Planning and control method for landing of unmanned aerial vehicle in complex ground environment

Abstract

The invention belongs to the technical field of unmanned aerial vehicles, and particularly relates to a planning and control method for landing of an unmanned aerial vehicle in a complex ground environment, fuselage information of the unmanned aerial vehicle is obtained through different types of sensors, then the ground condition is judged through a depth camera, and when the ground is a flat ground or an inclined plane with the inclination angle of 0-5 degrees, the unmanned aerial vehicle is landed. A plane landing strategy is adopted, namely, the lift force of propellers is reduced at a constant speed, the speed and acceleration during descending are kept without calling a special strategy, when the ground is an inclined plane with the inclination angle larger than 5 degrees, a PID-MPC model prediction control landing strategy is selected, and an accelerator for controlling an unmanned aerial vehicle motor is output, so that the unmanned aerial vehicle stably lands on the complex ground. The PID-MPC control algorithm is introduced in the landing control process, the stability of the control process is enhanced, and corresponding landing strategies are given according to different ground conditions; the descending strategy can be adjusted according to needs, the descending strategy is dynamically adjusted according to different ground conditions, and meanwhile PID parameters are changed; and the flexibility is high.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned aerial vehicles, and in particular relates to a method for planning and controlling the landing of an unmanned aerial vehicle in a complex ground environment. Background Art

[0002] With the continuous expansion of drone application areas, drone patrols have begun to show their prowess in the fields of military, public security, border defense, fire protection, environmental protection, urban management, etc. During drone patrols, drones can be equipped with various sensors and equipment, such as high-definition cameras, infrared thermal imagers, optical / radar remote sensing equipment, meteorological instruments, etc., to obtain images, videos, data and environmental information of the target area in real time. The safe flight and stable landing of drones are key factors to ensure the smooth completion of flight missions, so drone landing technology has also received extensive attention and research.

[0003] In the prior art, for example, a Chinese patent with patent number CN111506099A discloses an intelligent control system and method for the height of a UAV, which includes the following steps: Step 1, coupling the expected height and expected speed of the UAV with the expected error to obtain the corrected expected height and expected speed; Step 2, determining the total energy and energy balance of the UAV according to the corrected expected height and expected speed of the UAV, as well as the current height and speed, determining the corrected expected throttle according to the total energy, and determining the corrected expected pitch angle according to the energy balance; Step 3, outputting the servo control and throttle control according to the corrected expected throttle and expected pitch angle to complete the height control of the UAV.

[0004] For another example, a Chinese patent with patent number CN114675663A discloses a method for planning and controlling the landing of a UAV in a crosswind environment, including: step S1. collecting parameters of the aircraft's lateral and heading motion state, parameters of the longitudinal motion state, and environmental parameters through airborne sensors; step S2. predicting the aircraft's landing time based on the data collected by the airborne sensors, and determining the time to start trajectory planning; step S3. using the UAV's lateral deviation distance, lateral deviation speed, roll angle, and heading angle at the beginning of trajectory planning as initial planning parameters, determining the aircraft's terminal state constraint parameters based on aircraft usage restrictions, and using the MPC planning algorithm to plan the UAV's lateral flight trajectory in combination with the aircraft's current wind speed and wind direction; step S4. using the planning results as instructions for the lateral and heading trajectory tracking and attitude controller, and designing a corresponding controller algorithm to track the reference instructions generated by the planning.

[0005] However, there are many existing methods for altitude control and landing control of drones, but none of them consider the ground conditions when the drone lands. They only consider the ground where the drone can land directly. However, in the context of drone patrols, there may be situations where it needs to land on a non-smooth surface, such as an inclined surface, an uneven ground, etc.; and most methods use model predictive control alone or PID control alone to complete the control of altitude and attitude. In the above special cases, the landing of the drone cannot be completed well. In the case of an inclined surface, it may even cause the drone to roll over due to a non-vertical collision with the ground when the drone descends because the speed and acceleration are not well controlled. In view of the problems that have not been considered by the defects of the existing methods, the present invention proposes a planning and control method for the landing of drones in complex ground environments. It can effectively solve the problem of landing drones in complex ground conditions. Summary of the invention

