evaluation. The initial MAE is 12ns. After 30 iterations, the MAE decreases to 3ns and remains stable for 5 consecutive iterations, indicating model convergence. The training is then completed and saved, resulting in a trained multipath delay spread prediction model. Now, when a predicted transmission distance is input, the multipath delay spread prediction model can output a predicted multipath delay spread based on the learned nonlinear correlation between the transmission distance and the multipath delay spread.

[0056] Finally, the predicted transmission distance is input into the pre-trained multipath delay spread prediction model, which predicts the multipath delay spread for a preset time period in the future. The predicted multipath delay spread is a key indicator reflecting the intensity of future multipath interference, and its value directly determines the configuration of subsequent adaptive filtering parameters. For example, the larger the delay spread, the higher-order equalization filter needs to be configured. For example, the predicted transmission distance [11.9, 12.0, 11.9, 11.8, 11.7] generated by S300 is input into the trained multipath delay spread prediction model. Based on the learned nonlinear correlation law, the multipath delay spread prediction model calculates and outputs the predicted multipath delay spread value for the preset time period in the future, i.e., the predicted multipath delay spread. The multipath delay spread prediction model outputs the predicted multipath delay spread for 5 moments in the next 10 seconds: [194ns, 197ns, 195ns, 191ns, 188ns].

[0057] In this embodiment, a nonlinear regression algorithm effectively captures the complex correlation between transmission distance and multipath delay spread, avoiding the simplification errors of linear models. The output predicted multipath delay spread directly quantifies the intensity of future multipath interference, providing a quantifiable interference basis for the subsequent dynamic adjustment of parameters such as the order and symbol period of the adaptive filter by the S500. By transforming distance changes into interference intensity prediction, a mapping from physical distance to interference characteristics is achieved, providing decision support for ultimately improving the anti-interference capability of sensor data transmission.

[0058] S500: Based on the predicted multipath delay spread, dynamically configure the order of the adaptive equalization filter, the transmission symbol period, and the modulation method of the adaptive filter to remotely transmit the sensor data.

[0059] In this embodiment, based on the predicted multipath delay spread, the order of the adaptive equalization filter, the transmission symbol period, and the modulation scheme of the adaptive filter are dynamically configured for remote transmission of sensor data. Multipath delay spread directly determines the intensity and characteristics of multipath interference. The larger the delay spread, the greater the time difference between the multipath signal and the main path signal, and the higher the risk of inter-symbol interference and signal distortion. The anti-interference effect of adaptive filtering highly depends on the matching degree between the parameters and the multipath environment: if the equalization filter order is insufficient, it cannot cover all multipath components; if the symbol period is too short, it will exacerbate the time overlap of multipath signals; if the modulation scheme has insufficient anti-interference capability, it will amplify signal distortion. Fixed parameters are difficult to adapt to dynamically changing multipath delay spread, resulting in unstable compensation effects. Therefore, S500 needs to dynamically adjust the order of the adaptive equalization filter, the transmission symbol period, and the modulation scheme according to the predicted multipath delay spread, so that the adaptive filtering parameters match the future multipath environment in real time, maximizing the cancellation of multipath interference and ensuring stable sensor data transmission.

[0060] Step S500 in the method provided in this application embodiment includes: Determine the time span of the multipath component corresponding to the predicted multipath delay spread, and calculate the order of the adaptive equalization filter based on the time span of the multipath component; Based on the multipath complexity of the current scenario, a preset multiple relationship between the symbol period and the predicted multipath delay spread is determined, and the transmission symbol period is calculated based on the predicted multipath delay spread and the preset multiple relationship. Based on the complexity of the multipath propagation in the current scenario, determine the multipath delay spread threshold range, identify the specific range within which the predicted multipath delay spread is located within the multipath delay spread threshold range, and determine the modulation method based on the anti-interference requirements corresponding to the specific range. The order, transmission symbol period, and modulation scheme of the adaptive equalization filter are configured into the transmission link, and multipath interference is compensated in real time through adaptive filtering to complete the remote transmission of sensor data.

