Unmanned aerial vehicle communication low-complexity beam prediction method and system based on control view angle

Abstract

The invention discloses an unmanned aerial vehicle communication low-complexity beam prediction method and system based on a control view angle, and aims to solve the problem of narrow beam alignment and tracking in a high mobility scene. According to the method, beam offset is decomposed into offset caused by attitude change of the unmanned aerial vehicle and offset caused by relative position change of the unmanned aerial vehicle and the base station. For the offset generated by the attitude change of the unmanned aerial vehicle, the variable quantity of the offset relative to the previous time slot is predicted, based on a command / control sequence provided by a flight control system, local dynamic characteristics are captured through exponential decay variation, a dynamic action space is constructed in combination with the beam offset variable quantity, and the precision and real-time performance of beam prediction are optimized. And the beam offset caused by the change of the relative position of the base station is directly calculated through a geometric method. The method is extremely low in calculation complexity, can be realized in real time, and has a relatively high prediction success rate, relatively low beam scanning overhead and relatively high data transmission performance in a high dynamic scene of the unmanned aerial vehicle.

Description

Technical Field

[0001] The present invention belongs to the field of wireless communication, relates to a millimeter-wave UAV communication system, and particularly relates to a low-complexity beam prediction method and system for UAV communication from a control perspective. Background Art

[0002] UAV (Unmanned Aerial Vehicle) communication shows great potential in future wireless networks due to its flexibility and line-of-sight (LoS)-dominant characteristics, and can support reliable connections with larger capacity and longer distance. Millimeter-wave communication utilizes the spectrum resources between 30 - 300 GHz, and thus can provide higher transmission bandwidth, which is a key technology to meet the demand for high-speed data transmission. However, the high-frequency characteristics of millimeter waves lead to significant path loss, and beam alignment based on high-gain narrow beams faces severe challenges in high-dynamic UAV movement scenarios. Combining UAV with millimeter-wave technology can significantly improve communication performance, but how to achieve efficient and stable beam alignment is still a technical problem to be solved urgently.

[0003] Existing technologies usually adopt beam training and tracking schemes. Beam training includes initial alignment (such as adaptive search or hierarchical search), but has high computational overhead and poor real-time performance; subsequent tracking relies on beam prediction to reduce the search range. Current beam prediction methods are mainly divided into three categories: model-based methods (such as Kalman filtering, which requires manual construction of an accurate model and has limited applicability), machine learning methods (such as supervised learning SL relying on a large amount of labeled data, and reinforcement learning RL having a slow convergence speed and high computational complexity), and perception-assisted methods (such as using GPS or IMU attitude information, but relying on instantaneous states and being difficult to predict future change trends). These methods generally have defects such as high data requirements, complex models, insufficient generalization ability, or poor real-time performance, and are difficult to meet the communication requirements of UAVs in complex flight tasks. Summary of the Invention

[0004] Object of the Invention: The object of the present invention is to provide a low-complexity beam prediction method and system for UAV communication from a control perspective, to achieve efficient and stable beam tracking, improve the communication performance between high-speed UAVs and base stations in a semi-autonomous flight mode, reduce computational overhead and data acquisition costs at the same time, and enhance the adaptability of the method in different dynamic environments.

[0005] Technical Solution: To achieve the above object of the invention, the present invention provides a low-complexity beam prediction method for UAV communication from a control perspective. For the communication scenario between a UAV and a base station, the total beam offset is divided into the offset caused by the UAV attitude change and the offset caused by the relative position with the base station, and different prediction or estimation methods are adopted for the two types of offsets; the method specifically includes:

[0006] For the beam offset caused by the change in the UAV attitude, divide the possible change range of the change amount relative to the previous time slot into multiple sub-intervals, and construct and initialize the corresponding action space for each sub-interval; define a threshold parameter for selecting the optimal action space; in each time slot, first calculate the Exponential Decay Variance (EDV), and then select the optimal action space according to the set threshold relationship; the EDV is calculated based on the historical Control / Command Sequence (CCS) of the Flight Control System (FCS), and a higher weight is given to the recent changes by introducing an exponential decay factor; after selecting the action space, further select a specific action to determine the change amount of the beam offset caused by the change in the UAV attitude.

