A formation flight control method suitable for UAV advertising

By calculating the similarity relationship and offset weight of the multidimensional curve data of the drone, combined with the Markov clustering algorithm, the problem of inaccurate clustering results in the drone formation is solved, and the accuracy and control effect of the flight path are improved.

CN118466584BActive Publication Date: 2024-11-15AERIAL LANDMARK (ZHEJIANG) TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410541486.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-04-30
Publication Date
2024-11-15
Estimated Expiration
2044-04-30

AI Technical Summary

Technical Problem

The existing drone formation control method in a multi-dimensional environment due to external factors and errors in different sensor data, resulting in inaccurate clustering results, which reduces the accuracy of data prediction and flight control effect.

Method used

By obtaining the multi-dimensional curve data of the drone, calculate the similarity relationship weight and offset weight of the curve dimensions, combined with the Markov clustering algorithm, calculate the similarity and transfer probability between drones, and perform clustering and path prediction control. , accurately adjust the flight path of the drone.

Benefits of technology

It improves the accuracy and clustering accuracy of the drone formation flight path, and enhances the accuracy of subsequent data prediction and flight control effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118466584B_ABST
    Figure CN118466584B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of non-electric variable control, and in particular to a formation flight control method suitable for UAV advertising, comprising: obtaining several curves of UAVs in a UAV formation, obtaining a curve dimension set and a curve set of UAVs according to the several curves; obtaining a similarity relationship weight of each curve dimension; obtaining the similarity between any two UAVs at each moment according to the similarity relationship weight and the distance between the UAVs; obtaining the offset weight of each UAV at each moment; obtaining the transfer probability between different UAVs according to the similarity and the offset weight; obtaining the predicted path offset value of each UAV at the next moment; and controlling the flight path of the UAV according to the path offset value at the current moment and the predicted path offset value. The present invention improves the centralized control effect of UAV formation flight.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of non-electric variable control, and in particular to a formation flight control method suitable for unmanned aerial vehicle advertising. Background Art

[0002] Traditional advertising methods are often limited by the limitations of ground-based perspectives, while drones’ aerial perspectives bring entirely new possibilities to advertising.

[0003] Usually, the control method for UAV formation combines the relevant data of the dimensions between each UAV to predict the error trajectory that may be generated at the next moment, and combines centralized control to compensate for the possible trajectory deviation. Since the deviation caused by external factors on each UAV in the UAV formation process is small, the conventional clustering algorithm cannot highlight the differences between each UAV. The Markov clustering algorithm can highlight the differences between different data points to improve the clustering accuracy. However, in the process of clustering each UAV using this algorithm, since the Markov chain is only based on the direct relationship and distance between different data points, it cannot reflect the multi-dimensional environment and its own data collected by the UAV, making the clustering conditions deviate too much from the actual situation. At the same time, different sensor data have different degrees of influence on the errors generated during the UAV flight process at different times. Therefore, there is a certain error in the results of obtaining the similarity of the flight environment between different UAVs, resulting in the final clustering not being able to accurately represent the similarity between UAVs, reducing the accuracy of subsequent data prediction, and thus making the centralized control of UAV formation flight poor. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a formation flight control method suitable for UAV advertising.

[0005] The present invention is a formation flight control method suitable for UAV advertising, which adopts the following technical solution:

[0006] An embodiment of the present invention provides a formation flight control method suitable for UAV advertising, the method comprising the following steps:

[0007] Acquire several curves of each drone in the drone formation, wherein the several curves at least include: a path deviation curve and a speed curve, wherein the curves include several moments, and acquire the distance between different drones at each moment;

[0008] According to several curves, a curve dimension set and a curve set of each UAV are obtained; according to the curve set of each UAV, a similarity relationship weight of each curve dimension in the curve dimension set is obtained; according to the similarity relationship weight and the distance between different UAVs at each moment, the similarity between any two UAVs at each moment is obtained;

[0009] Obtain the preset flight path and actual flight path of each drone in the drone formation, and map the preset flight path and actual flight path of each drone in the drone formation into a three-dimensional space model; obtain the preset flight path vector of each drone at each moment and the actual flight path vector at each moment in the three-dimensional space model; obtain the offset weight of each drone at each moment according to the preset flight path vector of the drone at each moment and the actual flight path vector at each moment;

[0010] The transfer probability between different UAVs is obtained according to the similarity and offset weight; all UAVs are clustered according to the transfer probability between different UAVs combined with the Markov clustering algorithm to obtain several clusters, and the predicted path offset value of each UAV in each cluster at the next moment is obtained; the flight path of the UAV is controlled according to the path offset value at the current moment and the predicted path offset value.