[0061] First, determine the time span of the multipath component corresponding to the predicted multipath delay spread, and calculate the order of the adaptive equalization filter based on the time span of the multipath component.

[0062] The process includes determining the time span of the multipath components corresponding to the predicted multipath delay spread, and calculating the order of the adaptive equalization filter based on the time span of the multipath components, including: Obtain the time difference range between the multipath signal and the main path signal corresponding to the predicted multipath delay spread, and use it as the time span of the multipath component; The basic order of the adaptive equalization filter is set with reference to the time span of the multipath components. The ratio of the predicted multipath delay spread to the average multipath delay spread of similar historical scenarios is calculated and used as the order adjustment factor. The order of the adaptive equalizer is obtained by multiplying the base order of the adaptive equalizer by the order adjustment coefficient.

[0063] First, the time difference range between the multipath signal and the main path signal corresponding to the predicted multipath delay spread is obtained as the multipath component time span. The multipath delay spread itself reflects the maximum time difference between the multipath signal and the main path signal; therefore, the numerical range of the predicted multipath delay spread is directly used as the multipath component time span, i.e., the time interval from the earliest to the latest arrival of the multipath signal. The multipath component time span is the time distribution range of the multipath signal relative to the main path signal. The larger the span, the more multipath components need to be covered, and the higher the filter order needs to be. For example, if the predicted multipath delay spread is [194ns, 197ns, 195ns, 191ns, 188ns], its numerical range is 188ns-197ns; therefore, the multipath component time span is 188ns-197ns.

[0064] Secondly, using the multipath component time span as a reference, the base order of the adaptive equalizer is set. The order of the adaptive equalizer is the number of taps in the filter; each tap corresponds to a multipath signal at a specific time point. A higher order compensates for more multipath components. The base order is an initial value set based on a typical time span to ensure coverage of common multipath components. Using the median value of the multipath component time span as a reference, combined with the sampling frequency, the base order is set. The base order must cover at least the number of sampling points within the time span, where the number of sampling points = time span / sampling interval. For example, with a sampling frequency of 1 GHz and a sampling interval of 1 ns, the median value of the multipath component time span is (188 + 197) / 2 = 192.5 ns, corresponding to approximately 193 sampling points. Therefore, the base order is set to 200, slightly higher than the number of sampling points, to allow for redundancy.

[0065] Simultaneously, the ratio of the predicted multipath delay spread to the average multipath delay spread of similar historical scenarios is calculated and used as the order adjustment factor. The order adjustment factor = predicted multipath delay spread / average multipath delay spread of similar historical scenarios. The order adjustment factor is a proportional factor for dynamically adjusting the base order, reflecting the relative strength of the current multipath interference compared to similar historical scenarios. A value greater than 1 indicates stronger current interference, requiring an increase in order; a value less than 1 indicates weaker interference, requiring a decrease in order. For example, the average predicted multipath delay spread = (194 + 197 + 195 + 191 + 188) / 5 = 193 ns, and the average multipath delay spread of similar historical scenarios is 190 ns. The order adjustment factor = 193 / 190 ≈ 1.016.

[0066] Further, the adaptive equalizer order is obtained by multiplying the base order of the adaptive equalizer by the order adjustment coefficient. The adaptive equalizer order = base order × order adjustment coefficient. The order must be an integer, and the result is rounded up to ensure the filter can cover the currently predicted multipath components. The final adaptive equalizer order is the dynamically adjusted number of taps, preserving the general coverage capability of the base order while adapting to the current interference intensity through the adjustment coefficient, avoiding insufficient order or excessive redundancy. For example, with a base order of 200 and an order adjustment coefficient of 1.016, 200 × 1.016 ≈ 203.2. After rounding up, the adaptive equalizer order is 204.

[0067] Secondly, based on the multipath complexity of the current scenario, a preset multiple relationship between the symbol period and the predicted multipath delay spread is determined, and the transmission symbol period is calculated based on the predicted multipath delay spread and the preset multiple relationship.

