Multipath component clustering and tracking method for channel estimation of unmanned aerial vehicle
By using multipath component clustering and tracking methods in the UAV channel, dynamically adjusting the clusters and using classification models to predict cluster labels, the problems of high computational complexity, poor real-time performance and insufficient clustering accuracy of channel dynamic clustering in the prior art are solved, and high accuracy and high adaptability channel modeling is achieved.
Patent Information
- Application Number
- CN202411989470.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-31
- Publication Date
- 2025-04-04
AI Technical Summary
The existing dynamic clustering method of drone channel has high computational complexity and is difficult to guarantee real-time. It has insufficient clustering accuracy and is susceptible to noise interference, so it cannot effectively characterize the time-varying characteristics of drone channels.
The multipath component clustering and tracking method is used to cluster the multipath components at the initial moment through the KPD algorithm, and the cluster is dynamically adjusted at each moment. The classification model is used to predict the cluster label of the unmatched multipath components to ensure that the boundary and density threshold of the cluster are dynamically adjusted to adapt to channel changes.
It improves the accuracy and stability of channel clustering, enhances the adaptability to time-varying channels, avoids noise interference, and ensures real-time and efficient channel modeling.
Smart Images
Figure CN119766367A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of communication technology and relates to a multipath component clustering and tracking method for unmanned aerial vehicle channel estimation. Background Art
[0002] With their low cost and high flexibility, drones have great application potential and market demand in military communications, disaster relief, logistics and transportation, and smart cities. In order to meet the high reliability and high performance requirements of drone communications in diverse scenarios, accurate modeling of drone channel characteristics has become a key link in the research of wireless communication systems. Channel modeling is not only the basis of communication system design, but also directly affects the reliability, coverage and spectrum utilization efficiency of the system. Therefore, channel modeling research for drones, especially dynamic channel modeling in complex environments, is an important technical guarantee to ensure the stable and efficient operation of drone communication systems.
[0003] In recent years, machine learning has received extensive attention in channel modeling research due to its powerful nonlinear fitting capabilities and data-driven characteristics. Machine learning technology can learn features from a large amount of observational data and explore implicit laws in complex propagation environments, providing more accurate models and prediction capabilities for channel modeling. In the field of UAV channel modeling, the introduction of machine learning algorithms can effectively solve the problem of insufficient generalization ability faced by traditional channel modeling methods when dealing with diverse and dynamic environments. By training massive channel data, modeling methods based on machine learning can build models with high adaptability and high precision, further improving the reliability and efficiency of channel simulation. At present, many international scientific research institutions and scholars are committed to combining machine learning with channel modeling, and have achieved initial results in the field of UAV communications.
[0004] In a real propagation environment, signal propagation is affected by physical phenomena such as reflection, scattering, and diffraction, resulting in the formation of multiple paths of signal propagation, namely the multipath effect. Multipath signals often appear in the form of clusters, which can be described as a group of multipath signals with similar delay and angle characteristics. Cluster-based channel modeling methods can better reflect the fading characteristics, delay spread, and spatial distribution characteristics of the channel by grouping multipath signals, thereby achieving more realistic results in the simulation of communication systems. Given its significant modeling advantages, most communication standards (such as 3GPP and COST models) have incorporated cluster-based channel modeling as an important component when designing channel models.
[0005] However, most clustering algorithms are still mainly based on static channel models, and fail to fully consider the dynamic characteristics of the channel. In the UAV channel environment, due to the mobility of the UAV and the presence of dynamic obstacles in the propagation environment, the channel has significant time-varying characteristics, and the traditional static model cannot accurately describe these changes. In addition, although some algorithms take time-varying characteristics into consideration, their computational complexity is high and real-time performance is difficult to guarantee. At the same time, the accuracy of cluster recognition in each time snapshot is insufficient, resulting in the dynamic tracking of clusters being susceptible to noise interference, which affects the actual application effect.
