CN109743093B - Unmanned aerial vehicle millimeter wave communication beam tracking method - Google Patents

Unmanned aerial vehicle millimeter wave communication beam tracking method Download PDF

Info

Publication number
CN109743093B
CN109743093B CN201811602233.3A CN201811602233A CN109743093B CN 109743093 B CN109743093 B CN 109743093B CN 201811602233 A CN201811602233 A CN 201811602233A CN 109743093 B CN109743093 B CN 109743093B
Authority
CN
China
Prior art keywords
unmanned aerial
aerial vehicle
predicted
prediction
coordinate system
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201811602233.3A
Other languages
Chinese (zh)
Other versions
CN109743093A (en
Inventor
许文俊
张婧琳
冯志勇
高晖
张平
张治�
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Posts and Telecommunications
Original Assignee
Beijing University of Posts and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing University of Posts and Telecommunications filed Critical Beijing University of Posts and Telecommunications
Priority to CN201811602233.3A priority Critical patent/CN109743093B/en
Publication of CN109743093A publication Critical patent/CN109743093A/en
Application granted granted Critical
Publication of CN109743093B publication Critical patent/CN109743093B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Radar Systems Or Details Thereof (AREA)

Abstract

The invention discloses a millimeter wave communication beam tracking method for an unmanned aerial vehicle, wherein the time slot structure of millimeter wave communication of the unmanned aerial vehicle is designed into an exchange time slot and T tracking time slots, and the process in the exchange time slot comprises the step of receiving MSI feedback; calculating an actual beam forming vector according to the actual position and attitude information of the receiving-end unmanned aerial vehicle in the MSI feedback, and predicting the position and attitude information of the receiving-end unmanned aerial vehicle in the future T tracking time slots; transmitting information through an actual beamforming vector; the process in the tracking time slot comprises calculating a predicted beamforming vector according to the predicted position and attitude information; information is transmitted by means of a predicted beamforming vector. The invention realizes the effective tracking of the position and the posture of the unmanned aerial vehicle, can improve the millimeter wave communication spectrum efficiency of the unmanned aerial vehicle under the condition that the unmanned aerial vehicle moves rapidly, does not need the feedback of channel state information of a pilot frequency and a receiving end, and can reduce the time delay error between the unmanned aerial vehicle and the real motion state.

