Multi-rotor unmanned aerial vehicle and flight method thereof

Abstract

The invention discloses a multi-rotor unmanned aerial vehicle and a flight method thereof. The flight method comprises the following steps: collecting disturbance data; obtaining disturbance influence by using the disturbance data, and adding the disturbance influence into a linear prediction model to form a probability-based disturbance influence prediction model; and outputting a control sequence by using the disturbance influence prediction model based on the probability, and realizing accurate control flight of the multi-rotor unmanned aerial vehicle based on the control sequence.

Description

Technical Field

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

[0002] Quadrotors, due to their simple structure and flexible control, are widely used in aerial photography, mapping, rescue, and other fields. Their flight control systems typically require real-time processing of sensor data to achieve stable flight and precise target tracking.

[0003] However, in terms of quadcopter flight control technology, common algorithms such as PID control have limited effectiveness in complex environments. PID (Proportional-Integral-Derivative) control, which uses proportional, integral, and derivative control principles, is known as proportional-integral-derivative control. It uses proportional, integral, and derivative control strategies to adjust the output of the flight control system to achieve a setpoint.

[0004] Therefore, how to achieve stable flight and precise control of a multi-rotor (such as a quad-rotor) UAV in a complex environment is a technical problem to be solved by the present invention. Summary of the Invention

[0005] The present invention aims to provide a multi-rotor drone and a flight method thereof. This multi-rotor drone combines a Gaussian mixture model (GMM) with an optimized control strategy (SMPC) and introduces stochastic model predictive control based on a perturbation probability model. This creates a more robust and adaptable multi-rotor drone control scheme, effectively improving the drone's flight performance and safety in complex environments.

[0006] According to one aspect of the present invention, at least one embodiment provides a flight method applicable to a multi-rotor unmanned aerial vehicle, comprising: collecting disturbance data; using the disturbance data to obtain disturbance effects, and adding the disturbance effects to a linear prediction model to form a probability-based disturbance effect prediction model; using the probability-based disturbance effect prediction model to output a control sequence, and achieving precise control flight of the multi-rotor unmanned aerial vehicle based on the control sequence.

[0007] According to another aspect of the present invention, at least one embodiment further provides a multi-rotor drone, comprising: a processor, suitable for implementing various instructions; and a memory, suitable for storing multiple instructions, wherein the instructions are suitable for being loaded and executed by the processor: the above-mentioned flight method of the present invention.

[0008] According to another aspect of the present invention, at least one embodiment further provides a computer-readable non-volatile storage medium storing computer program instructions, which, when the computer executes the program instructions, performs: the above-mentioned flight method of the present invention.

[0009] Through the above-mentioned embodiments of the present invention, the disturbance effect brought by the disturbance data is added to the linear prediction model to form a probability-based disturbance effect prediction model, and the control sequence output by the probability-based disturbance effect prediction model is used to realize the control of the multi-rotor UAV. In the case of random disturbance and uncertainty, it can not only maintain flight stability but also optimize energy consumption and task execution efficiency, thereby ensuring stable flight and precise control of the multi-rotor UAV in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0011] Figure 1 is a schematic diagram of a multi-rotor drone according to an embodiment of the present invention;

[0012] Figure 2 is a flow chart of a flying method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0014] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0015] Currently, the flight control technologies for multi-rotor UAV flight control systems to deal with wind speed disturbances mainly include the following: (1) PID control. The PID controller offsets the impact of wind on the attitude and position of the multi-rotor UAV by adjusting the control output (such as rotor speed). Limitations: Basic PID control may not respond quickly or accurately to sudden or continuous wind speed changes, especially when wind direction and wind speed change in a complex manner. (2) Adaptive control. Common methods include model reference adaptive control or self-correcting control, which adjusts control parameters during flight to adapt to external disturbances such as wind speed. Limitations: This method may require an accurate model or online parameter estimation and may perform poorly when wind speed changes drastically or the pattern is complex. (3) Robust control. Such as H∞ control, this method is designed to take into account the worst-case disturbance to ensure that the system remains stable within the preset disturbance range. Limitations: The design may be too conservative, resulting in suboptimal performance in the absence of disturbance or small disturbance. (4) Model predictive control. Use a model of the flight control system to predict future behavior and minimize future errors by optimizing a series of control actions. Some model predictive control schemes consider wind as a measurable or predictable disturbance. Limitations: Relying on accurate predictions of wind speed, computationally intensive, and model accuracy significantly impacting control effectiveness.