[0011] Furthermore, the method of obtaining the similarity relationship weight of each curve dimension in the curve dimension set according to the curve set of each drone includes the following specific steps:

[0012] Any drone in the drone formation is recorded as a target drone, the first curve in the curve set of the target drone is recorded as the first curve, the path offset curve of the target drone is recorded as the fixed curve, the Pearson correlation coefficient between the first curve and the fixed curve is recorded as the initial similarity relationship weight of the first curve of the target drone, and the set consisting of the initial similarity relationship weights of the first curve in the curve set of all drones is recorded as the initial similarity relationship weight set;

[0013] All initial similarity relationship weights in the initial similarity relationship weight set are linearly normalized, and the result is used as the normalized similarity relationship weight of the first curve in the curve set of each drone. The accumulated value of the normalized similarity relationship weights of the first curve in the curve set of all drones is used as the similarity relationship weight of the first curve dimension in the curve dimension set.

[0014] Furthermore, the similarity between any two drones at each moment is obtained according to the similarity relationship weight and the distance between different drones at each moment, and the specific steps include the following:

[0015] Record any moment as the target moment;

[0016]

[0017] Where N is the number of curve dimensions in the curve dimension set, μ i is the similarity relationship weight of the i-th curve dimension in the curve dimension set, T x,i is the parameter data of the i-th curve in the curve set of the x-th UAV at the target time, T y,i is the parameter data of the i-th curve in the curve set of the y-th UAV at the target time, || is the absolute value, D x,y is the distance between the xth UAV and the yth UAV at the target time, S x,y is the similarity between the x-th UAV and the y-th UAV at the target moment, and exp[] is an exponential function with a natural constant as the base.

[0018] Furthermore, the specific method of obtaining the preset flight path vector of each drone at each moment and the actual flight path vector at each moment is as follows:

[0019] Record any moment as the target moment, record the next moment of the target moment as the target adjacent moment, and record any UAV in the UAV formation as the target UAV;

[0020] The direction and distance between the target moment and the target adjacent moment in the preset flight path of the target UAV are obtained in the three-dimensional space model, and the vector determined by the direction and distance between the target moment and the target adjacent moment in the preset flight path of the target UAV is used as the preset flight path vector of the target UAV at the target moment. The direction and distance between the target moment and the target adjacent moment in the actual flight path of the target UAV is obtained in the three-dimensional space model, and the vector determined by the direction and distance between the target moment and the target adjacent moment in the actual flight path of the target UAV is used as the actual flight path vector of the target UAV at the target moment.

[0021] Furthermore, the specific steps of obtaining the offset weight of each drone at each moment according to the preset flight path vector of the drone at each moment and the actual flight path vector at each moment are as follows:

[0022] Any drone in the drone formation is recorded as the target drone;

[0023]

[0024] Where N j is the number of moments in the time range of the jth moment, d j,k is the actual flight path vector of the target UAV at the kth moment within the time range of the jth moment, d j,k+1is the actual flight path vector of the target UAV at the k+1th moment within the time range of the jth moment, v j,k is the actual speed of the target UAV at the kth moment within the time range of the jth moment, v j,k+1 is the actual speed of the target UAV at the k+1th moment within the time range of the jth moment, and the actual speed is obtained through the speed curve in the curve set of the target UAV, N0 is the total duration of the preset flight path, d1 t is the preset flight path vector of the target UAV at the tth moment within the total duration of the preset flight path, d1 t+1 v1 is the preset flight path vector of the target UAV at the t+1th moment within the total duration of the preset flight path. t v1 is the preset speed of the target UAV at the tth moment within the total duration of the preset flight path, t+1 is the preset speed of the target UAV at the t+1th moment within the total duration of the preset flight path. The preset speed is the modulus of the preset flight path vector at the corresponding moment. ΔG j is the path deviation value of the target UAV at the jth moment, ΔG max,j is the maximum value of the path deviation value of the target UAV at all times within the time range of the jth moment, and the path deviation value is obtained through the path deviation curve in the curve set of the target UAV, ΔG j,k is the path deviation value of the target UAV at the kth moment within the time range of the jth moment, ΔG j,k+1 is the path deviation value of the target UAV at the k+1th moment within the time range of the jth moment, ‖‖ is the binary norm, || is the absolute value, and u j is the offset weight of the target UAV at the jth moment.

[0025] Furthermore, the specific method for obtaining the time range of the j-th moment is as follows:

[0026] The range from the j-TH1th moment to the jth moment is taken as the time range of the jth moment, and TH1 is a preset first value.