Description

Unmanned aerial vehicle millimeter wave communication beam tracking method
Technical Field
The invention relates to the technical field of millimeter wave beam tracking of unmanned aerial vehicles, in particular to a millimeter wave communication beam tracking method of an unmanned aerial vehicle.
Background
Millimeter wave beam tracking is a technique for maintaining millimeter wave communication links. The millimeter wave communication adopts narrow beams to compensate high path loss of millimeter wave transmission, obtains corresponding analog beam forming vectors through a beam tracking technology, keeps millimeter wave beam alignment, can fully utilize antenna array gain, keeps a millimeter wave communication link, and improves millimeter wave communication performance.
At present, the millimeter wave beam tracking method mainly uses auxiliary information such as position and the like or directly tracks the arrival angle/departure angle of a signal. For example, in Robust beam-tracking for mm wave mobile communications, the arrival angle of a signal is directly tracked by an extended kalman filtering method on the basis of the assumption that the arrival angle of the signal changes slowly, but the method may have a large tracking error in a high-speed mobile environment, and thus the method is not suitable for beam tracking between unmanned aerial vehicles. Channel tracking with flight control system for UAV mmWave MIMO communications proposes a scheme for tracking a millimeter wave Channel of an unmanned aerial vehicle on the ground by using self position and attitude information of the unmanned aerial vehicle, but the method does not consider the prediction tracking of non-self position and attitude information of the unmanned aerial vehicle, so that the performance loss is caused when the unmanned aerial vehicles of both communication sides have attitude position changes. There is a chinese patent application No. 201711015266 that discloses an electronic assisted beam alignment method for use with unmanned aerial vehicles. According to the method, the position and attitude information of the unmanned aerial vehicle is acquired through sensors such as a GPS (global positioning system), then the beam is mechanically adjusted to be aligned to a ground target, and finally a secondary alignment is carried out by adopting an analog beam forming algorithm. The method does not consider the acquisition of the position and attitude information of the unmanned aerial vehicle, and is not suitable for beam tracking among the unmanned aerial vehicles.
The applicant has found that the existing solutions present at least the following problems:
in the existing scheme, only the utilization of the position information of the receiving and transmitting ends or the position and attitude information of the transmitting ends and the direct tracking of the arrival angle/the departure angle of the signal are considered, and the influence of attitude change in the movement of the unmanned aerial vehicle at the receiving and transmitting ends is not considered, when the movement speed of the unmanned aerial vehicle is high, the tracking of the departure angle/the arrival angle of the signal can be obviously delayed or even misaligned, the millimeter wave beam alignment is influenced, and the communication speed is reduced; when the posture change of the unmanned aerial vehicle at the receiving and transmitting end is obvious, good tracking effect is difficult to obtain only by considering the position information assistance/transmitting end position and the posture information assistance millimeter wave beam tracking, and the millimeter wave communication performance between the unmanned aerial vehicles is influenced.
Disclosure of Invention
According to the defects of the prior art, the invention provides the millimeter wave communication beam tracking method for the unmanned aerial vehicle, which can realize efficient millimeter wave beam tracking under the conditions that the unmanned aerial vehicle moves at a high speed and has posture change.
The technical problem to be solved by the invention is realized by adopting the following technical scheme:
a millimeter wave communication beam tracking method for an unmanned aerial vehicle is characterized in that a time slot structure of millimeter wave communication of the unmanned aerial vehicle is designed into an exchange time slot and T tracking time slots,
the process in the switching time slot comprises
Receiving Motion State Information (MSI) feedback;
calculating an actual beam forming vector according to the actual position and attitude information of the receiving-end unmanned aerial vehicle in the MSI feedback, and predicting the position and attitude information of the receiving-end unmanned aerial vehicle in the future T tracking time slots;
transmitting information through an actual beamforming vector;
the process of tracking the time slot comprises
Calculating a predicted beamforming vector according to the predicted position and attitude information;
information is transmitted by means of a predicted beamforming vector.
As an optional implementation manner of the present invention, the step of predicting the position and posture information of the T tracking slots includes
Step 21, training a Gaussian process model by using historical MSI to obtain corresponding hyper-parameters;
step 22, comparing t0And TmaxWherein t is0-exchanging time slots, TmaxMaximum number of slots for inter-drone communication, if t0＜TmaxGo to step 23; otherwise, the flow is terminated;
step 23, exchange-basedTime slot t0Previous MSI and time slot [ t ]0,t]Predicting the unmanned aerial vehicle in the time slot section [ T, T + T ] by utilizing the prediction result of the internal MSI and the trained Gaussian process modelf]Position and attitude information of the inner part, where T-time slot in which prediction has been completed after last prediction, TfIndicates the number of time slots predicted backward from the time slot in which the last prediction is completed, and if the prediction is the first time, t is t0；
Step 24, after the prediction is completed, let T be T + Tf；
Step 25, if t < t0+ T, that is, the position and attitude information in the T tracking time slots is not predicted, returning to step 23 to continue prediction; otherwise, let t0＝t0+ T, go back to step 22 for the next prediction round.
As an optional implementation manner of the present invention, the position information prediction method includes
Constructing inputs i consisting of respective position coordinatesx/y/zAnd an output ox/y/zForming training set from historical information
Figure BDA0001922790190000031
Forming test sets from information to be predicted
Figure BDA0001922790190000032
Selecting a kernel function K ═ K for the Gaussian process1+K2In which K is1Is a linear kernel function, K2Is a square exponential kernel function;
maximizing the edge likelihood function on the training set to obtain the optimal estimation of the hyperparameters in the kernel function
Figure BDA0001922790190000033
Obtaining corresponding output prediction distribution on test set
Figure BDA0001922790190000034
Wherein the content of the first and second substances,
Figure BDA0001922790190000035
is the average of the results of the prediction,
Figure BDA0001922790190000036
is the predicted variance;
will be provided with
Figure BDA0001922790190000037
As
Figure BDA0001922790190000038
Wherein the predicted values of the position coordinates of the future t' slots are respectively
Figure BDA0001922790190000039
Thereby obtaining the position predicted value of the unmanned aerial vehicle
