An interference discrimination and multipath tracking method for dynamic UAV channel measurement

Abstract

The invention discloses an interference discrimination and multipath tracking method for dynamic unmanned aerial vehicle channel measurement, including: constructing an interference discrimination initial network, the input of the interference discrimination initial network is channel impulse response sampling data of length L, and the output of the interference discrimination initial network is 1 or 0; constructing an interference discrimination data set, and using the interference discrimination data set to train the interference discrimination network; constructing a channel multipath tracking network; inputting the channel impulse response slices obtained at the instantaneous moment in the actual channel measurement into the trained interference discrimination network, and outputting the discrimination result; according to the discrimination result, using a segmented short-time training method, the channel multipath tracking network is trained, and the channel impulse response of the interfered channel is multipath tracked. The present invention can realize the efficient and accurate acquisition of multipath parameters of real-time measurement of unmanned aerial vehicle dynamic channels in the presence of random burst interference.

Description

Technical Field

[0001] The present invention relates to the field of wireless information transmission, and in particular to an interference discrimination and multipath tracking method for dynamic unmanned aerial vehicle channel measurement, and is particularly suitable for multipath parameter estimation applications in high-dynamic unmanned aerial vehicle channel measurement in the presence of random burst interference. Background Art

[0002] With the rapid development and widespread application of drones, the number of drones operating simultaneously in the air will increase dramatically in the future. In addition, under the vision of building the sixth-generation mobile communication system, drones will also become an important part of accessing the mobile Internet. In this application scenario of multi-machine operation and multi-node network, the effective received signal is easily interfered by other transmission links. For communication, normal communication can still be guaranteed when the strongest transmission path is not overwhelmed by interference. For channel measurement, even if the strongest propagation path is still identifiable, some multipaths weaker than the strongest path may still be overwhelmed by interference, which seriously affects the accuracy of the channel estimation results. Therefore, it is urgent to study the drone channel measurement technology in the presence of interference.

[0003] One solution is to restore the original signal as much as possible through specific technical means, which mainly includes two directions. One direction is to use signal processing methods (such as filtering) or machine learning methods to weaken the interference signal, but such methods require a certain understanding of the characteristics of the interference signal or require a large amount of training data of the interference signal type. The other direction is to design special signals or encoding methods. This method can reduce the impact of the interference signal to a certain extent and has good applicability for communication itself, but it still cannot effectively restore all multipath information for channel measurement.

[0004] Another solution is to start from the application results, identify the data of the interfered fragments, and directly compensate and optimize the interfered results. Taking the channel impulse response obtained by channel measurement as an example, the interfered data is restored or predicted by using the characteristics of the surrounding undisturbed data. Among them, a common method is to use the valid data before and after the interfered data to complete and restore the current interference data, but this method requires the provision of historical and future data at the current moment, which is not suitable for channel estimation in application scenarios such as UAV dynamic measurement or interaesthesia integration with high real-time requirements. Summary of the invention

[0005] The purpose of the present invention is to provide an interference discrimination and multipath tracking method for dynamic UAV channel measurement, which combines the interference intelligent discrimination technology and the multipath tracking technology based on the channel evolution law, and can judge whether there is interference without knowing the interference signal characteristics in advance and take different channel parameter processing methods accordingly, so as to realize the efficient and accurate acquisition of multipath parameters of real-time measurement of UAV dynamic channels in the presence of random burst interference.

[0006] In order to achieve the above technical objectives, the technical solution adopted by the present invention is:

[0007] An interference discrimination and multipath tracking method for dynamic unmanned aerial vehicle channel measurement, the interference discrimination and multipath tracking method comprising the following steps:

[0008] Step A, constructing an initial interference discrimination network, the input of the initial interference discrimination network is channel impulse response sampling data of length L, and the output of the initial interference discrimination network is 1 or 0, 0 corresponds to non-interfered data, and 1 corresponds to interfered data;

[0009] Step B, constructing an interference discrimination data set, the interference discrimination data set consists of non-interfered data and interfered data in a 1:1 ratio, and using the interference discrimination data set to train the interference discrimination network until convergence;

[0010] Step C: construct a channel multipath tracking network. The input data dimension of the channel multipath tracking network is L cir ×Ts eq , where L cir is the length of the impulse response of a slice channel, T seq is the number of required history slices, and the output dimension of the channel multipath tracking network is 1×L cir ;

[0011] Step D: Input the channel impulse response slice obtained at the instant in the actual channel measurement into the trained interference discrimination network, and output the discrimination result as I predict ;

[0012] Step E: Based on the determination result of step D, a segmented short-time training method is used to train the channel multipath tracking network constructed in step C, and multipath tracking is performed on the channel impulse response of the interfered channel.