[0016] That is to say, although existing flight control technologies can cope with simple environmental changes to a certain extent, they still have significant limitations when dealing with the variable factors and uncertainties in complex environments. (1) Insufficient robustness: They are often based on simplified environmental models and cannot fully handle multimodal random disturbances in complex environments (such as gusts and airflow changes). They lack effective means to deal with random disturbances (such as wind speed changes in different environments), resulting in poor robustness in actual complex environments. (2) Weak scene generalization ability: When migrating from one environment to another, a large amount of recalibration or redesign of the controller is usually required. They lack the ability to generalize from a set of specific conditions to other conditions, which is not conducive to the rapid deployment of multi-rotor drones in various mission environments. (3) Large amount of calculation and insufficient real-time performance: Especially algorithms such as MPC, whose multi-step prediction and optimization lead to large amount of calculation, may not meet the real-time control requirements. Under high-speed motion or frequent disturbances, the system's calculation delay will reduce the control effect.

[0017] Due to the limitations of existing flight control technology, current multi-rotor (such as quad-rotor) drones have problems with stable flight and precise control in complex environments, and their robustness and adaptability are low. In order to solve the robustness and adaptability problems of multi-rotor (such as quad-rotor) drones in the face of complex environments and uncertainties, at least one embodiment of the present invention provides a multi-rotor drone, such as a quad-rotor drone, and in particular provides a flight control system for the multi-rotor drone, such as Figure 1As shown, including:

[0018] The processor 101 is adapted to implement various instructions. The processor 101 may be any suitable processor, such as a central processing unit, a microprocessor, an embedded processor, etc., and may adopt an architecture such as X86 or ARM.

[0019] The memory 103 is suitable for storing a plurality of instructions, and the instructions are suitable for being loaded by the processor 101 and executing the flying method developed by the present invention (this article will be elaborated in detail later and will not be described in detail here). The memory 103 can be various applicable storage devices, such as non-volatile storage devices, including but not limited to magnetic storage devices, semiconductor storage devices, optical storage devices, etc., and can be arranged as a single storage device, a storage device array or a distributed storage device, and the embodiments of the present invention do not limit these. Optionally, corresponding to the memory 103, at least one embodiment of the present invention further provides a computer-readable non-volatile storage medium that stores computer program instructions. When the computer executes the program instructions, it executes: the flying method developed by the present invention (this article will be elaborated in detail later and will not be described in detail here).

[0020] It will be understood by those skilled in the art that the above multi-rotor drone structure is only for illustration and does not limit the structure of the multi-rotor drone. Figure 1 The transmission device is used to receive or send data via a network. In one embodiment, the transmission device is a radio frequency (RF) module that is used to communicate wirelessly with other multi-rotor drones, various sensors, or other devices.

[0021] Through the above-mentioned method of the present invention, a multi-rotor UAV equipped with a flight control system is built, thereby achieving stable flight and precise control of the multi-rotor UAV in complex environments.

[0022] In the operating environment of the above-mentioned aircraft of the present invention, at least one embodiment of the present invention provides a flying method, which can also be called an optimized multi-rotor (quad-rotor) UAV control scheme, or a multi-rotor (quad-rotor) UAV control method based on random model predictive control, or a SMPC (random model predictive control) control method based on a Gaussian mixture perturbation probability model, which is loaded and executed by the processor 101. It can enhance the performance of the aircraft in dynamic and uncertain environments based on the actual flight environment of the multi-rotor UAV, and realize stable flight and precise control of the multi-rotor UAV in complex environments. Figure 2As shown, the flying method may include the following steps. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0023] Step S202: Collect disturbance data

[0024] Step S204: using the disturbance data to obtain the disturbance effect, and adding the disturbance effect to the linear prediction model to form a probability-based disturbance effect prediction model;

[0025] Step S206: Use the probability-based disturbance impact prediction model to output a control sequence, and achieve precise flight control of the multi-rotor UAV based on the control sequence.