Figure BDA00019227901900000310
As an optional implementation manner of the present invention, the attitude information prediction method includes
Obtaining the speed and acceleration of the unmanned aerial vehicle from the position information prediction result, and constructing an input consisting of an attitude angle, the speed and the acceleration
Figure BDA00019227901900000311
And output
Figure BDA00019227901900000312
Forming training set by historical information
Figure BDA00019227901900000313
Forming test sets from information to be predicted
Figure BDA00019227901900000314
Selecting a kernel function K ═ K for the Gaussian process1+K2+K3Wherein, K is1Is a linear kernel function, K2Is a square exponential kernel function, K3Is a quadratic rational quadratic kernel function;
maximizing the edge likelihood function on the training set to obtain the optimal estimation of the hyperparameters in the kernel function
Figure BDA00019227901900000315
Obtaining corresponding output prediction distribution on test set
Figure BDA0001922790190000041
Figure BDA0001922790190000042
Is the average of the results of the prediction,
Figure BDA0001922790190000043
is the predicted variance;
will be provided with
Figure BDA0001922790190000044
As
Figure BDA0001922790190000045
Wherein the attitude coordinate prediction values of the future t' slots are respectively
Figure BDA0001922790190000046
t′∈[t，t+Tf]Thereby obtaining the predicted value of the unmanned aerial vehicle attitude
Figure BDA0001922790190000047
Where ψ is the angle of rotation, θ is the pitch angle, φ is the yaw angle.
As an optional implementation manner of the present invention, the method for calculating the predicted beamforming vector includes
Step 41, calculating a coordinate system transformation matrix;
42, obtaining the position coordinates of uniform planar antenna arrays UPA of the receiving unmanned aerial vehicle and the transmitting unmanned aerial vehicle in the same coordinate system through coordinate system transformation;
step 43, calculating a corresponding beam pitch angle and azimuth angle;
and step 44, outputting a corresponding predicted beamforming vector.
As an alternative embodiment of the present invention, the coordinate system transformation matrix includes a rotation matrix and a translation matrix.
As an optional implementation manner of the present invention, a transformation process of a coordinate system of a UPA of the receiving-end unmanned aerial vehicle is as follows:
converting the local coordinate system of the receiving unmanned aerial vehicle into a global coordinate system through a coordinate system transformation matrix;
converting the coordinate system transformation matrix into a local coordinate system of the unmanned aerial vehicle at the transmitting end under the global coordinate system;
and converting the transformation matrix into a UPA coordinate system of the sending-end unmanned aerial vehicle through a coordinate system transformation matrix under the local coordinate system of the sending-end unmanned aerial vehicle.
According to the millimeter wave communication beam tracking method for the unmanned aerial vehicle, the position and the attitude information of the unmanned aerial vehicle are provided by the sensor through the sending unmanned aerial vehicle, the position and the attitude information fed back by low frequency are used, the position and the attitude information of the receiving unmanned aerial vehicle in the next T time slots are predicted through a machine learning method based on the Gaussian process, corresponding coordinate transformation is carried out based on the prediction result, and the corresponding beam angle and the beam forming vector are calculated through the geometric relation, so that the position and the attitude of the unmanned aerial vehicle are effectively tracked. In addition, the millimeter wave beam tracking based on position and attitude prediction does not need pilot frequency and channel state information feedback of a receiving end, can reduce the time delay error between the real motion state and the millimeter wave beam tracking, and is more suitable for a high-speed moving scene.
Drawings
The invention is further illustrated with reference to the following figures and examples.
Fig. 1 and fig. 2 are both usage scenario diagrams of the present embodiment;
FIG. 3 is a schematic diagram of a timeslot structure according to this embodiment;
FIG. 4 is a logic diagram of the prediction step for tracking the position and orientation information of the timeslot according to the present embodiment;
fig. 5 is a logic diagram of a method for calculating a predicted beamforming vector according to the present embodiment;
FIG. 6 is a coordinate system relationship diagram in a usage scenario of the present embodiment;
fig. 7 is a logic block diagram of a coordinate system transformation process of the UPA of the receiving drone according to the present embodiment.
In the figure, 1-originating drone, 11-UPA of originating drone, 2-receiving drone, 3-MSI feedback, 4-beamforming vector.
Detailed Description
The invention aims to provide self position and attitude information through a sensor of an unmanned aerial vehicle at a transmitting end, predict the position and attitude information of the unmanned aerial vehicle at a receiving end in the next T time slots by using the position and attitude information fed back by low frequency through a machine learning method based on the Gaussian process, perform corresponding coordinate transformation based on a prediction result, calculate corresponding beam angles and beam forming vectors by using geometric relations, and realize efficient millimeter wave beam tracking of the unmanned aerial vehicle under the conditions of high-speed motion and attitude change.
The following embodiments are provided to describe the embodiments of the present invention, and to further describe the detailed description of the embodiments of the present invention, such as the shapes, configurations, mutual positions and connection relationships of the components, the functions and operation principles of the components, the manufacturing processes and operation methods, etc., so as to help those skilled in the art to more fully, accurately and deeply understand the inventive concept and technical solutions of the present invention.
As an embodiment of the present invention, there is provided a method for tracking millimeter wave communication beam of an unmanned aerial vehicle, where a time slot structure of millimeter wave communication of the unmanned aerial vehicle is designed as an exchange time slot and T tracking time slots, as shown in fig. 3, a process in the exchange time slot includes
Step S1, receiving MSI feedback;
step S2, calculating an actual beam forming vector according to the actual position and attitude information of the receiving unmanned aerial vehicle in MSI feedback, and simultaneously predicting the position and attitude information of the receiving unmanned aerial vehicle in the future T tracking time slots;
step S3, information is transmitted through the actual beamforming vector;
the process of tracking the time slot comprises
Step S4, calculating a predicted beamforming vector according to the predicted position and attitude information;
step S5, information is transmitted by the predicted beamforming vector.
