Large fixed-wing unmanned aerial vehicle high-precision delivery system

Abstract

The application discloses a large fixed-wing unmanned aerial vehicle high-precision delivery system, which comprises an unmanned aerial vehicle platform, a delivery device, a navigation and sensing unit and a control unit; the control unit is used for executing a dynamic air delivery window and flight attitude cooperative control method, which comprises the following steps: based on the unmanned aerial vehicle state information, the environmental wind field information and the pre-stored target point geographic information acquired by the navigation and sensing unit in real time, a dynamic air delivery starting window is calculated in real time through a preset air delivery trajectory model; after entering the air delivery starting window, the control unit synchronously executes the following operations: the unmanned aerial vehicle platform is controlled to execute a predetermined attitude maneuvering action, so as to stabilize the initial condition when the goods are separated; meanwhile, in the process of executing the attitude maneuvering action, the optimal delivery time is continuously evaluated according to the delivery quality factor Q calculated in real time. The application can realize high-precision delivery of the unmanned aerial vehicle.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of unmanned aerial vehicle control, in particular to a large fixed-wing unmanned aerial vehicle high-precision delivery system. BACKGROUND

[0002] With the large-scale application of unmanned aerial vehicle technology in the fields of material delivery, emergency rescue, battlefield supply, etc., large fixed-wing unmanned aerial vehicles have become the core carriers for long-distance and large-scale delivery tasks due to their long endurance, strong load capacity and stable flight speed. However, their flight characteristics also bring inherent challenges in delivery: compared with small unmanned aerial vehicles or helicopters, large fixed-wing unmanned aerial vehicles have faster flight speed and larger body inertia, and in complex environments such as low altitude and mountainous areas, they are easily affected by vertical wind shear, terrain disturbance airflow, etc., resulting in significant fluctuations in flight attitude. These factors will directly affect the initial conditions when the goods are separated, and pose natural obstacles to precise delivery.

[0003] To meet the task requirements of precise delivery of goods into the target tolerance range, a large fixed-wing unmanned aerial vehicle needs to be equipped with a high-precision delivery system, which usually includes four core modules: unmanned aerial vehicle platform, delivery device, navigation and sensing unit, and control unit. However, existing high-precision delivery systems still have problems such as rough calculation of air-drop window, poor adaptability of attitude maneuver, insufficient wind field sensing accuracy, and unresolved issues of goods separation impact and residual disturbance in actual application. Therefore, a large fixed-wing unmanned aerial vehicle high-precision delivery system is proposed. SUMMARY

[0004] The present application solves the above technical problems by the following technical solutions: the present application includes an unmanned aerial vehicle platform, a delivery device, a navigation and sensing unit, and a control unit.

[0005] The control unit is used to execute a dynamic air-drop window and flight attitude cooperative control method, which includes:

[0006] Based on the real-time acquisition of unmanned aerial vehicle state information, environmental wind field information, and pre-stored target point geographic information by the navigation and sensing unit, a dynamic air-drop launch window is calculated in real time through a pre-set air-drop trajectory model. The air-drop launch window is a space-time region, which is a collection of all flight states that can ensure the goods to fall within the target tolerance range by adjusting the flight attitude and triggering the delivery within a continuous time period.

[0007] After entering the air-drop launch window, the control unit synchronously performs the following operations: controlling the unmanned aerial vehicle platform to execute the predetermined attitude maneuver action to stabilize the initial conditions when the goods are separated.

[0008] Meanwhile, during the execution of the attitude maneuver, the optimal release time is continuously evaluated according to the real-time calculated release quality factor Q, and the release device is triggered to release when the release quality factor Q meets the predetermined condition.

[0009] Further, the airdrop starting window is determined by the following calculation process:

[0010] Firstly, based on the UAV dynamics and the preset control law, the UAV flight path in the future period of time from the current time is predicted, which is composed of a plurality of discrete future state points , wherein is the position, is the speed, is the time, is the height;

[0011] Secondly, for each future state point , it is regarded as a potential release point, and the cargo drop point is estimated by using the airdrop trajectory model, and the specific process of estimating the cargo drop point by using the airdrop trajectory model includes:

[0012] Step a: calculating the total descent time of the cargo in the air , considering that there is an initial vertical velocity after the release of the cargo and air resistance, the following simplified model is used for estimation:

[0013] ;

[0014] wherein, is the vertical velocity at the instant of release of the cargo, which is determined by the UAV attitude and speed, g is the acceleration of gravity, and k is a dimensionless adjustment coefficient related to the aerodynamic characteristics of the cargo , the value of k for a streamlined cargo is closer to 1, and the value of k for a high-resistance cargo is reduced, and corresponding values can be preset in advance according to different cargos;

[0015] Step b: calculating the net displacement of the cargo in the horizontal direction , which is composed of the displacement of the cargo drifting in the wind field and its own inertial displacement:

[0016] ;

[0017] wherein, the wind field drift displacement is obtained by integrating the wind field during the descent process:

[0018] ;

[0019] Here, is the wind speed vector at the height , This represents the real-time height of the cargo during its descent.

[0020] Inertial displacement The initial velocity of the cargo upon release is generated under windless conditions:

[0021] ;

[0022] Next, calculate the total descent time of the cargo in the air. Net displacement of the object in the horizontal direction Estimated cargo landing point Specifically:

[0023] ;

[0024] Finally, filter all that meet the criteria. state point Where R is the allowable landing radius, and the spatiotemporal region formed by these state points is the airdrop initiation window. The geographic coordinates of the target landing point for the mission.