[0007] For the beam offset caused by the relative position between the UAV and the base station, it is directly calculated through spatial geometric relationships; by synthesizing the sum of the two types of beam offsets, determine the subspace containing the optimal beam direction for the current time slot, find the optimal beam by scanning this subspace, and use the optimal beam for data transmission.

[0008] Preferably, construct multiple action spaces and each action space corresponds to a range of wave speed change rates.

[0009] Preferably, divide the entire beam change range into at least two sub-ranges of faster and slower, and the corresponding action spaces are represented as follows:

[0010]

[0011] Among them, and correspond to the ranges of fast change and slow change respectively, w represents the change amount of the beam offset caused by the change in the UAV attitude between two adjacent time slots, b depicts the size of the interval containing this beam offset change amount, S and K represent the sizes of the two action spaces respectively, and the beam offset change amount corresponding to each action in the space is greater than the beam offset change amount corresponding to each action in the space ; when the EDV value is greater than the threshold, select otherwise select

[0012] Preferably, for each time slot t n , calculate the EDV according to the following formula:

[0013]

[0014] Among them, γ∈(0,1) is the exponential decay factor, ΔT = MΔt is the detection window length, M is the historical window length, Δt is the sampling interval or the length of each time slot, R(·) is the control / command function, R(t n-i ) and R(t n-i-1) respectively represent the sampling values corresponding to the control / command functions for time slots t n-i and t n-i-1 .

[0015] Preferably, after the action space is selected, the epsilon-greedy, Boltzmann or UCB strategy is further used to select a specific action, and the change amount of the beam offset caused by the attitude change is predicted.

[0016] Preferably, the steps of selecting a specific action and determining the change amount of the beam offset caused by the attitude change of the UAV include:

[0017] Initialize the parameters of the selected strategy, and then execute the strategy at each time slot to obtain a specific action, denoted as a n =(w n , b n ), where w n represents the change amount of the beam offset corresponding to the action, and b n represents the size of the interval containing the change amount of the beam offset;

[0018] Let u0 represent the width of each beam in the pitch direction. For the selected action a n =(w n , b n ), the range of the change amount of the beam offset caused by the attitude change of the UAV is ((w n -b n / 2)u0, (w n +b n / 2)u0);

[0019] Let θ n-1，1 represent the beam offset on the pitch channel at time t n-1 . Then, in time slot t n , the range of the beam offset caused by the attitude change is

[0020]

[0021] Let φ n-1，1 represent the beam offset on the roll channel at time t n-1 . Then, in the roll direction, the range of the beam offset corresponding to time slot t n is

[0022]

[0023] where a' n =(w' n , b' n ) is the action on the roll channel, and v0 represents the width of each beam in the roll direction.

[0024] Preferably, the beam offset caused by the relative displacement is calculated through spatial geometric relations, and then the subspace containing the optimal beam direction is determined by synthesizing the two types of beam offsets. The optimal beam is found through local scanning, including:

[0025] Assume that the heading of the UAV is parallel to the x-axis, and the influence of the earth's curvature is ignored. Let the relative position coordinates of the UAV and the base station at time slot t n be P n =(x n , y n , z n ) and P BS =(0, 0, 0). Then, the beam offsets in the pitch channel and roll channel caused by the relative position of the UAV and the base station are respectively

[0026]

[0027] Calculate the ranges of the total beam offsets in the pitch direction and roll direction, which are respectively and The optimal beam direction can be found by scanning the space formed by these two intervals;

[0028] Use the found optimal beam for data transmission, and then switch to the next time slot.

[0029] Based on the same inventive concept, the present invention provides a low-complexity beam prediction system for UAV communication from a control perspective. For the UAV and base station communication scenario, the total beam offset is divided into the offset caused by the UAV attitude change and the offset caused by the relative position with the base station. Different prediction or estimation methods are used for the two types of offsets; the system includes:

[0030] The first beam offset prediction module is used for the beam offset caused by the UAV attitude change. The possible change range of the change amount relative to the previous time slot is divided into multiple sub-intervals, and the corresponding action space is constructed and initialized for each sub-interval; a threshold parameter is defined to select the optimal action space; in each time slot, the exponentially decaying variance (EDV) is first calculated, and then the optimal action space is selected according to the set threshold relationship; the EDV is calculated based on the historical control / command sequence (CCS) of the flight control system (FCS), and a higher weight is given to the recent changes by introducing an exponentially decaying factor; after the action space is selected, a specific action is further selected to determine the change amount of the beam offset caused by the UAV attitude change;

[0031] The second beam offset prediction module is used to directly calculate the beam offset caused by the relative position of the UAV and the base station through spatial geometric relations;

[0032] The optimal beam prediction module is used to synthesize the sum of two types of beam offsets, determine the subspace containing the optimal beam direction of the current time slot, find the optimal beam by scanning this subspace, and use the optimal beam for data transmission.