[0027] Furthermore, the transfer probability between different drones is obtained according to the similarity and the offset weight, and the specific steps include the following:

[0028]

[0029] Where N0 is the total duration of the preset flight path, S1(x1,y1) a is the similarity between the x1th UAV and the y1th UAV at the ath moment within the total duration of the preset flight path, u x1,ais the offset weight of the x1th UAV at the ath moment, d1 x1,a is the preset flight path vector of the x1th UAV at the ath moment within the total duration of the preset flight path, u y1,a is the offset weight of the y1th UAV at the ath moment, d1 y1,a is the preset flight path vector of the y1th UAV at the ath moment within the total duration of the preset flight path, exp() is an exponential function with a natural constant as the base, ‖‖ is the second norm, P x1,y1 is the transfer probability from the x1th UAV to the y1th UAV.

[0030] Furthermore, the specific method for obtaining the predicted path deviation value of each drone in each cluster at the next moment is as follows:

[0031] The last moment within the total duration of the preset flight path is recorded as the current moment, and the moment after the current moment is recorded as the next moment. The path offset value of each UAV in each cluster at the next moment is obtained through the WMA algorithm. Any cluster is recorded as the target cluster, and the average value of the path offset values ​​of all UAVs in the target cluster at the next moment is used as the predicted path offset value of each UAV in the target cluster at the next moment.

[0032] Furthermore, the control of the flight path of the UAV according to the path deviation value at the current moment and the predicted path deviation value includes the following specific steps:

[0033] Any drone in the drone formation is recorded as the target drone;

[0034] The last moment within the total duration of the preset flight path is recorded as the current moment, and the moment after the current moment is recorded as the next moment. The path offset value of the target UAV at the current moment is obtained. According to the path offset value of the target UAV at the current moment and the predicted path offset value at the next moment, the preset flight path of the target UAV is adjusted in combination with the PID controller.

[0035] Furthermore, the step of obtaining a curve dimension set and a curve set of each drone according to a plurality of curves includes the following specific steps:

[0036] Any UAV in the UAV formation is recorded as the target UAV; the set consisting of the dimensions of the remaining curves in the target UAV except the path offset curve is recorded as the curve dimension set of all UAVs; the set consisting of the remaining curves in the target UAV except the path offset curve is recorded as the curve set of the target UAV.

[0037] The beneficial effect of the technical solution of the present invention is as follows: after obtaining a number of curves of each drone in a drone formation, the present invention reflects the flight-related data of the drone in multiple dimensions, and obtains the similarity relationship weight of each curve dimension in the curve dimension set. The similarity relationship weight reflects the correlation between the flight data of each dimension and the path deviation curve. The distance between different drones at each moment is weighted by the similarity relationship weight to obtain the similarity between any two drones at each moment. The similarity reflects the similarity relationship between the two drones. The higher the similarity relationship, the higher the degree of directness of the path. Calculating the similarity avoids the large error in the clustering result caused by directly measuring the optimal path between each drone based on the distance, thereby improving the accuracy of clustering, and thus This makes the flight path control of the UAV more accurate, and then the offset weight of each UAV at each moment is obtained through the preset flight path vector of the UAV at each moment and the actual flight path vector at each moment. The similarity is weighted by the offset weight to obtain the transfer probability between different UAVs. By correcting the transfer probability, the error compensation demand caused by environmental problems and the difference in the difficulty of tasks performed by different UAVs in the formation on the direct impact of the Markov clustering algorithm are avoided, the accuracy of subsequent data prediction is improved, and the predicted path offset value of each UAV at the next moment is obtained. Finally, the flight path of the UAV is controlled according to the path offset value at the current moment and the predicted path offset value, thereby improving the centralized control effect of UAV formation flight. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 A flowchart of a formation flight control method suitable for UAV advertising provided by an embodiment of the present invention;

[0040] Figure 2 A flow chart of controlling the flight path of a UAV provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a formation flight control method suitable for UAV advertising proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0042] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0043] The following is a detailed description of a specific scheme of a formation flight control method suitable for UAV advertising provided by the present invention in conjunction with the accompanying drawings.

[0044] See also Figure 1 and Figure 2 , which shows a flowchart of a formation flight control method suitable for UAV advertising provided by an embodiment of the present invention and a flowchart of controlling the flight path of the UAV, the method comprising the following steps:

[0045] Step S001: Obtain several curves of each drone in the drone formation, and obtain the distance between different drones at each moment.

[0046] Specifically, several curves of each drone in the drone formation within the last hour are obtained, and the several curves include: wind speed curve, wind direction curve, flight altitude curve, temperature curve, speed curve, path deviation curve and other curves in multiple dimensions; it should be noted that the curve of each dimension is a two-dimensional curve, the horizontal axis of the curve corresponds to time, and the vertical axis corresponds to parameter data of each dimension. For example, the parameter data corresponding to a certain time in the wind speed curve is the wind speed data. It is an existing method to obtain the wind speed at each moment and then obtain the wind speed curve and obtain other curves, which will not be repeated in this embodiment. The speed curve reflects the flight speed of the drone, and the path deviation curve reflects the deviation of the drone from the preset route at each time; the distance between different drones at each moment is obtained; it should be noted that the collection frequency of all data in this embodiment is once per second.