[0068] Specifically, based on the multipath complexity of the current scenario, a preset multiple relationship is determined between the symbol period and the predicted multipath delay spread, and the transmission symbol period is calculated based on the predicted multipath delay spread and the preset multiple relationship, including: The obstacle distribution density in the current scene is collected using environmental sensors; The ratio of obstacle distribution density to the standard obstacle distribution density of similar historical scenarios is used as the multipath complexity of the current scenario. Based on the complexity of the multipath, a preset multiple relationship is determined between the symbol period and the predicted multipath delay spread; The transmission symbol period is obtained by multiplying the preset multiplication factor relationship with the predicted multipath delay spread.

[0069] First, the obstacle distribution density of the current scene is collected using environmental sensors. The obstacle distribution density is the number of obstacles per unit area in the current scene. Obstacle distribution density is an indicator of the number of multipath signal sources in the scene; the higher the density, the more multipath components are reflected / refracted, and the more complex the multipath environment. For example, if an infrared sensor measures 3.2 obstacles per 10 square meters in an industrial workshop, the current obstacle distribution density is 3.2 obstacles per 10 square meters.

[0070] Secondly, the ratio of obstacle distribution density to the standard obstacle distribution density of similar historical scenarios is calculated as the multipath complexity of the current scenario. Multipath complexity = Current obstacle distribution density / Standard obstacle distribution density of similar historical scenarios. Multipath complexity is a relative indicator reflecting the complexity of the current multipath environment and directly affects the matching relationship between symbol period and multipath delay spread. The more complex the scenario, the longer the symbol period is required to avoid overlap. For example, if the current obstacle distribution density is 3.2 obstacles per 10 square meters, and the standard obstacle distribution density of similar historical scenarios is 3 obstacles per 10 square meters, the multipath complexity = 3.2 / 3 ≈ 1.067, indicating that the current multipath environment is slightly more complex than the historical average.

[0071] Furthermore, based on the multipath complexity, a preset multiple relationship is determined between the symbol period and the predicted multipath delay spread. The symbol period refers to the duration of a single symbol from transmission to completion in digital communication. The preset multiple relationship is the ratio of the symbol period to the multipath delay spread, ensuring that the symbol period is long enough so that the time difference of the multipath signals is less than 1 / 5 of the symbol period, reducing inter-symbol interference. The multiple relationship is set according to the multipath complexity: when the multipath complexity is ≤1, the multiple is 5, meaning that in a typical complex scenario, the symbol period is 5 times the delay spread; when 1 < multipath complexity ≤1.2, the multiple is 6, indicating a longer period is needed for more complex scenarios; when the multipath complexity >1.2, the multiple is 7, indicating a longer period is needed for highly complex scenarios. For example, if the current multipath complexity is 1.067, and 1 < 1.067 ≤ 1.2, then the preset multiple relationship is determined to be 6.

[0072] Finally, the preset multiplication factor is multiplied by the predicted multipath delay spread to obtain the transmission symbol period. Transmission symbol period = preset multiplication factor × average predicted multipath delay spread. The transmission symbol period is the duration of a single symbol. The longer the period, the less overlap of multipath signals in time, and the weaker the inter-symbol interference, but the transmission rate will be slightly reduced. For example, if the average predicted multipath delay spread is 193 ns and the preset multiplication factor is 6, the transmission symbol period = 6 × 193 = 1158 ns ≈ 1.16 μs.

[0073] Next, based on the complexity of the multipath propagation in the current scenario, the multipath delay spread threshold range is determined, the specific range in which the predicted multipath delay spread is located within the multipath delay spread threshold range is determined, and the modulation method is determined based on the anti-interference requirements corresponding to the specific range.

[0074] Specifically, based on the multipath complexity of the current scenario, a multipath delay spread threshold range is determined; the specific interval within which the predicted multipath delay spread falls within the multipath delay spread threshold range is identified; and the modulation scheme is determined based on the anti-interference requirements corresponding to the specific interval, including: Based on the complexity of the multipath, determine the multipath delay spread threshold range; The predicted multipath delay spread is compared with the multipath delay spread threshold range to determine the specific range in which the predicted multipath delay spread is located. The modulation method is selected based on the anti-interference requirements corresponding to the specific interval.