[0006] In summary, the existing channel dynamic clustering methods have high computational complexity and are difficult to ensure real-time performance. In addition, the clustering accuracy is insufficient and they are easily affected by noise. Summary of the invention
[0007] In order to solve the above-mentioned prior art problems, the present invention adopts a multipath component clustering and tracking method for UAV channel estimation, comprising:
[0008] S1. Obtain the UAV channel data at the initial moment and extract the characteristic parameter S of the multipath component in the channel data at the initial moment 1 ; According to the characteristic parameter S of the multipath component 1 The KPD algorithm is used to cluster the multipath components at the initial moment to obtain multiple clusters at the initial moment. Among them, k is the label of the cluster;
[0009] S2. Obtain the characteristic parameter S of the multipath component at the previous moment m-1 m-1 and Cluster Get the UAV channel data at the current time m, and extract the characteristic parameter S of the multipath component in the channel data at the current time m ;
[0010] S3, according to the characteristic parameter S m-1 and S m Optimally match the multipath components at the previous time m-1 and the current time m to obtain the matched multipath components and the unmatched multipath components at the current time m;
[0011] S4. Directly assign the multipath component matched at the current moment m to the cluster corresponding to the multipath component matched with it. In the classification model, the multipath components that are not matched at the current moment m are assigned to the corresponding clusters In the cluster
[0012] S5. Based on cluster Calculate boundary threshold and density threshold De m , according to the boundary threshold and density threshold Dem Cluster Perform dynamic adjustments to obtain multiple clusters at the current moment
[0013] Characteristic parameters of the multipath component MPC at time m in, is the characteristic parameter of the nth multipath component at time m, represents the phase of the multipath component, represents the transmitting azimuth, represents the receiving azimuth, represents the launch elevation angle, represents the receiving elevation angle, represents the delay of the multipath component, and N is the number of multipath components.
[0014] Clustering the multipath components at the initial moment includes:
[0015] Set the number of clusters K and calculate the density of each multipath component at the initial moment According to the density of each multipath component Calculate the relative density of each multipath component The relative density The largest K multipath components are used as the initial clustering centers. The distances between the remaining multipath components and the initial clustering centers at the initial moment are calculated. The remaining multipath components are assigned to the cluster corresponding to the initial clustering center with the shortest distance to obtain multiple clusters at the initial moment.
[0016] According to the characteristic parameter S m-1 and S m The optimal matching of the multipath components at the previous time m-1 and the current time m includes: according to the characteristic parameters S of the multipath components m-1 and S m Construct the probability correlation matrix cost of the multipath components at the previous moment m-1 and the current moment m m-1,m , according to the probability association matrix cost m-1,m The multipath components at the previous time m-1 and the current time m are optimally matched to obtain the matched multipath components and the unmatched multipath components at the current time m.
[0017] Probability association matrix Each element of the matrix represents the pairing probability of the multipath component at the previous moment m-1 and the current moment m It represents the distance between the n-th multipath component at the previous time m-1 and the n′th multipath component at time m, where N′ is the number of multipath components.
[0018] According to the probability association matrix cost m-1,mThe optimal matching of the multipath components at the previous time m-1 and the current time m includes:
[0019] According to the probability association matrix cost m-1,m Constructing the objective function U is a set of multipath components; the Kuhn-Munkres algorithm is used to solve the objective function to obtain the matched multipath components and the unmatched multipath components at the current moment m; wherein the Kuhn-Munkres algorithm is a Hungarian algorithm.
[0020] Assign unmatched multipath components to clusters Included:
[0021] According to the characteristic parameter S of the multipath component at the last moment m-1 m-1 Train the classification model, input the characteristic parameters of the unmatched multipath components into the trained classification model, obtain the cluster labels of the unmatched multipath components, and assign the unmatched multipath components to the corresponding clusters according to the cluster labels. middle.
[0022] According to cluster Calculate boundary threshold Including: setting the number threshold Q of multipath components, if the cluster There are Q multipath components at time m to the centroid of the cluster If the distances are the same, then Q multipath components are placed at the centroid of the cluster. The distance is used as the boundary threshold Otherwise, the boundary threshold Set to off-cluster The centroid The multipath component and centroid at the farthest time m distance.
[0023] According to cluster Calculate the density threshold De m include:
[0024]
[0025] in, Cluster A collection of Cluster The number of multipath components at time m, Cluster area.
[0026] Using Thresholds and threshold value De m Cluster The dynamic adjustment includes: calculating each multipath component of m at the current time to the corresponding cluster The centroid The distance is greater than the threshold The multipath components from the cluster Eliminate from the clusters and get and the multipath components to be removed; set the threshold value Q of the number of multipath components, sort all the multipath components to be removed from small to large according to the distance to the centroid of the cluster to which they belong, select the multipath components with the smallest distance, calculate the density De′ of the selected Q multipath components, if De′ is greater than the threshold value De m , all the removed multipath components are aggregated into a new cluster; otherwise, all the removed multipath components are marked as noise.
