Aircraft control method and device, aircraft and storage medium

Abstract

The invention discloses an aircraft control method and device, an aircraft and a storage medium. The method comprises the following steps: acquiring environment data and first control data of the aircraft; based on the environment data and the first control data, determining a first operation state of the aircraft by using a first machine learning model; determining a second operation state of the aircraft by using a second prediction model based on the first control data; determining a target operation state of the aircraft based on the first operation state and the second operation state, wherein the target operation state at least comprises one or more of the speed, the position and the attitude of the aircraft during operation; determining target control data of the aircraft based on the target operation state; the target control data is used for controlling the aircraft, and the target control parameters at least comprise the speed and / or the rotation angle of the aircraft.

Description

Technical Field

[0001] The present application relates to the technical field of aircraft control, and particularly to an aircraft control method, device, aircraft and storage medium. Background Art

[0002] As a new type of transportation vehicle, an aircraft has the ability to fly freely in the air, which can effectively alleviate the ground traffic congestion problem and achieve efficient point-to-point transportation. However, during the flight process, the aircraft faces complex and variable external environmental interferences, such as sudden changes in wind speed, air flow disturbances, temperature changes, etc. These factors pose severe challenges to the stability, safety and control accuracy of the aircraft.

[0003] Traditional flight control methods are mainly divided into two categories: the control method based on proportional-integral-derivative (PID) and the control method based on model predictive control (MPC).

[0004] PID control method: PID control is a classical control algorithm that realizes the control of the system by adjusting the proportional, integral and derivative parameters. Although the PID control structure is simple and easy to implement, its robustness is poor, and it is difficult to cope with the multivariable coupling and nonlinear interference in a complex environment. For example, when suddenly encountering a strong crosswind, the PID controller may not be able to quickly adjust the flight attitude, resulting in trajectory deviation or even out of control. In addition, the parameter adjustment of PID control depends on experience, and the applicable range is narrow, making it difficult to meet the control requirements of the aircraft in a high-dynamic environment.

[0005] MPC control method: MPC is a model-based control method that constructs a state transition model of the system, predicts the system behavior in the future for a period of time, and optimizes the control input to achieve the target state tracking. The core advantage of MPC lies in its ability to explicitly handle the control problems of multivariables and multiple constraints. However, traditional MPC methods highly rely on accurate prediction models, and the environmental interferences (such as turbulence, temperature gradient, etc.) suffered by the aircraft during actual flight are usually difficult to be described by accurate prediction models. Therefore, the control performance of traditional MPC is limited in a complex environment, and it is difficult to ensure the stability and safety of the aircraft.

[0006] In summary, in the prior art, the control methods of aircraft mainly face the following problems: traditional PID control has poor robustness and is difficult to cope with complex environmental interferences; traditional MPC methods rely on accurate mathematical models and cannot effectively handle the uncertainties brought by environmental changes. Therefore, there is an urgent need for an aircraft control method that can improve its control accuracy and robustness in a complex environment. Summary of the Invention

[0007] The present application provides a flight vehicle control method, apparatus, flight vehicle, and storage medium to at least solve the above technical problems existing in the prior art.

[0008] According to a first aspect of the present application, a flight vehicle control method is provided. The method includes:

[0009] Obtain environmental data and first control data of the flight vehicle;

[0010] Based on the environmental data and the first control data, use a first machine learning model to determine a first operating state of the flight vehicle;

[0011] Based on the first control data, use a second prediction model to determine a second operating state of the flight vehicle;

[0012] Based on the first operating state and the second operating state, determine a target operating state of the flight vehicle. The target operating state includes at least one or more of the speed, position, and attitude of the flight vehicle during operation;

[0013] Based on the target operating state, determine target control data of the flight vehicle; the target control data is used to control the flight vehicle, and the target control parameters include at least the speed and / or turning angle of the flight vehicle.

[0014] In an implementable manner, the determining the target operating state of the flight vehicle based on the first operating state and the second operating state includes:

[0015] Perform a weighted sum operation on a first operating parameter corresponding to the first operating state and a second operating parameter corresponding to the second operating state to obtain a target operating parameter; the target operating parameter corresponds to the target operating state.

[0016] In an implementable manner, the performing a weighted sum operation on a first operating parameter corresponding to the first operating state and a second operating parameter corresponding to the second operating state to obtain a target operating parameter includes:

[0017] Multiply the first operating parameter by a first weight corresponding to the first operating parameter to obtain a first operation result;

[0018] Multiply the second operating parameter by a second weight corresponding to the second operating parameter to obtain a second operation result;

[0019] Determine the sum of the first operation result and the second operation result as the target operating parameter.

[0020] In an implementable manner, the determining the target control data of the flight vehicle based on the target operating state includes:

[0021] Construct an objective function for evaluating the operating state of the aircraft;

[0022] Construct the constraint conditions for the operation of the aircraft;

[0023] Solve the objective function based on the constraint conditions to obtain the target control data of the aircraft.