[0026] Through the above-mentioned embodiments of the present invention, the disturbance effect brought by the disturbance data is added to the linear prediction model to form a probability-based disturbance effect prediction model, and the control sequence output by the probability-based disturbance effect prediction model is used to realize the control of the multi-rotor UAV. In the case of random disturbance and uncertainty, it can not only maintain flight stability but also optimize energy consumption and task execution efficiency, thereby ensuring stable flight and precise control of the multi-rotor UAV in complex environments.

[0027] In step S202, disturbance data is collected, including but not limited to wind speed and direction, and other uncertain external disturbances (such as temperature, humidity, etc.). For example, the GPS, IMU, and other sensors of the multi-rotor drone are used to measure data in different flight environments (such as sunny days and storms), and disturbance data such as wind speed, wind direction, temperature, humidity, and other uncertain external disturbances in the flight environment are collected.

[0028] In step S204, the disturbance data is used to obtain the disturbance effect, and the disturbance effect is added to the linear prediction model to form a probability-based disturbance effect prediction model. The present invention draws on the advantages of stochastic model predictive control (SMPC) technology and uses the disturbance data to establish a stochastic disturbance Gaussian mixture model (GMM) to capture the uncertainty in the environment. The disturbance effect is sampled from the established GMM, obtained and added to the linear prediction model to form a probability-based disturbance effect prediction model, thereby enhancing the adaptability of the linear prediction model to random disturbances.

[0029] That is to say, in view of the uncertainty and disturbance of traditional control technology (such as PID controller), in some cases it can meet the basic control requirements, but in the face of complex environments and dynamic changes, especially when affected by disturbances, it is often difficult to maintain good performance; that is, the PID controller has advantages in terms of fast dynamic response and strong stability of the system, but its robustness to nonlinearity and external disturbances is relatively weak, which may cause the system to perform poorly in a rapidly changing environment. With reference to the stochastic model predictive control (SMPC) technology, it can effectively deal with the advantages of uncertainty and randomness in the system. By introducing the modeling of the prediction error and using probabilistic methods to deal with uncertainty factors such as wind speed changes, sensor errors and load changes, it is possible to consider the prediction error at each time section and improve the robustness and performance of the system through multi-scenario selection and rolling optimization; SMPC can adapt to dynamic environments more flexibly, ensuring the stable flight of multi-rotor drones under various conditions, especially in complex urban environments or changing natural environments. SMPC can effectively improve control accuracy and responsiveness. Therefore, the present invention draws on the advantages of stochastic model predictive control (SMPC) technology:

[0030] First, use the perturbation data to get the perturbation effect. The perturbation data is w T , It represents the disturbance of the i-th predicted horizon multirotor UAV at time t within the T time period, It represents the disturbance of the Nth predicted horizon multirotor UAV at time t within the T time period. For example, using Get the perturbation sample sequence W, where W=[w1,w2,…w i …w N ], N is the number of predicted horizons, wi represents the disturbance of the multi-rotor drone in the i-th predicted horizon, w N Represents the disturbance of the Nth predicted horizon multi-rotor drone; Based on W, a random perturbation Gaussian mixture model is established, where the probability density function of the random perturbation Gaussian mixture model is π k is the weight of the kth Gaussian component of the random perturbation Gaussian mixture model, with a total of K Gaussian components, μ k is the mean vector of the kth Gaussian component of the random perturbation Gaussian mixture model, ∑k is the covariance matrix of the kth Gaussian component of the random perturbation Gaussian mixture model, R(W|μ k ,∑k) represents the random perturbation Gaussian mixture model at the kth μ k The probability density function of the multivariate Gaussian distribution of W under the condition of ; sample p(W) to obtain the disturbance effect w′.