[0033] The present invention also provides a computer system, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, the steps of the low-complexity beam prediction method for UAV communication based on the control perspective are implemented.

[0034] The present invention also provides a computer program product, including a computer program. When the computer program is executed by the processor, the steps of the low-complexity beam prediction method for UAV communication based on the control perspective are implemented.

[0035] Beneficial effects: The present invention proposes a low-complexity beam prediction method and system for UAV communication based on the control perspective. Combining the command sequence (CCS) of the flight control system (FCS), it can efficiently and stably achieve beam prediction and tracking in complex high-dynamic flight tasks, significantly improving communication performance. This method makes full use of the high-level future information provided by CCS, captures local dynamic characteristics through exponential decay variation (EDV), constructs a dynamic action space by combining the beam offset change amount, and optimizes the accuracy and real-time performance of beam prediction. In addition, the present invention decomposes the beam offset into a local offset caused by attitude change and a global offset caused by relative position change, and completes prediction and adjustment only relying on CCS and simple geometric calculations, without the need for large-scale data or complex model training, significantly reducing the computational complexity and data acquisition cost.

[0036] Compared with the prior art, the significant effect of the present invention is that it overcomes the defects of the traditional beam prediction method, such as the dependence on large-scale data, high computational complexity, and insufficient real-time performance. It realizes efficient prediction through lightweight algorithm design, achieves excellent tracking performance and communication efficiency while maintaining low complexity, and meets the real-time and robustness requirements of high-speed UAV communication in different dynamic environments. Description of the Drawings

[0037] Figure 1 It is a flowchart of the method according to an embodiment of the present invention.

[0038] Figure 2 It is a diagram of the device used in an embodiment of the present invention.

[0039] Figure 3 It is a diagram showing the relationship between CCS and the UAV attitude angle (RollAngle) changing with time in an embodiment of the present invention, and is used to illustrate the characteristics of CCS as future information.

[0040] Figure 4This is a comparison chart of the EDV and the traditional total variation (TV) in the analysis of the function change behavior in the embodiments of the present invention, which is used to illustrate the advantages of the EDV in capturing local change behaviors.

[0041] Figure 5 This is a performance comparison chart of the beam prediction and tracking (BPT) algorithm based on the EDV and the equal-weight variation (EWV) in terms of the probability of successful alignment (PSA) in the embodiments of the present invention, which is used to verify the effectiveness of the EDV metric.

[0042] Figure 6 This is a performance curve chart of the probability of successful beam alignment between the embodiments of the present invention and the comparative algorithm. Detailed implementation manners

[0043] The technical solution of the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0044] The embodiments of the present invention disclose a low-complexity beam prediction method for UAV communication from a control perspective. The future change information of the UAV attitude and position is captured through the command sequence (CCS) of the flight control system (FCS), and the beam prediction is optimized based on the exponentially decaying variation (EDV) metric. In this embodiment, for the UAV-to-base station communication scenario, the total beam offset is divided into two parts (i.e., the offset caused by the UAV attitude change and the offset caused by the relative position with the base station), and different prediction or estimation methods are adopted for the two types of offsets. For the offset caused by the UAV attitude change, the change amount relative to the previous time slot is predicted. Assuming that the yaw angle of the UAV is 0 degrees (corresponding to the headless mode in the UAV), the 3D beam prediction problem is reduced to predicting the offsets of the pitch channel and the roll channel. Due to the symmetry of the UAV body and the fact that the flight control system independently controls each channel, in order to reduce the computational complexity, this embodiment independently predicts the beam offsets on the pitch and roll channels.

[0045] In this embodiment, the low-complexity beam prediction method for UAV communication from a control perspective is as Figure 1 shown, and specifically includes the following steps. The operations in steps 1-3 are the same for the prediction of the beam offsets on the pitch and roll channels.