[0047] At this point, several curves of each drone in the drone formation and the distances between different drones at each moment are obtained.

[0048] Step S002: obtain a curve dimension set and a curve set of each UAV based on a number of curves; obtain a similarity relationship weight of each curve dimension in the curve dimension set based on the curve set of each UAV; obtain the similarity between any two UAVs at each moment based on the similarity relationship weight and the distance between different UAVs at each moment.

[0049] It should be noted that when using the Markov algorithm to predict data, the main step is to obtain the direct probability between each data point, and then obtain the transfer probability of the Markov chain and each node; the direct access involved in this process means that the path between the data points exists and follows the shortest path; but for the drone formation in this embodiment, there must be intercommunication paths between each drone, so directly based on the distance measurement of the optimal path between each drone, it will lead to large errors in the clustering results, which is not conducive to subsequent data prediction; therefore, the multi-dimensional data of each drone should be combined for analysis, first initialize the relevant feature values ​​of each drone, and obtain the similarity relationship.

[0050] It should be noted that data of different dimensions have different impacts on error offset. Directly obtaining the data differences between dimensions and then obtaining similarity relationships can easily make the distances between different nodes in the Markov chain too close, resulting in large errors in the final clustering results. Therefore, the impact relationship of the corresponding errors in each dimension should be obtained based on the data changes in different dimensions in time series and the size of the path offset, and then the similarity relationship weights of each dimension should be obtained.

[0051] Specifically, a curve dimension set and a curve set for each drone are obtained based on several curves, as follows:

[0052] Any UAV in the UAV formation is recorded as the target UAV; the set consisting of the dimensions of the remaining curves in the target UAV except the path offset curve is recorded as the curve dimension set of all UAVs; the set consisting of the remaining curves in the target UAV except the path offset curve is recorded as the curve set of the target UAV.

[0053] Furthermore, according to the curve set of each drone, the similarity relationship weight of each curve dimension in the curve dimension set is obtained, as follows:

[0054] The first curve in the curve set of the target UAV is recorded as the first curve, the path offset curve of the target UAV is recorded as the fixed curve, the Pearson correlation coefficient between the first curve and the fixed curve is recorded as the initial similarity relationship weight of the first curve of the target UAV, and the set consisting of the initial similarity relationship weights of the first curve in the curve set of all UAVs is recorded as the initial similarity relationship weight set.

[0055] It should be noted that the order of the curves in the curve set is arranged according to the description order of the curves in step S001, and the curve dimension set is also arranged according to the description order. For example, if the first curve in the curve set is the wind speed curve, then the first curve dimension in the curve dimension set is the wind speed dimension. Obtaining the Pearson correlation coefficient between two curves is an existing method and will not be repeated in this embodiment.

[0056] All initial similarity relationship weights in the initial similarity relationship weight set are linearly normalized, and the result is used as the normalized similarity relationship weight of the first curve in the curve set of each drone. The accumulated value of the normalized similarity relationship weights of the first curve in the curve set of all drones is used as the similarity relationship weight of the first curve dimension in the curve dimension set.

[0057] It should be noted that the similarity relationship weight of the curve dimension represents the similarity relationship weight of the parameter data in the curve corresponding to the dimension. The larger the similarity relationship weight, the higher the weight.

[0058] It should be noted that the multidimensional data collected in step S001 obtains several parameter data of each drone. The parameter data is distributed in time series and will change with the change of time. Based on the Euclidean norm combined with the shortest path in the Markov algorithm, the preliminary similarity relationship between any two drones is obtained, thereby obtaining the directness of the paths between the two drones.

[0059] Specifically, according to the similarity relationship weight and the distance between different drones at each moment, the similarity between any two drones at each moment is obtained, as follows:

[0060] Record any moment as the target moment.

[0061]

[0062] Where N is the number of curve dimensions in the curve dimension set, μ i is the similarity relationship weight of the i-th curve dimension in the curve dimension set, T x,i is the parameter data of the i-th curve in the curve set of the x-th UAV at the target time, T y,i is the parameter data of the i-th curve in the curve set of the y-th UAV at the target time, || is the absolute value, D x,y is the distance between the xth UAV and the yth UAV at the target time, S x,yis the similarity between the x-th UAV and the y-th UAV at the target moment, exp[] is an exponential function with a natural constant as the base, and this embodiment adopts the exp[-U] model to present the inverse proportional relationship and normalization processing, U is the input of the model, and the implementer can set the inverse proportional function and normalization function according to the specific implementation situation.