[0024] In an implementable embodiment, the method further includes:

[0025] Construct a training sample set based on historical environmental data and historical control data of the aircraft;

[0026] Input the training sample set into a first machine learning model to obtain a first predicted operating state of the aircraft output by the first machine learning model;

[0027] Adjust the parameters of the first machine learning model based on the first predicted operating state and the historical true operating state of the aircraft corresponding to the training sample set.

[0028] In an implementable embodiment, the method further includes:

[0029] Based on the historical control data, use the second prediction model to determine a second predicted operating state of the aircraft;

[0030] Determine a first residual value between the first predicted operating state and the historical true operating state of the aircraft;

[0031] Determine a second residual value between the second predicted operating state and the historical true operating state of the aircraft;

[0032] Adjust a first weight and a second weight for determining the target operating state based on the first residual value and the second residual value; wherein, the first weight corresponds to the first predicted operating state, and the second weight corresponds to the second predicted operating state.

[0033] In an implementable embodiment, the adjusting the first weight and the second weight for determining the target operating state based on the first residual value and the second residual value includes:

[0034] Calculate a first function result of an exponential function with e as the base and the opposite of the first residual value as the exponent;

[0035] Determine the first function result as a first candidate weight;

[0036] Calculate a second function result of an exponential function with e as the base and the opposite of the second residual value as the exponent;

[0037] Determine the second function result as a second candidate weight;

[0038] Normalize the first candidate weight to obtain the first weight;

[0039] Normalize the second candidate weight to obtain the second weight.

[0040] In an implementable embodiment, the method further includes:

[0041] Determine the actual operating state corresponding to the target operating state;

[0042] Determine the third residual value between the first operating state and the actual operating state;

[0043] Determine the fourth residual value between the second operating state and the actual operating state;

[0044] Based on the third residual value and the fourth residual value, adjust the first weight and the second weight used to determine the target operating state.

[0045] According to the second aspect of the present application, there is provided an aircraft control device, the device includes:

[0046] An acquisition module, configured to acquire environmental data and first control data of the aircraft;

[0047] A first determination module, configured to determine the first operating state of the aircraft by using a first machine learning model based on the environmental data and the first control data; determine the second operating state of the aircraft by using a second prediction model based on the first control data; determine the target operating state of the aircraft based on the first operating state and the second operating state, where the target operating state includes at least one or more of the speed, position, and attitude of the aircraft during operation;

[0048] A second determination module, configured to determine the target control data of the aircraft based on the target operating state; the target control data is used to control the aircraft, and the target control parameter includes at least the speed and / or turning angle of the aircraft.

[0049] In an implementable embodiment, the first determination module is specifically configured to perform a weighted sum operation on the first operating parameter corresponding to the first operating state and the second operating parameter corresponding to the second operating state to obtain a target operating parameter; the target operating parameter corresponds to the target operating state.

[0050] In an implementable embodiment, the first determination module is specifically configured to multiply the first operating parameter by the first weight corresponding to the first operating parameter to obtain a first operation result;

[0051] The second operation parameter is multiplied by a second weight corresponding to the second operation parameter to obtain a second operation result;

[0052] It is determined that the sum of the first operation result and the second operation result is the target operation parameter.

[0053] In an implementable manner, the second determination module is specifically configured to construct an objective function for evaluating the operating state of the aircraft; construct constraint conditions for the operation of the aircraft; solve the objective function based on the constraint conditions to obtain target control data of the aircraft.

[0054] In an implementable manner, the device further includes a training module, which is specifically configured to construct a training sample set based on historical environment data and historical control data of the aircraft;

[0055] The training sample set is input into a first machine learning model to obtain a first predicted operating state of the aircraft output by the first machine learning model;

[0056] Based on the first predicted operating state and the historical true operating state of the aircraft corresponding to the training sample set, the parameters of the first machine learning model are adjusted.

[0057] In an implementable manner, the training module is further configured to determine a second predicted operating state of the aircraft based on the historical control data by using the second prediction model;

[0058] Determine a first residual value between the first predicted operating state and the historical true operating state;

[0059] Determine a second residual value between the second predicted operating state and the historical true operating state;

[0060] Based on the first residual value and the second residual value, the first weight and the second weight for determining the target operating state are adjusted; wherein, the first weight corresponds to the first predicted operating state, and the second weight corresponds to the second predicted operating state.

[0061] In an implementable manner, the training module is specifically configured to calculate a first function result of an exponential function with e as the base and the opposite of the first residual value as the exponent;

[0062] Determine the first function result as a first candidate weight;

[0063] Calculate a second function result of an exponential function with e as the base and the opposite of the second residual value as the exponent;

[0064] Determine the second function result as a second candidate weight;

[0065] Normalize the first candidate weight to obtain the first weight;

[0066] Normalize the second candidate weight to obtain the second weight.

[0067] In an implementable manner, the device further includes: an adjustment module configured to determine an actual operating state corresponding to the target operating state;

[0068] Determine a third residual value between the first operating state and the actual operating state;

[0069] Determine a fourth residual value between the second operating state and the actual operating state;

[0070] Based on the third residual value and the fourth residual value, adjust the first weight and the second weight for determining the target operating state.