[0027] Beneficial effects:
[0028] 1. The present invention dynamically trains a classification model at each moment, and uses the classification model to predict the cluster labels of unmatched multipath components, and allocates the unmatched multipath components to corresponding clusters according to the cluster labels, avoiding interference from noise points, ensuring the accuracy of channel clustering, and enhancing the adaptability to time-varying channels; 2. After allocating corresponding multipath components to corresponding clusters at each moment, the present invention dynamically adjusts the boundary threshold and density threshold of the cluster, and uses the boundary threshold and density threshold to dynamically adjust the cluster to prevent mutations of multipath components and interference from noise points, ensuring the stability and accuracy of channel clustering, and enhancing the adaptability to time-varying channels. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 A flowchart of a multipath component clustering and tracking method for UAV channel estimation provided by an embodiment of the present invention;
[0030] Figure 2 A flow chart of dynamic adjustment of clusters provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0032] like Figure 1 As shown, the embodiment of the present invention adopts a multipath component clustering and tracking method for unmanned aerial vehicle channel estimation, including:
[0033] S1. Obtain the UAV channel data at the initial moment and extract the characteristic parameter S of the multipath component in the channel data at the initial moment 1 ; According to the characteristic parameter S of the multipath component1 The KPD algorithm (Kernel Power Density Based Algorithm) is used to cluster the multipath components at the initial moment to obtain multiple clusters at the initial moment. Among them, k is the label of the cluster;
[0034] Extracting characteristic parameters of multipath components in channel data includes: extracting characteristic parameters of each multipath component MPC from channel data by high-resolution parameter estimation algorithm in, is the characteristic parameter of the nth multipath component at time m, represents the phase of the multipath component, represents the transmitting azimuth, represents the receiving azimuth, represents the launch elevation angle, represents the receiving elevation angle, represents the delay of the multipath component, and N is the number of multipath components at time m.
[0035] According to the characteristic parameter S of the multipath component 1 The KPD algorithm is used to cluster the multipath components at the initial moment, including:
[0036] S11. Set the number of clusters K and calculate the density of each multipath component at the initial moment According to the density of each multipath component Calculate the relative density of each multipath component The relative density The largest K multipath components are used as the initial cluster centers; the value range of K is 2 to 30.
[0037] Specifically, density and relative density The calculation formula is as follows:
[0038]
[0039] in, is the set of Z multipath components that are closest to the nth multipath component at time m. The value range of Z is 1 to 20. The specific value needs to be obtained according to the actual situation of the multipath component. n′ is the index of the multipath component. is the standard deviation of multipath component delay at time m, is the standard deviation of the multipath component transmission azimuth at time m, is the standard deviation of the multipath component receiving azimuth angle at time m, is the standard deviation of the multipath component transmission elevation angle at time m, is the standard deviation of the multipath component receiving elevation angle at time m.
[0040] S12, calculating the distance between the remaining multipath components and the initial cluster center at the initial moment, and assigning the remaining multipath components to the cluster corresponding to the initial cluster center with the shortest distance thereto, thereby forming multiple multipath component clusters This step provides a preliminary cluster structure for subsequent dynamic adjustments.
[0041] S2. Obtain the characteristic parameter S of the multipath component at the previous moment m-1 m-1 and Cluster Get the UAV channel data at the current time m, and extract the characteristic parameter S of the multipath component in the channel data at the current time m ;
[0042] S3, according to the characteristic parameter S m-1 and S m Optimally match the multipath components at the previous time m-1 and the current time m to obtain the matched multipath components and the unmatched multipath components at the current time m;
[0043] According to the characteristic parameter S m-1 and S m The optimal matching of the multipath components at the previous time m-1 and the current time m includes:
[0044] S31, according to the characteristic parameter S of the multipath component m-1 and S m Construct the probability correlation matrix cost of the multipath components at the previous moment m-1 and the current moment m m-1,m :
[0045]
[0046]
[0047] in, represents the distance between the nth multipath component at the previous moment m-1 and the n′th multipath component at the moment m; wherein N′ is the number of multipath components; preferably, D() represents the Euclidean distance.
[0048] S32, according to the probability association matrix cost m-1,m Constructing the objective function U is the set of multipath components;
[0049] S33. Use the Kuhn-Munkres algorithm to solve the objective function and obtain the matched multipath components and the unmatched multipath components at the current moment m; wherein the Kuhn-Munkres algorithm is the Hungarian algorithm.
[0050] This step determines the optimal trajectory of the MPC at different times, ensuring continuous tracking of multipath components.
