Unmanned aerial vehicle offshore platform landing method based on autoregression model prediction

Abstract

The invention relates to an unmanned aerial vehicle offshore platform landing method based on autoregression model prediction. The method comprises the following steps that environment data are collected based on a sensor arranged on an unmanned aerial vehicle; relative position information between the unmanned aerial vehicle and the target landing platform and real-time position information of the unmanned aerial vehicle are determined through the collected environment data, and the unmanned aerial vehicle is controlled to fly to a first pre-landing position and keep stable hovering based on the position information; based on an inertial navigation system equipped on the target landing platform, collecting motion attitude data of the target landing platform, and performing future time period prediction of the attitude of the target landing platform through autoregression model prediction; and judging whether a preset landing condition is satisfied based on the predicted attitude of the target landing platform and the attitude of the unmanned aerial vehicle, the attitude including position information and attitude angle information, transmitting a landing instruction to the unmanned aerial vehicle when the preset landing condition is satisfied, and controlling the unmanned aerial vehicle to complete landing operation. Compared with the prior art, the method has the advantages of improving the landing reliability and stability of the unmanned aerial vehicle and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of offshore platforms, and in particular to a method for landing an unmanned aerial vehicle on an offshore platform based on autoregressive model prediction. Background Art

[0002] With the widespread application of drones and the increasing demand for offshore drones, the problem of drone landing on offshore platforms has become a key issue in the field of offshore platform technology.

[0003] At present, the technical problems existing in the traditional UAV landing method include: the interference of platform activities on the landing of the UAV, the posture prediction problem of offshore platform activities, and the timing selection problem of the landing of the UAV on the offshore platform. For example, patent CN108227751A discloses a landing method and system for a UAV, which uses visual recognition and combines the GPS position information sent by the mobile carrier for combined navigation, thereby effectively dealing with the problem that the mobile carrier is too fast and causes the visual system to lose the target and cannot track quickly; the position-time relationship and posture-time relationship of the landing platform are estimated by the digital information filtering fusion algorithm, and the prediction function of the mobile carrier on "posture, position-time" is obtained, so that the flight target of the UAV is updated in real time according to the prediction function of the mobile carrier on "posture, position-time", and the flight target includes position, speed, and posture, so as to accurately control the landing point, landing speed and landing time of the UAV, thereby solving the problem that the ship floats on the water surface and causes a three-dimensional offset, and the landing platform on the deck surface will also continuously tilt left and right, causing the UAV to land inaccurately or even unable to land stably. However, this solution only uses visual recognition, the prediction accuracy of the posture is low, and the stability and reliability of the UAV landing are not considered. Summary of the invention

[0004] The purpose of the present invention is to provide a method for landing a UAV on an offshore platform based on autoregressive model prediction, so as to achieve reliable and smooth landing control of the UAV.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] A method for landing a UAV on an offshore platform based on autoregressive model prediction includes the following steps:

[0007] S1, collects environmental data based on the sensors equipped by the drone;

[0008] S2, using the environmental data collected by the sensor to determine the relative position information between the UAV and the target landing platform and the real-time position information of the UAV, and based on the position information, controlling the UAV to fly to a first pre-landing position and maintain a stable hover;

[0009] S3, based on the inertial navigation system equipped with the target landing platform, collect its motion posture data, and predict the future time period of the target landing platform posture through autoregressive model prediction;

[0010] S4, based on the predicted posture of the target landing platform and the posture of the drone collected by the sensor in step S1, it is determined whether the preset landing conditions are met, and the posture includes position information and posture angle information. When the preset landing conditions are met, a landing instruction is transmitted to the drone to control the drone to complete the landing operation.

[0011] The sensor includes a GPS and a depth camera. In step S1, collecting environmental data specifically includes: using the depth camera to search for the position of the target landing platform, and determining the relative position of the target landing platform in the body coordinate system with the drone body as the center; obtaining the attitude angle information of the drone based on the inertial navigation system equipped in the depth camera; and determining the real-time coordinate position information of the drone in the GPS coordinate system during flight based on the GPS.