[0071] According to a third aspect of the present application, there is provided an aircraft control device, the device including:

[0072] At least one processor; and

[0073] A memory communicatively connected to the at least one processor; wherein,

[0074] The memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to execute the aircraft control method of the present application.

[0075] According to a fourth aspect of the present application, there is provided a computer-readable storage medium storing computer instructions, the computer instructions being used to cause a computer to execute the method of the present application.

[0076] According to a fifth aspect of the present application, there is provided a computer program product, including a computer program or instructions, storing executable instructions, and when being executed by a processor, implementing the aircraft control method of the present application.

[0077] According to a sixth aspect of the present application, there is provided a computer program product, the computer program product including a computer program / instructions, and when the computer program / instructions are executed by a processor, implementing the aircraft control of the present application.

[0078] The present application provides a flight vehicle control method, apparatus, flight vehicle, and storage medium. The method includes: obtaining environmental data and first control data of the flight vehicle; determining a first operating state of the flight vehicle by using a first machine learning model based on the environmental data and the first control data; determining a second operating state of the flight vehicle by using a second prediction model based on the first control data; determining a target operating state of the flight vehicle based on the first operating state and the second operating state; determining target control data of the flight vehicle based on the target operating state; and using the target control data to control the flight vehicle. By fusing the first operating state predicted by the first machine learning model with the second operating state predicted by the second prediction model to obtain the target operating state, the problems of inaccurate prediction results and poor robustness of the operating state of the flight vehicle under the condition of environmental interference are solved; thereby, the stability and safety of the flight vehicle are improved.

[0079] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] By referring to the drawings and reading the detailed description below, the above and other objects, features, and advantages of the exemplary embodiments of the present application will become easily understandable. In the drawings, several embodiments of the present application are shown in an exemplary rather than restrictive manner, where:

[0081] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts.

[0082] Figure 1 shows a schematic processing flow diagram of a flight vehicle control method provided by an embodiment of the present application;

[0083] Figure 2 shows an optional schematic processing flow diagram for determining a target operating state provided by an embodiment of the present application;

[0084] Figure 3 shows a specific implementation flow diagram for determining the target control data of the flight vehicle based on the target operating state provided by an embodiment of the present application;

[0085] Figure 4 shows a training flow diagram of a first machine learning model provided by an embodiment of the present application;

[0086] Figure 5 shows a specific flow diagram for adjusting a first weight and a second weight provided by an embodiment of the present application;

[0087] Figure 6Shows a schematic diagram of a specific implementation process for adjusting the first weight and the second weight used to determine the target operating state based on the first residual value and the second residual value provided by an embodiment of the present application;

[0088] Figure 7 Shows another schematic diagram of a specific implementation process for adjusting the first weight and the second weight used to determine the target operating state based on the first residual value and the second residual value provided by an embodiment of the present application;

[0089] Figure 8 Shows a schematic diagram of the composition structure of an aircraft control device provided by an embodiment of the present application;

[0090] Figure 9 Shows a schematic diagram of a hardware structure of a flight control device provided by an embodiment of the present application. Detailed implementation manners

[0091] To make the objectives, features, and advantages of the present application more obvious and understandable, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0092] To make the objectives, technical solutions, and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limitations on the present application. All other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0093] In the following descriptions, "some embodiments" are involved, which describe subsets of all possible embodiments. However, it can be understood that "some embodiments" can be the same subsets or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0094] In the following descriptions, the terms "first / second" involved are only used to distinguish similar objects and do not represent a specific order for the objects. It can be understood that "first / second" can be interchanged with a specific order or sequence when allowed, so that the embodiments of the present application described here can be implemented in an order other than that illustrated or described here.

[0095] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which this application belongs. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0096] It should be understood that in various embodiments of this application, the magnitude of the sequence number of each implementation process does not mean the sequence of execution order. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of this application.

[0097] A schematic diagram of a processing flow of the aircraft control method provided by an embodiment of this application is shown as Figure 1 shown, and at least includes the following steps:

[0098] Step S101, obtain environmental data and the first control data of the aircraft.

[0099] In some embodiments, parameters such as the position of real-time obstacles, wind speed, and turbulence intensity (such as wind speed and obstacle position) in the environment can be obtained through an on-board lidar, a weather sensor, etc.

[0100] In some embodiments, the first control data may be the speed and throttle amount at the current moment, and the value of the throttle amount is between 0-100%; the turning angle and pitch angle of the aircraft, and the pitch angle can be between -15° and +15°; the roll angle of the aircraft, and the roll angle can be between -20° and +20°.

[0101] Step S102, based on the environmental data and the first control data, use a first machine learning model to determine the first operating state of the aircraft.

[0102] In some embodiments, the first machine learning model may be a time series prediction model based on LSTM; the specific form of the first machine learning model in the embodiments of this application is not limited.

[0103] In some embodiments, the environmental data and the first control data are input into the first machine learning model, and the first machine learning model outputs the first operating state of the aircraft. Among them, the first operating state output by the first machine learning model is the operating state of the aircraft predicted by the first machine learning model according to the environmental data and the first control data.

[0104] In some embodiments, the first operating state may include one or more of the predicted speed, position, and attitude of the aircraft.