[0013] Furthermore, in step A, the interference discrimination initial network includes an input layer, a first convolutional layer, a second convolutional layer, a third convolutional layer, a pooling layer, a fully connected layer and an output layer connected in sequence; wherein, the input size of the input layer is 1×1024, the step sizes of the first convolutional layer, the second convolutional layer and the third convolutional layer are all 1, and the convolution kernel sizes are 1×5×128, 1×3×64, and 1×3×64 respectively; the step size of the pooling layer is 1, and the pooling kernel size is 2×2×64; the fully connected layer includes 50 neurons.

[0014] Furthermore, in step B, the process of constructing the interference discrimination data set includes the following steps:

[0015] Step B1, obtaining a certain amount of channel impulse response data with a slice length of L through software simulation or field measurement in an interference frequency band;

[0016] Step B2, selecting a number of non-interfered data and interfered data in a 1:1 ratio from the channel impulse response data obtained in step B1 to construct an interference discrimination data set, and labeling the channel impulse response data in the interference discrimination data set in the form of 0 and 1, where 0 corresponds to non-interfered data and 1 corresponds to interfered data, and the label value set is recorded as I;

[0017] Step B3: randomly split the interference discrimination data set constructed in step B2 into a training data set D in a ratio of m:n train And the verification data set D val , and the label set corresponding to the training data set is recorded as I train , the label set corresponding to the verification data set is denoted as I val ;

[0018] Step B4: The training data set D in step B3 is train Input the initial interference discrimination network constructed in step A, combine the output of the initial interference discrimination network and the label set I in step B3 train Feedback is given to the initial interference discrimination network to optimize the parameter set of the initial interference discrimination network. The optimized parameter set R0 of the initial interference discrimination network is expressed as:

[0019]

[0020] Where R is the parameter set of the initial network for interference discrimination, C(·) is the loss function, and f(·) is the network transmission function;

[0021] Step B5: Verify the data set D in step B3 val Input the interference discrimination initial network trained and optimized in step B4, and compare the output result of the interference discrimination initial network with the label set I in step B3. valCompare and count the verification set discrimination accuracy. If the discrimination accuracy is lower than the threshold Thr, repeat steps B4 to B5 until the verification set discrimination accuracy meets the requirement.

[0022] Furthermore, the channel multipath tracking network includes five neural networks connected in sequence; the input data dimension of each neural network is a×b, where a is the length of a certain slice channel impulse response, b is the required number of historical slices, the network output dimension is 1×a, and the connection operation of adjacent neural networks is expressed as:

[0023] x = [x1, x2];

[0024] In the formula, x1 and x2 are the vectors to be connected, and x is the connection result; the activation function tanh(z) is expressed as:

[0025]

[0026] In the formula, z is the input variable of the activation function;

[0027] The transfer function f of each neural network j It is expressed as:

[0028] f j =A j (w j x+b j );

[0029] Where j is the neural network number, A j is the activation function of neural network j, w j is the weight matrix of the neuron in neural network j, x is the neural network input vector, b j The bias matrix of the neurons in neural network j; where,

[0030]

[0031]

[0032] Furthermore, in step E, according to the determination result of step D, a segmented short-time training method is adopted to train the channel multipath tracking network constructed in step C, and the process of performing multipath tracking on the channel impulse response of the interfered channel includes the following steps:

[0033] Step E1, at time t i The judgment result I predict (t i ) to judge, if time t i The judgment result I predict (t i ) is 0, then the current channel impulse response slice h(t i), directly execute step E5, otherwise continue to execute step E2;

[0034] Step E2, at time t i-1 The judgment result I predict (t i-1 ) to make a judgment, if I predict (t i-1 ) is 0, then continue to step E3, otherwise directly go to step E4;

