Multi-rotor wind field estimation method

Abstract

The invention discloses a multi-rotor wind field estimation method, and relates to the technical field of unmanned aerial vehicles. S1, parameter definition and data acquisition; s2, real-time quality estimation of complete flight of the unmanned aerial vehicle; s3, establishing a kinetic equation; s3-1, calculating a horizontal thrust component; s3-2, calculating an airspeed direction angle delta a; s3-3, air resistance D is calculated; s3-4, the vacuum speed Va is calculated; s3-5, calculating a wind speed component; s4, estimating a basic wind field; s5, carrying out takeoff stage attitude error compensation estimation; and S6, carrying out high-precision estimation with resistance coefficient CD calibration. The multi-rotor wind field estimation method mainly depends on an IMU (Inertial Measurement Unit), a GPS (Global Positioning System), a motor controller and the like of an existing sensor of the unmanned aerial vehicle, the algorithm calculation steps are clear, the method is suitable for being realized by an embedded system, and an estimation result has a relatively good reference value.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicles (UAVs), and in particular to a multi-rotor wind field estimation method. Background Art

[0002] When multi-rotor drones perform tasks such as aerial photography, inspections, and logistics, their flight performance, track tracking accuracy, and flight safety are significantly affected by the ambient wind field. Accurate wind field information is crucial for autonomous control, navigation, and mission planning of drones.

[0003] Existing wind field measurement methods typically rely on installing specialized wind speed sensors (such as pitot tubes and anemometers) on drones. These sensors increase the weight, power consumption, and cost of the drone, and may affect its aerodynamic characteristics. Furthermore, some methods based on complex airflow simulations are computationally intensive, making real-time estimation difficult to achieve onboard.

[0004] Therefore, developing a method that does not rely on or relies less on dedicated wind speed sensors and can use the drone's own status information to perform real-time and accurate wind field estimation has important practical application value. Summary of the Invention

[0005] The purpose of the present invention is to provide a multi-rotor wind field estimation method to solve the problem that the wind field measurement method proposed in the above background technology mainly relies on wind speed sensors, which increases the weight and cost of the drone and makes it difficult to achieve onboard real-time estimation.

[0006] To achieve the above-mentioned object, the present invention provides the following technical solutions: a multi-rotor wind field estimation method, comprising S1, parameter definition and data acquisition;

[0007] S2,real-time quality estimation of the complete UAV flight;

[0008] S3. Establish the kinetic equation;

[0009] S3-1. Calculate the horizontal thrust component;

[0010] S3-2, calculate the airspeed direction angle (Δ a );

[0011] S3-3. Calculate air resistance (D);

[0012] S3-4. Calculate the true airspeed (V a );

[0013] S3-5. Calculate wind speed components;

[0014] S4. Basic wind field estimation;

[0015] S5, with attitude error compensation estimation during takeoff phase;

[0016] S6, with resistance coefficient (C D ) calibration with high accuracy.

[0017] Furthermore, the real-time state parameters in S1 include: aircraft mass m, gravitational acceleration g, total engine thrust T, angle β between the pulling force vector and the vertical direction (pulling force inclination angle), angle Δ between the pulling force projection on the horizontal plane and the geographic north direction, acceleration components of the aircraft on the x, y, and z axes in the aircraft body coordinate system x , y , z (or related items calculated from the speed change), the ground speed vector of the aircraft (including the north component of the ground speed V gN and the east component V gE ), provided by GPS or Inertial Navigation System (INS).

[0018] Furthermore, the complete flight of the UAV in S2 includes ground preparation, take-off and climb, enroute flight, descent and landing.