[0105] Step S103, based on the first control data, use a second prediction model to determine the second operating state of the aircraft.

[0106] In some embodiments, the second prediction model may be a state transition model or the like; the specific form of the second prediction model in the embodiments of the present application is not limited. As an example, the second prediction model may be a quadrotor kinematic model, and the quadrotor kinematic model may be as follows:

[0107]

[0108] Where x, y, and z are the position coordinates of the aircraft in three-dimensional space. φ, θ, and ψ are the roll angle (Roll), pitch angle (Pitch), and yaw angle (Yaw) of the aircraft, usually in radians (rad). And Are the linear accelerations of the aircraft in three directions, usually in meters per second squared (m / s 2 ) as the unit. And Are the angular accelerations of the aircraft in three directions, usually in radians per second squared (rad / s 2 ) as the unit. U1 is the total thrust, usually in Newtons (N). U2, U3, and U4 respectively represent the torques about the x-axis, y-axis, and z-axis, usually in Newton meters (N·m). m represents the mass of the aircraft, usually in kilograms (kg). Ixx, Iyy, and Izz are the moments of inertia of the aircraft about the x-axis, y-axis, and z-axis, usually in kilograms per square meter. p, q, and r respectively represent the angular velocities about the x-axis, y-axis, and z-axis, usually in radians per second (rad / s). J RPQ Represents a parameter related to the rotor moment of inertia, usually in kilograms per square meter. Ω is the total rotational speed of the rotor, usually in radians per second (rad / s). g is the acceleration due to gravity, usually taken as 9.81 meters per second squared (m / s 2 ).

[0109] In some embodiments, the first control data is input into the second prediction model, and the second prediction model outputs the second operating state of the aircraft. Among them, the second operating state output by the second prediction model is the operating state of the aircraft predicted by the second prediction model according to the first control data.

[0110] Step S104, determining the target operating state of the aircraft based on the first operating state and the second operating state.

[0111] In some embodiments, the first operating state output by the first machine learning model corresponds to the first operating parameter of the aircraft, and the second operating state output by the second prediction model corresponds to the second operating parameter of the aircraft. A weighted sum operation is performed on the first operating parameter corresponding to the first operating state and the second operating parameter corresponding to the second operating state to obtain a target operating parameter; the target operating parameter corresponds to the target operating state. Wherein, the weight of the first operating parameter is the first weight, and the weight of the second operating parameter is the second weight.

[0112] In some embodiments, a schematic diagram of an alternative processing flow for determining the target operating state is shown in Figure 2 and includes at least the following steps:

[0113] Step S104a: Multiply the first operating parameter by the first weight corresponding to the first operating parameter to obtain a first operation result.

[0114] In some embodiments, if the first weight is represented by w1, then:

[0115] First operation result = first operating parameter * w1.

[0116] Step S104b: Multiply the second operating parameter by the second weight corresponding to the second operating parameter to obtain a second operation result.

[0117] In some embodiments, if the second weight is represented by w2, then:

[0118] Second operation result = second operating parameter * w2.

[0119] Step S104c: Determine the sum of the first operation result and the second operation result as the target operating parameter.

[0120] In some embodiments, target operating parameter = first operating parameter * w1 + second operating parameter * w2.

[0121] Step S105: Determine the target control data of the aircraft based on the target operating state; the target control data is used to control the aircraft.

[0122] In some embodiments, a schematic diagram of the specific implementation process for determining the target control data of the aircraft based on the target operating state is shown in Figure 3 and includes at least the following steps:

[0123] Step S105a: Construct a target function for evaluating the operating state of the aircraft.

[0124] In some embodiments, the target function can be:

[0125]

[0126] Where Δu is the change in the control quantity, and λ is the smoothing coefficient.

[0127] The above formula (1) is only an example of the objective function. In specific implementation, the objective function can be other forms of functions constructed according to the actual application scenario and actual application requirements.

[0128] Step S105b, construct the constraint conditions for the operation of the aircraft.

[0129] In some embodiments, the constraint conditions for the operation of the aircraft may at least include one or more of the following: safety distance, maximum pitch angle, energy consumption limit, flight altitude, maximum turning angle, and speed limit.

[0130] As an example, the maximum safety distance can be: the distance from the obstacle ≥ 5 meters. The maximum pitch angle can be: ±15°. The energy consumption limit can be: the total throttle change ≤ 30%. The flight altitude can be greater than or equal to 50m to avoid low-altitude obstacles. The maximum turning angle can be 30°, and the speed range can be between 5m / s and 25m / s.

[0131] The numerical values of the above constraint conditions are only an example. In specific implementation, they can be set to different values according to the actual application scenario and actual application requirements, and the embodiments of the present application do not make limitations.

[0132] Step S105c, solve the objective function based on the constraint conditions to obtain the target control data of the aircraft.

[0133] In some embodiments, the quadratic programming (QP) algorithm can be used to solve the optimal control sequence; specifically, the objective function can be converted into the standard QP form and solved using the interior point method to obtain the optimal control quantity for a future period of time, and this optimal control quantity is the target control data of the aircraft, such as the speed adjustment amount, turning angle command, etc.