[0051] S4. The matched multipath components are directly assigned to the clusters corresponding to the matched multipath components. In the classification model, the multipath components that are not matched at the current moment m are assigned to the corresponding clusters In the cluster And update the cluster The centroid of According to cluster Calculate boundary threshold and density threshold De m ;
[0052] Use the classification model to assign the multipath components that are not matched at the current moment m to the corresponding clusters Including: according to the characteristic parameter S of the multipath component at the previous moment m-1 m-1 Train the classification model, input the characteristic parameters of the unmatched multipath components into the trained classification model, obtain the cluster labels of the unmatched multipath components, and assign the unmatched multipath components to the corresponding clusters according to the cluster labels. In the example above, we get the cluster at the current moment.
[0053] According to the characteristic parameter S of the multipath component at the last moment m-1 m-1 Training the classification model includes: taking the characteristic parameter S of the multipath component at the previous moment m-1 m-1 As a training set, the cluster to which the multipath component belongs The number k is used as the true label of the multipath component to train the classification model.
[0054] Preferably, the classification model is a support vector machine.
[0055] Centroid The update formula is:
[0056]
[0057] in, Representative cluster The power of the i-th multipath component at time m is, Cluster The characteristic parameters of the i-th multipath component with power removed at time m.
[0058] S5. Based on cluster Calculate boundary threshold and density threshold De m , according to the boundary threshold and density threshold De m Cluster Perform dynamic adjustments to obtain multiple clusters at the current moment
[0059] In order to adapt to the dynamic changes of the channel, this embodiment uses the boundary threshold and the density threshold as the initial basis for the dynamic adjustment of the cluster structure, ensuring that the boundary range and density distribution of the cluster remain stable at different times.
[0060] Specifically, according to the cluster Calculate boundary threshold Includes: Threshold To satisfy the following formula:
[0061]
[0062] in, Representation Cluster The centroid of the medium to cluster The distance is The number of multipath components at the moment m is Q, which is the threshold of the number of multipath components and has a value range of 10-30.
[0063] If not present If the above formula is satisfied, the boundary threshold Set to the number of multipath components at time m that is farthest from the cluster center.
[0064] According to cluster Calculate the density threshold De m include:
[0065]
[0066] in, Cluster A collection of It is a cluster The number of multipath components at time m, Representative cluster area, is the time m and the center of mass The multipath component with the greatest distance.
[0067] like Figure 2 As shown, according to the boundary threshold and density threshold De m Cluster The dynamic adjustment includes: calculating each multipath component of m at the current time to the corresponding cluster The centroid Distance Distance Greater than threshold The multipath components from the cluster Eliminate from the clusters and get And update the cluster The centroid of All the removed multipath components are regarded as new clusters or marked as noise points;
[0068] Centroid The update formula is:
[0069]
[0070] in, Representative cluster The power of the i-th multipath component at time m is, Cluster The characteristic parameters of the i-th multipath component with power removed at time m.
[0071] All the removed multipath components are treated as new clusters or marked as noise points, including:
[0072] According to the distance to the centroid of the cluster Sort all the eliminated multipath components from small to large, select the first Q multipath components with the smallest distance, and calculate the density De′ based on the selected Q multipath components. The calculation formula is as follows: Where A′ is the area of the selected Q multipath components. If De′ exceeds the threshold value De m , all the removed multipath components are aggregated into a new cluster; otherwise, all the removed multipath components are marked as noise.
[0073] The calculation process of the area of Q multipath components includes: taking Q multipath components as a cluster and calculating the centroid of the cluster, then the area is in, is the distance from the multipath component farthest from the centroid to the centroid.
[0074] The above embodiments further illustrate the purpose, technical solutions and advantages of the present invention in detail. It should be understood that the above embodiments are only preferred implementation modes of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made to the present invention within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A multipath component clustering and tracking method for unmanned aerial vehicle channel estimation, characterized in that: include: S1. Obtain the UAV channel data at the initial moment and extract the characteristic parameter S of the multipath component in the channel data at the initial moment 1 ; According to the characteristic parameter S of the multipath component 1 The KPD algorithm is used to cluster the multipath components at the initial moment to obtain multiple clusters at the initial moment. Among them, k is the label of the cluster, and the KPD algorithm is the kernel power density algorithm; S2. Obtain the characteristic parameter S of the multipath component at the previous moment m-1 m-1 and Cluster Get the UAV channel data at the current time m, and extract the characteristic parameter S of the multipath component in the channel data at the current time m ; S3, according to the characteristic parameter S m-1 and S m Optimally match the multipath components at the previous time m-1 and the current time m to obtain the matched multipath components and the unmatched multipath components at the current time m; S4. Directly assign the multipath component matched at the current moment m to the cluster corresponding to the multipath component matched with it. In the classification model, the multipath components that are not matched at the current moment m are assigned to the corresponding clusters In the cluster S5. Based on cluster Calculate boundary threshold and density threshold De m , according to the boundary threshold and density threshold De m Cluster Perform dynamic adjustments to obtain multiple clusters at the current moment 2. The multipath component clustering and tracking method for unmanned aerial vehicle channel estimation according to claim 1, characterized in that: Characteristic parameters of the multipath component MPC at time m in, is the characteristic parameter of the nth multipath component at time m, represents the phase of the multipath component, represents the transmitting azimuth, represents the receiving azimuth, represents the launch elevation angle, represents the receiving elevation angle, represents the delay of the multipath component, and N is the number of multipath components.