[0012] In the step S2, the UAV is controlled to fly to the first pre-landing position based on the position information. Specifically, based on the relative position coordinates (x, y, z) of the target landing platform in the body coordinate system obtained in step S1, the coordinates are adjusted to the preset coordinates (x0, y0, z0) by controlling the UAV to fly, that is, based on the real-time position information of the UAV, the UAV is controlled to fly in the GPS coordinate system according to the vector direction r = (x0-x, y0-y, z0-z) and finally move to the first pre-landing position.

[0013] In step S3, the autoregressive model is a multi-step prediction model constructed based on the motion posture data of the target landing platform, and is used to predict the platform posture within a preset time range in the future. The posture data of the target landing platform includes three-dimensional spatial position information and posture angle information.

[0014] In step S3, the autoregressive model prediction process includes the following steps:

[0015] S31, data reading and preprocessing;

[0016] S32, data standardization;

[0017] S33, autoregressive model training;

[0018] S34, Future state prediction.

[0019] The step S31 comprises the following steps:

[0020] S311, data reading: read the inertial navigation system data located on the target landing platform, including the data of the accelerometer and gyroscope equipped in the inertial navigation, and obtain the acceleration data a of the target landing platform in three directions in the spatial coordinate system x (t),a y (t),a z (t) and angular velocity data ω x (t),ω y (t),ω z (t);

[0021] S312, data preprocessing: by processing the acceleration data a x (t),a y (t),a z (t) is subjected to a second cumulative integration to obtain the spatial position information x(t), y(t), z(t):

[0022] x(t)=∫∫a(t)dtdt

[0023] By angular velocity data ω x (t),ω y (t),ω z (t) Perform a cumulative integration to obtain the attitude angle information θ roll (t),θ pitch (t),θ yaw (t):

[0024] θ(t)=∫ω(t)dt

[0025] Through data preprocessing, the three-dimensional spatial position and attitude angle information of the target landing platform are obtained.

[0026] In step S32, for the data standardization step, in order to eliminate the dimensional differences between different features, the position and attitude data of the target landing platform are processed using a standardization method:

[0027]

[0028] Among them, x is the original data, μ is the mean of the data set, and σ is the standard deviation of the data set.

[0029] Specifically, step S33 is to fit the position information and posture information observed in the past based on the autoregressive model to predict the position information and posture information in the future:

[0030]

[0031] Among them, y(t) is the position information or posture information observed at the corresponding time, φ1, φ2,…, φ pis the model coefficient, p is the lag order, ∈(t) is the noise term;

[0032] According to the different postures, they are the spatial position information x(t), y(t), z(t) and the attitude angle information θ of the target landing platform. roll (t),θ pitch (t),θ yaw (t) Establish their respective autoregressive models.

[0033] In step S34, for the future state prediction step, the historical data of the previous 5 seconds is used as the initial input, and the position and attitude changes of the drone in the next 10 seconds are predicted through the trained autoregressive model, and the prediction results are denormalized to restore the original dimension:

[0034] x=σx′+μ

[0035] During the prediction process, the prediction formula is consistent with the model form in the training process. The state at each moment is gradually predicted in a recursive manner, and the model parameters are adaptively adjusted according to the deviation between the actual information and the predicted information during the prediction process.

[0036] In step S4, at a time Δt from the current time t, when the attitude prediction result of the target landing platform and the attitude information of the UAV both meet the following conditions, it is considered that the preset landing condition is met:

[0037]

[0038] Among them, f(t) represents the predicted target landing platform or the spatial position information or attitude angle information of the collected drone, t0 is the minimum time interval when observing data, n0 is a preset integer not less than 1 and makes n0t0 not less than the minimum stable time required for landing, and eps is the preset stability judgment threshold.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] (1) The present invention uses GPS and a depth camera to locate the position of the UAV and determine the position of the target landing platform, thereby improving the accuracy of the UAV pre-flight operation and landing operation;