[0134] The above has described in detail the specific implementation process of the aircraft control method provided by the embodiments of the present application. Next, the first machine learning model adopted by the embodiments of the present application will be described.

[0135] In the embodiments of the present application, the schematic diagram of the training process of the first machine learning model is as Figure 4 shown, and at least includes the following steps:

[0136] Step S201, construct a training sample set based on historical environmental data and historical control data of the aircraft.

[0137] In some embodiments, the training sample set includes 10,000 sets of data; each set of data includes historical environment data and historical control data; wherein, the historical environment data may include wind speed and wind direction; the historical control data may include specific data of throttle and steering angle. The training sample set further includes the historical true operating state of the aircraft, such as Global Positioning System (GPS) coordinates and Inertial Measurement Unit (IMU) data, and the IMU data may include the acceleration and angular velocity of the aircraft.

[0138] Step S202: Input the training sample set into the first machine learning model to obtain the first predicted operating state of the aircraft output by the first machine learning model.

[0139] In some embodiments, the data in the training sample set can be sequentially input into the first machine learning model, and the first machine learning model predicts the corresponding operating state for each set of training samples.

[0140] Step S203: Based on the first predicted operating state and the historical true operating state of the aircraft corresponding to the training sample set, adjust the parameters of the first machine learning model.

[0141] In some embodiments, for each set of training samples, compare the difference between the first predicted operating state predicted by the first machine learning model for the training sample and the historical true operating state corresponding to the training sample; based on the difference, adjust the parameters of the first machine learning model until the difference between the first predicted operating state and the corresponding historical true operating state meets a preset condition, then the training of the first machine learning model is completed.

[0142] In the embodiments of the present application, the first weight and the second weight can be adjusted based on the same training sample set.

[0143] In the embodiments of the present application, the specific process schematic diagram for adjusting the first weight and the second weight is as Figure 5 shown, and at least includes the following steps:

[0144] Step S301: Based on the historical environment data and the historical control data of the aircraft, use the first machine learning model to determine the first predicted operating state of the aircraft.

[0145] Step S302: Based on the historical control data, use the second prediction model to determine the second predicted operating state of the aircraft.

[0146] Step S303: Determine the first residual value between the first predicted operating state and the historical true operating state.

[0147] In some embodiments, take the sliding time window as 5 seconds and the time step Δt as 0.1 second as an example.

[0148] The first residual value e1 can be expressed as:

[0149]

[0150] where x 1t is the first operating state at time t, and x 实际，t is the actual operating state at time t.

[0151] Step S304, determine the second residual value between the second predicted operating state and the historical true operating state.

[0152] In some embodiments, take the sliding time window as 5 seconds and the time step Δt as 0.1 second as an example.

[0153] The first residual value e2 can be expressed as:

[0154]

[0155] where x 2t is the second operating state at time t, and x 实际，t is the actual operating state at time t.

[0156] Step S305, adjust the first weight and the second weight used to determine the target operating state based on the first residual value and the second residual value.

[0157] In some embodiments, a schematic diagram of a specific implementation process for adjusting the first weight and the second weight used to determine the target operating state based on the first residual value and the second residual value is shown as Figure 6 shown, and at least includes the following steps:

[0158] Step S401, calculate the first function result of the exponential function with e as the base and the opposite number of the first residual value as the exponent, and determine the first function result as the first candidate weight.

[0159] In some embodiments, the first candidate weight is represented by w1′.

[0160] w1′ = exp(-e1) (4)

[0161] Step S402, calculate the second function result of the exponential function with e as the base and the opposite number of the second residual value as the exponent, and determine the second function result as the second candidate weight.

[0162] In some embodiments, the second candidate weight is represented by w2′.

[0163] w2′ = exp(-e2) (5)

[0164] Step S403, perform normalization processing on the first candidate weight to obtain the first weight.

[0165] In some embodiments, the first weight is denoted by w1.

[0166] w1 = w1′ / (w1′ + w2′) (6)

[0167] Step S404, perform normalization processing on the second candidate weight to obtain the second weight.

[0168] In some embodiments, the second weight is denoted by w2.

[0169] w2 = w2′ / (w1′ + w2′) (7)

[0170] As an example, if e1 = 0.5 and e2 = 0.2, then:

[0171] w1′ ≈ 0.6065, w2′ ≈ 0.8187; after normalization, w1 ≈ 0.43, w2 ≈ 0.57.

[0172] In some embodiments, based on the first residual value and the second residual value, another specific implementation flowchart for adjusting the first weight and the second weight used to determine the target operating state is as follows Figure 7 As shown, after performing step S105, at least the following steps are executed:

[0173] Step S501, determine the actual operating state corresponding to the target operating state.

[0174] In some embodiments, after predicting the target operating state in step S105, the actual operating state of the aircraft can be determined. Specifically, the actual operating state of the aircraft can be collected every 0.05 seconds.

[0175] Step S502, determine the third residual value between the first operating state and the actual operating state.

[0176] In some embodiments, the third residual value can be expressed as:

[0177] r1 = ||x 1t -x 实际，t || (8)

[0178] where x 1t is the first operating state at time t, and x 实际，t is the actual operating state at time t.