[0035] Step E3, select T seq Channel impulse response slices As a short-term training data set, the channel multipath tracking network is trained to obtain the channel multipath tracking network trained in the current segment;

[0036] Step E4, T seq Channel impulse response slices Input into the channel multipath tracking network trained in the current segment to obtain the tracking prediction t i The channel impulse response slice h at time predict (t i ), and use it as the output result;

[0037] Step E5, repeating steps E1 to E4, and performing tracking and prediction on the channel impulse response slice of the interfered channel.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] First, the interference discrimination and multipath tracking method for dynamic UAV channel measurement of the present invention combines intelligent interference discrimination and intelligent multipath tracking technology to achieve tracking and prediction of multipath information of the interfered channel without predicting the characteristics of the interference signal and eliminating the interference signal.

[0040] Second, the interference discrimination and multipath tracking method for dynamic UAV channel measurement of the present invention takes into account the short-time correlation characteristics of UAV channel evolution and proposes a segmented short-time multipath tracking network training using the interference intelligent discrimination results, which solves the problems of low efficiency and high redundancy of long-time dependent training and can improve the multipath tracking accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a flow chart of the interference discrimination and multipath tracking method for dynamic UAV channel measurement according to an embodiment of the present invention;

[0042] Figure 2 This is the result diagram of interference discrimination network training;

[0043] Figure 3 It is a schematic diagram of the multipath tracking network structure;

[0044] Figure 4 is a data graph containing impulse slices of the interfered and undisturbed channels;

[0045] Figure 5 This is a comparison chart of multipath tracking results. DETAILED DESCRIPTION

[0046] The embodiments of the present invention are further described in detail below with reference to the accompanying drawings.

[0047] The embodiment of the present invention discloses an interference discrimination and multipath tracking method for dynamic UAV channel measurement. The overall process of the scheme is as follows: Figure 1 As shown, the following steps are included:

[0048] Step 1: Construct an initial interference discrimination network. The input of the initial interference discrimination network is the channel impulse response sampling data of length L. The output of the initial interference discrimination network is 1 or 0 (corresponding to two states of yes or no).

[0049] Step 2: Construct an interference discrimination data set. The interference discrimination data set consists of uninterrupted data and interfered data in a 1:1 ratio. Train the interference discrimination network until convergence. The specific implementation steps are as follows:

[0050] 2.1) Obtain a certain amount of channel impulse response data with a slice length of L through software simulation or field measurement in an interference frequency band.

[0051] 2.2) Select an appropriate amount of undisturbed data and interfered data in a 1:1 ratio from the channel impulse response data obtained in step 2.1 to construct a data set, and label the channel impulse response data in the form of 0 and 1, where 0 corresponds to undisturbed data and 1 corresponds to interfered data, and the label value set is denoted as I.

[0052] 2.3) Randomly split the interference discrimination dataset constructed in step 2.2 into a training dataset D in a ratio of m:n train And the verification data set D val , and the label set corresponding to the training data set is recorded as I train , the label set corresponding to the verification data set is denoted as I val .

[0053] 2.4) The training data set D in step 2.3 train Input the initial interference discrimination network constructed in step 1, and combine the output of the initial interference discrimination network with the label set I in step 2.3 train Feedback is given to the initial interference discrimination network to optimize the network parameter set of the initial interference discrimination network, which can be expressed as:

[0054]

[0055] Where R is the network parameter set, C(·) is the loss function, and f(·) is the network transmission function.

[0056] 2.5) The verification data set D in step 2.3 val Input the interference discrimination initial network trained and optimized in step 2.4, and compare the output result of the interference discrimination initial network with the label set I in step 2.3 val Compare and count the verification set discrimination accuracy. If the discrimination accuracy is lower than the threshold Thr, repeat steps 2.4 to 2.5 until the verification set discrimination accuracy reaches the requirement and end step 2.

[0057] Step 3: Construct a channel multipath tracking network. The input data dimension of the channel multipath tracking network is L cir ×T seq , where L cir is the length of the impulse response of a slice channel, T seq is the number of required history slices, and the output dimension of channel multipath tracking is 1×L cir .