[0075] First, based on the multipath complexity, the multipath delay spread threshold range is determined. This threshold range is the critical value range for classifying interference intensity levels. The range is dynamically adjusted according to the multipath complexity to ensure that the interference level judgment is more closely aligned with the current scenario. For example, the multipath complexity determines the threshold range: when complexity ≤ 1, the threshold range is [0, 180ns), [180ns, 200ns), [200ns, ∞), corresponding to weak interference, medium interference, and strong interference, respectively; when 1 < complexity ≤ 1.2, the threshold range tightens to [0, 170ns), [170ns, 190ns), [190ns, ∞), corresponding to weak interference, medium interference, and strong interference, respectively. Currently, the multipath complexity is 1.067, and 1 < 1.067 ≤ 1.2, therefore the threshold range is [0, 170ns), [170ns, 190ns), [190ns, ∞).

[0076] Secondly, the predicted multipath delay spread is compared with the multipath delay spread threshold range to determine the specific range in which the predicted multipath delay spread is located. The average value of the predicted multipath delay spread is compared with the threshold range to determine whether it belongs to weak, moderate, or strong interference. The specific range corresponds to the interference intensity level; the higher the level, the higher the requirement for the anti-interference capability of the modulation method. For example, if the average predicted multipath delay spread is 193 ns, compared with the threshold ranges [0, 170 ns), [170 ns, 190 ns), and [190 ns, ∞), 193 ns ≥ 190 ns, placing it in the strong interference range.

[0077] Furthermore, the modulation method is selected based on the anti-interference requirements corresponding to the specific interval. Modulation is the method of converting digital signals into analog signals. Low-order modulation carries 1 bit of information per symbol, offering strong noise and interference resistance; high-order modulation carries 4 bits of information per symbol, offering high speed but weak anti-interference, and the choice must be based on the interference intensity. In the low-threshold interval, multipath interference is weak, so high-order modulation is used, relying on adaptive filtering to accurately compensate for weak interference and improve the transmission rate; in the middle-threshold interval, multipath interference is moderate, so middle-order modulation is used to balance the compensation pressure of adaptive filtering with transmission efficiency; in the high-threshold interval, multipath interference is strong, so low-order anti-interference modulation is used, leveraging its strong anti-interference characteristics. For example, if we are currently in a strong interference interval with high anti-interference requirements, then a low-order anti-interference modulation method is chosen.

[0078] Finally, the order, transmission symbol period, and modulation scheme of the adaptive equalizer filter are configured in the transmission link. Adaptive filtering compensates for multipath interference in real time, completing the remote transmission of sensor data. The calculated adaptive equalizer filter order, transmission symbol period, and modulation scheme are written into the parameter configuration module of the transmission link. The adaptive filtering module dynamically adjusts the filter coefficients based on the real-time multipath signal and the configured parameters to compensate for multipath interference, ultimately completing the remote transmission of sensor data. The transmission link is the communication channel connecting the sensor node and the receiving end. The parameter configuration module is responsible for storing and distributing filtering, modulation, and other parameters. The adaptive filtering module adjusts the coefficients in real time to cancel the interference of multipath signals on the main path signal. For example, if the adaptive equalizer filter order is 204, the transmission symbol period is 1.16μs, and the modulation scheme is low-order modulation, after the central control system receives the parameter configuration, the adaptive equalizer filter covers the multipath components with a 204th-order tap and transmits data with a 1.16μs symbol period. Low-order modulation is used to enhance anti-interference capability, and the temperature sensing data is ultimately transmitted to the central control system with a low bit error rate.

[0079] In this embodiment, the adaptive equalization filter order accurately covers the current multipath component time span, avoiding undercompensation of multipath signals due to insufficient order; the transmission symbol period is matched with the multipath delay spread multiple to reduce the time overlap of multipath signals and reduce inter-symbol interference; the low-order modulation method provides stronger anti-distortion capability for strong interference scenarios. The synergistic effect of these three elements enables the adaptive filter to adapt to future multipath environment changes in real time, effectively offsetting multipath interference caused by dynamic changes in transmission distance, ultimately effectively reducing the transmission bit error rate, improving the anti-interference capability, accuracy, and reliability of remote transmission of sensor data, and completely closing the technical link of prediction, optimization, and compensation.