3. The multipath component clustering and tracking method for unmanned aerial vehicle channel estimation according to claim 2 is characterized in that: Clustering the multipath components at the initial moment includes: Set the number of clusters K and calculate the density of each multipath component at the initial moment According to the density of each multipath component Calculate the relative density of each multipath component The relative density The largest K multipath components are used as the initial clustering centers. The distances between the remaining multipath components and the initial clustering centers at the initial moment are calculated. The remaining multipath components are assigned to the cluster corresponding to the initial clustering center with the shortest distance to obtain multiple clusters at the initial moment.
4. The multipath component clustering and tracking method for unmanned aerial vehicle channel estimation according to claim 2, characterized in that: According to the characteristic parameter S m-1 and S m The optimal matching of the multipath components at the previous time m-1 and the current time m includes: according to the characteristic parameter S of the multipath component m-1 and S m Construct the probability correlation matrix cost of the multipath components at the previous moment m-1 and the current moment m m-1,m , according to the probability association matrix cost m-1,m The multipath components at the previous time m-1 and the current time m are optimally matched to obtain the matched multipath components and the unmatched multipath components at the current time m.
5. The multipath component clustering and tracking method for unmanned aerial vehicle channel estimation according to claim 4, characterized in that: Probability association matrix Each element of the matrix represents the pairing probability of the multipath component at the previous moment m-1 and the current moment m It represents the distance between the n-th multipath component at the previous time m-1 and the n′th multipath component at time m, where N′ is the number of multipath components.
6. The multipath component clustering and tracking method for unmanned aerial vehicle channel estimation according to claim 5, characterized in that: According to the probability association matrix cost m-1,m The optimal matching of the multipath components at the previous time m-1 and the current time m includes: According to the probability association matrix cost m-1,m Constructing the objective function U is a set of multipath components; the Kuhn-Munkres algorithm is used to solve the objective function to obtain the matched multipath components and the unmatched multipath components at the current moment m; wherein the Kuhn-Munkres algorithm is a Hungarian algorithm.
7. The multipath component clustering and tracking method for unmanned aerial vehicle channel estimation according to claim 1, characterized in that: Assign unmatched multipath components to clusters Included: According to the characteristic parameter S of the multipath component at the last moment m-1 m-1 Train the classification model, input the characteristic parameters of the unmatched multipath components into the trained classification model, obtain the cluster labels of the unmatched multipath components, and assign the unmatched multipath components to the corresponding clusters according to the cluster labels. middle.
8. The multipath component clustering and tracking method for unmanned aerial vehicle channel estimation according to claim 1, characterized in that: According to cluster Calculate boundary threshold Including: setting the number threshold Q of multipath components, if the cluster There are Q multipath components at time m to the centroid of the cluster The distances of Q multipath components to the centroid of the cluster are the same. The distance is used as the boundary threshold Otherwise, the boundary threshold Set to off-cluster The centroid The multipath component and centroid at the farthest time m distance.
9. The multipath component clustering and tracking method for unmanned aerial vehicle channel estimation according to claim 1, characterized in that: According to cluster Calculate the density threshold De m include: in, Cluster A collection of Cluster The number of multipath components at time m, Cluster area.
10. The multipath component clustering and tracking method for unmanned aerial vehicle channel estimation according to claim 1, characterized in that: Using Thresholds and threshold value De m Cluster The dynamic adjustment includes: calculating each multipath component of m at the current time to the corresponding cluster The centroid The distance is greater than the threshold The multipath components from the cluster Eliminate from the clusters and get and the multipath components to be removed; set the threshold value Q of the number of multipath components, sort all the multipath components to be removed from small to large according to the distance to the centroid of the cluster to which they belong, select the multipath components with the smallest distance, calculate the density De′ of the selected Q multipath components, if De′ is greater than the threshold value De m , all the removed multipath components are aggregated into a new cluster; otherwise, all the removed multipath components are marked as noise.