[0179] Step S503, determine the fourth residual value between the second operating state and the actual operating state.

[0180] In some embodiments, the fourth residual value can be expressed as:

[0181] r2 = ||x 2t -x 实际，t || (9)

[0182] where x 2t is the second operating state at time t, and x 实际，t is the actual operating state at time t.

[0183] Step S504, based on the third residual value and the fourth residual value, adjust the first weight and the second weight used to determine the target operating state.

[0184] In some embodiments, the residuals of the most recent 10 time steps can be accumulated to obtain the third residual e3 and the fourth residual e4.

[0185]

[0186] In some embodiments, after calculating the third residual and the fourth residual according to the above steps, the first weight and the second weight can be adjusted based on Figure 6 the weight adjustment method described above.

[0187] It should be noted that in the above embodiments of the present application, the first weight and the second weight are adjusted based on the residuals; in specific implementation, other methods can also be used to adjust the first weight and the second weight, and the specific methods for adjusting the first weight and the second weight are not limited in the embodiments of the present application.

[0188] In the embodiments of the present application, based on the residuals between the first operating state and the second operating state predicted by the first machine learning model and the second prediction model and the target operating state respectively, the first weight corresponding to the first operating state and the second weight corresponding to the second operating state are adaptively adjusted, further improving the control accuracy of the aircraft and ensuring the stable operation of the aircraft in a dynamic environment.

[0189] Next, the implementation manners of the aircraft control method provided in the embodiments of the present application in different application scenarios will be illustrated by examples.

[0190] Application scenario 1, an aircraft control method in an urban air traffic scenario.

[0191] In a dense urban air traffic environment, an aircraft needs to respond to dynamic obstacles (such as other aircraft, drones) and complex airflow disturbances in real time, while maintaining the tracking accuracy of a predetermined flight path. When implementing an aircraft control method, environmental data such as the real-time position of obstacles, wind speed, and turbulence intensity can be obtained through on-board lidar and meteorological sensors. The control data includes: the throttle amount (0 - 100%) at the current moment, the pitch angle (-15° to +15°), and the roll angle (-20° to +20°).

[0192] The first machine learning model can utilize the environmental data and control data to predict the first operating state (position, velocity, and attitude) of the aircraft within the next 5 seconds based on the existing quadrotor dynamics equations. Among them, the first machine learning model can adopt an LSTM network, with the input being the historical control data, historical environmental data, and historical operating state in the past 3 seconds, and the output being the prediction sequence of the operating state within the next 5 seconds. The LSTM network structure can be 3-layer LSTM (128 units per layer).

[0193] The second prediction model can utilize the control data to predict the first operating state (position, velocity, and attitude) of the aircraft within the next 5 seconds.

[0194] Perform a weighted sum of the first operating state and the second operating state to obtain a fused operating state; the position in the target operating state to be solved can be represented by x 目标 and the velocity in the target operating state can be represented by v 目标 Based on this, construct an objective function, which is expressed as:

[0195]

[0196] where Δu is the adjustment amount of the throttle and angle.

[0197] The constraint conditions of this objective function are: (1) Safety distance: the distance from obstacles ≥ 5 meters; (2)

[0198] Maximum pitch angle: ±15°; (3) Energy consumption limit: the total throttle change amount ≤ 30%.

[0199] Solving the objective function based on this constraint condition can obtain an optimal control data sequence for a future period of time. Any one in the optimal control sequence can be selected as the current control instruction; or the first item in the optimal control sequence can be selected as the current control instruction to achieve rolling optimization.

[0200] Through the embodiments of this application, in the scenario where the aircraft suddenly encounters a crosswind, the trajectory offset amount can be smaller than that when only using the second prediction model to determine the target operating state.

[0201] Application scenario two, real-time control in an emergency obstacle avoidance scenario.

[0202] When the aircraft encounters dynamic obstacles (such as a flock of birds) during low-altitude flight, it needs to generate an emergency obstacle avoidance path within 0.5 seconds. The amount of data corresponding to emergency obstacle avoidance in the training dataset used by the first machine learning model may be small, and it fails to well predict the changes in future states; while the state transition equation in the second prediction model can achieve better prediction results. Therefore, the values of the first weight and the second weight can be adjusted, and then the control data is preferentially generated depending on the operating state output by the second prediction model to improve the success rate of obstacle avoidance.

[0203] Based on Application Scenario 1 and Application Scenario 2, it can be seen that this application can adjust the first weight and the second weight in real time according to the actual environmental data and control data to improve the robustness.

[0204] The aircraft described in the embodiments of this application refers to a device capable of flying in the air, including but not limited to flying cars, drones, jet packs, flapping-wing aircraft, etc.