[0031] (1)Use Get the disturbance sample sequence W. For example, obtain the state transfer matrix A, control input matrix B, output matrix C and disturbance input matrix F of the multi-rotor drone. These matrices are the parameters of the linear discrete state space model and need to be obtained in advance through the flight control system identification or physical modeling of the multi-rotor drone; obtain the state sequence x (including position, speed, attitude angle and angular velocity) of the multi-rotor drone in different flight states through the sensors of the multi-rotor drone (such as GPS, IMU, etc.), record the input sequence u (i.e., rotor rotation angular velocity) of the multi-rotor drone in different flight states through the flight control system of the multi-rotor drone, arrange the state sequence x and input sequence u in chronological order, and use Get the perturbation sample sequence W=[w1,w2,…w i …w N ],in, Indicates the flight status of the Nth predicted horizon multi-rotor drone at time t within the T time period, It represents the flight status of the i-th predicted horizon multi-rotor UAV at time t within the T time period, represents the flight state of the multi-rotor drone at the 0th predicted horizon at time t within the T time period, x t Indicates the flight status of the multi-rotor drone at time t, x, y, and z are the three-axis positions of the multi-rotor drone. are the three-axis speeds of the multi-rotor drone, φ, θ, They are the pitch angle, rotation angle and yaw angle of the multi-rotor drone respectively. They are the pitch rate, roll rate and yaw rate of the multi-rotor drone respectively. represents the input of the Nth predicted horizon multirotor drone at time t within the T time period, represents the input of the i-th predicted horizon multi-rotor drone at time t within the T time period, ω1, ω2, ω3, and ω4 are the angular velocities of the quadrotors of the multi-rotor drone.

[0032] (2) Establish a random perturbation Gaussian mixture model based on W. Given that Gaussian mixture model (GMM) is widely used in speech recognition, image recognition, chaotic system control and other fields, GMM is a probability density estimation method that combines the advantages of parametric model and non-parametric model and is not limited to a specific probability density function form assumption. Therefore, the present invention obtains the perturbation sample sequence W = [w1, w2, ... w i …w N ], and then use GMM to establish the probability density function of the disturbance.

[0033] First, the number of Gaussian components K in the randomly perturbed Gaussian mixture model is evaluated using information criteria, where information criteria include AIC and / or BIC (Bayesian Information Criterion). In other words, AIC or BIC is used to evaluate the performance of the randomly perturbed Gaussian mixture model under different K values. AIC or BIC can help determine the balance between model complexity and goodness of fit. The specific steps for determining the K value are as follows: train the randomly perturbed Gaussian mixture model for different K values ​​(such as K = 1, 2, 3, ..., M); calculate the AIC and BIC values ​​corresponding to each K value; compare the AIC or BIC values ​​under different K values, and select the K value with the lowest AIC or BIC value as the optimal randomly perturbed Gaussian mixture model.

[0034] Secondly, the expectation maximization algorithm is used to estimate the weight π of the kth Gaussian component of the random perturbation Gaussian mixture model k , k≤K, where L(z=k|w i ) represents the random perturbation Gaussian mixture model in w i

[0035] Under the condition, each prediction horizon i belongs to the posterior probability of the kth Gaussian component of the independent variable z. That is, the present invention uses the expectation maximization (EM) algorithm to estimate the parameters of the randomly perturbed Gaussian mixture model. The relevant steps are as follows: ① Initialization, randomly select the initial parameters of the Gaussian distribution (mean, covariance and weight). ② Expectation step (E step), calculate the posterior probability of each data point belonging to each Gaussian distribution, d represents the dimension of the observed data, det(∑k) represents the determinant of the covariance matrix ∑k, -1 Represents the inverse matrix of the covariance matrix ∑k. ③ Maximization step (M step), use the posterior probability calculated in the E step to update the model parameters and update the mean μ k : Update the covariance matrix: Update weights: ④ Repeat the E-step and M-step until convergence and the parameters of the randomly perturbed Gaussian mixture model no longer change significantly.