[0019] Furthermore, the real-time mass estimation in S2 involves estimating the total thrust T during the drone's constant-speed climb phase using the motor controller output, and then inferring the vehicle mass m from the total thrust. By processing the "rotor total thrust T," model uncertainty is avoided. The total thrust T of a multirotor is difficult to accurately model, as it depends on multiple factors, such as motor efficiency, battery voltage, air density, and propeller aging. Directly inferring the thrust from the motor PWM can introduce significant errors. Rather than directly using the motor model to calculate the thrust T, the algorithm utilizes the vertical dynamics equation (T = (mg - maz) / cosβ) to indirectly and in real time infer the total thrust T at the current moment. This inversely calculated T is a "result quantity" that already incorporates all factors affecting the thrust and is more accurate than inferring it from the "cause" (motor speed). In this way, the algorithm successfully avoids the complex and imprecise modeling of the rotor power system, significantly improving its robustness and practical accuracy. This is a significant advantage over many other methods that rely on precise motor models.

[0020] Furthermore, the dynamic equations established in S3 include the vertical dynamic equation, the horizontal X-direction dynamic equation (aircraft axis, along the horizontal projection direction of the thrust), and the horizontal Y-direction dynamic equation (aircraft axis, perpendicular to the horizontal projection direction of the thrust); direction first: by dividing the two horizontal dynamic equations, the unknown quantity "drag D" is cleverly eliminated, thereby directly and analytically (rather than iteratively) solving the direction angle Δ of "airspeed" a, which is extremely efficient in calculation and avoids the problem that the iterative algorithm may not converge; Post-size: After accurately obtaining the airspeed direction, substitute it back into any horizontal dynamic equation to calculate the size of the "drag D", and finally calculate the size of the "airspeed" V through the drag formula a The decoupling idea of ​​"solving the direction first and then determining the size" decomposes a complex multi-variable coupling problem into two independent and directly solvable simple problems. This is the core ingenuity and creativity of this method in mathematical processing.

[0021] Furthermore, the airspeed direction angle Δ in S3-2 and S3-3 a It is the angle between the airspeed direction (the direction of the aircraft's velocity vector relative to the air) and the geographic north direction. The air resistance D is the air resistance experienced by the aircraft.

[0022] Furthermore, the S4 basic wind field estimation system uses an IMU to obtain attitude angles (roll, pitch, and yaw), an accelerometer to obtain three-axis acceleration, and GPS to obtain ground speed and geographic heading. The calculation module calculates the tension tilt angle and the angle Δ between the tension horizontal projection and the north direction based on the attitude and heading angles. The wind field estimation core module executes and outputs the total wind speed W and wind direction. This creates a complete, low-cost, and practical systematic solution, forming a closed-loop technical solution:

[0023] Clear input: Only uses standard drone sensor data such as IMU, GPS, and attitude calculation module.

[0024] Clear process: Follow the clear logical steps of "acquire data → calculate the pull direction and inclination → decouple and calculate the airspeed vector → apply wind triangle to calculate wind speed".

[0025] Strong practicality: It is a pure software algorithm that can be seamlessly integrated into the existing flight control system as a firmware module without any hardware changes. It has extremely low implementation cost and extremely high commercial promotion value.

[0026] Furthermore, the estimation with attitude error compensation in the takeoff phase in S5 includes adding attitude dead zone compensation in the hovering / take-off and landing phases. The estimation of attitude error compensation in the takeoff phase does not belong to estimation, but is a logical restriction. In the absence of horizontal speed and a small angle attitude of the aircraft, the output result of the wind estimation algorithm is directly made 0 (equivalent to the algorithm output being 0 within the dead zone, and the algorithm outputting the estimation result normally outside the dead zone).

[0027] Furthermore, the S6 has a resistance coefficient C D The high-precision estimation of calibration includes optimizing the accuracy of the aerodynamic database through real-time flight data and optimizing the accuracy of the aerodynamic database through CFD simulation.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] 1. Core Innovation: The most critical innovation is the decoupled analytical calculation method for the direction and magnitude of "airspeed". Existing technologies usually require iterative solutions to complex nonlinear equations or rely on methods such as Kalman filtering for estimation.