[0006] The purpose of the present invention is to provide a planning and control method for the landing of a UAV in a complex ground environment. First, the fuselage information of the UAV is obtained through different types of sensors, and then the ground situation is judged by a depth camera. When the ground is flat or an inclined plane with an inclination angle of 0-5 degrees, a plane landing strategy is adopted, that is, the propeller lift is reduced at a uniform speed, and no special strategy is called to maintain the speed and acceleration during descent. When the ground is an inclined plane with an inclination angle greater than 5 degrees, a PID-MPC model predictive control landing strategy is selected, and the throttle of the UAV motor is output to control the UAV so that the UAV can land smoothly on the complex ground.

[0007] The technical solution adopted by the present invention is as follows:

[0008] A method for planning and controlling the landing of a UAV in a complex ground environment comprises the following steps:

[0009] Step 1: Get the drone's body information;

[0010] Step 2: Select the drone landing strategy based on the ground conditions;

[0011] When the ground is flat or an inclined plane with an inclination angle of 0-5 degrees, the flat landing strategy is directly adopted. When the ground is an inclined plane with an inclination angle greater than 5 degrees, the PID-MPC model predictive control landing strategy is selected;

[0012] Step 3: Select the PID-MPC model predictive control landing strategy; specifically including the following steps:

[0013] Step 301: Perform steady-state optimization on target throttle, lift, and attitude, and obtain a system prediction model;

[0014] Step 302: Generate an optimal reference trajectory of throttle change according to the target altitude;

[0015] Step 303: Establishing an optimal control objective function;

[0016] Step 304: Find the optimal solution of the objective function, and use the optimal solution to control the throttle output and the motor to complete the control of the drone landing;

[0017] Step 305: Perform PID correction on the optimal solution to control the throttle output, use the prediction model of the PID error compensation model prediction control, and perform hybrid optimization prediction to complete the update of the system prediction model.

[0018] Preferably, in step 1, obtaining the drone's fuselage information specifically involves obtaining the drone's altitude information, the drone's attitude information, and initial ground condition data through sensors.

[0019] Preferably, in step 2, a depth camera is used to determine whether the ground has an inclination angle greater than 5 degrees, and when a flat landing strategy is adopted, the propeller lift is uniformly reduced, and no special strategy is called to maintain the speed and acceleration during descent.

[0020] Preferably, in step 3, the target throttle, lift, and attitude are optimized in a steady state, the step response of the system is obtained, and sampling is performed. According to the system requirements, we select ten sampling data points, (iT, a i ), i = 1, 2, ... 10, according to the homogeneous and superposition properties of linear systems, at time ki (k ≥ i) there is an input u (ki), then the effect of the input increment Δu (k-1) on the output y (k) is:

[0021] Thus, the prediction model of system y(k+1) is obtained:

[0022]

[0023] Perform prediction n-step control m-step modeling and finally obtain the system prediction model in is the n-step prediction of the system, A is an n×m-dimensional constant matrix, which is the data points we sampled, and ΔU is the input increment of m steps.

[0024] Preferably, in step 4, a filtering method is designed to optimize the height change curve, and the reference trajectory adopts an exponential change form:

[0025] ω(k+i)=a i y(k)+(1-a i )y r ;

[0026] in T sis the sampling period, T is the time constant of the reference trajectory, y(k) is the actual measured value at the current moment, y r This prevents the control system from experiencing drastic changes in input and output, allowing the aircraft to land smoothly and prevent rollover.

[0027] Preferably, in step 5, an objective function of optimal control is established, and this objective function has two constraints. The first is to use the square of the error as the objective function to track the generated optimal throttle trajectory, and the second is to minimize the total square of the throttle change output each time to minimize the energy. Therefore:

[0028]

[0029] So the comprehensive objective function is:

[0030] J=J1Q+J2R

[0031] Among them, Q and R are the weight coefficients of the objective function, which are used to adjust the objective function.

[0032] Preferably, in step 6, the optimal solution can be found by using a gradient descent method or a direct solution method; if the direct solution method is used, the analytical solution of the objective function is:

[0033] Δu(k)=(A T QA+R) -1 A T (W-Y0).