[0036] Finally, based on π k Establish a random perturbation Gaussian mixture model, where the random perturbation Gaussian mixture model satisfies That is, the present invention collects disturbance data such as wind speed and wind direction by measuring data in different flight environments (such as sunny days and storms), and uses these disturbance data to establish a random disturbance Gaussian mixture model (GMM) to capture the uncertainty in the environment. TWhen there are multiple peaks or modes, the present invention uses a random perturbation Gaussian mixture model to describe it, and its probability density function is expressed as

[0037] (3) Sample p(W) to generate disturbance effects w′ that can represent various weather factors, wind speed changes, wind direction, acceleration, and angular velocity. For example, divide p(W) into n equal probability intervals, that is, divide [0,1] into n equal probability intervals, and the length of each equal probability interval is The i-th interval is: Cumulative distribution function of p(W) ω is the integral variable; the midpoint or random point of each probability interval is selected as the sampling point. Assume that the selected point is a, the midpoint The random points are That is, the random point is a random variable uniformly distributed in each probability interval; the disturbance effect w′ corresponding to the sampling point is obtained by the inverse transformation of the probability distribution function p(W)=W, where w′=F -1 (a), F -1 (a) represents the inverse transformation of the midpoint or random point a in the cumulative distribution function F(W), and a represents the selected midpoint or random point.

[0038] Second, the disturbance effect is added to the linear prediction model to form a probability-based disturbance effect prediction model. The linear prediction model is described as follows: The present invention assumes that when the multi-rotor UAV is in an approximate hovering state, the dynamic model of the multi-rotor UAV can be approximated by a linear prediction model. A first-order Taylor expansion is used near the hovering equilibrium point to convert it into a linear discrete state space model. The flight state x of the multi-rotor UAV is t (including three-axis position, speed and attitude angle), input u t (such as rotor angular velocity), output y t They are defined as: u t =[ω1,ω2,ω3,ω4], u t ∈R m ,y t ∈R q In the finite time optimal control, the present invention assumes that the linear prediction model of the multi-rotor UAV solves the optimization problem from t to t+N, where N is the prediction horizon, and establishes the state prediction model of the optimal control in the finite time: x = Ax t +Bu+Fw,y=Cx,A,B,F,C are the state transfer matrix, control input matrix, disturbance input matrix, output matrix respectively,A∈R n×n , B∈R n×mThe system matrix is ​​known and Lyapunov stable, C∈R q×n , F∈R n×p .x t+1 Indicates the flight status of the multi-rotor drone at the next moment, which can be obtained from the current known flight status x t With the predicted input u t , disturbance w t The calculation shows that the compact model of flight state x, input u, output y and disturbance w is as follows: At the same time, given that SMPC (Stochastic-Model Predictive control) is a model-based predictive control algorithm that takes uncertainty into account, it describes the dynamic behavior of the system by establishing a random model, and uses an optimization algorithm to consider various possible interferences and uncertainties during prediction, thereby achieving a more robust control effect. Therefore, in order to perform multi-scenario selection, the present invention samples from the established random disturbance Gaussian mixture model (GMM) to obtain the disturbance effect w′, and adds it to the above-mentioned linear prediction model to form a probability-based disturbance effect prediction model, thereby enhancing the model's adaptability to random disturbances.

[0039] (1) Use a clustering algorithm to select representative samples w″ from w′. For example, the K-means clustering algorithm is used to select representative samples w″, which effectively reduces the number of samples while retaining the main features of the sample data. First, the original sample n is clustered using m cluster centers, and the m cluster centers are initialized. Each sample is assigned to the nearest cluster center according to the Euclidean distance, and the cluster center is updated to the average value of all samples in its cluster. The above assignment and update actions are repeated until the cluster center no longer changes significantly. Secondly, for each cluster, a central sample or the sample closest to the center is selected as the representative point, and a new sample set is formed with the center point of each cluster as the representative. In this way, the samples can be generated and reduced by K-means, and representative samples w″ (or disturbance combinations) of multi-rotor drones under different weather or environmental conditions can be obtained in turn. These disturbance combinations can include different configurations of various weather factors, wind speed changes, wind direction, acceleration and angular velocity.