[0030] Direction first: By dividing the two horizontal dynamic equations, the unknown quantity "drag D" is cleverly eliminated, thereby directly and analytically (rather than iteratively) solving the direction angle Δ of "airspeed" a , which is computationally extremely efficient and avoids the problem that iterative algorithms may not converge.

[0031] Post-scaling: After accurately obtaining the airspeed direction, substitute it back into any horizontal dynamics equation to inversely calculate the magnitude of the "drag D", and finally calculate the magnitude of the "airspeed V" through the drag formula a The decoupling idea of ​​"solving the direction first and then determining the size" decomposes a complex multi-variable coupling problem into two independent and directly solvable simple problems. This is the core ingenuity and creativity of this method in mathematical processing.

[0032] 2. By addressing the "total rotor thrust T," model uncertainty is avoided. The total thrust T of a multirotor is difficult to accurately model. It depends on multiple factors, including motor efficiency, battery voltage, air density, and propeller aging. Directly inferring the thrust from the motor PWM can introduce significant errors. Instead of using the motor model to calculate the thrust T directly, the algorithm uses the vertical dynamics equation (T = (mg - maz) / cosβ) to indirectly and in real time infer the total thrust T at the current moment. This inversely calculated T is a "result quantity" that incorporates all factors affecting the thrust and is more accurate than inferring it from the "cause" (motor speed). This approach successfully avoids the complex and imprecise modeling of the rotor power system, significantly improving its robustness and accuracy. This is a significant advantage over many other methods that rely on precise motor models.

[0033] 3. A complete, low-cost, and practical system solution has been built, which constitutes a closed-loop technical solution:

[0034] Clear input: Only uses standard drone sensor data such as IMU, GPS, and attitude calculation module.

[0035] Clear process: Follow the clear logical steps of "acquire data → calculate the pull direction and inclination → decouple and calculate the airspeed vector → apply wind triangle to calculate wind speed".

[0036] Strong practicality: It is a pure software algorithm that can be seamlessly integrated into the existing flight control system as a firmware module without any hardware changes. It has extremely low implementation cost and extremely high commercial promotion value.

[0037] The present invention mainly relies on the UAV's own existing sensors (IMU, GPS, motor controller, etc.), reducing additional hardware costs, weight and power consumption. The algorithm calculation steps are clear, suitable for embedded system implementation, and can perform real-time wind field estimation. Especially in level flight or when the wind speed is high, the estimation results have good reference value and can provide a basis for flight safety decisions (such as determining whether to continue flight or whether to exceed the wind speed limit). By introducing improvements such as real-time mass estimation, more accurate calibration and modeling, the estimation accuracy and robustness can be further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 Schematic diagram of the workflow of the present invention. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0040] Example 1: Figure 1 The technical solution shown establishes the dynamic projection equation of the UAV in the body coordinate system in the plane perpendicular to the total thrust vector of the UAV (i.e., the "tangent normal direction"), and assumes that the wind mainly acts in the horizontal plane. The horizontal wind speed and direction are estimated by inversely calculating the aerodynamic force acting on the aircraft, including S1, parameter definition and data acquisition.

[0041] The real-time status parameters in S1 include: aircraft mass m, gravitational acceleration g, total engine thrust T, angle β between the pulling force vector and the vertical direction (pulling force inclination angle), angle Δ between the pulling force projection on the horizontal plane and the geographic north direction, and acceleration components of the aircraft's x, y, and z axes in the aircraft's body coordinate system. x , y , z (or related items calculated from the speed change), the ground speed vector of the aircraft (including the north component of the ground speed V gN and the east component V gE ), provided by GPS or Inertial Navigation System (INS).