[0034] Preferably, after control, due to the actual situation, the system will have an error Here e(k+1) is the error, y(k+1) represents the actual output of the system, Represents the system output predicted by MPC; use this deviation to correct the system's prediction At this time, PID is introduced to correct it. E(k+1) is first passed through the PID controller to obtain E(k+1), and then the following formula is used for correction:

[0035] Roll the model predictions forward Perform a prediction update, where S is the shift matrix.

[0036] The technical effects achieved by the present invention are:

[0037] The present invention proposes a method for controlling the landing of a UAV under complex ground conditions, which fills the gap of the current lack of a method for landing a UAV under complex ground conditions; the present invention proposes an improved PID-MPC control algorithm in the process of controlling the landing of a UAV under complex ground conditions, which enhances the control performance of the traditional MPC algorithm and makes the throttle curve of the UAV during landing smoother; the present invention targets a variety of complex ground conditions, including inclined surfaces and uneven surfaces, and has different strategies corresponding to different conditions and different PID parameters.

[0038] Compared with the existing UAV landing method, the present invention is more targeted at complex ground conditions and ensures that the UAV can land more safely; the present invention introduces a PID-MPC control algorithm in the landing control process, enhances the stability of the control process, and provides corresponding descent strategies for different ground conditions; the present invention can adjust the descent strategy as needed, dynamically adjust the descent strategy according to different ground conditions, and change the PID parameters at the same time; it has strong flexibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a flow chart of a method for planning and controlling the landing of a UAV in a complex ground environment according to the present invention;

[0040] Figure 2 It is a flow chart of the PID-MPC control algorithm in a method for planning and controlling the landing of an unmanned aerial vehicle in a complex ground environment according to the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose and advantages of the present invention more clearly understood, the present invention is specifically described below in conjunction with embodiments. It should be understood that the following text is only used to describe one or several specific embodiments of the present invention, and does not strictly limit the scope of protection of the specific claims of the present invention.

[0042] like Figure 1-2 As shown, a method for planning and controlling the landing of a UAV in a complex ground environment includes the following steps:

[0043] Step 1: Get the drone's body information;

[0044] Specifically, the drone body information is obtained by using sensors to obtain the drone height information, drone attitude information and initial ground condition data. More specifically, the airborne ultrasonic sensor, laser sensor and depth camera are used to obtain the height information of the body and the initial ground condition data. Then, the raw data is filtered by combining the decision and fusion algorithms to obtain the height information that can be actually used, including whether the ground is flat. At the same time, the attitude information of the body is obtained through the gyroscope. The decision-level fusion algorithm process is as follows:

[0045] First: Get the decision of the first sensor and the second sensor (D is the sensor decision, x s is the value of the sensor's processed output) D1 = x s1 ; D2 = x s2 ; Then assign weights to the decisions of each sensor (WD is the assigned value, W is the weight)

[0046] WD1=W1*D1;WD2=W2*D2;Finally: calculate the fused decision output (DO is the decision output)

[0047]

[0048] Step 2: Select the drone landing strategy based on the ground conditions;

[0049] Use the depth camera to determine whether the ground is tilted at an angle greater than 5 degrees. When the ground is flat or a slope with an angle of 0-5 degrees, directly adopt the flat landing strategy, that is, reduce the propeller lift at a uniform speed, and do not call special strategies to maintain the speed and acceleration during descent; this method is the simplest; when the ground is a slope with an angle greater than 5 degrees, select the PID-MPC model predictive control landing strategy;

[0050] Step 3: Select the PID-MPC model predictive control landing strategy; specifically including the following steps:

[0051] Step 301: Perform steady-state optimization on target throttle, lift, and attitude, and obtain a system prediction model;

[0052] Specifically: For the target throttle, lift, and attitude, steady-state optimization is performed to obtain the step response of the system and sample it. According to the system requirements, we select ten sampling data points, (iT, a i ),i=1,2,…10, according to the homogeneity and superposition properties of linear systems,