The time slot structure comprises two time slots of an exchange time slot and a tracking time slot, wherein the exchange time slot comprises four stages of Motion State Information (MSI) exchange, beam alignment, prediction and data transmission, and the tracking time slot only comprises two stages of beam alignment and data transmission. And T tracking time slots are arranged between every two switching time slots, namely the unmanned aerial vehicle carries out MSI switching and position attitude prediction once every T time slots. By introducing the position and attitude information prediction of the unmanned aerial vehicle into the millimeter wave beam tracking technology of the unmanned aerial vehicle and utilizing the prediction information to calculate and obtain the corresponding predicted beam forming vector, the millimeter wave beam tracking of the unmanned aerial vehicle with high efficiency is realized, and the purpose of improving the frequency spectrum efficiency of millimeter wave communication between the unmanned aerial vehicles is achieved. The invention realizes the effective tracking of the position and the posture of the unmanned aerial vehicle, and compared with other comparison schemes, the invention can improve the millimeter wave communication spectrum efficiency of the unmanned aerial vehicle under the condition that the unmanned aerial vehicle moves rapidly. In addition, millimeter wave beam tracking based on position attitude prediction does not need pilot frequency and channel state information feedback of a receiving end, can reduce time delay error between the millimeter wave beam tracking and a real motion state, and is more suitable for a high-speed moving scene.
Fig. 1 and 2 are usage scene diagrams of the present invention, in which two dotted arrows respectively represent the moving directions of the originating drone 1 and the terminating drone 2. In fig. 1, it is shown that in the exchange time slot, the receiving-side unmanned aerial vehicle 2 feeds back the motion state information (i.e., MSI feedback 3) of itself to the sending-side unmanned aerial vehicle 1 by using the sensor, and after receiving the MSI feedback 3, the sending-side unmanned aerial vehicle 1 calculates to obtain a beamforming vector 4, and then transmits information to the receiving-side unmanned aerial vehicle 2 by using the beamforming vector 4. Fig. 2 shows that in tracking time slots, the transmitting-side drone 1 obtains a predicted beamforming vector 4 through calculation under the condition of no MSI feedback, and then transmits information to the receiving-side drone 2 by using the beamforming vector 4.
As an alternative embodiment of the present invention, as shown in fig. 4, in the step S2, the step of predicting the position and posture information of the T time slots in the future includes
Step S21, training a Gaussian process model by using the historical MSI to obtain corresponding hyperparameters;
step S22, comparison t0And TmaxWherein t is0MSI exchange slots, TmaxMaximum number of slots for inter-drone communication, if t0＜TmaxProceeding to step S23; otherwise, the flow is terminated;
step S23, based on exchanging time slot t0Previous MSI and time slot [ t ]0,t]Predicting the unmanned aerial vehicle in the time slot section [ T, T + T ] by utilizing the prediction result of the internal MSI and the trained Gaussian process modelf]Position and attitude information of the inner part, where T-time slot in which prediction has been completed after last prediction, TfIndicates the number of time slots predicted backward from the time slot in which the last prediction is completed, and if the prediction is the first time, t is t0；
Step S24, after the prediction is completed, let T be T + Tf；
Step S25, if t < t0+ T, that is, the position and attitude information in T slots is not predicted, and go back to step S23 to continue prediction; otherwise, let t0＝t0+ T, go back to step S22, and perform the next round of prediction.
Further, the main task of the method for training the Gaussian process model to obtain the hyperparameters is to learn the mapping relation between input and output from empirical data (historical MSI of the unmanned aerial vehicle), and accordingly, a prediction model is established for future MSI of the unmanned aerial vehicle.
MSI of the sending unmanned aerial vehicle and the receiving unmanned aerial vehicle mainly comprises two parts of position information and attitude information, MSI prediction is position and attitude information prediction, and the positions of the sending unmanned aerial vehicle and the receiving unmanned aerial vehicle are respectively expressed as Xt＝(xt，yt，zt) And Xr＝(xr，yr，zr) The attitude of the drone is defined by a rotation angle psi, a pitch angle theta and yawThe angle phi represents, so the attitudes of the originating unmanned aerial vehicle and the terminating unmanned aerial vehicle are respectively recorded as thetat＝(ψt，θt，φt) And Θr＝(ψr，θr，φt)。
Wherein the position information prediction method comprises
Constructing inputs i consisting of respective position coordinatesx/y/zAnd an output ox/y/zForming training set from historical information
Figure BDA0001922790190000071
Forming test sets from information to be predicted
Figure BDA0001922790190000072
Selecting a kernel function K ═ K for the Gaussian process1+K2In which K is1Is a linear kernel function, K2Is a square exponential kernel function;
maximizing the edge likelihood function on the training set to obtain the optimal estimation of the hyperparameters in the kernel function
Figure BDA0001922790190000073
Obtaining corresponding output prediction distribution on test set
Figure BDA0001922790190000074
Wherein the content of the first and second substances,
Figure BDA0001922790190000075
is the average of the results of the prediction,
Figure BDA0001922790190000076
is the predicted variance;
will be provided with
Figure BDA0001922790190000081
As
Figure BDA0001922790190000082
Wherein does notThe predicted values of the position coordinates from the t' time slot are respectively
Figure BDA0001922790190000083
Thereby obtaining the position predicted value of the unmanned aerial vehicle
Figure BDA0001922790190000084
For convenience of description, a specific prediction method of the position information is described by taking an x coordinate of the position of the drone as an example. For convenience of presentation, it is assumed that the position and attitude observations obtained by the drone are noiseless, but it should be noted that the learning method based on gaussian processes can also handle predictions of noisy observations.
The input and output of the receiving-end unmanned aerial vehicle position x coordinate prediction are respectively past x coordinate information ix＝{xr(t-T)，...，xr(t-1) } and future x-coordinate information ox＝{xr(t)，...，xr(t+Tf) Where T isfIndicating the number of slots predicted backward from the time slot t. The relationship between the output and input for the x coordinate of position can be expressed as ox＝f(ix) Where f is a potential gaussian function. According to the Gaussian process definition, f (i)x) Is a gaussian process determined by mean and variance, expressed as follows:
Figure BDA0001922790190000085
wherein m (i)x) Is the mean function, k (i)x，ix') is a variance (kernel) function. The training set with the size of K and formed by the x coordinate historical information of the position of the unmanned aerial vehicle is recorded as
Figure BDA0001922790190000086
Test set containing x-coordinate future information
Figure BDA0001922790190000087
The input and output of the training set are respectively aggregated into a matrix I ═ Ix，1，...，ix，K]TAnd O ═ Ox，1，...，ox，K]T。
Figure BDA0001922790190000088
And
Figure BDA0001922790190000089
defined as the form of a matrix formed in the same way by the inputs and outputs of the test set, respectively. Known input IxThe distribution of the output can be written as