[0040] (2) The representative sample w″ is added to the above linear prediction model to form a probability-based disturbance impact prediction model. The probability-based disturbance impact prediction model optimization objective function includes three parts: the accuracy of the state approximation target, the smoothness and energy saving of the control input, and the minimization of the disturbance impact, that is,

[0041] x t+i+1represents the future flight state of the multi-rotor drone based on the i-th prediction horizon at time t, x t+i+1 ref represents the target flight state of the multi-rotor UAV based on the i-th predicted horizon at time t, Q, R, and S represent the penalty factors of the multi-rotor UAV flight state, input, and disturbance, respectively, and u t+i represents the input of the i-th prediction horizon multi-rotor drone at time t, w″ t+i It represents the disturbance effect of the representative sample w″ on the i-th prediction horizon multirotor UAV at time t, T′ represents the transposition, and the dynamic constraint of the probability-based disturbance impact prediction model is x t+i+1 =Ax t+i +Bu t+i +Fw″ t+i The control input constraint of the probability-based disturbance effect prediction model is Umin≤U≤Umax, where U represents the control sequence of the multi-rotor UAV that will be output by the probability-based disturbance effect prediction model, Umin represents the control sequence of the multi-rotor UAV with the minimum output of the probability-based disturbance effect prediction model, and Umax represents the control sequence of the multi-rotor UAV with the maximum output of the probability-based disturbance effect prediction model. The flight state constraint of the probability-based disturbance effect prediction model is x min ≤x t+i ≤x max , x min The probability-based disturbance impact prediction model hopes that the multi-rotor UAV will have the minimum flight state, x max It represents the maximum flight state of the multi-rotor UAV expected by the probability-based disturbance impact prediction model.

[0042] In step S206, a probability-based disturbance effect prediction model is used to output a control sequence, and precise control flight of the multi-rotor UAV is achieved based on the control sequence. The control sequence is U = {u(t), u(t+1),…, u(t+E-1)}, where u(t) represents the control input to the multi-rotor UAV at the planned time t, u(t+E-1) represents the control input to the multi-rotor UAV at the planned time t+E-1, and E is the time domain length predicted by the probability-based disturbance effect prediction model.

[0043] First, the present invention transforms the optimization problem of the probability-based disturbance impact prediction model into a standard quadratic programming problem, solving the optimal control input sequence U = {u(t), u(t+1), ..., u(t+E-1)} within the prediction horizon. The objective function J is transformed into a quadratic programming problem J(U), where U is the decision variable (combination of input sequences), T' represents the transpose. H matrix of the objective function J, H = B T′ QB+R+F T′SF. The linear term f of the objective function J takes into account the linear impact of the current flight state and the predicted disturbance on the objective function, f = B T′ Q(x k -x k ref )+Ru(t),x t represents the flight status of the multi-rotor drone at time t, x t ref Represents the reference flight state of the multi-rotor drone at time t. Equality constraint: x i+1 =Ax i +Bu i +Fw i . Inequality constraints: Control input constraints A u U≤b u , State Constraint A x X≤b x ,

[0044] Second, the transformed quadratic programming problem J(U) is input into a quadratic programming solver to obtain the optimal control sequence U. Specifically, the present invention optimizes the objective function J to minimize state deviation and energy consumption, taking into account the impact of future disturbances. By solving a quadratic programming problem, the optimal control sequence U is obtained to ensure the stability and efficiency of the multi-rotor UAV. The optimized control sequence is then applied to the multi-rotor UAV to adjust the rotor speed to achieve the desired motion trajectory and flight performance.

[0045] Third, time t is advanced by one step to t+1, and the optimization window is rolled to the next time period: from [t, t+N] to [t+1, t+N+1], and J(U) is re-solved to obtain a new control sequence U*={u(t+1), u(t+2),…, u(t+N)}; the above process is repeated continuously, and at each time t, the new first input U* is applied to update the state of the multi-rotor drone to achieve precise flight control of the multi-rotor drone. In other words, the present invention performs rolling time domain optimization, shifts the optimization problem forward by one time step, and performs a new round of optimization based on the updated state to ensure the real-time performance of the control strategy. Using the updated state and disturbance information, a linear prediction model is established to predict the state of the multi-rotor drone at future moments, ensuring that the linear prediction model can effectively cope with uncertainty.