[0053] At time ki (k ≥ i), there is an input u (ki), then the effect of the input increment Δu (k-1) on the output y (k) is:

[0054]

[0055] Thus, the prediction model of system y(k+1) is obtained:

[0056]

[0057] Perform prediction n-step control m-step modeling and finally obtain the system prediction model in is the n-step prediction of the system, A is the constant matrix of n×m dimensions, that is, the data points we sampled, and ΔU is the input increment of m steps;

[0058] Step 302: Generate an optimal reference trajectory of throttle change according to the target altitude;

[0059] Specifically, the optimal reference trajectory of throttle change is generated according to the target altitude. When the altitude is too high, the first-order derivative of the output curve of the system will be large, which may cause the aircraft to descend too fast and lead to instability. In this embodiment, a filtering method is designed to optimize the altitude change curve. The reference trajectory adopts an exponential change form:

[0060] ω(k+i)=a i y(k)+(1-a i )y r

[0061] in T s is the sampling period, T is the time constant of the reference trajectory, y(k) is the actual measured value at the current moment, y r is the set value; thereby avoiding drastic changes in input and output of the control system, allowing the aircraft to land smoothly and prevent rollover.

[0062] Step 303: Establishing an optimal control objective function;

[0063] Specifically: Establish the objective function of optimal control. This objective function has two constraints. The first is to track the generated optimal throttle trajectory as much as possible. Here, the square of the error is used as the objective function. The second is to minimize the energy. Because the output increments are positive and negative, in order to ensure the minimum energy, the total square of the throttle change output each time is minimized. This can ensure that power is saved on the basis of throttle smoothness.

[0064]

[0065] So the comprehensive objective function is:

[0066] J=J1Q+J2R

[0067] Wherein Q and R are weight coefficients of the objective function, which are used to adjust a more accurate objective function of this embodiment.

[0068] Step 304: Find the optimal solution of the objective function, use this optimal solution to control the throttle output, control the motor, and complete the control of the drone landing; find the optimal solution Δu(k) of the objective function, and use this optimal solution to control the throttle output. The optimal solution can be found by using the gradient descent method or the direct solution method. Here, the direct solution method is used to directly give the analytical solution of the objective function:

[0069] Δu(k)=(AT QA+R) -1 A T (W-Y0);

[0070] like Figure 2 As shown, step 305: PID correction is performed on the optimal solution to control the throttle output, a prediction model of the PID error compensation model prediction control is used, and a hybrid optimization prediction quantity is performed to complete the update of the system prediction model;

[0071] After control, due to the actual situation, the system will have an error Here e(k+1) is the error, y(k+1) represents the actual output of the system, Represents the system output predicted by MPC; this embodiment uses this deviation to correct the predicted value of the system At this time, PID is introduced to correct it. E(k+1) is first passed through the PID controller to obtain E(k+1), and then the following formula is used for correction: Here, different landing strategies will result in different PID controller parameters, thereby improving the robustness and anti-interference ability of the system;

[0072] Roll the model predictions forward Update the prediction quantity, which is also one of the key steps of MPC, where S is the shift matrix.

[0073] It is worth mentioning that the present invention proposes a UAV landing control method for complex ground environments, which fills the gap of the current lack of UAV landing methods for complex ground conditions; the present invention proposes an improved PID-MPC control algorithm in the process of controlling the landing of the UAV in complex ground conditions, which enhances the control performance of the traditional MPC algorithm and makes the throttle curve of the UAV during landing smoother; the present invention targets a variety of complex ground conditions, including slopes and uneven surfaces, and has different strategies for different conditions and different PID parameters.

[0074] It has to be said that compared with the existing UAV landing methods, the present invention is more targeted at complex ground conditions and ensures that the UAV can land more safely; the present invention introduces the PID-MPC control algorithm in the landing control process, enhances the stability of the control process, and provides corresponding descent strategies for different ground conditions; the present invention can adjust the descent strategy as needed, dynamically adjust the descent strategy according to different ground conditions, and change the PID parameters at the same time; it has strong flexibility.

[0075] The above is only a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be considered as the protection scope of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention shall be implemented according to the conventional means in the art unless otherwise specified and limited.