[0041] (2) The present invention uses autoregressive model prediction technology to predict and analyze the platform attitude and adds adaptive adjustment to improve the accuracy of the prediction data and increase the reliability of the landing reference;

[0042] (3) The present invention incorporates landing condition judgment and decision-making, thereby increasing the stability of the landing operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1is a flow chart of the method of the present invention;

[0044] Figure 2 A schematic diagram of the sensors provided in the embodiment of the present invention;

[0045] Figure 3 This is a flow chart of target landing platform posture prediction in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0047] This embodiment provides a method for landing a UAV on an offshore platform based on autoregressive model prediction. Figure 1 As shown, the following steps are included:

[0048] S1, collects environmental data based on the sensors equipped by the drone.

[0049] At the beginning of the landing process, the environment data needs to be sensed and collected through the relevant sensors. Specifically, the sensor requirements are as follows: Figure 2 As shown, includes GPS and depth camera.

[0050] The specific environmental data collected are:

[0051] S11, using a depth camera to search for the target landing platform position, and determining the relative position of the target landing platform in the body coordinate system with the drone body as the center; Specifically, the depth camera can obtain the coordinate position information p of the target landing platform in the camera coordinate system. camera This information can be determined through the depth information, and then converted to obtain the relative position p of the target landing platform in the body coordinate system drone =(x,y,z):

[0052] p drone =R camera→drone ·p camera +t camera→drone

[0053] Among them, R camera→drone is the rotation matrix from the camera to the body coordinate system, t camera→drone is the translation vector from the camera to the body coordinate system.

[0054] S12, obtaining attitude angle information of the drone based on the inertial navigation system equipped in the depth camera;

[0055] S13, determining the real-time coordinate position information of the UAV in the GPS coordinate system during flight based on the GPS.

[0056] S2, using the environmental data collected by the sensor to determine the relative position information between the UAV and the target landing platform and the real-time position information of the UAV, and based on the position information, controlling the UAV to fly to a first pre-landing position and maintain a stable hover.

[0057] Controlling the UAV to fly to the first pre-landing position based on the position information is specifically as follows: based on the relative position coordinates (x, y, z) of the target landing platform in the body coordinate system obtained in step S1, the coordinates are adjusted to the preset coordinates (x0, y0, z0) by controlling the UAV to fly, that is, based on the real-time position information of the UAV, the UAV is controlled to fly in the GPS coordinate system according to the vector direction r = (x0-x, y0-y, z0-z) and finally move to the first pre-landing position.

[0058] S3, based on the inertial navigation system equipped with the target landing platform, collects its motion posture data, and predicts the future time period of the target landing platform posture through the autoregressive model prediction.

[0059] The autoregressive model is a multi-step prediction model constructed based on the motion posture data of the target landing platform, and is used to predict the platform posture within a preset time range in the future. The posture data of the target landing platform includes three-dimensional spatial position information and posture angle information.

[0060] like Figure 3 As shown, the autoregressive model prediction process includes the following steps:

[0061] S31, data reading and preprocessing.

[0062] S311, data reading: read the inertial navigation system data located on the target landing platform, including the data of the accelerometer and gyroscope equipped in the inertial navigation, and obtain the acceleration data a of the target landing platform in three directions in the spatial coordinate system x (t),a y (t),a z (t) and angular velocity data ω x (t),ω y (t),ω z (t);

[0063] S312, data preprocessing: by processing the acceleration data a x (t),a y (t),a z (t) is subjected to a second cumulative integration to obtain the spatial position information x(t), y(t), z(t):

[0064] x(t)=∫∫a(t)dtdt

[0065] By angular velocity data ω x (t),ω y (t),ω z (t) Perform a cumulative integration to obtain the attitude angle information θ roll (t),θ pitch (t),θ yaw (t):

[0066] θ(t)=∫ω(t)dt

[0067] Through data preprocessing, the three-dimensional spatial position and attitude angle information of the target landing platform are obtained.