[0046] The above-mentioned method of the present invention has the following advantages compared with the existing technology: enhanced robustness. By introducing a probability-based disturbance impact prediction model, SMPC can effectively handle random disturbances, allowing multi-rotor drones to maintain stable flight in complex environments; improved adaptability and generalization capabilities. There is no need to frequently redesign the controller of the multi-rotor drone flight control system, and it can adapt to changes in different environments and achieve cross-scenario generalization, thereby accelerating the deployment of multi-rotor drones in various mission scenarios; reduced computational complexity and guaranteed real-time performance. When performing multi-scenario rolling optimization, SMPC adopts strategies such as scene reduction to reduce the amount of calculation and ensure real-time control capabilities in complex environments.

[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A flying method, applicable to a multi-rotor drone, characterized in that: include: Collect disturbance data; Using the disturbance data to obtain a disturbance effect, and adding the disturbance effect to a linear prediction model to form a probability-based disturbance effect prediction model; The probability-based disturbance impact prediction model is used to output a control sequence, and precise control flight of a multi-rotor UAV is achieved based on the control sequence.

2. The method according to claim 1, characterized in that The disturbance data includes wind speed and wind direction, and collecting the disturbance data includes: Collect wind speed and direction of the flight environment.

3. The method according to claim 1, characterized in that The disturbance data is w T , It represents the disturbance of the i-th predicted horizon multirotor UAV at time t within the T time period, It represents the disturbance of the Nth predicted horizon multi-rotor UAV at time t within the time period T. The disturbance impact obtained by using the disturbance data includes: use Get the disturbance sample sequence W, where W=[w1,w2,…w i …w N ], N is the number of predicted horizons, wi represents the disturbance of the multi-rotor drone in the i-th predicted horizon, w N Indicates the disturbance to which the Nth predicted horizon multirotor drone is subjected; A random perturbation Gaussian mixture model is established based on W, wherein the probability density function of the random perturbation Gaussian mixture model is π k is the weight of the kth Gaussian component of the random perturbation Gaussian mixture model, with a total of K Gaussian components, μ k is the mean vector of the kth Gaussian component of the random perturbation Gaussian mixture model, ∑k is the covariance matrix of the kth Gaussian component of the random perturbation Gaussian mixture model, R(W|μ k ,∑k) represents the random perturbation Gaussian mixture model at the kth μ k The probability density function of the multivariate Gaussian distribution of W under the condition of ; The p(W) is sampled to obtain the disturbance effect w′.

4. The method according to claim 3, characterized in that use The perturbation sample sequence W is obtained including: Obtain the state transfer matrix A, control input matrix B, output matrix C and disturbance input matrix F of the multi-rotor drone; use Get the perturbation sample sequence W=[w1,w2,…·w i …w N ],in, Indicates the flight status of the Nth predicted horizon multi-rotor drone at time t within the T time period, It represents the flight status of the i-th predicted horizon multi-rotor UAV at time t within the T time period, represents the flight state of the multi-rotor drone at the 0th predicted horizon at time t within the T time period, x t Indicates the flight status of the multi-rotor drone at time t, x, y, and z are the three-axis positions of the multi-rotor drone. are the three-axis speeds of the multi-rotor drone, φ, θ, They are the pitch angle, rotation angle and yaw angle of the multi-rotor drone respectively. They are the pitch rate, roll rate and yaw rate of the multi-rotor drone respectively. represents the input of the Nth predicted horizon multi-rotor drone at time t within the T time period, represents the input of the i-th predicted horizon multirotor drone at time t within the T time period, Represents the input of the multi-rotor drone with the 0th predicted horizon at time t within the time period T.