[0046] Step 1: In order to predict or estimate the change amount of the beam offset caused by the UAV attitude change, the possible change range is divided into multiple sub-intervals (set as K), and the corresponding action space is constructed and initialized for each sub-interval. Further, K-1 threshold parameters are defined to facilitate the algorithm to select the optimal action space.

[0047] Step 2: For each time slot, first calculate the Exponential Decay Variance (EDV), and then select the optimal action space according to the set threshold relationship; the EDV is calculated based on the historical Control / Command Sequence (CCS) of the Flight Control System (FCS), and a higher weight is given to recent changes by introducing an exponential decay factor.

[0048] Step 3: After selecting the action space, any feasible strategy can be used to further select a specific action, based on which the change amount of the beam offset caused by the UAV attitude change can be determined (which is also approximately equal to the change amount of the beam direction).

[0049] Step 4: For the beam offset caused by the relative position between the UAV and the base station, it can be directly calculated through spatial geometric relationships. Then, by synthesizing the sum of the two types of beam offsets, a subspace containing the optimal beam direction in the current time slot is determined, and the optimal beam can be found by scanning this subspace, and this optimal beam is used for data transmission.

[0050] Specifically, in Step 1, multiple action spaces need to be constructed, and each action space corresponds to a range of wave speed change rates. If the entire beam change range is divided into two sub-ranges, fast and slow, the action spaces can be expressed as follows:

[0051]

[0052] Among them, and correspond to the fast change and slow change ranges respectively, w represents the change amount of the beam offset caused by the UAV attitude change between two adjacent time slots, b depicts the size of the interval containing this beam offset change amount, S and K represent the sizes of the two action spaces respectively, and the beam offset change amount corresponding to each action in the space is greater than the beam offset change amount corresponding to each action in the space ; when the EDV value is greater than the threshold, is selected, otherwise

[0053] In Step 2, for each time slot t n (the current time slot), calculate the EDV according to the following formula: <[

[0054]

[0055] Among them, γ ∈ (0, 1) is the exponential decay factor, ΔT = MΔt is the detection window length, M is the historical window length, Δt is the sampling interval or the length of each time slot, R(·) is the control / command function (from the remote controller), R(t i ) represents the sampling value of the control / command function corresponding to the time slot t i , and R(t i-1) has a similar definition.

[0056] Select the action space according to the calculated EDV value. For the case of two action spaces (i.e., and ), if the EDV value is greater than the threshold C, it indicates that the beam offset caused by the UAV attitude change is relatively large. At this time, the action space should be selected. Otherwise, select the action space For the case of multiple action spaces, the operation is similar.

[0057] In step 3, after selecting the action space, further use strategies including but not limited to epsilon-greedy, Boltzmann, or UCB to select specific actions, and predict the change amount of the beam offset caused by the attitude change, that is:

[0058] 1) First, initialize the parameters of the selected strategy, and then execute the strategy in each time slot to obtain the specific action, denoted as a n =(w n , b n ), where w n represents the beam offset change amount corresponding to the action, and b n represents the size of the interval containing the beam offset change amount.

[0059] 2) Let u0 represent the width of each beam in the pitch direction. For the selected action a n =(w n , b n ), the range of the beam offset change amount caused by the UAV attitude change is ((w n -b n / 2)u0, (w n +b n / 2)u0).

[0060] 3) Let θ n-1，1 represent the beam offset on the pitch channel at time t n-1 (caused by the UAV attitude change). Then, at time slot t n the range of the beam offset caused by the attitude change is

[0061]

[0062] 4) For the roll channel, use exactly the same method. Let φ n-1，1 represent the beam offset on the roll channel at time t n-1 (caused by the UAV attitude change). Then, in the roll direction, at time slot t n the corresponding range of the beam offset is

[0063]

[0064] where a' n =(w' n , b' n ) is an action on the roll channel, and v0 represents the width of each beam in the roll direction.

[0065] In step 4, the beam offset caused by the relative displacement is calculated through spatial geometric relations, and then the subspace containing the optimal beam direction is determined by integrating the two types of beam offsets. The optimal beam is found through local scanning, that is:

[0066] 1) Without loss of generality, assume that the heading of the UAV is parallel to the x-axis and the influence of the earth's curvature is ignored. Let the relative position coordinates of the UAV and the base station at time slot t n be P n =(x n , y n , z n ) and P BS =(0, 0, 0). Then the beam offsets on the pitch channel and the roll channel caused by the relative position of the UAV and the base station are respectively

[0067]

[0068] 2) Calculate the ranges of the total beam offsets in the pitch direction and the roll direction, which are respectively and The optimal beam direction can be found by scanning the space formed by these two intervals.