Figure BDA00019227901900000810
Wherein m (I)x) Is the mean matrix of the training set, the ith element is m (i)x，i)，K(Ix，Ix) Is a covariance matrix whose elements are expressed as K (i, j) ═ K (i)x，i，ix，j). Since a zero mean function is typically used, the training set outputs oxAnd test set output
Figure BDA00019227901900000811
The joint prior distribution of (a) is represented as follows:
Figure BDA00019227901900000812
wherein
Figure BDA00019227901900000813
Representing the covariance of the training and test points in the K x S dimension.
Figure BDA00019227901900000814
And
Figure BDA00019227901900000815
the definition of (A) is similar to that of (B). The predicted distribution of the test set output is represented as follows:
Figure BDA00019227901900000816
mean value of outputs
Figure BDA0001922790190000091
Can be used as a predictor, the above covariance
Figure BDA0001922790190000092
The statistical uncertainty of the prediction is represented. The predicted mean and variance of the test set output are related to the kernel function K and are therefore set to conform to the data pattern of the drone location. For unmanned plane position prediction, a linear kernel function K1 is selected to fit a linear relation of adjacent time slot positions caused by inertia, and the (i, j) th element of the linear relation
Figure BDA0001922790190000093
Modeling the position smooth change caused by mechanical adjustment by using a square exponential kernel function K2, wherein the (i, j) th element
Figure BDA0001922790190000094
The existing kernel functions may be formed into new kernel functions by addition, multiplication, and convolution. Thus the kernel function for drone location prediction is denoted K ═ K1+K2。
L in the kernel function constitutes the hyperparameter thetaGPAnd can be obtained through model training. First, the optimal hyper-parametric estimation is obtained by maximizing the edge likelihood function on a specific data set (obtained by historical information of the x coordinate of the unmanned aerial vehicle position)
Figure BDA0001922790190000095
Thereby obtaining the output distribution after training
Figure BDA0001922790190000096
And predicted distribution of outputs (future drone position x coordinates) on a specific test set
Figure BDA0001922790190000097
The process can be realized by a Gaussian process tool kit (GPML), and the training set and the test set are constructed according to the method and then inputAnd the GPML tool box selects the kernel function, the mean function and the Gaussian likelihood function optimally used by the hyper-parameter, can complete model training to obtain the optimal hyper-parameter, and performs one-step prediction to obtain the prediction distribution and the mean value output by the test set
Figure BDA0001922790190000098
Future unmanned aerial vehicle position x coordinate can be through the mean value of prediction distribution
Figure BDA0001922790190000099
Estimate, i.e. ox＝{xr(t)，...，xr(t+Tf) Aggregated matrix Ox＝[ox，1，...，ox，K]TIs estimated as
Figure BDA00019227901900000910
The estimated value of the x coordinate of the future t' time slot is recorded as
Figure BDA00019227901900000911
The prediction method of the y coordinate and the z coordinate is similar to the same, and then the method can obtain
Figure BDA00019227901900000912
And
Figure BDA00019227901900000913
to sum up, the predicted value of the position of the UAV at the future t' time slot is the predicted value
Figure BDA00019227901900000914
The attitude information prediction method comprises the following steps:
obtaining the speed and acceleration of the unmanned aerial vehicle from the position information prediction result, and constructing an input consisting of an attitude angle, the speed and the acceleration
Figure BDA0001922790190000101
And output
Figure BDA0001922790190000102
Forming training set by historical information
Figure BDA0001922790190000103
Forming test sets from information to be predicted
Figure BDA0001922790190000104
Selecting a kernel function K ═ K for the Gaussian process1+K2+K3Wherein, K is1Is a linear kernel function, K2Is a square exponential kernel function, K3Is a quadratic rational quadratic kernel function;
maximizing the edge likelihood function on the training set to obtain the optimal estimation of the hyperparameters in the kernel function
Figure BDA0001922790190000105
Obtaining corresponding output prediction distribution on test set
Figure BDA0001922790190000106
Figure BDA0001922790190000107
Is the average of the results of the prediction,
Figure BDA0001922790190000108
is the predicted variance;
will be provided with
Figure BDA0001922790190000109
As
Figure BDA00019227901900001010
Wherein the attitude coordinate prediction values of the future t' slots are respectively
Figure BDA00019227901900001011
t′∈[t，t+Tf]Thereby obtaining the predicted value of the unmanned aerial vehicle attitude
Figure BDA00019227901900001012
Where ψ is the angle of rotation, θ is the pitch angle, φ is the yaw angle.
For convenience of description, a specific attitude information prediction method is described by taking the yaw angle phi of the position of the unmanned aerial vehicle as an example.
The current pose of the drone is not only related to its past pose, but also to its position coordinates. Therefore, for the attitude prediction of the unmanned aerial vehicle, taking the yaw angle as an example, the input and output of the unmanned aerial vehicle are column vectors iφ＝{φr(t-T)，...，φr(t-1)，vr(t)，ar(t) } and oφ＝{φr(t)，...，φr(t+Tf) Therein of
Figure BDA00019227901900001013
And
Figure BDA00019227901900001014
representing velocity and acceleration vectors, respectively, may be obtained from the results of the position prediction. dt represents the slot length. Similar to the gaussian process model adopted for unmanned plane position prediction, the predicted distribution of future unmanned plane yaw angles is expressed as follows:
Figure BDA00019227901900001015
mean value of outputs
Figure BDA00019227901900001016
Can be used as a predictor, the above covariance
Figure BDA00019227901900001017
The statistical uncertainty of the prediction is represented. The mean and covariance are still closely related to the kernel function, and a proper kernel function needs to be selected to conform to the posture change data pattern.
For unmanned aerial vehicle attitude prediction, a proportional quadratic kernel function K3 is used to fit relative irregularities in the attitude data while using a linear kernel function K1 and a squared exponential kernel function K2Mode of law, its (i, j) th element
Figure BDA0001922790190000111
This is because the relative position, attitude may change immediately upon application of a force, resulting in a relatively irregular data pattern. Thus the kernel function for unmanned aerial vehicle attitude prediction is denoted as K ═ K1+K2+K3。
Similar to the location prediction, a and l in the kernel function described above constitute the hyper-parameter θGP. The optimal hyper-parameter and the corresponding output (future unmanned aerial vehicle pitch angle) prediction distribution on a specific test set can be obtained through training
Figure BDA0001922790190000112
Future unmanned aerial vehicle attitude yaw angle can pass through mean value of prediction distribution
Figure BDA0001922790190000113
Estimate, i.e. oφ＝{φr(t)，...，φr(t+Tf) An estimated value of
Figure BDA0001922790190000114
Recording the estimated value of the yaw angle of the future t' time slot as the estimated value
Figure BDA0001922790190000115