5. The method according to claim 3, characterized in that Establishing a random perturbation Gaussian mixture model based on W includes: Evaluating the number K of Gaussian components of the randomly perturbed Gaussian mixture model using an information criterion, wherein the information criterion includes AIC and / or BIC; The weight π of the kth Gaussian component of the random perturbation Gaussian mixture model is estimated using the expectation maximization algorithm k , k≤K, where L(z=k|w i ) represents the random perturbation Gaussian mixture model in w i The posterior probability that each prediction horizon i belongs to the kth Gaussian component of the independent variable z under the condition; Based on the π k Establish a random perturbation Gaussian mixture model, wherein the random perturbation Gaussian mixture model satisfies 6. The method according to claim 3, characterized in that Sampling the p(W) includes: Divide the p(W) into n equal probability intervals, where the cumulative distribution function of the p(W) is ω is the integration variable; The midpoint or random point of each probability interval is selected as the sampling point, and the disturbance effect w′ corresponding to the sampling point is obtained by the inverse transformation of the probability distribution function p(W)=W, where w′=F -1 (a), F -1 (a) represents the inverse transformation of the midpoint or random point a in the cumulative distribution function F(W), and a represents the selected midpoint or random point.

7. The method according to claim 6, characterized in that Adding the disturbance effect to the linear prediction model to form a probability-based disturbance effect prediction model includes: Use clustering algorithm to select representative samples w″ in w′; The w″ is added to the linear prediction model to form a probability-based disturbance impact prediction model, wherein the optimization objective of the probability-based disturbance impact prediction model is x t+i+1 represents the future flight state of the multi-rotor drone based on the i-th prediction horizon at time t, x t+i+1 ref represents the target flight state of the multi-rotor UAV based on the i-th predicted horizon at time t, Q, R, and S represent the penalty factors of the multi-rotor UAV flight state, input, and disturbance, respectively, and u t+i represents the input of the i-th prediction horizon multi-rotor drone at time t, w″ t+i It indicates that the i-th prediction horizon multi-rotor UAV is affected by the disturbance of the representative sample w″ at time t, T ′ Represents transposition, the dynamic constraint of the probability-based disturbance impact prediction model is x t+i+1 =Ax t+i +Bu t+i +Fw″ t+i The control input constraint of the probability-based disturbance effect prediction model is Umin≤U≤Umax, where U represents the control sequence of the multi-rotor UAV that will be output by the probability-based disturbance effect prediction model, Umin represents the control sequence of the multi-rotor UAV with the minimum output of the probability-based disturbance effect prediction model, and Umax represents the control sequence of the multi-rotor UAV with the maximum output of the probability-based disturbance effect prediction model. The flight state constraint of the probability-based disturbance effect prediction model is x min ≤x t+i ≤x max , x min The probability-based disturbance impact prediction model hopes that the multi-rotor UAV will have the minimum flight state, x max It represents the maximum flight state of the multi-rotor UAV expected by the probability-based disturbance impact prediction model.

8. The method according to claim 7, characterized in that The control sequence is U={u(t), u(t+1), ..., u(t+E-1)}, where u(t) represents the planned control input to the multi-rotor UAV at time t, u(t+E-1) represents the planned control input to the multi-rotor UAV at time t+E-1, and E is the time domain length predicted by the probability-based disturbance impact prediction model. Outputting the control sequence using the probability-based disturbance impact prediction model includes: The optimization objective J of the probability-based disturbance impact prediction model is transformed into a quadratic programming problem J(U), where H=B T′ QB+R+F T′ SF, f = B T′ Q(x t -x t ref )+Ru(t),x t ref represents the reference flight state of the multirotor UAV at time t; The J(U) is input into the quadratic programming solver to obtain the control sequence U.

9. The method according to claim 8, characterized in that The precise control of the multi-rotor UAV flight based on the control sequence includes: Advance the time t by one step to t+1, and re-solve J(U) to obtain a new control sequence U*={u(t+1),u(t+2),…,u(t+N)}; At each moment t, U* is applied to update the state of the multi-rotor UAV to achieve precise flight control of the multi-rotor UAV.

10. A multi-rotor drone, characterized in that: include: A processor, adapted to implement each instruction; and a memory, adapted to store a plurality of instructions, wherein the instructions are adapted to be loaded and executed by the processor: a flying method according to any one of claims 1 to 9.

Images