[0080] The embodiments of this application, through the specific implementation methods described above, achieve the following technical effects: This application provides a method and system for optimizing the remote transmission effect of sensor data based on adaptive filtering. It comprehensively collects historical transmission distance sequences, current scene types, and motion patterns, and performs standardized processing to form a complete and model-adaptive standard feature vector, providing a high-quality input foundation for subsequent prediction. A model is constructed based on an LSTM network combined with the principle of motion continuity to predict the basic transmission distance that conforms to physical laws, accurately capturing the temporal correlation of distance changes. By correcting for deviations in similar historical scenes and constraining the standard distance range of the current scene, the basic prediction results are optimized to obtain a high-precision predicted transmission distance. A model is constructed using a nonlinear regression algorithm to accurately predict multipath delay spread based on the predicted transmission distance, quantifying the future multipath interference intensity. By dynamically configuring the order of the adaptive equalization filter, the transmission symbol period, and the modulation method, multipath interference is compensated in real time, reducing inter-symbol interference and signal distortion. This forms a complete closed loop of feature input, distance prediction, accuracy optimization, interference quantification, and parameter adaptation, effectively improving the anti-interference capability, accuracy, and reliability of remote sensor data transmission, and adapting to the dynamic transmission needs of multiple scenarios.

[0081] Example 2, as Figure 2 As shown, this application provides a system for optimizing the remote transmission effect of sensor data based on adaptive filtering. The system includes: The feature vector generation module 11 is used to collect the transmission distance sequence of the most recent N times, the scene type and motion mode at the current time, and after standardizing the transmission distance data, bind it with the scene type and motion mode to form a standard feature vector. The basic distance prediction module 12 is used to input the standard feature vector into the pre-trained transmission distance prediction model and predict and output the basic predicted transmission distance for a future preset time period. The prediction distance correction module 13 is used to correct the basic prediction transmission distance based on the prediction deviation data of similar historical scenarios and the standard distance parameters of the current scenario, so as to obtain the prediction transmission distance. Multipath delay prediction module 14 is used to input the predicted transmission distance into a pre-trained multipath delay extension prediction model and predict and output the predicted multipath delay extension for a future preset time period. The parameter configuration transmission module 15 is used to dynamically configure the order of the adaptive equalization filter, the transmission symbol period, and the modulation method of the adaptive filter according to the predicted multipath delay spread, and to remotely transmit the sensor data.

[0082] In one embodiment, the feature vector generation module 11 is further configured to: Historical transmission distance data is collected at a preset sampling interval, and the transmission distance data of the N most recent moments in history is extracted to form a transmission distance sequence; Environmental features are collected using environmental sensors to identify the current scene type; The velocity change rate is calculated based on the transmission distance sequence, and the motion mode is determined according to the velocity change rate, wherein the motion mode includes stationary, uniform, and variable speed. The transmission distance sequence is normalized to obtain a standard transmission distance sequence; The standard transmission distance sequence, scene type, and motion mode are concatenated in sequence to form a standard feature vector with dimension N+2.

[0083] In one embodiment, the basic distance prediction module 12 is further configured to: The process of constructing the transmission distance prediction model includes: Multiple historical standard feature vectors are collected from historical data to form a sample feature sequence set; Obtain the actual transmission distance for each historical standard feature vector in the sample feature sequence set corresponding to a future preset time period, and label multiple such actual transmission distances as a sample transmission distance set; Constructing a transmission distance prediction model based on LSTM networks; The transmission distance prediction model is trained and validated in a supervised manner using the sample feature sequence set and the sample transmission distance set, and training is completed after convergence.