[0042] By addressing the "total rotor thrust T," model uncertainty is avoided. The total thrust T of a multirotor is difficult to accurately model. It depends on multiple factors, including motor efficiency, battery voltage, air density, and propeller aging. Directly inferring the thrust from the motor PWM results in significant errors. Instead of using the motor model to calculate the thrust T directly, the algorithm uses the vertical dynamics equation (T = (mg - maz) / cosβ) to indirectly and in real time infer the total thrust T at the current moment. This inversely calculated T is a "result quantity" that incorporates all factors affecting the thrust and is more accurate than inferring it from the "cause" (motor speed). This approach successfully avoids the complex and imprecise modeling of the rotor power system, significantly improving its robustness and accuracy. This is a significant advantage over many other methods that rely on precise motor models.

[0043] S2. Real-time quality estimation of a complete UAV flight.

[0044] The complete flight of the UAV in S2 includes ground preparation, takeoff and climb, route flight, descent and landing. Real-time mass estimation exists in the takeoff and climb phase. At this time, the UAV usually goes through accelerated climb (takeoff), constant speed climb and decelerated climb (about to enter the route). During the constant speed climb, the aircraft is usually stable in attitude and speed. By counting the total pulling force of the aircraft within a certain period of time in this phase, and then using The current mass of the aircraft can be estimated. It should be noted that this assumes that the aircraft's attitude and speed are stable. In actual flight environments, there are often various external disturbances that do not meet this assumption, and attitude and tension compensation must be performed based on actual conditions.

[0045] S3. Establish the kinetic equation.

[0046] Based on the following assumptions: the wind speed is mainly in the horizontal plane, and the aerodynamic lift generated by the aircraft is ignored (or its effect is considered to be small under certain conditions).

[0047] The dynamic equation is established in the plane perpendicular to the total thrust direction and projected onto the body axis:

[0048] 1. Vertical dynamic equations.

[0049]

[0050] 2. Horizontal X-direction dynamic equation (body axis, along the horizontal projection direction of thrust).

[0051]

[0052] 3. Horizontal Y-direction dynamic equation (body axis, perpendicular to the horizontal projection direction of thrust).

[0053]

[0054] The airspeed direction angle Δa in S3-2 and S3-3 is the angle between the airspeed direction (the direction of the aircraft's velocity vector relative to the air) and the geographic north direction, and the air resistance D is the air resistance experienced by the aircraft.

[0055] S3-1. Calculate the horizontal thrust component.

[0056] From the vertical dynamic equation we can get

[0057] The horizontal thrust component is Expressed as:

[0058] S3-2. Calculate the airspeed direction angle (Δa).

[0059] Substitute the expression obtained in the above steps into the horizontal dynamics equation to calculate the tangent of the airspeed direction angle (Δa): , (Δa) is calculated by the inverse tangent function.

[0060] S3-3. Calculate the air resistance (D).

[0061] Using the data obtained in the above steps and the horizontal dynamic equation (such as the equation in the X direction), calculate the air resistance D:

[0062]

[0063] S3-4. Calculate the true airspeed (Va).

[0064] air resistance The physical equation is ,in is the air density, is the reference area, is the resistance coefficient, in the present invention, Obtained from a pre-built aerodynamic database that stores various tension and tilt angles The corresponding value.

[0065] According to the calculated And the table , calculate the true airspeed :

[0066]

[0067] S3-5. Calculate the wind speed components.

[0068] According to the wind triangle relationship ( ), calculate the wind speed vector:

[0069] Northward component of airspeed .

[0070] Northward component of airspeed .

[0071] Northerly component of wind speed .

[0072] Northerly component of wind speed .

[0073] Total wind speed , the wind direction angle can be obtained from Calculated (in meteorology, wind direction is usually defined as the direction from which the wind is blowing, but the actual output of the algorithm is the direction from which the wind is blowing, so a conversion is needed).

[0074] Example 2: Figure 1 In the technical solution shown, real-time mass estimation is to estimate the total thrust through the motor controller output during the uniform climb phase of the UAV, and then calculate the aircraft mass based on the total thrust. To avoid the impact of excessive errors in a single or several inverse calculation results, the inverse calculation results over a continuous period of time can be averaged to estimate the mass.