[0063] It should be noted that |T x,i -T y,i | represents the difference in parameter data between two drones at the same time in the same dimension,

[0064] |T x,i -T y,i The smaller | is, the smaller the difference in parameter data is at the same time in the current dimension, and the better the similarity is. At the same time, the similarity relationship weights of each curve dimension obtained above are used to weight the data and traverse all curve dimensions to obtain And use this value to weight the distance between the two drones, thus obtaining S x,y , S x,y The larger the value, the higher the similarity relationship between the two drones and the higher the directness of the path.

[0065] It should be noted that by obtaining the correlation between multidimensional data and path offset, the preliminary similarity relationship between two UAVs at different times is obtained, which avoids large errors in the clustering results caused by directly measuring the optimal path between each UAV based on distance, and improves the accuracy of clustering.

[0066] At this point, the similarity between any two drones at each moment is obtained.

[0067] Step S003, mapping the preset flight path and actual flight path of each UAV in the UAV formation into the three-dimensional space model; obtaining the preset flight path vector of each UAV at each moment and the actual flight path vector at each moment in the three-dimensional space model; obtaining the offset weight of each UAV at each moment according to the preset flight path vector of the UAV at each moment and the actual flight path vector at each moment.

[0068] It should be noted that, as the environmental factors change over time during the advertising process of the drone formation, the error offsets generated for any drone at different times are different; and the control system has different requirements for error compensation for each drone in the drone formation due to the above-mentioned environmental problems during the advertising process. Therefore, the similarity relationship obtained only based on the above-mentioned calculation of the multidimensional data differences corresponding to the current moment is relatively one-sided, and should be comprehensively evaluated in combination with historical data; at the same time, different drones in the formation perform tasks of different difficulty, and the path offsets and frequencies they generate are different. Therefore, the transfer probability of the drone at the current moment should be corrected in combination with historical data.

[0069] Specifically, the preset flight path and the actual flight path of each drone in the drone formation are obtained, and the preset flight path and the actual flight path of each drone in the drone formation are mapped to the three-dimensional space model, and the preset flight path vector and the actual flight path vector of each drone at each moment are obtained in the three-dimensional space model, as follows:

[0070] It should be noted that obtaining the preset flight path and the actual flight path of each drone in the drone formation is an existing method, which will not be described in detail in this embodiment; the preset flight path and the actual flight path of each drone in the drone formation are mapped to the three-dimensional space model through the geographic information system software ArcGIS Pro, and the next moment of the target moment is recorded as the target adjacent moment. The direction and distance between the target moment and the target adjacent moment in the preset flight path of the target drone are obtained in the three-dimensional space model, and the vector determined by the direction and distance between the target moment and the target adjacent moment in the preset flight path of the target drone is used as the preset flight path vector of the target drone at the target moment, and the direction and distance between the target moment and the target adjacent moment in the actual flight path of the target drone is obtained in the three-dimensional space model, and the vector determined by the direction and distance between the target moment and the target adjacent moment in the actual flight path of the target drone is used as the actual flight path vector of the target drone at the target moment; it should be noted that each position of the path in the three-dimensional space model after mapping corresponds to a time point and corresponds one-to-one with the actual moment. If there is no next moment at the target moment, the path vector is not obtained.

[0071] Specifically, according to the preset flight path vector of the drone at each moment and the actual flight path vector at each moment, the offset weight of each drone at each moment is obtained. As an embodiment, the specific calculation method is as follows:

[0072]

[0073] Where N jis the number of moments in the time range of the jth moment, and the specific method for obtaining the time range of the jth moment is as follows: the range from the j-TH1th moment to the jth moment is taken as the time range of the jth moment, TH1 is a preset first value, and this embodiment is described as TH1=5; it should be noted that if there is no TH1 moment before a certain moment, the time range is not obtained, and no subsequent related analysis is performed; d j,k is the actual flight path vector of the target UAV at the kth moment within the time range of the jth moment, d j,k+1 is the actual flight path vector of the target UAV at the k+1th moment within the time range of the jth moment, v j,k is the actual speed of the target UAV at the kth moment within the time range of the jth moment, v j,k+1 is the actual speed of the target UAV at the k+1th moment within the time range of the jth moment, and the actual speed is obtained through the speed curve in the curve set of the target UAV, N0 is the total duration of the preset flight path, d1 t is the preset flight path vector of the target UAV at the tth moment within the total duration of the preset flight path, d1 t+1 v1 is the preset flight path vector of the target UAV at the t+1th moment within the total duration of the preset flight path. t v1 is the preset speed of the target UAV at the tth moment within the total duration of the preset flight path, t+1 is the preset speed of the target UAV at the t+1th moment within the total duration of the preset flight path, and the preset speed is the modulus of the preset flight path vector at the corresponding moment; it should be noted that since the parameter data is collected once per second, the modulus is directly used as the preset speed here; ΔG j is the path deviation value of the target UAV at the jth moment, ΔG max,j is the maximum value of the path deviation value of the target UAV at all times within the time range of the jth moment, and the path deviation value is obtained through the path deviation curve in the curve set of the target UAV, ΔG j,k is the path deviation value of the target UAV at the kth moment within the time range of the jth moment, ΔG j,k+1 is the path deviation value of the target UAV at the k+1th moment within the time range of the jth moment, ‖‖ is the binary norm, || is the absolute value, and u j is the offset weight of the target UAV at the jth moment.