[0084] In one embodiment, the predicted distance correction module 13 is further configured to: Calculate the similarity between the standard feature vector and the historical standard feature vector, filter out similar historical standard feature vectors that meet the preset similarity threshold, and take the scene corresponding to the similar historical standard feature vector as the historical similar scene. Obtain multiple historical predicted transmission distances and historical actual transmission distances corresponding to similar historical scenarios, calculate the difference between each historical predicted transmission distance and the corresponding historical actual transmission distance, and average the multiple differences to obtain the average prediction deviation data. Using the scene type and motion mode in the current standard feature vector as retrieval constraints, the standard distance parameter range of the current scene is obtained from the pre-stored scene standard parameter library; Using the standard distance parameter range as a boundary constraint, the basic predicted transmission distance is fused with the average prediction deviation data to obtain the predicted transmission distance.

[0085] In one embodiment, the multipath delay prediction module 14 is further configured to: The construction process of the multipath delay spread prediction model includes: From historical data, we collect multiple historical transmission distances associated with historical scene types and motion patterns to form a sample transmission distance set; For each historical transmission distance in the sample transmission distance set, obtain the real multipath delay spread for a future preset time period, and label multiple real multipath delay spreads as a sample multipath delay spread label set; A multipath delay spread prediction model is constructed based on a nonlinear regression algorithm; The multipath delay spread prediction model is trained and validated in a supervised manner using the sample transmission distance set and the sample multipath delay spread label set, and training is completed after convergence.

[0086] In one embodiment, the parameter configuration transmission module 15 is further configured to: Determine the time span of the multipath component corresponding to the predicted multipath delay spread, and calculate the order of the adaptive equalization filter based on the time span of the multipath component; Based on the multipath complexity of the current scenario, a preset multiple relationship between the symbol period and the predicted multipath delay spread is determined, and the transmission symbol period is calculated based on the predicted multipath delay spread and the preset multiple relationship. Based on the complexity of the multipath propagation in the current scenario, determine the multipath delay spread threshold range, identify the specific range within which the predicted multipath delay spread is located within the multipath delay spread threshold range, and determine the modulation method based on the anti-interference requirements corresponding to the specific range. The order, transmission symbol period, and modulation scheme of the adaptive equalization filter are configured into the transmission link, and multipath interference is compensated in real time through adaptive filtering to complete the remote transmission of sensor data.

[0087] The process includes determining the time span of the multipath components corresponding to the predicted multipath delay spread, and calculating the order of the adaptive equalization filter based on the time span of the multipath components, including: Obtain the time difference range between the multipath signal and the main path signal corresponding to the predicted multipath delay spread, and use it as the time span of the multipath component; The basic order of the adaptive equalization filter is set with reference to the time span of the multipath components. The ratio of the predicted multipath delay spread to the average multipath delay spread of similar historical scenarios is calculated and used as the order adjustment factor. The order of the adaptive equalizer is obtained by multiplying the base order of the adaptive equalizer by the order adjustment coefficient.

[0088] Specifically, based on the multipath complexity of the current scenario, a preset multiple relationship is determined between the symbol period and the predicted multipath delay spread, and the transmission symbol period is calculated based on the predicted multipath delay spread and the preset multiple relationship, including: The obstacle distribution density in the current scene is collected using environmental sensors; The ratio of obstacle distribution density to the standard obstacle distribution density of similar historical scenarios is used as the multipath complexity of the current scenario. Based on the complexity of the multipath, a preset multiple relationship is determined between the symbol period and the predicted multipath delay spread; The transmission symbol period is obtained by multiplying the preset multiplication factor relationship with the predicted multipath delay spread.

[0089] Specifically, based on the multipath complexity of the current scenario, a multipath delay spread threshold range is determined; the specific interval within which the predicted multipath delay spread falls within the multipath delay spread threshold range is identified; and the modulation scheme is determined based on the anti-interference requirements corresponding to the specific interval, including: Based on the complexity of the multipath, determine the multipath delay spread threshold range; The predicted multipath delay spread is compared with the multipath delay spread threshold range to determine the specific range in which the predicted multipath delay spread is located. The modulation method is selected based on the anti-interference requirements corresponding to the specific interval.