[0068] S32, data standardization.

[0069] In order to eliminate the dimensional differences between different features, the position and attitude data of the target landing platform are processed using a standardized method:

[0070]

[0071] Among them, x is the original data, μ is the mean of the data set, and σ is the standard deviation of the data set.

[0072] S33, autoregressive model training.

[0073] Based on the autoregressive model, the position information and posture information observed in the past are fitted to predict the future position information and posture information:

[0074]

[0075] Among them, y(t) is the position information or posture information observed at the corresponding time, φ1, φ2,…, φ p is the model coefficient, p is the lag order, ∈(t) is the noise term;

[0076] According to the different postures, they are the spatial position information x(t), y(t), z(t) and the attitude angle information θ of the target landing platform. roll (t),θ pitch (t),θ yaw (t) Establish their respective autoregressive models.

[0077] S34, Future state prediction.

[0078] For the future state prediction step, the historical data of the previous 5 seconds is used as the initial input, and the position and attitude changes of the drone in the next 10 seconds are predicted through the trained autoregressive model. In order to make the prediction results more realistic, the prediction results are denormalized to restore the original dimensions:

[0079] x=σx′+μ

[0080] During the prediction process, the prediction formula is consistent with the model form in the training process. The state at each moment is gradually predicted in a recursive manner, and the model parameters are adaptively adjusted according to the deviation between the actual information and the predicted information during the prediction process.

[0081] S4, receiving the predicted attitude of the target landing platform and the attitude of the drone collected by the sensor in step S1 through the relay computer, specifically, Figure 2 As shown, the platform attitude information is transmitted through the Bluetooth module, and the drone status information is transmitted through the digital transmission WiFi. Based on the transmitted information, it is determined whether the preset landing conditions are met. When the preset landing conditions are met, the landing command is transmitted to the drone onboard computer through the digital transmission WiFi to control the drone to complete the landing operation.

[0082] At time Δt from the current time t, when the attitude prediction result of the target landing platform and the attitude information of the UAV meet the following conditions, the preset landing conditions are considered to be met:

[0083]

[0084] Among them, f(t) represents the predicted target landing platform or the spatial position information or attitude angle information of the collected drone, t0 is the minimum time interval when observing data, n0 is a preset integer not less than 1 and makes n0t0 not less than the minimum stable time required for landing, and eps is the preset stability judgment threshold.

[0085] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning, or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A method for landing a UAV on an offshore platform based on autoregressive model prediction, characterized in that: The following steps are involved: S1, collects environmental data based on the sensors equipped by the drone; S2, using the environmental data collected by the sensor to determine the relative position information between the UAV and the target landing platform and the real-time position information of the UAV, and based on the position information, controlling the UAV to fly to a first pre-landing position and maintain a stable hover; S3, based on the inertial navigation system equipped with the target landing platform, collect its motion posture data, and predict the future time period of the target landing platform posture through autoregressive model prediction; S4, based on the predicted posture of the target landing platform and the posture of the drone collected by the sensor in step S1, it is determined whether the preset landing conditions are met, and the posture includes position information and posture angle information. When the preset landing conditions are met, a landing instruction is transmitted to the drone to control the drone to complete the landing operation.

2. The method for landing a UAV on an offshore platform based on autoregressive model prediction according to claim 1, characterized in that: The sensor includes a GPS and a depth camera. In step S1, collecting environmental data specifically includes: using the depth camera to search for the position of the target landing platform, and determining the relative position of the target landing platform in the body coordinate system with the drone body as the center; obtaining the attitude angle information of the drone based on the inertial navigation system equipped in the depth camera; and determining the real-time coordinate position information of the drone in the GPS coordinate system during flight based on the GPS.

3. The method for landing a UAV on an offshore platform based on autoregressive model prediction according to claim 2 is characterized in that: In the step S2, the UAV is controlled to fly to the first pre-landing position based on the position information. Specifically, based on the relative position coordinates (x, y, z) of the target landing platform in the body coordinate system obtained in step S1, the coordinates are adjusted to the preset coordinates (x0, y0, z0) by controlling the UAV to fly, that is, based on the real-time position information of the UAV, the UAV is controlled to fly in the GPS coordinate system according to the vector direction r = (x0-x, y0-y, z0-z) and finally move to the first pre-landing position.