The prediction method of the rotation angle and the pitch angle is similar to the yaw angle, and the predicted value of the rotation angle can be obtained
Figure BDA0001922790190000116
And pitch angle prediction
Figure BDA0001922790190000117
The predicted value of the attitude of the unmanned aerial vehicle at the future t' time slot is
Figure BDA0001922790190000118
The result of the prediction of the position and attitude of the drone is therefore
Figure BDA0001922790190000119
And
Figure BDA00019227901900001110
as an alternative embodiment of the present invention, as shown in fig. 5, the method for calculating the predicted beamforming vector includes
Step S41, calculating a coordinate system transformation matrix;
step S42, obtaining the balance of the receiving unmanned aerial vehicle and the sending unmanned aerial vehicle under the same coordinate system through coordinate system transformation
Position coordinates of a uniform plane antenna array UPA;
step S43, calculating the corresponding beam pitch angle and azimuth angle;
and step S44, outputting the corresponding predicted beamforming vector.
The predicted unmanned aerial vehicle position attitude data and the beam angle required to be obtained are not in the same reference frame, and in order to introduce the relationship between the beam angle and the unmanned aerial vehicle position attitude information, the reference frame needs to be transformed, the related reference frame is shown as an attached drawing 6, and O in the attached drawingaDenotes the zero point of the a-coordinate system, ObZero point, O, of b-coordinate systemcZero point, O, of the c-coordinate systemgRepresenting the zero point of the g-coordinate system. The orientation of each coordinate system is:
(1) originating drone local coordinate system (a-coordinate system): the origin is located the center of gravity of unmanned aerial vehicle, its xa,ya,zaThe axes point to the front, left and below of the drone, respectively.
(2) Receiving end unmanned aerial vehicle local coordinate system (b-coordinate system): the origin is located the center of gravity of receiving end unmanned aerial vehicle, its xb,yb,zbThe axes point to the front, left and below of the drone, respectively.
(3) Uniform planar antenna array (UPA) coordinate system (c-coordinate system): the origin is located at the geometric center of the UPA11 of the originating drone, its xcAxis and xaSame, zcAxis perpendicular to the plane of the antenna, outward, yaPerpendicular to xc、zcThe axis is a plane.
(4) Global coordinate system (g-coordinate system): the origin of the projection is the projection (x) of the starting unmanned aerial vehicle track starting point on the groundt(0),yt(0),0). Its xg,ygAxis pointing north and east, zgThe axis is perpendicular to the ground.
Assuming that the UPA is located at a fixed position on the UAV, the relative a-coordinate system and b-coordinate system on the originating and receiving UAVs are recorded as
Figure BDA0001922790190000121
And
Figure BDA0001922790190000122
similarly, the UPA attitude is denoted as ΘPt(r)＝(ψPt(r)，θPt(r)，φPt(r)). The transmit beam angle relative to the plane of the array antenna can be calculated in a c-coordinate system. However, most of the drone position and attitude information obtained by the sensors is relative to the g-coordinate system, and therefore, coordinate transformation between different coordinate systems is required. Specifically, the coordinate system transformation matrix includes a rotation matrix and a translation matrix. The rotation of the local coordinate system relative to the global coordinate system is determined by the attitude of the unmanned aerial vehicle, and the rotation matrix between the b-coordinate system and the g-coordinate system is recorded as
Figure BDA0001922790190000123
Figure BDA0001922790190000124
Figure BDA0001922790190000125
Figure BDA0001922790190000126
Translation matrix is noted
Figure BDA0001922790190000131
Position coordinate of UPA of receiving-end unmanned aerial vehicle under g-coordinate system
Figure BDA0001922790190000132
Specifically, as shown in fig. 7, the transformation process of the UPA of the receiving-end unmanned aerial vehicle is as follows:
converting the local coordinate system of the receiving unmanned aerial vehicle into a global coordinate system through a coordinate system transformation matrix;
converting the coordinate system transformation matrix into a local coordinate system of the unmanned aerial vehicle at the transmitting end under the global coordinate system;
and converting the transformation matrix into a UPA coordinate system of the sending-end unmanned aerial vehicle through a coordinate system transformation matrix under the local coordinate system of the sending-end unmanned aerial vehicle.
The position of UPA of the receiving unmanned aerial vehicle in the c-coordinate system of the sending unmanned aerial vehicle is recorded as
Figure BDA0001922790190000133
UPA of the unmanned aerial vehicle at the origin is at the c-coordinate system position
Figure BDA0001922790190000134
The corresponding predicted angle for beam alignment in the c-coordinate system for the originating drone is calculated as follows:
Figure BDA0001922790190000135
satisfy f in the beam forming vector and the combining vector respectively*(t)＝A(αt，βt) And w*(t)＝A(αr，βr) The instantaneous power of the signal reaches a maximum. Wherein A (alpha)t，βt) Is an antenna array response vector, expressed as follows:
Figure BDA0001922790190000136
thus, based on the predictionThe beam angle, the corresponding beamforming vector and combining vector are respectively
Figure BDA0001922790190000137
And
Figure BDA0001922790190000138
the invention designs a new time slot structure which comprises an exchange time slot and T tracking time slots, wherein in the exchange time slot, a low-frequency auxiliary frequency band receives MSI feedback of a receiving-end unmanned aerial vehicle, a current actual beam forming vector is calculated according to the received MSI feedback, then the actual beam forming vector is used for data transmission, meanwhile, a machine learning method based on a Gaussian process is used for predicting the position and attitude information of the receiving-end unmanned aerial vehicle of the T time slots in the future according to the MSI feedback, in the T tracking time slots in the future, a predicted beam forming vector is calculated according to a prediction result, and then the predicted beam forming vector is used for data transmission. According to the invention, the position and the posture of the unmanned aerial vehicle can be effectively tracked, and compared with other comparison schemes, the millimeter wave communication spectrum efficiency of the unmanned aerial vehicle can be improved under the condition that the unmanned aerial vehicle moves rapidly. In addition, millimeter wave beam tracking based on position attitude prediction does not need pilot frequency and channel state information feedback of a receiving end, can reduce time delay error between the millimeter wave beam tracking and a real motion state, and is more suitable for a high-speed moving scene.
The invention has been described in an illustrative manner, and it is to be understood that the invention is not limited to the precise form disclosed, and that various insubstantial modifications of the inventive concepts and solutions, or their direct application to other applications without such modifications, are intended to be covered by the scope of the invention. The protection scope of the present invention shall be subject to the protection scope defined by the claims.