[0069] 3) Use the found optimal beam for data transmission, and then switch to the next time slot.

[0070] In this embodiment, the selected historical information is from the CCS in the FCS. As a kind of future information, the CCS has the following characteristics: The CCS is directly provided by the FCS and contains a sequence of control commands (such as real-time control of the pitch, roll, and throttle channels) input by remote control, reflecting the future attitude adjustment intention of the UAV, rather than only depending on the current or past instantaneous states (such as position and attitude angle); The CCS is generated through the control loop of the FCS. For example, based on the output of a proportional-integral-derivative (PID) controller, compared with the instantaneous measurement data (such as GPS position or IMU attitude) used in traditional methods, it can predict the movement / change trend of the UAV in advance and has time foresight.

[0071] When calculating the EDV, the historical sequence of CCS is utilized to capture the dynamic characteristics of future changes through weighted differences, thereby providing a more accurate basis for environmental perception in the selection of the action space. The acquisition method of CCS includes real-time extraction through the serial port, USB interface, or Ethernet interface of the FCS. For example, based on the parsing of the Mavlink protocol, the data sampling frequency (such as 50Hz) is matched with the dynamic response of the UAV to ensure the real-time and reliability of the information.

[0072] The EDV adopted in this embodiment has the following differences and advantages compared with the traditional total variation: The EDV introduces an exponential decay factor γ, which assigns higher weights to recent changes and gradually attenuates the influence on earlier historical data, while the total variation treats the changes in all time slots equally; The EDV can more accurately characterize the local change behavior. Especially in the high-dynamic UAV scenario, by highlighting the change rate near the current time slot, it can effectively distinguish fast and slow change patterns, while the total variation reflects the global changes in the entire time interval and cannot focus on local dynamics; The computational complexity of the EDV is comparable to that of the total variation, both being (n is the number of time slots within the window). By adjusting γ and ΔT, a better balance can be achieved between real-time performance and robustness. Experimental verification shows that the EDV is 10%-15% higher than the total variation in terms of the probability of successful alignment (PSA).

[0073] As Figure 2 shown, to further verify the practical value of the method of the present invention, the widely used F450 quadcopter UAV equipped with Pixhawk hardware and PX4 software as the flight control platform is adopted in the experiment, and the proposed algorithm is run on a low-power single-board computer (ARM dual-core Cortex A7 processor). In the experimental setup, the CCS information is transmitted from the FCS to the computing platform through the unoccupied serial port (Serial4), ensuring the easy implementation of the scheme. The test results show that the control-enabled or control-assistant beam prediction and tracking scheme (CBPT) of the present invention has better PSA and EAR performance than the existing technologies (such as equal-weighted variation or total variation EWV, stochastic bandit learning SBL, and Gaussian process regression GPR) in both local (see Figure 5 ) and global (see Figure 6 ) scenarios. Especially in the global scenario, by decoupling the attitude and position change factors, the generalization ability of the method is further improved.

[0074] As Figure 3As shown, the horizontal axis represents time (unit: second), and the vertical axis represents the original value. Among them, the blue curve (Remote Input) represents the CCS (remote control input) obtained through FCS, and the orange curve (RollAngle) represents the roll angle of the UAV (Roll Angle), which directly corresponds to the change in the beam direction of the corresponding channel. It can be observed from the figure that the change in CCS precedes the change in the attitude angle. This phenomenon indicates that CCS can reflect the future attitude change / adjustment intention of the UAV in advance, showing time foresight. Based on this important feature, the present invention uses CCS as an input and designs an efficient beam prediction algorithm through exponential decay variation (EDV) and beam offset change amount, significantly improving the prediction accuracy and real-time performance of millimeter-wave UAV communication.