[0205] The embodiments of this application also provide an aircraft control device. The schematic structural diagram of the composition of the aircraft control device is as Figure 8 shown and includes:

[0206] An acquisition module 601, configured to acquire environmental data and first control data of the aircraft;

[0207] A first determination module 602, configured to determine a first operating state of the aircraft by using a first machine learning model based on the environmental data and the first control data; determine a second operating state of the aircraft by using a second prediction model based on the first control data; and determine a target operating state of the aircraft based on the first operating state and the second operating state, where the target operating state includes at least one or more of the speed, position, and attitude of the aircraft during operation;

[0208] A second determination module 603, configured to determine target control data of the aircraft based on the target operating state; the target control data is used to control the aircraft, and the target control parameters include at least the speed and / or turning angle of the aircraft.

[0209] In an implementable manner, the first determination module 602 is specifically configured to perform a weighted sum operation on a first operating parameter corresponding to the first operating state and a second operating parameter corresponding to the second operating state to obtain a target operating parameter; the target operating parameter corresponds to the target operating state.

[0210] In an implementable manner, the first determination module 602 is specifically configured to multiply the first operating parameter by a first weight corresponding to the first operating parameter to obtain a first operation result;

[0211] The second operating parameter is multiplied by a second weight corresponding to the second operating parameter to obtain a second operating result;

[0212] It is determined that the sum of the first operation result and the second operation result is the target operating parameter.

[0213] In an implementable manner, the second determination module 603 is specifically configured to construct an objective function for evaluating the operating state of the aircraft; construct constraint conditions for the operation of the aircraft; solve the objective function based on the constraint conditions to obtain target control data of the aircraft.

[0214] In an implementable manner, the device further includes a training module (not shown in the figure), which is specifically configured to construct a training sample set based on historical environment data and historical control data of the aircraft;

[0215] The training sample set is input into a first machine learning model to obtain a first predicted operating state of the aircraft output by the first machine learning model;

[0216] Based on the first predicted operating state and the historical true operating state of the aircraft corresponding to the training sample set, the parameters of the first machine learning model are adjusted.

[0217] In an implementable manner, the training module is further configured to determine a second predicted operating state of the aircraft by using the second prediction model based on the historical control data;

[0218] Determine a first residual value between the first predicted operating state and the historical true operating state;

[0219] Determine a second residual value between the second predicted operating state and the historical true operating state;

[0220] Based on the first residual value and the second residual value, the first weight and the second weight for determining the target operating state are adjusted; wherein, the first weight corresponds to the first predicted operating state, and the second weight corresponds to the second predicted operating state.

[0221] In an implementable manner, the training module is specifically configured to calculate a first function result of an exponential function with e as the base and the opposite of the first residual value as the exponent;

[0222] Determine the first function result as the first candidate weight;

[0223] Calculate a second function result of an exponential function with e as the base and the opposite of the second residual value as the exponent;

[0224] Determine that the second function result is the second candidate weight;

[0225] Normalize the first candidate weight to obtain the first weight;

[0226] Normalize the second candidate weight to obtain the second weight.

[0227] In an implementable embodiment, the device further includes an adjustment module (not shown in the figure) for determining the actual operating state corresponding to the target operating state;

[0228] Determine the third residual value between the first operating state and the actual operating state;

[0229] Determine the fourth residual value between the second operating state and the actual operating state;

[0230] Based on the third residual value and the fourth residual value, adjust the first weight and the second weight for determining the target operating state.

[0231] According to an embodiment of the present application, the present application further provides an aircraft control device. Wherein, the aircraft control device includes at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the aircraft control method described in the present application. The computer instructions are used to cause the computer to execute the aircraft control method described in the present application.

[0232] An embodiment of the present application further provides an aircraft, on which a flight control device can be provided. A schematic hardware structure diagram of the flight control device is shown as Figure 9 shown, the flight control device at least includes: a processor 410, a memory 450, and a bus 440; each module in the flight control device is coupled together through the bus 440. It can be understood that the bus 440 is used to realize the connection and communication between these modules. In addition to the data bus, the bus 440 also includes a power bus, a control bus, and a status signal bus. However, for the sake of clarity, in Figure 9 all kinds of buses are labeled as bus 440. The processor 410 has the ability to process signals, such as a general-purpose processor, a digital signal processor (DSP, Digital Signal Processor), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware models, etc. Among them, the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0233] The memory 450 can be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid state memories, hard disk drives, optical disk drives, etc. The memory 450 optionally includes one or more storage devices that are physically remote from the processor 410.

[0234] In some embodiments, the memory 450 is capable of storing data to support various operations, examples of which include programs, modules, and data structures, or subsets or supersets thereof.

[0235] In some embodiments, the flight control device may further include:

[0236] An operating system 451, including system programs for handling various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and handling hardware-based tasks;

[0237] A network communication module 452, for reaching other computing devices via one or more (wired or wireless) network interfaces 420. Exemplary network interfaces 420 include: Bluetooth, Wireless Fidelity (WiFi), and Universal Serial Bus (USB), etc.

[0238] Embodiments of the present application provide a computer-readable storage medium storing executable instructions, where the executable instructions, when executed by a processor, will trigger the processor to execute the aircraft control method provided by the embodiments of the present application. For example, as Figures 1 to 7 shown in the aircraft control method.

[0239] In some embodiments, the computer-readable storage medium may be a ferroelectric random access memory (FRAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a flash memory, a magnetic surface memory, an optical disk, or a CD-ROM, etc.; or it may be various devices including one or any combination of the above memories.