[0058] Step 4: Input the channel impulse response slice obtained at the instantaneous moment in the actual channel measurement into the interference discrimination network trained in step 2, and output the discrimination result as I predict .

[0059] Step 5: According to the judgment result of step 4, the segmented short-time training method is used to train the channel multipath tracking network constructed in step 3, and realize the multipath tracking of the impulse response of the interfered channel. The specific implementation steps are as follows:

[0060] 5.1) If at some time t i Judgment result I predict (t i ) is 0, then the current channel impulse response slice h(t i ), proceed directly to step 5.5, otherwise continue to step 5.2.

[0061] 5.2) If the judgment result I predict (t i ) is 1, indicating that the current channel impulse response slice is interfered. predict (t i-1 ) is 0, continue to step 5.3, otherwise skip step 5.3 and go directly to step 5.4.

[0062] 5.3) Select T seq Channel impulse response slices As a short-term training data set, the multipath tracking network is trained to obtain the multipath tracking network trained in the current segment.

[0063] 5.4) T seq Channel impulse response slices Input into the network trained in the current segment to obtain the tracking prediction t i The channel impulse response slice h at time predict (t i ) as the output result.

[0064] 5.5) Repeat steps 5.1 to 5.4 to implement tracking and prediction of the interfered channel impulse response slice.

[0065] Examples

[0066] The following is a clear and complete description of the technical solution in conjunction with a specific implementation example. This implementation example includes the following steps:

[0067] Step 1: Construct an initial interference discrimination network based on the convolutional neural network architecture. The specific parameters of the initial interference discrimination network are shown in Table 1. The input of the initial interference discrimination network is the channel impulse response sampling data with a length of 1024, and the output of the initial interference discrimination network is two values ​​of 1 or 0 (corresponding to the two states of being interfered or not being interfered).

[0068] Table 1 Specific parameter design of initial network for interference discrimination

[0069] Layer Index Layer Type Related parameters 1 Input Layer Input size is 1×1024 2 The first convolutional layer The convolution kernel size is 1×5×128, the stride is 1, and there is no padding. 3 The second convolutional layer The convolution kernel size is 1×3×64, the stride is 1, and there is no padding. 4 The third convolutional layer The convolution kernel size is 1×3×64, the stride is 1, and there is no padding. 5 Pooling Layer The pooling kernel size is 2×2×64, the stride is 1, and there is no padding. 6 Fully connected layer 50 neurons 7 Output Layer 2 categories

[0070] Step 2: Construct an interference discrimination data set. The interference discrimination data set consists of uninterrupted data and interfered data in a 1:1 ratio. Train the interference discrimination network until convergence. The specific implementation steps are as follows:

[0071] 2.1) This implementation case obtained some channel impulse response data with a slice length of 1024 samples by conducting field measurements in an interference frequency band.

[0072] 2.2) The channel impulse response data obtained in step 2.1 are selected from the non-interfered data and the interfered data in a 1:1 ratio to construct an interference discrimination data set, with a total of 200 slices, and the data are labeled in the form of 0 and 1, 0 corresponds to the non-interfered data, 1 corresponds to the interfered data, and the label value set is denoted as I.

[0073] 2.3) The interference discrimination data set constructed in step 2.2 is randomly split into the training data set D in a ratio of 8:2 train And the verification data set D val , and the label set corresponding to the training data set is recorded as I train , the label set corresponding to the verification data set is denoted as I val .

[0074] 2.4) The training data set D in step 2.3 train Input the initial interference discrimination network constructed in step 1, and combine the output of the initial interference discrimination network with the label set I in step 2.3 train Feedback is given to the initial interference discrimination network to optimize the network parameter set of the initial interference discrimination network, which can be expressed as:

[0075]

[0076] Where R is the network parameter set, C(·) is the loss function, and f(·) is the network transmission function.

[0077] 2.5) The verification data set D in step 2.3 val Input the interference discrimination initial network trained and optimized in step 2.4, and compare the network output result with the label set I in step 2.3 val Compare and count the accuracy of the verification set. In this implementation case, the accuracy threshold is set to 0.95. If the accuracy is lower than 0.95, repeat steps 2.4 to 2.5 until the accuracy reaches the requirement. Step 2 is ended. Figure 2 It can be seen that after 40 network trainings, the discrimination accuracy of this implementation case is stabilized at above 0.95.