[0075] As Figure 4 shown, the horizontal axis represents the independent variable, and the vertical axis represents the function value. Among them, the blue curve represents the exponential function, and its change behavior is analyzed through traditional total variation (TV) and exponential decay variation (EDV) respectively. Two key points and their corresponding two time intervals are marked in the figure, which are used to compare the change characteristics of the function at different positions. Traditional total variation (TV) describes its behavior by calculating the cumulative change of the function in the entire interval. However, this method can only reflect the global change characteristics and cannot effectively distinguish the differences in the local change rates of the function at different time periods. In contrast, the EDV proposed by the present invention assigns higher weights to recent changes by introducing an exponential decay factor, weakening the influence of earlier historical data, so as to be able to more accurately capture the local dynamic behavior of the function near the decision-making moment. It can be observed from the figure that in the interval near the left side of the function peak, the function change rate is relatively fast, while in the interval on the right side of the peak, the change rate is significantly slower. The traditional total variation (TV) has the same evaluation result for the changes in these two intervals, so it fails to reflect the differences in local changes. However, EDV highlights the characteristics of the interval with a faster change rate through the exponential weighting mechanism, and draws a conclusion consistent with the actual change behavior. This feature indicates that EDV is more suitable for scenarios that require attention to local dynamics, such as the real-time response to the attitude change of the UAV in beam prediction. Based on this advantage, the present invention uses EDV to analyze the mutation characteristics of CCS and designs an efficient beam prediction algorithm to further improve the real-time performance and prediction accuracy of millimeter-wave UAV communication.

[0076] Figure 5 Shows the PSA performance of the EDV algorithm corresponding to different decay factors γ. The horizontal axis represents the signal-to-noise ratio (SNR), and the vertical axis represents the probability of successful beam alignment (PSA). Figure 5 Consider the local situation, that is, the flight range of the UAV is small, so the beam offset change caused by the relative position between the UAV and the base station can be ignored. Figure 5It contains multiple curves. The brown dashed line (corresponding to γ = 1) represents the performance of the beam prediction and tracking (BPT) algorithm based on equal-weighted variation or classical total variation (EWV), and the other colored solid lines (corresponding to different attenuation factors γ respectively) represent the performance of the BPT algorithm based on the exponential decay variation (EDV) proposed in the present invention. It can be observed from the figure that under various settings of the attenuation factor γ, the BPT algorithm based on EDV shows better performance than EWV under different signal-to-noise ratio conditions. Especially in the range of signal-to-noise ratio from low to high, the EDV method shows a significant advantage in the successful alignment probability. This result indicates that the EDV metric can more effectively improve the accuracy of beam prediction.

[0077] Figure 6 The beam success alignment probability performance curves of different beam prediction and tracking algorithms are provided. The hardware-in-the-loop simulation scenario considered is the global scenario, that is, the flight range of the unmanned aerial vehicle (UAV) is large, and the beam offset change caused by the relative position between the UAV and the base station cannot be ignored. CBPT is the control-empowered / assisted intelligent beam prediction and tracking algorithm proposed in the present invention, and the attenuation factor is 0.85. Figure 6 It can be seen that the algorithm of the present invention achieves the optimal beam tracking performance for two reasons: on the one hand, it is the advantage brought by the EDV-based beam variation mentioned above, and on the other hand, the present invention decouples the two beam variations caused by the UAV attitude and the relative position between the UAV and the base station, effectively solving the generalization problem of the AI algorithm caused by the geographical location change in UAV communication. In contrast, the algorithm based on Gaussian process regression cannot solve the generalization problem caused by the geographical location of the UAV flight. When the flight range exceeds the geographical location corresponding to the training data, the error probability of the algorithm extrapolation increases significantly, thus reducing the algorithm performance.

[0078] From the perspective of control theory, the present invention proposes an innovative beam prediction method for UAV millimeter-wave communication. Different from the traditional technical solutions that rely on instantaneous position or attitude information, for the first time, it deeply explores the command sequence (CCS) in the flight control system (FCS) as an easily accessible and forward-looking information source, which is particularly suitable for semi-autonomous flight modes. The CCS contains future motion intention information based on a proportional-integral-derivative (PID) controller, and can better reflect the dynamic trend of the UAV compared with instantaneous measurement data. The present invention also designs an efficient algorithm to identify the mutation features in the CCS based on total variation detection, and improves it to propose an exponential decay variation (EDV) method, which highlights local dynamic changes through exponential weighting and overcomes the limitation that traditional variation detection is overly sensitive to global changes. At the same time, a dynamic action space is constructed by combining the beam offset change amount to adapt to fast and slow UAV movement scenarios, further optimizing the beam prediction accuracy. The method of the present invention decomposes the beam offset into a local offset caused by attitude changes and a global offset caused by the relative position change between the base station and the UAV, and realizes real-time tracking only relying on CCS information and a simple geometric model, significantly reducing the computational complexity and data acquisition cost, solving the problems of complex models and insufficient real-time performance in the prior art, thereby improving the robustness and practicality of millimeter-wave UAV communication.