[0240] In some embodiments, the executable instructions may be in the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including being deployed as a stand-alone program or being deployed as a module, model, subroutine, or other unit suitable for use in a computing environment.

[0241] As an example, the executable instructions may be deployed to execute on one computing device, or on multiple computing devices located at one location, or alternatively, on multiple computing devices distributed at multiple locations and interconnected by a communication network.

[0242] An embodiment of the present application provides a computer program product, the computer program product includes a computer program / instructions, and when the computer program / instructions are executed by a processor, the aircraft control method described in the present application is implemented.

[0243] As described above, the above are only the specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for controlling an aircraft, characterized in that, The method includes: Obtaining environmental data and first control data of the aircraft; Based on the environmental data and the first control data, determining a first operating state of the aircraft using a first machine learning model; Based on the first control data, determining a second operating state of the aircraft using a second prediction model; Determining a target operating state of the aircraft based on the first operating state and the second operating state, where the target operating state includes at least one or more of the speed, position, and attitude of the aircraft during operation; Determining target control data of the aircraft based on the target operating state; the target control data is used to control the aircraft, and the target control parameters include at least the speed and / or turning angle of the aircraft.

2. The method according to claim 1, wherein The determining the target operating state of the aircraft based on the first operating state and the second operating state includes: Performing a weighted sum operation on a first operating parameter corresponding to the first operating state and a second operating parameter corresponding to the second operating state to obtain a target operating parameter; the target operating parameter corresponds to the target operating state.

3. The method according to claim 2, characterized in that, The performing a weighted sum operation on the first operating parameter corresponding to the first operating state and the second operating parameter corresponding to the second operating state to obtain a target operating parameter includes: Multiplying the first operating parameter by a first weight corresponding to the first operating parameter to obtain a first operation result; Multiplying the second operating parameter by a second weight corresponding to the second operating parameter to obtain a second operation result; Determining the sum of the first operation result and the second operation result as the target operating parameter.

4. The method according to claim 1, characterized in that The determining the target control data of the aircraft based on the target operating state includes: Constructing an objective function for evaluating the operating state of the aircraft; Constructing constraint conditions for the operation of the aircraft; Solving the objective function based on the constraint conditions to obtain the target control data of the aircraft.

5. The method according to claim 1, characterized in that The method further includes: Constructing a training sample set based on historical environmental data and historical control data of the aircraft; Inputting the training sample set into the first machine learning model to obtain a first predicted operating state of the aircraft output by the first machine learning model; Adjusting the parameters of the first machine learning model based on the first predicted operating state and the historical true operating state of the aircraft corresponding to the training sample set.

6. The method according to claim 5, characterized in that The method further includes: Determining a second predicted operating state of the aircraft using the second prediction model based on the historical control data; Determining a first residual value between the first predicted operating state and the historical true operating state; Determining a second residual value between the second predicted operating state and the historical true operating state; Adjusting a first weight and a second weight for determining the target operating state based on the first residual value and the second residual value; where the first weight corresponds to the first predicted operating state and the second weight corresponds to the second predicted operating state.

7. The method according to claim 6, wherein The adjusting the first weight and the second weight for determining the target operating state based on the first residual value and the second residual value includes: Calculate a first function result of an exponential function with base e and exponent being the opposite of the first residual value; Determine the first function result as a first candidate weight; Calculate a second function result of an exponential function with base e and exponent being the opposite of the second residual value; Determine the second function result as a second candidate weight; Normalize the first candidate weight to obtain the first weight; Normalize the second candidate weight to obtain the second weight.

8. The method according to claim 1, wherein The method further includes: Determine an actual operating state corresponding to the target operating state; Determine a third residual value between the first operating state and the actual operating state; Determine a fourth residual value between the second operating state and the actual operating state; Based on the third residual value and the fourth residual value, adjust the first weight and the second weight used to determine the target operating state.

9. An aircraft control device, characterized in that, The device includes: An acquisition module, configured to acquire environmental data and first control data of the aircraft; A first determination module, configured to determine a first operating state of the aircraft by using a first machine learning model based on the environmental data and the first control data; determine a second operating state of the aircraft by using a second prediction model based on the first control data; determine a target operating state of the aircraft based on the first operating state and the second operating state, where the target operating state includes at least one or more of the speed, position, and attitude of the aircraft during operation; A second determination module, configured to determine target control data of the aircraft based on the target operating state; the target control data is used to control the aircraft, and the target control parameters include at least the speed and / or turning angle of the aircraft.

10. An aircraft control device, characterized in that, The device includes: At least one processor; A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to execute the aircraft control method according to any one of claims 1 to 8.

11. An aircraft, characterized in that, Apply the aircraft control method according to any one of claims 1 to 8.

12. A computer-readable storage medium, characterized in that, Store executable instructions, which are used to implement the aircraft control method according to any one of claims 1 to 8 when executed by a processor.

13. A computer program product, characterized in that, The computer program product includes computer programs / instructions, and when the computer programs / instructions are executed by a processor, the aircraft control method according to any one of claims 1 to 8 is implemented.

Images