[0078] Step 3: Construct a channel multipath tracking network based on time recursion. The network structure of the channel multipath tracking network is as follows: Figure 3 As shown, the input data dimension of the channel multipath tracking network is 1024×5, where 1024 is the length of the channel impulse response of a slice, 5 is the required number of historical slices, and the output dimension of the channel multipath tracking network is 1×1024. Figure 3 The join operation can be expressed as:

[0079] x=[x1,x2] (2);

[0080] In the formula, x1 and x2 are the vectors to be connected; Figure 3 Where tanh is the activation function, which can be expressed as:

[0081]

[0082] Figure 3 The transfer function of neural network 1 to neural network 5 can be expressed as:

[0083] f j =A j (W j x+b j ) (4); where j is the neural network number, A j is the activation function of neural network j, w jis the weight matrix of the neuron in neural network j, x is the neural network input vector, b j The bias matrix of the neurons in neural network j; where:

[0084]

[0085]

[0086] Step 4: Input the channel impulse response slice obtained at the instantaneous moment in the actual channel measurement into the interference discrimination network trained in step 2, and output the discrimination result as I predict , some measured slices of the interference and non-interference channel impulse responses are shown in Figure 4 shown.

[0087] Step 5: According to the judgment result of step 4, the segmented short-time training method is used to train the channel multipath tracking network constructed in step 3, and realize the multipath tracking of the impulse response of the interfered channel. The specific implementation steps are as follows:

[0088] 5.1) If at some time t i Judgment result I predict (t i ) is 0, then the current channel impulse response slice h(t i ), proceed directly to step 5.5, otherwise continue to step 5.2.

[0089] 5.2) If the judgment result I predict (t i ) is 1, indicating that the current channel impulse response slice is interfered. If Ip redict (t i-1 ) is 0, continue to step 5.3, otherwise skip step 5.3 and go directly to step 5.4.

[0090] 5.3) Select 5 channel impulse response slices {h(t i-6 ), h(t i-5 ), ..., h(t i-2 )} is used as a short-term training data set to train the multipath tracking network and obtain the multipath tracking network trained in the current segment.

[0091] 5.4) Slice the impulse responses of the five channels {h(t i-5 ), h(t i-4 ), ..., h(t i-1 )} is input into the network trained in the current segment to obtain the tracking prediction t i The channel impulse response slice h at time predict (t i ), and use it as the output result, Figure 4Taking the impulse response slice of the mid-channel as an example, in order to facilitate the verification and comparison of the prediction results, this case tracked and predicted five slices that were not interfered with and compared them with the actual measurement results, such as Figure 5 shown.

[0092] 5.5) Repeat steps 5.1 to 5.4 to implement tracking and prediction of the interfered channel impulse response slice.

[0093] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiments of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0094] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0095] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0096] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0097] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0098] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. An interference discrimination and multipath tracking method for dynamic UAV channel measurement, characterized in that: The interference discrimination and multipath tracking method comprises the following steps: Step A, constructing an interference discrimination network, the input of the interference discrimination network is the channel impulse response sampling data of length L, and the output of the interference discrimination network is 1 or 0, 0 corresponds to non-interfered data, and 1 corresponds to interfered data; Step B, constructing an interference discrimination data set, the interference discrimination data set consists of non-interfered data and interfered data in a 1:1 ratio, and using the interference discrimination data set to train the interference discrimination network until convergence; Step C: construct a channel multipath tracking network. The input data dimension of the channel multipath tracking network is L cir ×T seq , where L cir is the length of the impulse response of a slice channel, T seq is the number of required history slices, and the output dimension of the channel multipath tracking network is 1×L cir ; Step D: Input the channel impulse response slice obtained at the instant in the actual channel measurement into the trained interference discrimination network, and output the discrimination result as I predict ; Step E: Based on the determination result of step D, a segmented short-time training method is used to train the channel multipath tracking network constructed in step C, and multipath tracking is performed on the channel impulse response of the interfered channel.