[0074] It should be noted that Indicates the normalized value of the path deviation value at the jth moment. The closer the value is to 1, the greater the mutation of the path deviation value at the jth moment, indicating that the tendency of the path deviation at the current position is higher. It indicates the degree of change of the offset at each moment within the time range of the jth moment. The larger the value, the higher the offset amplitude within the time range, which means the higher the offset trend of the local area within the time range. For the UAV flight process, the more drastic the change in the direction of the path, that is, ‖d j,k -d j,k+1 ‖ 2 The larger the value, the greater the frequency and difference of speed change during flight, that is, (v j,k -v j,k+1 ) 2 The larger it is, the more complex the flight mission performed by the UAV is. The actual flight path and the predicted flight path are traversed separately, and the Euclidean norm is used to convert the above two equations into the same dimension, which respectively represent the local and overall complexity. The two are expressed as fractions, which shows the complexity ratio of the current time range. The larger the fraction is, the more complex the flight mission performed by the UAV in the current time range is, and the higher the tendency of path deviation is. Therefore, the deviation weight of the UAV at each moment is obtained.

[0075] At this point, the offset weight of each UAV at each moment is obtained.

[0076] Step S004, obtain the transfer probability between different UAVs according to the similarity and the offset weight; cluster all UAVs according to the transfer probability between different UAVs in combination with the Markov clustering algorithm to obtain several clusters, and obtain the predicted path offset value of each UAV in each cluster at the next moment; control the flight path of the UAV according to the path offset value at the current moment and the predicted path offset value.

[0077] It should be noted that, for any UAV at any time, an offset weight can be obtained; at this time, for all flight data of all UAVs, their similarity relationship with the current UAV at the current moment can be obtained; and then by obtaining the difference in actual flight trajectories of two different UAVs and combining the calculated offset weight, the transfer probability of the Markov algorithm is corrected.

[0078] Specifically, the transfer probability between different drones is obtained according to the similarity and the offset weight. As an embodiment, the specific calculation method is as follows:

[0079]

[0080] Where N0 is the total duration of the preset flight path, S1(x1,y1) a is the similarity between the x1th UAV and the y1th UAV at the ath moment within the total duration of the preset flight path, u x1,a is the offset weight of the x1th UAV at the ath moment, d1x1,a is the preset flight path vector of the x1th UAV at the ath moment within the total duration of the preset flight path, u y1,a is the offset weight of the y1th UAV at the ath moment, d1 y1,a is the preset flight path vector of the y1th UAV at the ath moment within the total duration of the preset flight path, exp() is an exponential function with a natural constant as the base, and this embodiment adopts the exp(-U) model to present the inverse proportional relationship and normalization processing, U is the input of the model, and the implementer can set the inverse proportional function and normalization function according to the specific implementation situation, ‖‖ is the second norm, and P x1,y1 is the transfer probability from the x1th UAV to the y1th UAV.

[0081] It should be noted that ‖u x1,a ×d1 x1,a -u y1,a ×d1 y1,a ‖ represents the weighted difference in the flight path vectors of different drones at the same time. The smaller the difference, the closer the path offset between the two drones is. The higher the degree of directness of the path between the two drones is, the greater the transfer probability should be. The value after taking the norm is used to weight the similarity relationship and traverse all the time, thus obtaining the transfer probability P of the drone. x1,y1 , which is the corrected transition probability. In the Markov algorithm, the transition probability value ranges from [0,1].

[0082] It should be noted that by calculating the flight path offset weights at different times, combining the similarity relationship to obtain the correlation between the two UAVs, and correcting the transfer probability, this operation avoids the direct impact of the error compensation requirements caused by environmental problems and the differences in the difficulty of tasks performed by different UAVs in the formation on the algorithm, improves the clustering accuracy, and indirectly improves the accuracy of subsequent data predictions, thereby better controlling the flight of the UAV at the next moment.