[0075] S4. Basic wind field estimation;

[0076] i. The system obtains attitude angles (roll, pitch, and yaw) through the IMU, three-axis acceleration through the accelerometer, and ground speed and geographic heading through the GPS.

[0077] ii. The calculation module calculates the tension tilt angle based on the attitude angle and heading angle Angle between the horizontal projection of the tension and the north direction .

[0078] iii. iii~viii in the technical solution for implementing the core module of wind farm estimation:

[0079] 1. Calculation .

[0080] 2. Calculation , and then get .

[0081] 3. Calculation .

[0082] 4. From the pre-stored pneumatic database, according to the current Look up the table to get Assuming air density and reference area Known.

[0083] 5. Calculation .

[0084] 6. Calculate wind speed components and .

[0085] 7. Output total wind speed and wind direction.

[0086] S5, with attitude error compensation estimation during takeoff phase;

[0087] Based on real-time mass estimation and S4 basic wind field estimation, attitude dead zone compensation during hovering / take-off and landing phases is added.

[0088] When the drone is under the above working conditions, the tilt in a certain attitude direction is small. This tilt may be caused by sensor zero position error, installation bias, etc. The wind speed calculated by this small angle value has low credibility. It can be considered that the wind speed at this time is 0, which increases the stability of the algorithm output under this working condition; the attitude error compensation estimation during the takeoff phase is not an estimation, but a logical limitation. In the absence of horizontal speed and the small-angle attitude of the aircraft, the output result of the wind estimation algorithm is directly set to 0 (equivalent to the algorithm output being 0 within the dead zone, and the algorithm outputting the estimation result normally outside the dead zone).

[0089] S6, with resistance coefficient (C D ) High-precision estimation of calibration;

[0090] It can be calibrated in the following ways To improve the accuracy of the algorithm's estimation results:

[0091] Before leaving the factory or during regular maintenance, the drone is temporarily equipped with a high-precision airspeed meter to optimize the accuracy of the aerodynamic database through real-time flight data and CFD simulation.

[0092] The present invention mainly relies on the UAV's own existing sensors (IMU, GPS, motor controller, etc.), reducing additional hardware costs, weight and power consumption. The algorithm calculation steps are clear, suitable for embedded system implementation, and can perform real-time wind field estimation. Especially in level flight or when the wind speed is high, the estimation results have good reference value and can provide a basis for flight safety decisions (such as determining whether to continue flight or whether to exceed the wind speed limit). By introducing improvements such as real-time mass estimation, more accurate calibration and modeling, the estimation accuracy and robustness can be further improved.

[0093] Core innovation: The most critical innovation is the decoupled analytical calculation method for the direction and magnitude of "airspeed". Existing technologies usually require iterative solutions to complex nonlinear equations or rely on methods such as Kalman filtering for estimation.

[0094] Direction first: By dividing the two horizontal dynamic equations, the unknown quantity "drag D" is cleverly eliminated, thereby directly and analytically (rather than iteratively) solving the direction angle Δ of "airspeed" a , which is computationally extremely efficient and avoids the problem that iterative algorithms may not converge.

[0095] Post-scaling: After accurately obtaining the airspeed direction, substitute it back into any horizontal dynamics equation to inversely calculate the magnitude of the "drag D", and finally calculate the magnitude of the "airspeed V" through the drag formula a The decoupling idea of ​​"solving the direction first and then determining the size" decomposes a complex multi-variable coupling problem into two independent and directly solvable simple problems. This is the core ingenuity and creativity of this method in mathematical processing.

[0096] A complete, low-cost and practical system solution has been built, which constitutes a closed-loop technical solution:

[0097] Clear input: Only uses standard drone sensor data such as IMU, GPS, and attitude calculation module.

[0098] Clear process: Follow the clear logical steps of "acquire data → calculate the pull direction and inclination → decouple and calculate the airspeed vector → apply wind triangle to calculate wind speed".