2. The interference discrimination and multipath tracking method for dynamic UAV channel measurement according to claim 1 is characterized in that: In step A, the interference discrimination network includes an input layer, a first convolutional layer, a second convolutional layer, a third convolutional layer, a pooling layer, a fully connected layer and an output layer connected in sequence; wherein, the input size of the input layer is 1×1024, the step sizes of the first convolutional layer, the second convolutional layer and the third convolutional layer are all 1, and the convolution kernel sizes are 1×5×128, 1×3×64, and 1×3×64 respectively; the step size of the pooling layer is 1, and the pooling kernel size is 2×2×64; the fully connected layer includes 50 neurons.

3. The interference discrimination and multipath tracking method for dynamic UAV channel measurement according to claim 1 is characterized in that: In step B, the process of constructing the interference discrimination data set includes the following steps: Step B1, obtaining a certain amount of channel impulse response data with a slice length of L through software simulation or field measurement in an interference frequency band; Step B2, select a number of non-interfered data and interfered data in a 1:1 ratio from the channel impulse response data obtained in step B1 to construct an interference discrimination data set, and label the channel impulse response data in the interference discrimination data set in the form of 0 and 1, where 0 corresponds to non-interfered data and 1 corresponds to interfered data, and the label value set is recorded as 1; Step B3: randomly split the interference discrimination data set constructed in step B2 into a training data set D in a ratio of m:n train And the verification data set D val , and the label set corresponding to the training data set is recorded as I train , the label set corresponding to the verification data set is denoted as I val ; Step B4: The training data set D in step B3 is train Input the interference discrimination network constructed in step A, combine the output of the interference discrimination network and the label set I in step B3 train Feedback is given to the interference discrimination network to optimize the parameter set of the interference discrimination network. The optimized parameter set R0 of the interference discrimination network is expressed as: Where R is the parameter set of the interference discrimination network, C(·) is the loss function, and f(·) is the network transmission function; Step B5: Verify the verification data set D in step B3 val Input the interference discrimination network trained and optimized in step B4, and compare the output result of the interference discrimination network with the label set I in step B3. val Compare and count the verification set discrimination accuracy. If the discrimination accuracy is lower than the threshold Thr, repeat steps B4 to B5 until the verification set discrimination accuracy meets the requirement.

4. The interference discrimination and multipath tracking method for dynamic UAV channel measurement according to claim 1 is characterized in that: The channel multipath tracking network includes five neural networks connected in sequence; the input data dimension of each neural network is a×b, where a is the length of the impulse response of a certain slice channel, b is the required number of historical slices, and the network output dimension is 1×a. The connection operation of adjacent neural networks is expressed as: x=[x1,x2]; In the formula, x1 and x2 are the vectors to be connected, and x is the connection result; the activation function tanh(z) is expressed as: In the formula, z is the input variable of the activation function; The transfer function f of each neural network j It is expressed as: f j =A j (w j x+b j ); Where j is the neural network number, A j is the activation function of neural network j, w j is the weight matrix of the neuron in neural network j, x is the neural network input vector, b j The bias matrix of the neurons in neural network j; in, 5. The interference discrimination and multipath tracking method for dynamic UAV channel measurement according to claim 1 is characterized in that: In step E, according to the determination result of step D, a segmented short-time training method is adopted to train the channel multipath tracking network constructed in step C. The process of performing multipath tracking on the channel impulse response of the interfered channel includes the following steps: Step E1, at time t i The judgment result I predict (t i ) to judge, if time t i The judgment result I predict (t i ) is 0, then the current channel impulse response slice h(t i ), directly execute step E5, otherwise continue to execute step E2; Step E2, at time t i-1 The judgment result I predict (t i-1 ) to make a judgment, if I predict (t i-1 ) is 0, then continue to step E3, otherwise directly go to step E4; Step E3, select T seq Channel impulse response slices As a short-term training data set, the channel multipath tracking network is trained to obtain the channel multipath tracking network trained in the current segment; Step E4, T seq Channel impulse response slices Input into the channel multipath tracking network trained in the current segment to obtain the tracking prediction t i The channel impulse response slice h at time predict (t i ), and use it as the output result; Step E5, repeating steps E1 to E4, and performing tracking and prediction on the channel impulse response slice of the interfered channel.

Images