[0090] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0091] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0092] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for optimizing the effect of remote transmission of sensor data based on adaptive filtering, characterized in that, The method comprises: Collecting the transmission distance sequence of the last N time points, the scene type and the motion mode at the current time point, and binding the transmission distance data after standardization processing with the scene type and the motion mode to form a standard feature vector; Inputting the standard feature vector into a pre-trained transmission distance prediction model to predict and output the basic predicted transmission distance in a future preset period; Based on the prediction deviation data of the historical similar scene and the standard distance parameter of the current scene, the basic predicted transmission distance is corrected to obtain the predicted transmission distance; Inputting the predicted transmission distance into a pre-trained multipath time delay spread prediction model to predict and output the predicted multipath time delay spread in a future preset period; According to the predicted multipath time delay spread, the order of the adaptive equalization filter, the transmission symbol period and the modulation mode of the adaptive filter are dynamically configured for remote transmission of the sensing data. 2.The adaptive filter based sensor data remote transmission effect optimization method of claim 1, wherein, Collecting the transmission distance sequence of the last N time points, the scene type and the motion mode at the current time point, and binding the transmission distance data after standardization processing with the scene type and the motion mode to form a standard feature vector, comprising: Collecting historical transmission distance data at a preset sampling interval, and extracting transmission distance data of the last N time points to form a transmission distance sequence; Collecting environmental features through an environmental sensor to identify the current scene type; Calculating the speed change rate based on the transmission distance sequence, and determining the motion mode according to the speed change rate, wherein the motion mode includes static, uniform speed and variable speed; Normalizing the transmission distance sequence to obtain a standard transmission distance sequence; Concatenating the standard transmission distance sequence, the scene type and the motion mode in order to form a standard feature vector with a dimension of N+2. 3.The adaptive filter based sensor data remote transmission effect optimization method of claim 1, wherein, The construction process of the transmission distance prediction model comprises: Collecting a plurality of historical standard feature vectors from historical data to form a sample feature sequence set; Obtaining the real transmission distance corresponding to each historical standard feature vector in the sample feature sequence set in a future preset period, and labeling a plurality of the real transmission distances as a sample transmission distance set; Building a transmission distance prediction model based on an LSTM network; Using the sample feature sequence set and the sample transmission distance set to supervise the training and verification of the transmission distance prediction model, and completing the training after convergence.

4. The adaptive filtering based sensor data remote transmission effect optimization method according to claim 1, characterized in that, Based on the prediction deviation data of the historical similar scene and the standard distance parameter of the current scene, the basic predicted transmission distance is corrected to obtain the predicted transmission distance, comprising: Calculating the similarity of the standard feature vector and the historical standard feature vector, and screening out similar historical standard feature vectors that meet a preset similarity threshold, and taking the scene corresponding to the similar historical standard feature vector as a historical similar scene; Obtaining a plurality of historical predicted transmission distances and historical real transmission distances corresponding to the historical similar scene, calculating the difference between each historical predicted transmission distance and the corresponding historical real transmission distance, and averaging a plurality of difference values to obtain an average prediction deviation data; Taking the scene type and the motion mode in the current standard feature vector as a retrieval constraint to obtain the standard distance parameter range of the current scene from a pre-stored scene standard parameter library; The base predicted transmission distance is fused with the average predicted deviation data to obtain a predicted transmission distance, with the standard distance parameter range as a boundary constraint.

5. The adaptive filtering based sensor data remote transmission effect optimization method according to claim 1, characterized in that, The construction process of the multipath delay spread prediction model includes: From historical data, a plurality of historical transmission distances associated with historical scene types and motion modes are collected to form a sample transmission distance set; For each historical transmission distance in the sample transmission distance set, real multipath delay spreads of a future preset time period are obtained, and the plurality of real multipath delay spreads are labeled as a sample multipath delay spread label set; A multipath delay spread prediction model is constructed based on a nonlinear regression algorithm; The multipath delay spread prediction model is supervisedly trained and verified by using the sample transmission distance set and the sample multipath delay spread label set, and the training is completed after convergence.