[0079] Based on the same inventive concept, an embodiment of the present invention discloses a low-complexity beam prediction system for UAV communication from a control perspective. For the UAV and base station communication scenario, the total beam offset is divided into an offset caused by UAV attitude changes and an offset caused by the relative position with the base station, and different prediction or estimation methods are adopted for the two types of offsets; the system includes:

[0080] A first beam offset prediction module, for the beam offset caused by UAV attitude changes, divides the possible change range of the change amount relative to the previous time slot into multiple sub-intervals, constructs and initializes a corresponding action space for each sub-interval; defines a threshold parameter for selecting the optimal action space; in each time slot, first calculates the exponential decay variation (EDV), and then selects the optimal action space according to the set threshold relationship; the EDV is calculated based on the historical control / command sequence (CCS) of the flight control system (FCS), and higher weights are assigned to recent changes by introducing an exponential decay factor; after the action space is selected, specific actions are further selected to determine the change amount of the beam offset caused by UAV attitude changes;

[0081] A second beam offset prediction module, for directly calculating the beam offset caused by the relative position between the UAV and the base station through spatial geometric relationships;

[0082] The optimal beam prediction module is used to synthesize the sum of two types of beam offsets to determine a subspace containing the optimal beam direction for the current time slot, find the optimal beam by scanning this subspace, and use the optimal beam for data transmission.

[0083] For the specific implementation of each module, refer to the above method embodiments and will not be elaborated here.

[0084] An embodiment of the present invention also discloses a computer system, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is executed by the processor, the steps of the low-complexity beam prediction method for UAV communication based on the control perspective are implemented.

[0085] An embodiment of the present invention also discloses a computer program product, including a computer program. When the computer program is executed by the processor, the steps of the low-complexity beam prediction method for UAV communication based on the control perspective are implemented.

[0086] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to the processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing devices, so that when the program codes are executed by the processor or controller, the steps of the method of the present invention are implemented. The program codes can be executed entirely on the machine, partially on the machine, executed partially on the machine as an independent software package and partially on a remote machine, or executed entirely on a remote machine or server. Where the present invention is not elaborated, it is common knowledge to those skilled in the art.

Claims

1. A low-complexity beam prediction method for UAV communication based on a control perspective, characterized in that For the communication scenario between a drone and a base station, the total beam offset is divided into the offset caused by the change in the drone's attitude and the offset caused by the relative position with the base station. Different prediction or estimation methods are adopted for the two types of offsets. The methods include: For the beam offset caused by the change in the drone's attitude, the possible change range of the change amount relative to the previous time slot is divided into multiple sub-intervals, and the corresponding action space is constructed and initialized for each sub-interval; a threshold parameter is defined to select the optimal action space; in each time slot, the exponentially weighted moving variance (EWMA) is first calculated, and then the optimal action space is selected according to the set threshold relationship; the EWMA is calculated based on the historical control / command sequence (CCS) of the flight control system (FCS), and a higher weight is given to the recent changes by introducing an exponential decay factor; after the action space is selected, a specific action is further selected to determine the change amount of the beam offset caused by the change in the drone's attitude; For the beam offset caused by the relative position between the drone and the base station, it is directly calculated through the spatial geometric relationship; the sum of the two types of beam offsets is integrated to determine the subspace containing the optimal beam direction in the current time slot, and the optimal beam is found by scanning this subspace, and the optimal beam is used for data transmission.

2. The low-complexity beam prediction method for UAV communication based on the control perspective according to claim 1, wherein Construct multiple action spaces, and each action space corresponds to a range of wave speed change rates.