4. The method for landing a UAV on an offshore platform based on autoregressive model prediction according to claim 1, characterized in that: In step S3, the autoregressive model is a multi-step prediction model constructed based on the motion posture data of the target landing platform, and is used to predict the platform posture within a preset time range in the future. The posture data of the target landing platform includes three-dimensional spatial position information and posture angle information.

5. The method for landing a UAV on an offshore platform based on autoregressive model prediction according to claim 4 is characterized in that: In step S3, the autoregressive model prediction process includes the following steps: S31, data reading and preprocessing; S32, data standardization; S33, autoregressive model training; S34, Future state prediction.

6. The method for landing a UAV on an offshore platform based on autoregressive model prediction according to claim 5, characterized in that: The step S31 comprises the following steps: S311, data reading: read the inertial navigation system data located on the target landing platform, including the data of the accelerometer and gyroscope equipped in the inertial navigation, and obtain the acceleration data a of the target landing platform in three directions in the spatial coordinate system x (t), a y (t), a z (t) and angular velocity data ω x (t),ω y (t),ω z (t); S312, data preprocessing: by processing the acceleration data a x (t), a y (t), a z (t) is subjected to a second cumulative integration to obtain the spatial position information x(t), y(t), z(t): x(t)=∫∫ a(t)dtdt By angular velocity data ωx ( t),ω y (t),ω z (t) Perform a cumulative integration to obtain the attitude angle information θ roll (t), θ pitch (t), θ yaw (t): θ(t)=∫ω(t)dt Through data preprocessing, the three-dimensional spatial position and attitude angle information of the target landing platform are obtained.

7. The method for landing a UAV on an offshore platform based on autoregressive model prediction according to claim 5, characterized in that: In step S32, for the data standardization step, in order to eliminate the dimensional differences between different features, the position and attitude data of the target landing platform are processed using a standardization method: Among them, x is the original data, μ is the mean of the data set, and σ is the standard deviation of the data set.

8. The method for landing a UAV on an offshore platform based on autoregressive model prediction according to claim 5, characterized in that: Specifically, step S33 is to fit the position information and posture information observed in the past based on the autoregressive model to predict the position information and posture information in the future: Among them, y(t) is the position information or posture information observed at the corresponding time, φ1, φ2, …, φ p is the model coefficient, p is the lag order, ∈(t) is the noise term; According to the different postures, they are the spatial position information x(t), y(t), z(t) and the attitude angle information θ of the target landing platform roll (t), θ pitch (t), θ yaw (t) Establish their respective autoregressive models.

9. The method for landing a UAV on an offshore platform based on autoregressive model prediction according to claim 5, characterized in that: In step S34, for the future state prediction step, the historical data of the previous 5 seconds is used as the initial input, and the position and attitude changes of the drone in the next 10 seconds are predicted through the trained autoregressive model, and the prediction results are denormalized to restore the original dimension: x=σx′+μ During the prediction process, the prediction formula is consistent with the model form in the training process. The state at each moment is gradually predicted in a recursive manner, and the model parameters are adaptively adjusted according to the deviation between the actual information and the predicted information during the prediction process.

10. The method for landing a UAV on an offshore platform based on autoregressive model prediction according to claim 1, characterized in that: In step S4, at a time Δt from the current time t, when the attitude prediction result of the target landing platform and the attitude information of the UAV both meet the following conditions, it is considered that the preset landing condition is met: Among them, f(t) represents the predicted target landing platform or the spatial position information or attitude angle information of the collected drone, t0 is the minimum time interval when observing data, n0 is a preset integer not less than 1 and makes n0t0 not less than the minimum stable time required for landing, and eps is the preset stability judgment threshold.

Images