Claims

1. A method for planning and controlling the landing of an unmanned aerial vehicle in a complex ground environment, characterized in that: The following steps are involved: Step 1: Get the drone's body information; Step 2: Select the drone landing strategy based on the ground conditions; When the ground is flat or an inclined plane with an inclination angle of 0-5 degrees, the flat landing strategy is directly adopted. When the ground is an inclined plane with an inclination angle greater than 5 degrees, the PID-MPC model predictive control landing strategy is selected; Step 3: Select the PID-MPC model predictive control landing strategy; specifically including the following steps: Step 301: Perform steady-state optimization on target throttle, lift, and attitude, and obtain a system prediction model; Step 302: Generate an optimal reference trajectory of throttle change according to the target altitude; Step 303: Establishing an optimal control objective function; Step 304: Find the optimal solution of the objective function, and use the optimal solution to control the throttle output and the motor to complete the control of the drone landing; Step 305: Perform PID correction on the optimal solution to control the throttle output, use the prediction model of the PID error compensation model prediction control, and perform hybrid optimization prediction to complete the update of the system prediction model.

2. The method for planning and controlling the landing of a UAV in a complex ground environment according to claim 1, characterized in that: In the step 1, obtaining the drone's fuselage information specifically involves obtaining the drone's altitude information, the drone's attitude information, and initial ground condition data through sensors.

3. The method for planning and controlling the landing of a UAV in a complex ground environment according to claim 2, characterized in that: In step 2, the depth camera is used to determine whether the inclination angle of the ground is greater than 5 degrees. When a flat landing strategy is adopted, the propeller lift is uniformly reduced, and no special strategy is called to maintain the speed and acceleration during descent.

4. The method for planning and controlling the landing of a UAV in a complex ground environment according to claim 3, characterized in that: In step 3, the target throttle, lift, and attitude are optimized in a steady state to obtain the step response of the system and sample it. According to the system requirements, we select ten sampling data points, (iT, a i ), i = 1, 2, ... 10, according to the homogeneous and superposition properties of linear systems, at time ki (k ≥ i) there is an input u (ki), then the effect of the input increment Δu (k-1) on the output y (k) is: Thus, the prediction model of system y(k+1) is obtained: Perform prediction n-step control m-step modeling and finally obtain the system prediction model in is the n-step prediction of the system, A is an n×m-dimensional constant matrix, which is the data points we sampled, and ΔU is the input increment of m steps.

5. The method for planning and controlling the landing of a UAV in a complex ground environment according to claim 4, characterized in that: In step 4, a filtering method is designed to optimize the height change curve, and the reference trajectory adopts an exponential change form: ω(k+i)=a i y(k)+(1-a i )y r ; in T s is the sampling period, T is the time constant of the reference trajectory, y(k) is the actual measured value at the current moment, y r is the set value.

6. The method for planning and controlling the landing of a UAV in a complex ground environment according to claim 5, characterized in that: In step 5, the objective function of the optimal control is established. This objective function has two constraints. The first is to use the square of the error as the objective function to track the generated optimal throttle trajectory. The second is that the total square of the throttle change output each time is minimized to minimize the energy. Therefore: So the comprehensive objective function is: J=J1Q+J2R Among them, Q and R are the weight coefficients of the objective function, which are used to adjust the objective function.

7. The method for planning and controlling the landing of a UAV in a complex ground environment according to claim 6, characterized in that: In step 6, the optimal solution can be found by using the gradient descent method or the direct solution method; if the direct solution method is used, the analytical solution of the objective function is: Δu(k)=(A T QA+R) -1 A T (W-Y0)。 8. The method for planning and controlling the landing of a UAV in a complex ground environment according to claim 7, characterized in that: After control, due to the actual situation, the system will have an error Here e(k+1) is the error, y(k+1) represents the actual output of the system, Represents the system output predicted by MPC; use this deviation to correct the system's prediction At this time, PID is introduced to correct it. E(k+1) is first passed through the PID controller to obtain E(k+1), and then the following formula is used for correction: Roll the model predictions forward Perform a prediction update, where S is the shift matrix.

Images