[0083] Specifically, all drones are clustered according to the transfer probability between different drones combined with the Markov clustering algorithm to obtain several clusters, and the predicted path deviation value of each drone in each cluster at the next moment is obtained, as follows:

[0084] The path deviation value of each UAV in each cluster at the next moment is obtained by the WMA (Weighted Moving Average) algorithm; it should be noted that the next moment is the next moment of the total duration of the preset flight path, that is, a moment in the future. All UAVs are clustered according to the transition probability between different UAVs combined with the Markov clustering algorithm to obtain several clusters and obtain the path deviation value of each UAV in each cluster at the next moment by the WMA algorithm. This is an existing method and will not be repeated in this embodiment; any cluster is recorded as the target cluster, and the average value of the path deviation values ​​of all UAVs in the target cluster at the next moment is used as the predicted path deviation value of each UAV in the target cluster at the next moment.

[0085] Furthermore, the flight path of the UAV is controlled according to the path deviation value at the current moment and the predicted path deviation value, as follows:

[0086] The last moment within the total duration of the preset flight path is recorded as the current moment, and the path offset value of the target UAV at the current moment is obtained. According to the path offset value of the target UAV at the current moment and the predicted path offset value at the next moment, the preset flight path of the target UAV is adjusted in combination with the PID controller; it should be noted that adjusting the preset flight path of the target UAV according to the path offset value of the target UAV at the current moment and the predicted path offset value at the next moment in combination with the PID controller is an existing method, which will not be repeated in this embodiment.

[0087] At this point, by calculating the flight path offset weights at different times, combining the similarity relationship to obtain the correlation between the two drones, and correcting the transfer probability, the predicted path offset value at the next moment is obtained, and then the flight path of the drone is controlled to complete the formation flight control of drone advertising.

[0088] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A formation flight control method suitable for UAV advertising, characterized in that: The method comprises the following steps: Acquire several curves of each drone in the drone formation, wherein the several curves at least include: a path deviation curve and a speed curve, wherein the curves include several moments, and acquire the distance between different drones at each moment; According to several curves, a curve dimension set and a curve set of each UAV are obtained; according to the curve set of each UAV, a similarity relationship weight of each curve dimension in the curve dimension set is obtained; according to the similarity relationship weight and the distance between different UAVs at each moment, the similarity between any two UAVs at each moment is obtained; Obtain the preset flight path and actual flight path of each drone in the drone formation, and map the preset flight path and actual flight path of each drone in the drone formation into a three-dimensional space model; obtain the preset flight path vector of each drone at each moment and the actual flight path vector at each moment in the three-dimensional space model; obtain the offset weight of each drone at each moment according to the preset flight path vector of the drone at each moment and the actual flight path vector at each moment; The transfer probability between different drones is obtained according to the similarity and the offset weight; all drones are clustered according to the transfer probability between different drones combined with the Markov clustering algorithm to obtain several clusters, and the predicted path offset value of each drone in each cluster at the next moment is obtained; the flight path of the drone is controlled according to the path offset value at the current moment and the predicted path offset value; The specific method of obtaining the preset flight path vector of each drone at each moment and the actual flight path vector at each moment is as follows: Record any moment as the target moment, record the next moment of the target moment as the target adjacent moment, and record any UAV in the UAV formation as the target UAV; Obtaining the direction and distance between the target moment and the target adjacent moment in the preset flight path of the target UAV in the three-dimensional space model, using the vector determined by the direction and distance between the target moment and the target adjacent moment in the preset flight path of the target UAV as the preset flight path vector of the target UAV at the target moment, obtaining the direction and distance between the target moment and the target adjacent moment in the actual flight path of the target UAV in the three-dimensional space model, using the vector determined by the direction and distance between the target moment and the target adjacent moment in the actual flight path of the target UAV as the actual flight path vector of the target UAV at the target moment; The specific steps of obtaining a curve dimension set and a curve set for each drone according to a plurality of curves are as follows: Any UAV in the UAV formation is recorded as the target UAV; the set consisting of the dimensions of the remaining curves in the target UAV except the path offset curve is recorded as the curve dimension set of all UAVs; the set consisting of the remaining curves in the target UAV except the path offset curve is recorded as the curve set of the target UAV.

2. A formation flight control method suitable for UAV advertising according to claim 1, characterized in that: The specific steps of obtaining the similarity relationship weight of each curve dimension in the curve dimension set according to the curve set of each drone are as follows: Any drone in the drone formation is recorded as a target drone, the first curve in the curve set of the target drone is recorded as the first curve, the path offset curve of the target drone is recorded as the fixed curve, the Pearson correlation coefficient between the first curve and the fixed curve is recorded as the initial similarity relationship weight of the first curve of the target drone, and the set consisting of the initial similarity relationship weights of the first curve in the curve set of all drones is recorded as the initial similarity relationship weight set; All initial similarity relationship weights in the initial similarity relationship weight set are linearly normalized, and the result is used as the normalized similarity relationship weight of the first curve in the curve set of each drone. The accumulated value of the normalized similarity relationship weights of the first curve in the curve set of all drones is used as the similarity relationship weight of the first curve dimension in the curve dimension set.