Claims (4)

1. A millimeter wave communication beam tracking method for an unmanned aerial vehicle is characterized by comprising the following steps: the time slot of the millimeter wave communication of the unmanned aerial vehicle comprises an exchange time slot and T tracking time slots,
the process in the switching time slot comprises
Receiving MSI feedback;
calculating an actual beam forming vector according to the actual position and attitude information of the receiving-end unmanned aerial vehicle in the MSI feedback, and predicting the position and attitude information of the receiving-end unmanned aerial vehicle in the future T tracking time slots;
transmitting information through an actual beamforming vector;
the process of tracking the time slot comprises
Calculating a predicted beamforming vector according to the predicted position and attitude information;
transmitting information through a predicted beamforming vector;
wherein the step of predicting the position and posture information of the T tracking slots comprises:
step 21, training a Gaussian process model by using historical MSI to obtain corresponding hyper-parameters;
step 22, comparing t0And TmaxWherein t is0-exchanging time slots, Tmax-a maximum number of time slots for inter-drone communication,
if t0＜TmaxGo to step 23; otherwise, the flow is terminated;
step 23, based on the exchanging time slot t0Previous MSI and time slot [ t ]0，t]Predicting the unmanned aerial vehicle in the time slot section [ T, T + T ] by utilizing the prediction result of the internal MSI and the trained Gaussian process modelf]Position and attitude information of the inner part, where T-time slot in which prediction has been completed after last prediction, TfIndicates the number of time slots predicted backward from the time slot in which the last prediction is completed, and if the prediction is the first time, t is t0；
Step 24, after the prediction is completed, let T be T + Tf；
Step 25, if t < t0+ T, that is, the position and attitude information in the T tracking time slots is not predicted, returning to step 23 to continue prediction; otherwise, let t0＝t0+ T, go back to step 22 for the next prediction;
the method for calculating the predicted beamforming vector comprises the following steps:
step 41, calculating a coordinate system transformation matrix;
42, obtaining the position coordinates of uniform planar antenna arrays UPA of the receiving unmanned aerial vehicle and the transmitting unmanned aerial vehicle in the same coordinate system through coordinate system transformation;
step 43, calculating a corresponding beam pitch angle and azimuth angle;
step 44, outputting a corresponding predicted beamforming vector;
the transformation process of the UPA of the receiving unmanned aerial vehicle is as follows:
converting the local coordinate system of the receiving unmanned aerial vehicle into a global coordinate system through a coordinate system transformation matrix;
converting the coordinate system transformation matrix into a local coordinate system of the unmanned aerial vehicle at the transmitting end under the global coordinate system;
and converting the transformation matrix into a UPA coordinate system of the sending-end unmanned aerial vehicle through a coordinate system transformation matrix under the local coordinate system of the sending-end unmanned aerial vehicle.
2. The millimeter wave communication beam tracking method for unmanned aerial vehicles according to claim 1, wherein: the position information prediction method comprises
Constructing inputs i consisting of respective position coordinatesx/y/zAnd an output ox/y/zForming training set from historical information
Figure FDA0002490421840000021
Forming test sets from information to be predicted
Figure FDA0002490421840000022
Wherein, Ix/y/zRepresenting input vectors i from historyx/y/zFormed matrix, Ox/y/zRepresenting output vector O from historyx/y/zFormed matrix, Ix/y/z，*Representing an input vector i to be predictedx/y/z，*Formed matrix, Ox/y/z，*Representing the output vector o to be predictedx/y/z，*A matrix of formations;
selecting a kernel function K ═ K for the Gaussian process1+K2In which K is1Is a linear kernel function, K2Is a square exponential kernel function;
maximizing the edge likelihood function on the training set to obtain the optimal estimation of the hyperparameters in the kernel function
Figure FDA0002490421840000023
Wherein, thetaGPIs a hyper-parameter;
obtaining corresponding output prediction distribution on test set
Figure FDA0002490421840000024
Wherein the content of the first and second substances,
Figure FDA0002490421840000025
is the average of the results of the prediction,
Figure FDA0002490421840000026
is a predicted variance, wherein O*And I*Respectively representing the output and the input on the test set;
will be provided with
Figure FDA0002490421840000027
As Ox/y/z，*Wherein the predicted values of the position coordinates of the future t' slots are respectively
Figure FDA0002490421840000028
t′∈[t，t+Tf]To obtain the predicted value of the position of the unmanned aerial vehicle
Figure FDA0002490421840000029
3. The millimeter wave communication beam tracking method for unmanned aerial vehicles according to claim 2, wherein: the attitude information prediction method comprises the steps of
Obtaining the speed and the acceleration of the unmanned aerial vehicle from the position information prediction result, and constructing the attitude angle, the speed and the accelerationComposed input
Figure FDA0002490421840000031
And output
Figure FDA0002490421840000032
Forming training set by historical information
Figure FDA0002490421840000033
Forming test sets from information to be predicted
Figure FDA0002490421840000034
Wherein the content of the first and second substances,
Figure FDA0002490421840000035
representing input vectors from history
Figure FDA0002490421840000036
The matrix of the composition is formed by the following components,
Figure FDA0002490421840000037
representing vectors output by history
Figure FDA0002490421840000038
The matrix of the composition is formed by the following components,
Figure FDA0002490421840000039
representing input vectors derived from the input to be predicted
Figure FDA00024904218400000310
The matrix of the composition is formed by the following components,
Figure FDA00024904218400000311
representing output vectors from the block to be predicted
Figure FDA00024904218400000312
A matrix of formations;
selecting a kernel function K ═ K for the Gaussian process4+K5+K3Wherein, K is4Is a linear kernel function, K5Is a square exponential kernel function, K3Is a quadratic rational quadratic kernel function;
maximizing the edge likelihood function on the training set to obtain the optimal estimation of the hyperparameters in the kernel function
Figure FDA00024904218400000313
Wherein, thetaGPIs a hyper-parameter;
obtaining corresponding output prediction distribution on test set
Figure FDA00024904218400000314
Figure FDA00024904218400000315
Is the average of the results of the prediction,
Figure FDA00024904218400000316
is a predicted variance, wherein O*And I*Respectively representing the output and the input on the test set;
will be provided with
Figure FDA00024904218400000317
As
Figure FDA00024904218400000318
Wherein t is future*The attitude coordinate prediction values of the time slots are respectively
Figure FDA00024904218400000319
t′∈[t，t+Tf]Thereby obtaining the predicted value of the unmanned aerial vehicle attitude
Figure FDA00024904218400000320
Where ψ is the angle of rotation, θ is the pitch angle, φ is the yaw angle.
4. The millimeter wave communication beam tracking method for unmanned aerial vehicles according to claim 1, wherein: the coordinate system transformation matrix comprises a rotation matrix and a translation matrix.
CN201811602233.3A 2018-12-26 2018-12-26 Unmanned aerial vehicle millimeter wave communication beam tracking method Active CN109743093B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811602233.3A CN109743093B (en) 2018-12-26 2018-12-26 Unmanned aerial vehicle millimeter wave communication beam tracking method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811602233.3A CN109743093B (en) 2018-12-26 2018-12-26 Unmanned aerial vehicle millimeter wave communication beam tracking method