6. The adaptive filtering based sensor data remote transmission effect optimization method according to claim 1, characterized in that, According to the predicted multipath delay spread, an adaptive equalization filter order, a transmission symbol period, and a modulation mode of adaptive filtering are dynamically configured for remote transmission of sensing data, including: determining a multipath component time span corresponding to the predicted multipath delay spread, and calculating an adaptive equalization filter order based on the multipath component time span; determining a preset multiple relationship between the symbol period and the predicted multipath delay spread according to the multipath complexity of the current scene, and calculating the transmission symbol period according to the predicted multipath delay spread and the preset multiple relationship; determining a multipath delay spread threshold interval according to the multipath complexity of the current scene, judging a specific interval in which the predicted multipath delay spread is located in the multipath delay spread threshold interval, and determining a modulation mode according to the anti-interference requirement corresponding to the specific interval; configuring the adaptive equalization filter order, the transmission symbol period, and the modulation mode to a transmission link, and compensating multipath interference in real time through adaptive filtering to complete remote transmission of sensing data.

7. The adaptive filter based remote sensing data transmission effect optimization method of claim 6, wherein, determining a multipath component time span corresponding to the predicted multipath delay spread, and calculating an adaptive equalization filter order based on the multipath component time span, including: obtaining a time difference range between a multipath signal and a main path signal corresponding to the predicted multipath delay spread as a multipath component time span; setting a basic order of the adaptive equalization filter with reference to the multipath component time span; calculating a ratio of the predicted multipath delay spread to an average multipath delay spread of historical similar scenes as an order adjustment coefficient; multiplying the basic order of the adaptive equalization filter by the order adjustment coefficient to obtain the adaptive equalization filter order. 8.The adaptive filter based sensor data remote transmission effect optimization method of claim 6, wherein, determining a preset multiple relationship between the symbol period and the predicted multipath delay spread according to the multipath complexity of the current scene, and calculating the transmission symbol period according to the predicted multipath delay spread and the preset multiple relationship, including: obtaining an obstacle distribution density of the current scene through an environmental sensor; calculating a ratio of the obstacle distribution density to a standard obstacle distribution density of historical similar scenes as a multipath complexity of the current scene; determining a preset multiple relationship between the symbol period and the predicted multipath delay spread according to the multipath complexity; multiplying the preset multiple relationship by the predicted multipath delay spread to obtain the transmission symbol period.

9. The adaptive filtering based sensor data remote transmission effect optimization method according to claim 6, characterized in that, According to the multipath complexity of the current scene, a multipath delay spread threshold interval is determined, it is judged that the predicted multipath delay spread is located in a specific interval in the multipath delay spread threshold interval, and a modulation mode is determined according to the anti-interference requirement corresponding to the specific interval, including: According to the multipath complexity, a multipath delay spread threshold interval is determined; The predicted multipath delay spread is compared with the multipath delay spread threshold interval to determine the specific interval in which the predicted multipath delay spread is located; According to the anti-interference requirement corresponding to the specific interval, a modulation mode is selected.

10. A system for optimizing the effect of remote transmission of sensor data based on adaptive filtering, characterized in that, The system for implementing the adaptive filtering-based remote transmission effect optimization method of the sensor data according to any one of claims 1-9, the system comprising: A feature vector generation module for collecting the transmission distance sequence of the last N moments, the scene type and the motion mode at the current moment, and binding the standardized transmission distance data with the scene type and the motion mode to form a standard feature vector; A basic distance prediction module for inputting the standard feature vector into a pre-trained transmission distance prediction model to predict and output the basic predicted transmission distance in a future preset period; A predicted distance correction module for correcting the basic predicted transmission distance based on the prediction deviation data of the same type of historical scene and the standard distance parameter of the current scene to obtain the predicted transmission distance; A multipath delay prediction module for inputting the predicted transmission distance into a pre-trained multipath delay spread prediction model to predict and output the predicted multipath delay spread in a future preset period; A parameter configuration transmission module for dynamically configuring the adaptive equalization filter order, the transmission symbol period and the modulation mode of the adaptive filter according to the predicted multipath delay spread to remotely transmit the sensor data.

Images