3. A formation flight control method suitable for UAV advertising according to claim 1, characterized in that: The specific steps of obtaining the similarity between any two drones at each moment according to the similarity relationship weight and the distance between different drones at each moment are as follows: Record any moment as the target moment; Where N is the number of curve dimensions in the curve dimension set, μ i is the similarity relationship weight of the i-th curve dimension in the curve dimension set, T x,i is the parameter data of the i-th curve in the curve set of the x-th UAV at the target time, T y,i is the parameter data of the i-th curve in the curve set of the y-th UAV at the target time, || is the absolute value, D x,y is the distance between the xth UAV and the yth UAV at the target time, S x,y is the similarity between the x-th UAV and the y-th UAV at the target moment, and exp[] is an exponential function with a natural constant as the base.

4. A formation flight control method suitable for UAV advertising according to claim 1, characterized in that: The specific steps of obtaining the offset weight of each UAV at each moment according to the preset flight path vector of the UAV at each moment and the actual flight path vector at each moment are as follows: Any drone in the drone formation is recorded as the target drone; Where N j is the number of moments in the time range of the jth moment, d j,k is the actual flight path vector of the target UAV at the kth moment within the time range of the jth moment, d j,k+1 is the actual flight path vector of the target UAV at the k+1th moment within the time range of the jth moment, v j,k is the actual speed of the target UAV at the kth moment within the time range of the jth moment, v j,k+1 is the actual speed of the target UAV at the k+1th moment within the time range of the jth moment, and the actual speed is obtained through the speed curve in the curve set of the target UAV, N0 is the total duration of the preset flight path, d1 t is the preset flight path vector of the target UAV at the tth moment within the total duration of the preset flight path, d1 t+1 v1 is the preset flight path vector of the target UAV at the t+1th moment within the total duration of the preset flight path. t v1 is the preset speed of the target UAV at the tth moment within the total duration of the preset flight path, t+1 is the preset speed of the target UAV at the t+1th moment within the total duration of the preset flight path. The preset speed is the modulus of the preset flight path vector at the corresponding moment. ΔG j is the path deviation value of the target UAV at the jth moment, ΔG max,j is the maximum value of the path deviation value of the target UAV at all times within the time range of the jth moment, and the path deviation value is obtained through the path deviation curve in the curve set of the target UAV, ΔG j,k is the path deviation value of the target UAV at the kth moment within the time range of the jth moment, ΔG j,k+1 is the path deviation value of the target UAV at the k+1th moment within the time range of the jth moment, || || is the binary norm, || is the absolute value, and u j is the offset weight of the target UAV at the jth moment.

5. A formation flight control method suitable for UAV advertising according to claim 4, characterized in that: The specific method for obtaining the time range of the j-th moment is as follows: The range from the jth moment TH1 to the jth moment is taken as the time range of the jth moment, and TH1 is a preset first value.

6. The formation flight control method suitable for UAV advertising according to claim 1, characterized in that: The specific steps of obtaining the transfer probability between different drones according to the similarity and the offset weight are as follows: Where N0 is the total duration of the preset flight path, S1(x1, y1) a is the similarity between the x1th UAV and the y1th UAV at the ath moment within the total duration of the preset flight path, u x1,a is the offset weight of the x1th UAV at the ath moment, d1 x1,a is the preset flight path vector of the x1th UAV at the ath moment within the total duration of the preset flight path, u y1,a is the offset weight of the y1th UAV at the ath moment, d1 y1,a is the preset flight path vector of the y1th UAV at the ath moment within the total duration of the preset flight path, exp() is an exponential function with a natural constant as the base, ‖‖ is the second norm, P x1,y1 is the transfer probability from the x1th UAV to the y1th UAV.

7. The formation flight control method for UAV advertising according to claim 1, characterized in that: The specific method for obtaining the predicted path deviation value of each drone in each cluster at the next moment is as follows: The last moment within the total duration of the preset flight path is recorded as the current moment, and the moment after the current moment is recorded as the next moment. The path offset value of each UAV in each cluster at the next moment is obtained through the WMA algorithm. Any cluster is recorded as the target cluster, and the average value of the path offset values ​​of all UAVs in the target cluster at the next moment is used as the predicted path offset value of each UAV in the target cluster at the next moment.

8. The formation flight control method for UAV advertising according to claim 1, characterized in that: The specific steps of controlling the flight path of the UAV according to the path deviation value at the current moment and the predicted path deviation value are as follows: Any drone in the drone formation is recorded as the target drone; The last moment within the total duration of the preset flight path is recorded as the current moment, and the moment after the current moment is recorded as the next moment. The path offset value of the target UAV at the current moment is obtained. According to the path offset value of the target UAV at the current moment and the predicted path offset value at the next moment, the preset flight path of the target UAV is adjusted in combination with the PID controller.