[0099] Strong practicality: It is a pure software algorithm that can be seamlessly integrated into the existing flight control system as a firmware module without any hardware changes. It has extremely low implementation cost and extremely high commercial promotion value.

[0100] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A multi-rotor wind field estimation method, characterized by: The following steps are included: S1, parameter definition and data acquisition; S2,real-time quality estimation of the complete UAV flight; S3. Establish the kinetic equation; S3-1. Calculate the horizontal thrust component; S3-2. Calculate the airspeed direction angle Δ a ; S3-3. Calculate the air resistance D; S3-4. Calculate the true airspeed V a ; S3-5. Calculate wind speed components; S4. Basic wind field estimation; S5, with attitude error compensation estimation during takeoff phase; S6, with drag coefficient C D High-precision estimation of calibration.

2. A multi-rotor wind field estimation method according to claim 1, characterized in that: The real-time status parameters in S1 include: aircraft mass m, gravitational acceleration g, total engine thrust T, angle β between the pulling force vector and the vertical direction, angle Δ between the pulling force projection on the horizontal plane and the geographic north direction, and acceleration components of the aircraft's x, y, and z axes in the aircraft's body coordinate system. x , y , z or related terms calculated from velocity changes; The ground speed vector of the aircraft includes the north component of the ground speed V gN and the east component V gE , provided by GPS or inertial navigation system INS.

3. The multi-rotor wind field estimation method according to claim 2, characterized in that: The complete UAV flight in S2 includes ground preparation, take-off and climb, enroute flight, descent and landing.

4. The multi-rotor wind field estimation method according to claim 3, characterized in that: The real-time mass estimation in S2 includes estimating the total thrust T through the output of the motor controller during the uniform climbing phase of the UAV, and inversely calculating the total thrust to obtain the aircraft mass m.

5. The multi-rotor wind field estimation method according to claim 4, characterized in that: The dynamic equations established in S3 include the vertical dynamic equations, Dynamic equation in horizontal X direction: body axis, along the direction of horizontal projection of thrust and dynamic equation in horizontal Y direction: body axis, perpendicular to the direction of horizontal projection of thrust; Specifically, the wind speed is mainly in the horizontal plane, ignoring the aerodynamic lift generated by the body, and the dynamic equation projected onto the body axis in the plane perpendicular to the total thrust direction is: (1) Vertical dynamic equation ； (2) Horizontal X-direction dynamic equation: body axis, along the horizontal projection direction of thrust ； (3) Horizontal Y-direction dynamic equation: body axis, perpendicular to the horizontal projection direction of thrust 。 6. A multi-rotor wind field estimation method according to claim 5, characterized in that: The airspeed direction angle Δ in S3-2 and S3-3 a The airspeed direction is the angle between the velocity vector direction of the aircraft relative to the air and the geographic north direction. The air resistance D is the air resistance experienced by the aircraft.

7. The multi-rotor wind field estimation method according to claim 6, characterized in that: The S4 basic wind field estimation includes the system obtaining attitude angles (roll, pitch, and yaw) through the IMU, obtaining three-axis acceleration through the accelerometer, and obtaining ground speed and geographic heading through the GPS. The calculation module calculates the tension tilt angle and the angle Δ between the horizontal projection of the tension and the north direction based on the attitude angle and heading angle. The wind field estimation core module executes and outputs the total wind speed W and wind direction.

8. The multi-rotor wind field estimation method according to claim 7, characterized in that: The estimation with attitude error compensation in the take-off phase in S5 includes adding attitude dead zone compensation in the hovering / take-off and landing phases.

9. The multi-rotor wind field estimation method according to claim 8, characterized in that: The S6 has a drag coefficient C D The high-precision estimation of calibration includes optimizing the accuracy of the aerodynamic database through real-time flight data and optimizing the accuracy of the aerodynamic database through CFD simulation.

Images