3. The low-complexity beam prediction method for UAV communication based on the control perspective according to claim 2, wherein The entire beam change range is divided into at least two sub-ranges, namely faster and slower, and the corresponding action spaces are represented as follows: Among them, and correspond to the fast-changing and slow-changing ranges respectively. w represents the beam offset change amount generated by the UAV attitude change between two adjacent time slots. b depicts the size of the interval containing the beam offset change amount. S and K represent the sizes of two action spaces respectively. The beam offset change amount corresponding to each action in the space is greater than the beam offset change amount corresponding to each action in the space ; when the EDV value is greater than the threshold, select otherwise select 4. The low-complexity beam prediction method for UAV communication based on the control perspective according to claim 1, wherein For each time slot t n , calculate the EDV according to the following formula: where γ ∈ (0, 1) is the exponential decay factor, ΔT = MΔt is the detection window length, M is the historical window length, Δt is the sampling interval or the length of each time slot, and R(·) is the control / command function, R(t n-i ) and R(t n-i-1 ) denote the sampled values of the control / command function corresponding to time slots t n-i and t n-i-1 , respectively.

5. The low-complexity beam prediction method for UAV communication based on the control perspective according to claim 1, characterized in that, After the action space is selected, the epsilon-greedy, Boltzmann or UCB strategy is further used to select a specific action to predict the change amount of the beam offset caused by the attitude change.

6. The low-complexity beam prediction method for UAV communication based on the control perspective according to claim 5, wherein The steps of selecting a specific action to determine the change amount of the beam offset caused by the change in the drone's attitude include: Initialize the parameters of the selected policy, and then execute the policy in each time slot to obtain the specific action, denoted as a n =(w n ,b n ), where w n represents the beam offset change amount corresponding to the action, and b n represents the size of the interval containing the beam offset change amount; Let \(u_0\) denote the width of each beam in the elevation direction for the selected action \(a\). n =(w n , b n ), the range of the change in beam offset caused by the attitude change of the UAV is \(((w n - b n / 2)u_0, (w n + b n / 2)u_0)\); Let θ n-1，1 represent the beam offset on the pitch channel at time t n-1 . Then, in time slot t n , the range of the beam offset caused by the attitude change is Let φ n-1，1 represent the moment t n-1 (the beam offset on the roll channel, then in the roll direction the beam offset corresponding to time slot t n has a range of where a' n = (w' n , b' n ) is the action on the roll channel, and v0 represents the width of each beam in the roll direction.

7. The method for predicting a low-complexity beam of UAV communication based on a control perspective according to claim 6, wherein The beam offset caused by the relative displacement is calculated through the spatial geometric relationship, and then the two types of beam offsets are integrated to determine the subspace containing the optimal beam direction, and the optimal beam is found by local scanning, including: Assume that the heading of the UAV is parallel to the x-axis and the influence of the earth's curvature is ignored. Let the relative position coordinates of the UAV and the base station at time slot t n be P n =(x n , y n , z n ) and P BS =(0, 0, 0). Then the beam offsets in the pitch channel and roll channel caused by the relative position of the UAV and the base station are respectively Calculate the range of the total beam offset in the pitch direction and the roll direction, which are respectively and The optimal beam direction can be found by scanning the space formed by these two intervals; Use the found optimal beam for data transmission, and then switch to the next time slot.

8. A low-complexity beam prediction system for UAV communication based on the control perspective, characterized in that, For the communication scenario between a drone and a base station, the total beam offset is divided into the offset caused by the change in the drone's attitude and the offset caused by the relative position with the base station. Different prediction or estimation methods are adopted for the two types of offsets. The system includes: The first beam offset prediction module is used to, for the beam offset caused by the change in the drone's attitude, divide the possible change range of the change amount relative to the previous time slot into multiple sub-intervals, construct and initialize the corresponding action space for each sub-interval; define a threshold parameter to select the optimal action space; in each time slot, the exponentially weighted moving variance (EWMA) is first calculated, and then the optimal action space is selected according to the set threshold relationship; the EWMA is calculated based on the historical control / command sequence (CCS) of the flight control system (FCS), and a higher weight is given to the recent changes by introducing an exponential decay factor; after the action space is selected, a specific action is further selected to determine the change amount of the beam offset caused by the change in the drone's attitude; The second beam offset prediction module is used to directly calculate the beam offset caused by the relative position of the UAV and the base station through spatial geometric relationships; The optimal beam prediction module is used to synthesize the sum of the two types of beam offsets, determine the subspace containing the optimal beam direction in the current time slot, find the optimal beam by scanning this subspace, and use the optimal beam for data transmission.

9. A computer system, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the computer program is executed by a processor, it implements the steps of the low-complexity beam prediction method for UAV communication based on the control perspective according to any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the low-complexity beam prediction method for UAV communication based on the control perspective according to any one of claims 1-7.

Images