[0025] Furthermore, the predetermined attitude maneuver is a pull-up and attitude stabilization maneuver, specifically including: controlling the UAV to pitch upward around the lateral axis, increasing its angle of attack by a predetermined value to generate an additional upward lift component, which is used to partially offset the initial sinking caused by gravity and wind disturbance at the moment of cargo release.

[0026] Furthermore, the calculation process of the delivery quality factor Q integrates instantaneous delivery accuracy and flight attitude stability, specifically as follows:

[0027] ;

[0028] in, To estimate the coordinates of the cargo's landing point based on the current status, This is the current pitch angle of the drone. The target pitch angle for the predetermined attitude maneuver. and These are the weighting coefficients. To prevent small constants with a denominator of zero;

[0029] The control unit continuously calculates the Q value within the airdrop initiation window and triggers the airdrop when it reaches its maximum value.

[0030] Further, the navigation and sensing unit comprises a Doppler laser radar based wind field estimation module, which estimates the wind speed and direction at different altitudes between the UAV and the ground by emitting laser beams downward in front of the UAV and analyzing the echo signals, thereby obtaining vertical wind field profile data; the vertical wind field profile data is input into the air-drop trajectory model to accurately calculate the wind drift displacement . .

[0031] Further, the control unit comprises an adaptive controller for optimizing the parameters of the attitude maneuvering action in the sky above complex terrain;

[0032] The adaptive controller takes the environmental wind shear intensity and the terrain undulation as the input state, and takes the pitch angle increment and the attitude stabilization time as the output action, and its strategy is trained by reinforcement learning based on the drop success rate to maximize the expected value of the drop quality factor Q.

[0033] Further, the drop device adopts an active damping release mechanism, which, after receiving a trigger signal, first releases the main lock, and then applies a controllable resistance force to the cargo through a damping arm controlled by a linear servo motor, the calculation formula of the resistance force is as follows:

[0034] ;

[0035] Where c is the adjustable damping coefficient, and are the velocity vectors of the cargo and the UAV respectively; this process lasts for a very short time to smooth the speed separation process between the cargo and the UAV.

[0036] Further, during the operation of the active damping release mechanism, the control unit records the body angular velocity disturbance caused by the release , and after the damping release is completed, a compensation torque proportional to is introduced into the attitude control loop to quickly suppress the body oscillation caused by the release impact.

[0037] Compared with the prior art, the large fixed-wing unmanned aerial vehicle high-precision delivery system has the following advantages: through the cooperative work of the unmanned aerial vehicle platform, the delivery device, the navigation and sensing unit and the control unit, the control unit can calculate the dynamic air drop starting window based on the real-time obtained unmanned aerial vehicle state, environmental wind field and target point geographic information, combined with the air drop trajectory model, to ensure that the unmanned aerial vehicle adjusts the flight attitude within a continuous time period to trigger the delivery and still enables the goods to fall into the target tolerance range; after entering the air drop starting window, the control unit synchronously controls the unmanned aerial vehicle to perform the posture maneuvering action of pulling up and stabilizing the posture, which can stabilize the initial condition of the goods separation, and partially offset the initial sinking caused by gravity and wind disturbance at the moment of goods separation; at the same time, the optimal delivery time is evaluated and selected by calculating the delivery quality factor Q of the comprehensive instantaneous delivery accuracy and flight attitude stability in real time, to improve the delivery accuracy; the Doppler laser radar wind field estimation module in the navigation and sensing unit can obtain the vertical wind field profile data, to provide more accurate wind force prediction basis for the air drop trajectory model, and further improve the landing point prediction accuracy; the adaptive controller of the control unit can optimize the posture maneuvering parameters according to the environmental wind shear intensity and the terrain undulation degree, and the strategy thereof can maximize the expected value of the delivery quality factor Q through reinforcement learning training, to adapt to complex terrain environment; the active damping release mechanism of the delivery device smoothes the speed separation process of the goods and the unmanned aerial vehicle through the controllable resistance force, reduces the separation impact, and at the same time the control unit introduces a compensation torque after release, to quickly suppress the body oscillation, guarantee the flight stability of the unmanned aerial vehicle, and greatly improve the delivery accuracy and the ability of the system to adapt to complex environment, to optimize the delivery effect and the flight safety of the unmanned aerial vehicle. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 is a system block diagram of the present application. DETAILED DESCRIPTION

[0039] The embodiments of the present application will be described in detail below, and the embodiments are implemented on the premise of the technical solutions of the present application, and detailed implementation manners and specific operation processes are given, but the protection scope of the present application is not limited to the following embodiments.

[0040] As Figure 1 shown, the present embodiment provides a technical solution: a large fixed-wing unmanned aerial vehicle high-precision delivery system, comprising an unmanned aerial vehicle platform, a delivery device, a navigation and sensing unit and a control unit;

[0041] The control unit is used for executing a dynamic air drop window and flight attitude cooperative control method, comprising:

[0042] Based on the real-time state information of the UAV, the environmental wind field information and the pre-stored target point geographic information, a dynamic air-drop starting window is calculated in real time through a pre-set air-drop trajectory model, wherein the air-drop starting window is a space-time region, and is a collection of all flight states in which the UAV can trigger the drop by adjusting the flight attitude and within a continuous time period, and still ensure that the goods fall within the target tolerance range;

[0043] After entering the air-drop starting window, the control unit synchronously performs the following operations: controlling the UAV platform to perform a predetermined attitude maneuvering action to stabilize the initial conditions when the goods are separated;

[0044] Meanwhile, during the execution of the attitude maneuvering action, the optimal drop timing is continuously evaluated according to the real-time calculated drop quality factor Q, and the drop device is triggered to drop when the drop quality factor Q meets the predetermined condition.

[0045] The air-drop starting window is determined through the following calculation process:

[0046] Firstly, based on the UAV dynamics and a pre-set control law, the UAV flight path in the future period of time from the current time is predicted, and the flight path is composed of a plurality of discrete future state points , wherein is the position, is the speed, is the time, and is the height;

[0047] Secondly, for each future state point , it is regarded as a potential drop point, and the drop point of the goods is estimated by using the air-drop trajectory model , and the specific process of estimating the drop point of the goods by using the air-drop trajectory model includes:

[0048] Step a: calculating the total descent time of the goods in the air , considering that there is an initial vertical speed after the goods are released and air resistance, the following simplified model is used for estimation:

[0049] ;

[0050] wherein is the vertical speed at the instant when the goods are released, is determined by the attitude and speed of the UAV, g is the acceleration of gravity, and k is a dimensionless adjustment coefficient related to the aerodynamic characteristics of the goods , and for a streamlined goods, the value of k is closer to 1, and for a high-resistance goods, the value of k is reduced;

[0051] Step b: calculating the net displacement of the goods in the horizontal direction , which is composed of the displacement of the goods drifting in the wind field and the inertial displacement of the goods itself:

[0052] ;

[0053] wherein the wind field drift displacement By integrating the wind field during the falling process, we get:

[0054] ;

[0055] Here, is the wind speed vector at height interpolated from the vertical wind profile data, is the real-time height during the falling process of the cargo;

[0056] Inertial displacement Generated by the initial velocity when the cargo is released under windless conditions:

[0057] ;

[0058] Then, the total falling time of the cargo in the air and the net displacement of the cargo in the horizontal direction The estimated landing point of the cargo , specifically:

[0059] ;

[0060] Finally, filter all state points that satisfy where R is the allowed landing radius, and these state points constitute the air-drop starting window in space-time region, is the target landing point of the drop task;

[0061] First, based on the dynamics of the UAV and the preset control law, predict the UAV flight path composed of discrete state points in the future; then, by regarding each state point as a potential drop point, calculate the falling time of the cargo, synthesize the horizontal net displacement, and estimate the landing point of the cargo through the air-drop trajectory model; finally, filter all state points with a distance between the landing point and the target point ≤ the allowed landing radius R, to constitute a dynamic air-drop starting window.

[0062] Traditional air-drop systems often preset fixed drop points, but in actual flight, the speed and height of the UAV are easily disturbed by airflow fluctuations, and the environmental wind field also changes in real time, so fixed-point drop is prone to deviate from the target. This case designs the air-drop starting window as a dynamic space-time region by predicting discrete state points in the future, rather than relying on a single current state, which can adapt to changes such as UAV attitude fine-tuning and wind field disturbance in real time, ensuring that the drop decision always fits the actual flight scenario.

[0063] Traditional drop point estimation often ignores key factors such as air resistance and stratified wind field, resulting in large errors. By stratified modeling, the core influencing factors are covered: when calculating the falling time, the vertical speed at the instant of cargo release is taken into account (the attitude of the UAV determines) and air resistance (different cargo aerodynamic characteristics are adapted through dimensionless coefficient k); when calculating the horizontal net displacement, the inertial displacement of the cargo with the UAV and the drift displacement caused by the stratified wind field are considered; finally, the drop point is directly related to the net displacement through the state point position, and each step of calculation relies on physical principles, significantly reducing the estimation error.

[0064] For example, a large fixed-wing UAV performs a material delivery task, and the target point geographic information is pre-stored as (plane rectangular coordinate system), allowing a landing radius . The parameters of a future state point of the UAV are: position , horizontal speed (corresponding to x and y directions respectively), height , vertical speed at the instant of cargo release (negative for downward);

[0065] The dropped cargo is a streamlined material, and the dimensionless adjustment coefficient k = 0.9;

[0066] The acceleration of gravity ; the vertical wind field profile is obtained through navigation and sensing units, and the wind speed vector at different heights is:

[0067] The wind speed at a height of 500m is , the wind speed at a height of 300m is , the wind speed at a height of 100m is , and the wind speed at a height of 0m is .

[0068] The total falling time of the cargo in the air is calculated as :

[0069] ;

[0070] Substitute the parameters into the calculation: the numerator part is , first calculate the value inside the square root: , the square root result is about 94, so the numerator is ;

[0071] The denominator part is ; finally .

[0072] The net displacement of the cargo in the horizontal direction is calculated as :

[0073] According to the formula , which is calculated as follows:

[0074] Inertial displacement : generated by the initial velocity of the cargo when released under windless conditions, formula is ;

[0075] Substitute the parameters, the x-direction displacement is , and the y-direction displacement is , that is .

[0076] Wind field drift displacement : obtained by integrating the wind field during the falling process, as the wind field changes with height, the average wind speed is used for simplified calculation;

[0077] The cargo falls from 500 m to 0 m, with an average height of 250 m, and the corresponding wind speed is obtained by linear interpolation , formula is ;

[0078] Substitute the parameters, the x-direction displacement is , and the y-direction displacement is , that is .

[0079] Net displacement synthesis: .

[0080] Estimate the cargo drop point and filter:

[0081] The drop point formula is:

[0082] ;

[0083] Substitute the parameters to get .

[0084] Verify whether it meets the tolerance requirement: calculate the Euclidean distance between the drop point and the target point, formula is .

[0085] Substitute the numerical value to calculate, the part inside the square root is , and the square root result is about 27.57 m, because , that is, less than the allowed landing radius R, so the state point is included in the air drop starting window.

[0086] The dynamic flight path prediction adapts to the scenario where the wind field changes with height; multi-factor modeling (such as considering layered wind field and cargo aerodynamic characteristics) makes the drop point estimation error controlled within 27.57 m, which is much smaller than the allowed radius; the quantitative tolerance standard clearly defines the basis for state point screening; and the dimensionless coefficient k adapts to the aerodynamic characteristics of streamlined cargo, fully embodying its advantages in precision, adaptability, and reliability.

[0087] Through the technical path from dynamic trajectory prediction to multi-physical quantity modeling to quantitative tolerance screening, the problems of traditional air drop system static decision, low precision and poor adaptability are solved. The air drop starting window is upgraded from a static point to a dynamic controllable space-time region, which is the core technical support for realizing high-precision delivery of large fixed-wing unmanned aerial vehicles.

[0088] The predetermined attitude maneuver action is a pull-up and attitude stabilization action, specifically including: controlling the unmanned aerial vehicle to pitch upward around the horizontal axis, so that the attack angle increases by a predetermined value, to generate an additional upward lift component, which is used to partially offset the initial sinking caused by gravity and wind disturbance at the moment of cargo separation;

[0089] At the moment of cargo release from the unmanned aerial vehicle, an initial sinking will occur due to its own gravity (vertically downward) and environmental wind disturbance (such as downward airflow), resulting in an initial vertical speed of the cargo that is too large (the value is more negative when downward). The pull-up action generates an additional upward lift by increasing the attack angle, which can directly offset part of the downward force of gravity and wind disturbance, so that the vertical speed of the cargo at the moment of separation is closer to the ideal value, avoiding the error of falling ahead or deviating from the vertical direction due to too fast initial sinking.

[0090] In traditional air drop, the attitude of the unmanned aerial vehicle is easily affected by airflow fluctuations (such as small amplitude pitching and rolling), resulting in large differences in the initial attitude (such as the angle relative to the unmanned aerial vehicle) of cargo separation at different times of delivery, and high dispersion of the final landing point. The attitude stabilization action is a predetermined standardized operation that adjusts the attitude of the unmanned aerial vehicle to a unified delivery attitude reference regardless of small environmental disturbances, ensuring consistent initial attitude of cargo separation each time and significantly reducing precision fluctuations between different delivery tasks.

[0091] In complex wind field environments such as low altitude and mountainous areas, downward gusts (wind disturbance) can exacerbate the initial sinking of the cargo. The traditional delivery method without attitude maneuver is prone to serious deviation of the landing point due to wind disturbance. The additional lift is equivalent to adding wind resistance to the cargo separation. Even if a sudden downward wind disturbance occurs, the additional lift can partially offset the downward force of the wind disturbance, avoiding a sudden increase in the initial sinking speed and allowing the delivery precision to remain stable in complex wind fields.

[0092] For example, when a large fixed-wing unmanned aerial vehicle performs material delivery, the target point allows a landing radius ; the initial state of the unmanned aerial vehicle when it enters the air drop starting window is:

[0093] horizontal speed , height , initial attack angle , initial pitch angle ;

[0094] Cargo parameter: mass , streamlined k = 0.9

[0095] Environmental wind disturbance: sudden downward airflow, causing the cargo to be additionally subjected to downward wind disturbance force when it separates ;

[0096] Attitude maneuver parameter: predetermined angle of attack increment , i.e. target angle of attack , target pitch angle , steady attitude time ;

[0097] Auxiliary aerodynamic parameters: air density , reference area of UAV wing , lift line slope , lift coefficient increment corresponding to an increase of 1 radian in angle of attack

[0098] Calculate the additional lift generated by the pull-up action :

[0099] Lift coefficient is proportional to the angle of attack, and the angle of attack increases (in radians: , the lift coefficient increment is:

[0100] ;

[0101] Substitute the parameters to get: .

[0102] The formula for calculating the additional lift (positive upward) is:

[0103] ;

[0104] Where the flight speed of the UAV , substitute the parameters to get:

[0105] ;

[0106] Compare the initial sinking of the cargo with and without attitude maneuver:

[0107] At the moment of cargo separation, the downward resultant force is composed of the gravity component and the wind disturbance force, and the upward force is the support force of the UAV on the cargo (before separation). The additional lift is transmitted to the cargo through the UAV attitude, which offsets part of the downward resultant force:

[0108] Without attitude maneuver: total downward resultant force , substitute to get:

[0109] ;

[0110] Vertical acceleration before cargo separation ( negative sign means downward), after preparation time, initial vertical velocity ( sink velocity is large).

[0111] With attitude maneuver: additional lift upward, downward total force , substitute:

[0112] ;

[0113] Vertical acceleration

[0114] , after , initial vertical velocity ( sink velocity is significantly reduced).

[0115] Comparison of landing point error:

[0116] According to the descent time formula , respectively calculate the two cases and landing point:

[0117] No attitude maneuver: , substitute:

[0118] ;

[0119] Combined with horizontal net displacement calculation , landing point , but because of the large sink velocity, the actual vertical direction of additional offset resulting in horizontal landing point ahead, the final distance from the target point is about ( still in , but the accuracy is low).

[0120] With attitude maneuver , substitute:

[0121] ;

[0122] Horizontal net displacement due to change fine-tuning for , landing point , the distance from the target point is about , the accuracy is higher than no attitude maneuver.

[0123] Pull up and stable attitude action through additional lift the initial vertical velocity of the cargo from optimized for , directly offset about downward force, accounting for total downward forceAt the same time, the standardized attack angle increment and the attitude stabilization time ensure the consistency of attitude adjustment, and finally improve the landing point accuracy by nearly 20%, fully embodying the core value of stable initial conditions, wind disturbance resistance and accuracy improvement.

[0124] The calculation process of the delivery quality factor Q integrates the instantaneous delivery accuracy and flight attitude stability, specifically:

[0125] ;

[0126] Among them, is the estimated cargo drop point coordinates according to the current state, is the current pitch angle of the UAV, is the target pitch angle of the predetermined attitude maneuver, and are weight coefficients, is a small constant to prevent the denominator from being zero;

[0127] The control unit continuously calculates the Q value within the air drop starting window, and triggers the delivery when it reaches the maximum value;

[0128] The above process can avoid the risk of single-dimensional decision-making and ensure the reliability of delivery. Traditional air drop often falls into the decision-making error of "this or that": if only the "estimated landing point within the tolerance" (single accuracy indicator) triggers the delivery, it may deviate from the Too much, resulting in the actual drop point deviating from the estimated value due to the influence of body oscillation at the moment of cargo separation;

[0129] If only the attitude close to the target value (single attitude indicator) triggers the delivery, it may be delivered when the attitude is stable but the accuracy is poor, which is easy to exceed the tolerance due to sudden wind disturbance.

[0130] Forcing the coupling of the two dimensions, only when the accuracy is high (small denominator, large accuracy term value) and the attitude is stable (small deviation, large exponential term value), Q will be maximized, fundamentally avoiding task failure caused by single-dimensional misjudgment.

[0131] The air drop starting window only solves the problem of delivery, but the accuracy and attitude at different times within the window still differ; the setting of Q solves the problem of the best delivery, by continuously calculating the dynamic change of Q, locking the superposition time of the accuracy peak and the attitude stability peak, making the actual drop point deviation much smaller than the error of random delivery within the window.

[0132] The weight is flexible and adjustable, adapting to the task requirements of multiple scenes, and The design of the system has scene adaptability: for precise material delivery, the can be increased to prioritize accuracy, even if the attitude has a slight deviation, as long as the accuracy is high enough, Q can still meet the standard;

[0133] For complex terrain emergency delivery: can increase , priority to ensure attitude stability, to avoid attitude fluctuations caused by unmanned aerial vehicle out of control, while taking into account the basic accuracy. This flexibility does not need to modify the hardware, only by adjusting the parameters can cover multiple types of delivery task, reduce the system adaptation cost.

[0134] Continue the previous delivery task:

[0135] Target point , allow landing radius ;

[0136] Pull up and stable attitude target pitch angle ;

[0137] Set the calculation parameters of Q: (accuracy weight), (attitude weight), (prevent denominator to zero);

[0138] After the unmanned aerial vehicle enters the air drop start window, select three consecutive time , the key parameters of each time are as follows:

[0139] : estimated landing point , current pitch angle ;

[0140] : estimated landing point current pitch angle ;

[0141] : estimated landing point , current pitch angle .

[0142] Calculate the Q value of each time and select the optimal opportunity:

[0143] Calculate the accuracy term of each time :

[0144] The core of the accuracy term is "the Euclidean distance between the estimated landing point and the target point", the smaller the distance, the greater the accuracy term value:

[0145] Time:

[0146] First calculate the Euclidean distance:

[0147] :

[0148] Then substitute the accuracy term formula:

[0149] :

[0150] Time:

[0151] Euclidean distance calculation:

[0152] ;

[0153] Precision term formula substitution:

[0154] ;

[0155] Time:

[0156] Euclidean distance calculation:

[0157] ;

[0158] Precision term formula substitution:

[0159] ;

[0160] Calculate the attitude term at each time :

[0161] The core of the attitude term is the deviation of the current pitch angle and the target pitch angle. The smaller the deviation, the closer the exponential term is to 1, and the larger the value of the attitude term:

[0162] Time:

[0163] Pitch angle deviation: ;

[0164] Substitute the attitude term formula:

[0165] ;

[0166] Time:

[0167] Pitch angle deviation: ;

[0168] Substitute the attitude term formula:

[0169] ;

[0170] Time:

[0171] Pitch angle deviation: ;

[0172] Substitute the attitude term formula:

[0173] ;

[0174] Step 3: Calculate the Q value of each time and determine the optimal release time

[0175] According to , respectively:

[0176] : ;

[0177] : ;

[0178] : ;

[0179] Compare the Q values of the three moments, is the largest, so the control unit triggers the release device at moment - at this time, not only the estimated landing point is the smallest deviation from the target point, but also the deviation of the UAV attitude from the target pitch angle is the smallest, which is the optimal release time in the air drop starting window.

[0180] The navigation and sensing unit includes a wind field estimation module based on Doppler laser radar. The wind field estimation module estimates the wind speed and direction at different height layers between the UAV and the ground in real time by emitting laser beams to the lower front of the UAV and analyzing the echo signals, thereby obtaining vertical wind field profile data. The vertical wind field profile data is input into the air drop trajectory model to accurately calculate the wind field drift displacement ; ;

[0181] Traditional wind field sensing often ignores the change of wind speed with height. However, the present application obtains wind speed and direction at different height layers through Doppler laser radar to generate vertical wind field profile data, which can accurately match the wind speed corresponding to the real-time height during the falling process of goods, avoiding the calculation deviation of wind field drift displacement caused by the average wind speed at all heights.

[0182] Laser radar is a real-time active detection, which can dynamically update the vertical wind field profile. Even if there is a low-altitude gust or airflow shear during air drop, the wind speed data can be corrected in time to ensure that the calculation of wind field drift displacement always matches the actual situation. The vertical wind field profile data is directly used for wind field drift displacement calculation of the air drop trajectory model, which can reduce errors and directly reduce the estimated deviation of the landing point, providing key data support for high-precision release.

[0183] For example, the state of the UAV: a future state point , position , horizontal speed , height ;

[0184] Cargo parameter: streamlined material, dimensionless adjustment coefficient , calculated cargo descent time ;

[0185] Target point: ;

[0186] Wind field data: vertical wind profile detected by laser radar (updated in real time):

[0187] Height , wind speed vector (x negative for headwind, y positive for crosswind);

[0188] Height , wind speed vector ;

[0189] Height , wind speed vector ;

[0190] Height , wind speed vector ;

[0191] Cargo falling height change: simplified as linear change , t is the falling time.

[0192] Wind field drift displacement calculation based on vertical wind profile:

[0193] Wind field drift displacement is the integral of the wind speed vector during the cargo falling process, that is:

[0194] ;

[0195] Because linear change, combined with the stratified data of the vertical wind profile, the integral is segmented according to the height interval, and the average value of the interval end values is taken as the wind speed of each segment, and the time corresponds to the length of the height change:

[0196] Segment 1: :

[0197] Corresponding to the falling time , the average wind speed in this interval:

[0198] ;

[0199] The drift displacement of this segment is: ;

[0200] Segment 2: :

[0201] Corresponding to the falling time the average wind speed in the interval:

[0202] ;

[0203] the segment drift displacement:

[0204] ;

[0205] Segment 3: :

[0206] corresponding drop time the average wind speed in the interval:

[0207] );

[0208] the segment drift displacement:

[0209]

[0210] total wind field drift displacement:

[0211] ;

[0212] Drop point calculation and precision verification based on accurate wind field data:

[0213] According to the drop point formula, the cargo drop point is the sum of the UAV state point position, inertial displacement and wind field drift displacement, wherein the inertial displacement :

[0214] ;

[0215] ;

[0216] Substitute the data to calculate:

[0217] ;

[0218] Verify the distance between the drop point and the target point:

[0219] ;

[0220] The distance is much smaller than the allowed landing radius , and because the vertical wind field data is accurate, the drop point deviation is reduced compared to the case of ignoring the stratified wind field, directly reflecting the improvement of the precision of the present case.

[0221] The control unit comprises an adaptive controller for optimizing the parameters of the attitude maneuvering action in the complex terrain.

[0222] Adaptive controller takes environmental wind shear intensity and terrain roughness as input states, and pitch angle increment and stabilization time as output actions, whose strategy is trained by reinforcement learning based on drop success rate to maximize the expected value of drop quality factor Q.

[0223] This case takes environmental wind shear intensity and terrain roughness (maximum slope of target area) as input, and adjusts the key parameters of attitude maneuver in real time. The stronger the wind shear and the steeper the terrain, the larger the enhanced wind resistance, the longer stabilization of attitude, avoiding the failure of fixed parameters in complex environment; the controller strategy is trained by reinforcement learning driven by drop success rate, which can accumulate experience in multiple tasks and continuously correct and the matching rules of Q, so as to continuously improve the expected value of drop quality factor Q, rather than being limited to the initial setting.

[0224] Scenario and basic parameters continue:

[0225] Unmanned aerial vehicle core state: when entering the air drop start window, the initial pitch angle , horizontal speed , height ;

[0226] Target point: , landing radius ;

[0227] Q calculation parameters: ;

[0228] Adaptive controller rule: after reinforcement learning training, the output parameters and the simplified correlation formula (engineering commonly used linear approximation) of the input are:

[0229] Pitch angle increment: , unit: °, S is the wind shear intensity, unit ; T is the terrain roughness, unit °;

[0230] Stabilization time: (unit: s).

[0231] Adaptive control effect under two typical working conditions:

[0232] Working condition 1: small wind shear + gentle terrain (common environment):

[0233] Input parameters: wind shear intensity (wind speed changes gently with height), terrain roughness (target area is a plain);

[0234] Adaptive output calculation:

[0235] ;

[0236] ;

[0237] Attitude and Q value correlation: target pitch angle , because Adapt to steep terrain, actual pitch angle (deviation ); estimated landing point deviation , into Q formula:

[0238] ;

[0239] Case 2: strong wind shear + steep terrain (complex environment)

[0240] Input parameters: wind shear intensity (wind speed changes rapidly with height), terrain undulation (target area is mountainous);

[0241] Adaptive output calculation:

[0242] ;

[0243] ;

[0244] Attitude and Q value correlation: target pitch angle , because Extend the adaptation of steep terrain disturbance, actual pitch angle (deviation ); estimated landing point deviation , into Q formula:

[0245] ;

[0246] Comparison of the disadvantages of fixed parameters (highlight the benefits:

[0247] If the traditional fixed parameters :

[0248] In case 2, because , the actual pitch angle , the target attitude deviation ; estimated landing point deviation due to wind shear increased to , Q value at this time:

[0249] ;

[0250] Comparison can be seen: adaptive controller in case 2 (complex environment) , far higher than the fixed parameter , and the attitude deviation is always stable at ; if after multiple task reinforcement learning, the controller can further correct the formula coefficient (such as fine-tuning the coefficient of S from 0.8 to 0.85), so that is improved, which embodies the advantage of continuous optimization of strategy.

[0251] The delivery device adopts an active damping release mechanism. After receiving a trigger signal, the mechanism first releases the main lock, and then applies a controllable resistance force to the goods through a damping arm controlled by a linear servo motor. The resistance force The calculation formula is as follows:

[0252] ;

[0253] Where c is the adjustable damping coefficient, and are the velocity vectors of the goods and the unmanned aerial vehicle respectively. This process lasts for a very short time to smooth the speed separation process between the goods and the unmanned aerial vehicle;

[0254] By applying a controllable resistance force through a damping arm controlled by a linear servo motor, the speed difference between the goods and the unmanned aerial vehicle is forced to decrease slowly rather than instantaneously, avoiding structural damage or functional failure of high-value goods (such as precision equipment and fragile materials) due to impact. The instantaneous impact of rigid release will bring a reverse moment to the unmanned aerial vehicle, causing pitch / roll oscillation. The damping force disperses the impact energy by slowly offsetting the speed difference, significantly reducing the disturbance of the release process to the attitude of the unmanned aerial vehicle, ensuring the stability of subsequent flight. The adjustable damping coefficient c can be flexibly adjusted according to the mass and aerodynamic characteristics of the goods. For heavy goods (such as heavy materials), c can be increased to enhance the resistance effect. For light goods (such as light equipment), c can be reduced to avoid excessive resistance, without the need to replace the release mechanism to adapt to different delivery needs.

[0255] Scene and basic parameters continue:

[0256] Unmanned aerial vehicle state: when entering the delivery stage, the horizontal speed (x is the flight direction, y is the lateral direction), and the attitude is stable;

[0257] Goods parameters: mass (streamlined goods, as in the previous scenario), consistent with the speed of the unmanned aerial vehicle before release ;

[0258] Active damping parameters: after adaptation, the damping coefficient (according to the mass of the goods), and the damping action time (extremely short time, does not affect the air drop window);

[0259] Traditional rigid release: no damping force, speed drops from to instantly due to air resistance.

[0260] Active damping release: speed drops from to

[0261] slowly due to damping force. The damping force is calculated in real time:

[0262] The damping force is proportional to the speed difference between the cargo and the UAV, and is in the opposite direction:

[0263] ;

[0264] At the beginning of the release, the cargo is affected by air resistance, and its speed drops to , at which point the speed difference is:

[0265] ;

[0266] Substitute the formula to calculate the damping force:

[0267] ;

[0268] The smoothing effect of the speed separation process:

[0269] Comparison of active damping and traditional rigid release of cargo speed change (take x direction as an example):

[0270] Traditional rigid release: no damping force, speed drops from to instantly (at ), the speed change rate is:

[0271] ;

[0272] (negative sign indicates deceleration, and sharp deceleration produces a large impact);

[0273] Active damping release: under the action of damping force, the speed slowly drops, to at , to , the average speed change rate is:

[0274] ;

[0275] (deceleration amplitude is only 1 / 4 of the traditional one, impact energy is greatly dispersed);

[0276] Risk assessment of cargo impact damage:

[0277] Impact acceleration is proportional to the impact force on the cargo :

[0278] Traditional rigid release: , far beyond the impact threshold of precision goods (usually <2000N), easy to damage;

[0279] Active damping release: , below the impact threshold, the goods are intact.

[0280] Verification of the impact on the attitude of the UAV:

[0281] In traditional rigid release, the sudden change in speed of the goods will exert a reverse reaction force on the UAV (Newton's third law), causing instantaneous fluctuations in the x-direction speed of the UAV , causing pitch angle disturbance ;

[0282] In active damping release, the damping force acts slowly, and the peak value of the reaction force is only 1 / 4 of the traditional one , traditional reaction force ≈2400N), the x-direction speed of the UAV fluctuates , the pitch angle disturbance , which is far below the Q value requirement for attitude stability (deviation <0.5°), and does not affect the judgment of the optimal release timing.

[0283] During the operation of the active damping release mechanism, the control unit records the body angular velocity disturbance caused by the release , and after the damping release is completed, a compensation torque proportional to is introduced into the attitude control loop to quickly suppress the body oscillation caused by the release impact;

[0284] After active damping release, even if the impact has been reduced, the body angular velocity disturbance (such as a slight rotation in the pitch / roll direction) caused by the reaction force of the goods separation will still remain. The traditional system needs to rely on the body's own stability to slowly decay; while this mechanism actively counteracts the disturbance by introducing a compensation torque proportional to , the oscillation decay time is shortened to within 0.5 seconds, avoiding the disturbance from affecting subsequent flight; the residual oscillation will cause the pitch angle of the UAV to deviate from the target attitude , and thus affect the attitude term of the Q value; the compensation torque can quickly pull back to the target range, ensuring that the attitude deviation is always minimal when calculating the Q value, and not interfering with the optimal release timing judgment;

[0285] The compensation torque is achieved by recording , calculating the proportional torque, and then introducing it into the existing attitude control loop. No additional sensors or actuators are needed, and only control logic optimization is required to improve stability, making it suitable for existing systems and low cost.

[0286] UAV state: horizontal speed before release , attitude control loop gain is normal;

[0287] Release disturbance premise: execute active damping release After that, there is still a disturbance in the pitch direction angular velocity (Clockwise is positive, that is, the body has a slight tendency to lower);

[0288] Compensation parameters: after system calibration, the proportional coefficient of the compensation torque to (The proportion of the moment of inertia of the UAV determines it) and the compensation torque only acts on the angular velocity approaching 0;

[0289] Traditional comparison: when there is no compensation, the body relies on its own damping (damping coefficient ) to slowly decay the disturbance.

[0290] Compensation torque calculation and vibration suppression effect verification:

[0291] Compensation torque calculation:

[0292] Compensation torque is proportional to the residual angular velocity disturbance , and the direction is opposite (to offset the disturbance), the formula is:

[0293] ;

[0294] Substitute :

[0295] ;

[0296] (The negative sign indicates that the torque direction is counterclockwise, opposite to the clockwise , directly offsetting the tendency to lower);

[0297] Angular velocity decay process comparison (pitch direction):

[0298] No compensation (traditional method): the body angular velocity decays over time following the law of a first-order system, the formula is:

[0299] ;

[0300] Where the moment of inertia of the UAV in the pitch direction , substituting the decay coefficient .

[0301] When decays to 0.05 (the disturbance can be ignored), the solution time is:

[0302] ; ​

[0303] (Slow decay, long time residual disturbance)

[0304] After the compensation torque is introduced, the angular velocity decay formula becomes (compensation torque equivalent to increase damping):

[0305] ;

[0306] Substitute , the decay coefficient .

[0307] Similarly, when the decay is , the time is:

[0308] ;

[0309] If further required to decay to , with compensation , without compensation , the vibration suppression efficiency is improved;

[0310] Q value influence verification

[0311] Residual angular velocity disturbance will cause the pitch angle deviate from the target value , the attitude deviation and integration related:

[0312] Without compensation, the attitude deviation accumulates to in 0-3 seconds, and the attitude term is:

[0313] ;

[0314] With compensation, the attitude deviation is only in 0-3 seconds, and the attitude term is:

[0315] ;

[0316] Assuming the accuracy term is 11.98 (same as the previous example), then:

[0317] No compensation ;

[0318] With compensation .

[0319] As can be seen, the compensation mechanism ensures that the Q value is always in the optimal interval, avoiding the sudden drop of Q value due to residual oscillation, and does not affect the launch timing decision.

[0320] In addition, the terms "first", "second", etc. are used only for the purpose of description, and should not be understood as indicating or implying relative importance or a specific number of the technical features indicated. Therefore, the features defined as "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, for example, two, three, etc., unless otherwise specifically limited.

[0321] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms is not necessarily directed to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. Furthermore, the person skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples, without contradiction.

[0322] Although the embodiments of the present application have been shown and described above, it is understood that the above-described embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above-described embodiments within the scope of the present application.

Claims

1. A high-precision delivery system for large fixed-wing drones, characterized in that, The unmanned aerial vehicle platform, the delivery device, the navigation and sensing unit, and the control unit are included. The control unit is used to execute the dynamic air-drop window and flight attitude cooperative control method, which includes the following steps: Based on the real-time state information of the unmanned aerial vehicle, the environmental wind field information, and the pre-stored geographical information of the target point obtained by the navigation and sensing unit, the dynamic air-drop starting window is calculated in real time through a pre-set air-drop trajectory model. After entering the air-drop starting window, the control unit synchronously executes the following operations: controlling the unmanned aerial vehicle platform to execute a predetermined attitude maneuvering action to stabilize the initial conditions when the cargo is separated; At the same time, during the execution of the attitude maneuvering action, the optimal delivery time is continuously evaluated according to the real-time calculated delivery quality factor Q, and the delivery device is triggered to deliver when the delivery quality factor Q meets the predetermined condition. The air-drop starting window is determined by the following process: First, based on the UAV dynamics and a pre-set control law, a UAV trajectory for a future time period from the current time is predicted, the trajectory being composed of a number of discrete future state points wherein is the position, is the velocity, is the time, is the altitude; Secondly, for each future state point , it is regarded as a potential drop point, and the cargo drop point is estimated by using the air-drop trajectory model The specific process of estimating the cargo drop point by using the air-drop trajectory model includes: Step a: Calculate total descent time of the cargo in the air ; Step b: Calculate the net displacement of the cargo in the horizontal direction ; Afterwards, the total descent time of the cargo in the air is calculated the net displacement of the cargo in the horizontal direction the estimated landing point of the cargo ; Finally, all the state points satisfying are screened out where R is the allowed landing radius, and the space-time region formed by these state points is the airdrop initiation window. The predetermined attitude maneuvering action is a pull-up and stabilization action, which specifically includes: controlling the unmanned aerial vehicle to pitch upward around the horizontal axis to increase the attack angle by a predetermined value, so as to generate an additional upward lift component, which is used to partially offset the initial sinking caused by gravity and wind disturbance at the moment of cargo separation; The specific acquisition process of the delivery quality factor Q is as follows: ； wherein, is the estimated drop point coordinate of the cargo according to the current state, is the current pitch angle of the UAV, is the target drop point geographic coordinate of the delivery task, is the target pitch angle of the predetermined attitude maneuver action, and is the weight coefficient, is a small constant to prevent the denominator from being zero; The control unit continuously calculates the Q value within the air-drop starting window, and triggers the delivery when the Q value reaches the maximum value.

2. The large fixed-wing UAV high-precision delivery system according to claim 1, characterized in that: The navigation and sensing unit includes a wind field estimation module based on a Doppler laser radar, which estimates the wind speed and direction at different height layers between the unmanned aerial vehicle and the ground in real time by emitting laser beams downward in front of the unmanned aerial vehicle and analyzing the echo signals, thereby obtaining vertical wind field profile data. This vertical wind profile data is input into the air-drop trajectory model for calculating the wind drift displacement in the integrand function .

3. The large fixed-wing UAV high-precision delivery system according to claim 1, characterized in that: The control unit includes an adaptive controller for optimizing the parameters of the attitude maneuvering action in the complex terrain. An adaptive controller takes environmental wind shear intensity and terrain roughness as input states, and pitch angle increment and stabilization time as output actions, whose policy is trained by reinforcement learning based on drop success rate to maximize the expected value of drop quality factor Q.

4. The large fixed-wing UAV high-precision delivery system according to claim 1, characterized in that: The delivery device adopts an active damping release mechanism, which, after receiving the trigger signal, first releases the main lock, and then applies a controllable resistance force to the cargo through a damping arm controlled by a linear servo motor.

5. The large fixed-wing UAV high-precision delivery system according to claim 1, characterized in that: During the operation of the active damping release mechanism, the control unit records the disturbance of the body angular velocity caused by the release After the end of the damping release, a compensation torque proportional to is introduced into the attitude control loop for rapid damping of the body oscillations caused by the release impact.

Images