Publications (2)

Publication Number Publication Date
CN109743093A CN109743093A (en) 2019-05-10
CN109743093B true CN109743093B (en) 2020-12-04

Family

ID=66359998

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811602233.3A Active CN109743093B (en) 2018-12-26 2018-12-26 Unmanned aerial vehicle millimeter wave communication beam tracking method

Country Status (1)

Country Link
CN (1) CN109743093B (en)

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111988079A (en) * 2019-05-22 2020-11-24 远传融创(杭州)科技有限公司 Information processing terminal and wireless communication method between information processing terminals
CN110365386B (en) * 2019-07-10 2022-06-17 中国水利水电第四工程局有限公司 Communication enhancement method for unmanned aerial vehicle and unmanned aerial vehicle
CN110492911B (en) * 2019-07-10 2021-10-15 鹰潭泰尔物联网研究中心 Beam tracking method and system for unmanned aerial vehicle communication
CN110611527B (en) * 2019-09-17 2021-01-15 北京邮电大学 Array antenna, beam forming method and sensing and communication integrated system
CN110928328B (en) * 2019-12-17 2020-08-28 中国人民解放军火箭军工程大学 Time-varying channel tracking method for unmanned aerial vehicle system under beam tilt effect
CN111200456B (en) * 2019-12-18 2021-02-26 西安电子科技大学 Fast and low-consumption 3D beam forming method based on joint autonomous positioning
CN111372195B (en) * 2020-01-23 2022-07-01 鹏城实验室 Method, apparatus and storage medium for tracking position of mobile terminal in mobile communication network
CN111416647B (en) * 2020-04-07 2021-08-06 北京邮电大学 Beam tracking method, codebook generation method and device
CN112019251A (en) * 2020-07-21 2020-12-01 深圳市月光宝和科技有限责任公司 Communication method, communication device, and storage medium
CN112383336B (en) * 2020-11-17 2022-03-08 电子科技大学 Beam tracking method for long-distance high-speed flying object
CN113630887B (en) * 2021-09-16 2024-02-09 中南大学 Internet of vehicles communication method of millimeter wave network based on online learning
CN114222312B (en) * 2021-12-30 2023-10-17 西安电子科技大学 Moving target tracking method based on self-adaptive wave beam

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106464342A (en) * 2014-02-17 2017-02-22 优必奇奥姆有限责任公司 Broadband access system via drone/uav platforms
CN107094044A (en) * 2017-03-30 2017-08-25 中国民航大学 A kind of unmanned plane trunking traffic path planning method of space-time block code
CN108037523A (en) * 2017-10-26 2018-05-15 北京航空航天大学 A kind of electron assistant beam alignment applied to unmanned plane

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9780859B2 (en) * 2014-02-28 2017-10-03 Spatial Digital Systems, Inc. Multi-user MIMO via active scattering platforms
GB2536015A (en) * 2015-03-03 2016-09-07 Stratospheric Platforms Ltd Process and apparatus for communicating with a user antenna

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106464342A (en) * 2014-02-17 2017-02-22 优必奇奥姆有限责任公司 Broadband access system via drone/uav platforms
CN107094044A (en) * 2017-03-30 2017-08-25 中国民航大学 A kind of unmanned plane trunking traffic path planning method of space-time block code
CN108037523A (en) * 2017-10-26 2018-05-15 北京航空航天大学 A kind of electron assistant beam alignment applied to unmanned plane

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
《Position-Aided Compressive Channel Estimation and Tracking for Millimeter Wave Multi-User MIMO Air-to-Air Communications》;Javier Rodriguez-Fernandez等;《2018 IEEE International Conference on Communications Workshops》;IEEE;20180705;1-6 *
《Sum Rate Maximization in UAV-Enabled Mobile Relay Networks》;Zhen Xue；Qihui Wu；Zhiyong Feng等;《2018 10th International Conference on Wireless Communications and Signal Processing》;IEEE;20181203;1-6 *
《基于定向天线的无人机自组网路由协议研究》;董航；孙志刚;《电子测量技术》;20180930;第41卷(第18期);85-90 *

Also Published As

Publication number Publication date
CN109743093A (en) 2019-05-10

Similar Documents

Publication Publication Date Title
CN109743093B (en) Unmanned aerial vehicle millimeter wave communication beam tracking method
CN112738764B (en) Broadband millimeter wave beam tracking method based on vehicle motion track cognition
US9128186B2 (en) Target tracking device and target tracking method
CN110362075B (en) Unmanned ship output feedback formation control design method with preset performance
US10476584B1 (en) Systems and methods for autonomous operations of ground station networks
Chen et al. Edge computing assisted autonomous flight for UAV: Synergies between vision and communications
JP2016171458A (en) Tracking antenna system and tracking antenna device
Xiao et al. Trajectory prediction of UAV in smart city using recurrent neural networks
JP2016142705A (en) Tracking system, tracking method, and program
Kim Control laws to avoid collision with three dimensional obstacles using sensors
CN111551968B (en) Unmanned aerial vehicle alignment system and method based on deep learning motion prediction
WO2024082475A1 (en) Autonomous navigation system and method based on millimeter-wave integrated sensing and communication unmanned device
CN109918706B (en) Generalized dynamics-based satellite-antenna coupling system path planning algorithm
González‐Prelcic et al. High‐Capacity Millimeter Wave UAV Communications
Watza et al. Hybrid rf propagation model using itm and gaussian processes for communication-aware planning
JP7156464B2 (en) Vehicles and Programs
WO2022151900A1 (en) Channel estimation method based on neural network and communication apparatus
CN111342875B (en) Unmanned aerial vehicle communication robust beam forming method based on DoA estimation
CN114035154A (en) Motion parameter assisted single-station radio frequency signal positioning method
Mou et al. Millimeter wave wireless communication assisted three-dimensional simultaneous localization and mapping
Eskandari et al. SLAPS: Simultaneous localization and phase shift for a RIS-equipped UAV in 5G/6G wireless communication networks
Garnett et al. Optimal Control for Improved UAV Communication
CN115603789B (en) Method for generating and tracking high-dynamic millimeter wave directional narrow beam
Wan et al. Rigid body localization based on conformal sensor array in non-line-of-sight environment
US11916612B2 (en) Mobile terminal and communication quality prediction method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant