Hovercraft path tracking method, device and equipment based on reinforcement learning and storage medium

Abstract

The invention provides a hovercraft path tracking method and device based on reinforcement learning, equipment and a storage medium, and the method comprises the steps: obtaining the initial positions and motion information of a hovercraft and a dock landing ship, and determining the expected path of the hovercraft based on the initial positions and motion information; calculating a target heading angle of the hovercraft in the expected path based on a first guidance algorithm; taking the target heading angle as input of a reinforcement learning network and an anti-interference controller, generating a first control parameter, training the first control parameter based on a first neural network, and generating a target control parameter; and controlling the hovercraft to move according to the expected path based on the target control parameter.

Description

Technical Field

[0001] The present disclosure relates to the field of hovercraft motion control, and in particular to a hovercraft path tracking method, device, equipment and storage medium based on reinforcement learning. Background Art

[0002] At present, the application range of full-lift air cushion vehicles is relatively wide, but due to their special navigation mechanism, they are difficult to control. When turning, there is often insufficient centripetal force, which leads to an increase in the turning radius. In addition, because the navigation environment is changeable and the mission is complex, the control difficulty is further increased. And when performing long-distance ocean transportation missions, full-lift air cushion vehicles rely heavily on delivery vehicles such as dock landing ships, so research on air cushion vehicle recovery control is very important.

[0003] At present, the recovery motion control of hovercraft at home and abroad largely relies on the manual operation of the driver. However, the environmental interference is very serious during the recovery process, and manual operation is prone to large errors, which in turn leads to safety accidents. Therefore, the research on the recovery of full-lift hovercraft needs to be further deepened. Summary of the invention

[0004] The present disclosure provides a method, device, equipment and storage medium for hovercraft path tracking based on reinforcement learning, so as to at least solve the above technical problems existing in the prior art.

[0005] According to a first aspect of the present disclosure, a method for hovercraft path tracking based on reinforcement learning is provided, the method comprising: Acquire the initial position and motion information of the hovercraft and the dock landing ship, and determine the expected path of the hovercraft based on the initial position and motion information; Calculating a target heading angle of the hovercraft on the desired path based on a first guidance algorithm; Using the target heading angle as an input of a reinforcement learning network and an anti-disturbance controller to generate a first control parameter, and training the first control parameter based on a first neural network to generate a target control parameter; The hovercraft is controlled to move along the desired path based on the target control parameter.

[0006] In one possible implementation, determining the expected path of the hovercraft based on the initial position and motion information includes: The initial position includes a first initial position of the hovercraft and a second initial position of the dock landing ship, and the motion information includes first initial motion information of the hovercraft and second initial motion information of the dock landing ship; A north-east ground coordinate system and a hull coordinate system are constructed based on the first initial position and the second initial position, and an expected path of the hovercraft in the north-east ground coordinate system is constructed based on the first initial motion information and the second initial motion information.

[0007] In one possible implementation, the calculating the target heading angle of the hovercraft on the desired path based on the first guidance algorithm includes: constructing a guidance coordinate system corresponding to the first guidance algorithm based on the second initial position, and defining attitude error vectors of the north-east ground coordinate system and the guidance coordinate system respectively; Constructing a six-degree-of-freedom kinematic mathematical model of the hovercraft based on the north-east coordinate system and the hull coordinate system, and obtaining a vector of the attitude error derivative of the hovercraft in the guidance coordinate system based on the six-degree-of-freedom kinematic mathematical model and the attitude error vector; A position error vector is defined based on the guidance coordinate system, a Lyapunov function is constructed based on the position error vector, a dynamic foresight distance parameter is set based on the position error vector and a maximum value and a minimum value of a foresight distance parameter of the hovercraft, and the target heading angle is obtained based on the dynamic foresight distance parameter, the Lyapunov function and a vector of the attitude error derivative.

[0008] In one possible implementation manner, the training of the first control parameter based on the first neural network to generate the target control parameter includes: Controlling the operation of the anti-disturbance controller based on the first control parameter, and obtaining an observed value outputted when the anti-disturbance controller is in operation; The observation values ​​include input signal feedback observation values ​​and compensation coefficient feedback observation values, and the input signal feedback observation values ​​are input into the first neural network for training to obtain observation value optimization data; The observation value optimization data is combined with the compensation coefficient feedback observation value to obtain a control optimization coefficient, and the first control parameter is optimized based on the control optimization coefficient to obtain the target control parameter.

[0009] In one embodiment, controlling the hovercraft to move along the desired path based on the target control parameter includes: Inputting the target control parameter into the anti-disturbance controller to control the heading of the unmanned boat when it moves along the expected path; Inputting the first control parameter into the anti-disturbance controller to control the speed of the unmanned boat when it moves along the expected path; The lateral displacement of the unmanned boat when moving along the expected path is controlled based on the first control system.

[0010] According to a second aspect of the present disclosure, a hovercraft path tracking device based on reinforcement learning is provided, the device comprising: A construction unit is used to obtain the initial position and motion information of the hovercraft and the dock landing ship, and determine the expected path of the hovercraft based on the initial position and motion information; A path guidance unit, configured to calculate a target heading angle of the hovercraft on the desired path based on a first guidance algorithm; an optimization unit, configured to use the target heading angle as an input of a reinforcement learning network and an anti-disturbance controller to generate a first control parameter, and train the first control parameter based on a first neural network to generate a target control parameter; A control unit is used to control the hovercraft to move along the desired path based on the target control parameter.

[0011] In one embodiment, the construction unit is further configured to: the initial position includes a first initial position of the hovercraft and a second initial position of the dock landing ship, and the initial motion information includes first initial motion information of the hovercraft and second initial motion information of the dock landing ship; construct a north-east ground coordinate system and a hull coordinate system based on the first initial position and the second initial position, and construct an expected path of the hovercraft in the north-east ground coordinate system based on the first initial motion information and the second initial motion information; The path guidance unit is also used to construct a guidance coordinate system corresponding to the first guidance algorithm based on the second initial position, and define the attitude error vectors of the north-east ground coordinate system and the guidance coordinate system respectively; construct a six-degree-of-freedom kinematic mathematical model of the hovercraft based on the north-east ground coordinate system and the hull coordinate system, and obtain a vector of attitude error derivatives of the hovercraft in the guidance coordinate system based on the six-degree-of-freedom kinematic mathematical model and the attitude error vector; define a position error vector based on the guidance coordinate system, construct a Lyapunov function based on the position error vector, set a dynamic foresight distance parameter based on the position error vector and the maximum and minimum values ​​of the foresight distance parameter of the hovercraft, and obtain the target heading angle based on the dynamic foresight distance parameter, the Lyapunov function and the vector of the attitude error derivative.

[0012] In one possible implementation manner, the optimization unit is further used to control the operation of the anti-disturbance controller based on the first control parameter, and obtain the observation value output when the anti-disturbance controller is running; the observation value includes an input signal feedback observation value and a compensation coefficient feedback observation value, and the input signal feedback observation value is input into the first neural network for training to obtain observation value optimization data; the observation value optimization data is combined with the compensation coefficient feedback observation value to obtain a control optimization coefficient, and the first control parameter is optimized based on the control optimization coefficient to obtain the target control parameter; The control unit is also used to input the target control parameter into the anti-disturbance controller to control the bow direction of the unmanned boat when it moves along the expected path; input the first control parameter into the anti-disturbance controller to control the speed of the unmanned boat when it moves along the expected path; and control the lateral displacement of the unmanned boat when it moves along the expected path based on the first control system.

[0013] According to a third aspect of the present disclosure, there is provided an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described in the present disclosure.

[0014] According to a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute the method described in the present disclosure.

[0015] The disclosed method, device, equipment and storage medium for hovercraft path tracking based on reinforcement learning combine the guidance algorithm with the coordinate system and use the guidance algorithm with the foresight distance parameter to achieve tracking of straight and curved paths. The nonlinear self-anti-disturbance control and guidance method optimized by reinforcement learning and neural network is used to achieve high-precision recovery of the hovercraft. Under the premise of not significantly increasing the overshoot of the control curve, the control speed is accelerated and the ability of the system to resist unknown interference is enhanced. Accurate recovery of the hovercraft is achieved.

[0016] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and other objects, features and advantages of the exemplary embodiments of the present disclosure will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present disclosure are shown in an exemplary and non-limiting manner, in which: In the drawings, the same or corresponding reference numerals represent the same or corresponding parts.

[0018] Figure 1 The following is a schematic diagram showing the implementation process of a hovercraft path tracking method based on reinforcement learning in an embodiment of the present disclosure. Figure 1 ; Figure 2 The following is a schematic diagram showing the implementation process of a hovercraft path tracking method based on reinforcement learning in an embodiment of the present disclosure. Figure 2 ; Figure 3 A schematic diagram of a target ship and a curve in a north-east coordinate system in a LOS guidance method in an embodiment of the present disclosure is shown; Figure 4 A schematic diagram of the ADRC structure optimized by ANN in an embodiment of the present disclosure is shown; Figure 5 A schematic diagram of the value neural network structure of an embodiment of the present disclosure is shown; Figure 6 A schematic diagram of the structure of an action neural network in an embodiment of the present disclosure is shown; Figure 7 A schematic diagram of the BP neural network structure in an embodiment of the present disclosure is shown; Figure 8 A schematic diagram of a hovercraft path tracking device based on reinforcement learning in an embodiment of the present disclosure is shown; Fig. 9 A schematic diagram of the structure of an electronic device according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0019] In order to make the purpose, features, and advantages of the present disclosure more obvious and easy to understand, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present disclosure.

[0020] Figure 1 The following is a schematic diagram showing the implementation process of a hovercraft path tracking method based on reinforcement learning in an embodiment of the present disclosure. Figure 1 ,like Figure 1 As shown, the implementation process of a hovercraft path tracking method based on reinforcement learning in an embodiment of the present disclosure includes: Step 101, obtaining the initial position and motion information of the hovercraft and the dock landing ship, and determining the expected path of the hovercraft based on the initial position and motion information.

[0021] In the disclosed embodiment, the initial position includes the first initial position of the hovercraft and the second initial position of the dock landing ship, and the motion information includes the first initial motion information of the hovercraft and the second initial motion information of the dock landing ship, wherein the motion information of the hovercraft and the dock landing ship respectively includes their corresponding speed and heading. A northeast ground coordinate system and a hull coordinate system are constructed based on the first initial position and the second initial position, and an expected path of the hovercraft in the northeast ground coordinate system is constructed based on the first initial motion information and the second initial motion information.

[0022] Step 102: Calculate a target heading angle of the hovercraft on the desired path based on a first guidance algorithm.

[0023] In an embodiment of the present disclosure, a guidance coordinate system corresponding to the first guidance algorithm is constructed based on the second initial position, and attitude error vectors of the north-east ground coordinate system and the guidance coordinate system are defined respectively, wherein the first guidance algorithm is a Frenet guidance algorithm (SFLOS) in which the guidance coordinate system of the LOS (Line of sight) guidance algorithm is a Serret-Frenet coordinate system (SF coordinate system), and the attitude error vectors of the north-east ground coordinate system and the guidance coordinate system are defined according to a differential homeomorphism transformation to obtain a transformation matrix, and the attitude error vector of the north-east ground coordinate system is made to converge in the SF coordinate system.

[0024] Constructing a six-degree-of-freedom kinematic mathematical model of the hovercraft based on the north-east coordinate system and the hull coordinate system, and obtaining a vector of the attitude error derivative of the hovercraft in the guidance coordinate system based on a transformation matrix obtained by combining the six-degree-of-freedom kinematic mathematical model and the attitude error vector; A position error vector is defined based on the guidance coordinate system, a Lyapunov function is constructed based on the position error vector, the Lyapunov function is differentiated and the vector of the attitude error derivative is brought in, and the Lyapunov function is ensured to be stable. A dynamic foresight distance parameter is set based on the position error vector and the maximum and minimum values ​​of the foresight distance parameter of the hovercraft, and a rotation angle, a virtual control rate, a foresight distance parameter, a position tracking error, and a target heading angle of the North East coordinate system converted to the SF coordinate system are obtained based on the dynamic foresight distance parameter, the Lyapunov function, and the vector of the attitude error derivative.

[0025] Step 103: Using the target heading angle as an input of a reinforcement learning network and an anti-disturbance controller to generate a first control parameter, and training the first control parameter based on a first neural network to generate a target control parameter.

[0026] In the disclosed embodiment, the disturbance rejection controller adopts a nonlinear active disturbance rejection controller (ADRC), which is composed of four parts: a tracking differentiator (TD), an extended state observer (ESO), a nonlinear state error feedback control law (NLSEF), and a disturbance compensation device. The reinforcement learning network is a deterministic policy gradient (DDPG) in reinforcement learning. The target heading angle is a necessary condition for the hovercraft to navigate along the expected path, and is input into the ADRC to generate the initial speed control parameters and the initial heading control parameters for controlling the hovercraft to navigate. The DDPG reinforcement learning optimizes the initial speed control parameters and the initial heading control parameters on the basis of the initial speed control parameters and the initial heading control parameters to obtain the above-mentioned first control parameters, which are used to control the hovercraft to operate along the expected path. However, since the dock landing ship is in a state of motion (simulated according to a uniform motion state), the above-mentioned first control parameters lack targeted control and have certain limitations. Therefore, a first neural network, i.e., a feedforward neural network (Back Propagation, BP), is introduced to optimize the first control parameters, specifically: based on the first control parameter, the anti-disturbance controller is controlled to operate, and the observation value output when the anti-disturbance controller is running is obtained; the observation value includes an input signal feedback observation value and a compensation coefficient feedback observation value, and the input signal feedback observation value is input into the first neural network for training to obtain observation value optimization data; the observation value optimization data is combined with the compensation coefficient feedback observation value to obtain the control optimization coefficient, and the first control parameter is optimized based on the control optimization coefficient to obtain the target control parameter.

[0027] Step 104: Control the hovercraft to move along the desired path based on the target control parameter.

[0028] In the disclosed embodiment, the target control parameter is input into the anti-disturbance controller to control the bow direction of the unmanned boat when it moves along the expected path; the first control parameter is input into the anti-disturbance controller to control the speed of the unmanned boat when it moves along the expected path; because in the process of recovering the hovercraft, there are relatively strict requirements for the lateral displacement. Considering that the hovercraft has a slow sailing speed, in addition to the need to control the bow direction and speed, the lateral displacement of the hovercraft also needs to be accurately controlled. During the entire recovery process of the hovercraft, we can regard the movement of the dock landing ship as uniform linear motion. In addition, the bow angle of the hovercraft when it is about to enter the dock should be consistent with the bow angle of the dock landing ship to ensure the precise control of the ship alignment and lateral displacement during the recovery process. Therefore, the lateral displacement of the unmanned boat when it moves along the expected path is controlled based on the first control system, wherein the first control system is a PID (Proportion Integral Differential) control system.

[0029] In the disclosed embodiment, during the recovery process of the hovercraft, it is necessary to observe the dock landing ship and obtain its real information, and the complex and changeable marine environmental interference has a great impact on the observer, so it is necessary to use Kalman filtering to remove the interference that occurs during observation to obtain accurate dock landing ship information. The recovery process of the hovercraft requires high accuracy, and the Kalman filtering method can help eliminate environmental interference to a certain extent, provide accurate observation information, and thus improve the accuracy and safety of recovery.

[0030] Figure 2 The following is a schematic diagram showing the implementation process of a hovercraft path tracking method based on reinforcement learning in an embodiment of the present disclosure. Figure 2 ,like Figure 2 As shown, the implementation process of a hovercraft path tracking method based on reinforcement learning in an embodiment of the present disclosure includes: Step 201, obtaining the initial position and motion information of the hovercraft and the dock landing ship.

[0031] In the disclosed embodiment, the motion information of the hovercraft and the dock landing ship includes their corresponding speed and heading respectively. In this embodiment, in order to ensure the smooth recovery of the hovercraft, the dock landing ship is in a uniform linear motion state. According to the initial position and motion information of the hovercraft, a north-east ground coordinate system and a hull coordinate system are established, and the north-east ground coordinate system and the hull coordinate system are combined with each other to construct a conversion matrix. Specifically:

[0032] Therefore, according to the fixed coordinate system ( ) and the conversion relationship between the north-east coordinate system and the hull coordinate system, it can be concluded that the six-degree-of-freedom kinematic mathematical model of the full-cushion hovercraft in the fixed coordinate system is as follows:

[0033]

[0034] Among them, the various forces acting on the hovercraft are synthesized in the hull coordinate system, and the six-degree-of-freedom dynamic mathematical model of the full-lift hovercraft in the form of a differential equation group is established as follows:

[0035] The resultant force on each degree of freedom is as follows:

[0036] Where G is the mass of the hovercraft, , , , , , They are the longitudinal force, transverse force, vertical force, heeling moment, pitching moment, and bow turning moment of the hovercraft. , , They are the moments of inertia of the hovercraft around the x, y, and z axes, respectively. The subscript a represents aerodynamic force, m represents air momentum force, h represents hydrodynamic force, p represents propeller force, R represents rudder force, s represents side wind door force, and n represents bow nozzle force.

[0037]

[0038] Table 1, Hovercraft Symbol Correspondence Table Step 202, determining the desired path of the hovercraft.

[0039] In the disclosed embodiment, the expected path for the hovercraft to complete the landing process is calculated based on the speed and heading of the hovercraft and the dock landing ship.

[0040] Step 203: derive the target bow phase angle required for the hovercraft to track the desired path according to the SFLOS guidance algorithm.

[0041] In the disclosed embodiment, the principle of the LOS (Line of sight) guidance algorithm is to find a suitable desired point for the controlled ship at the current moment by analyzing the mathematical relationship between the desired route and the location of the ship, and to control the heading of the target ship to align with the "line of sight" direction of the desired point, so as to achieve the control of the navigation trajectory of the ship gradually converging to the desired path. This patent uses an improved LOS guidance algorithm with a forward-looking distance parameter. The SFLOS guidance algorithm is used to guide the ship's current position in front of the projection point on the target path. The position at the distance is taken as the expected point, and the expected heading angle is calculated according to the foresight distance parameter, so as to realize the tracking of the straight path and the curved path. Since the hovercraft cannot be controlled laterally during the recovery process, only the speed and heading control can be performed. Therefore, the goal of the hovercraft is to make the hovercraft close to the dock landing ship through the trajectory tracking algorithm and the heading and speed controller, and the lateral displacement difference between the hovercraft and the dock landing ship's driving trajectory is required to be within a certain range. Since the dock landing ship does not change significantly during the recovery of the hovercraft, the movement of the dock landing ship can be regarded as uniform linear motion, so if the expected path is set is a straight line, the angle between this straight line and the longitudinal axis is equal to the bow phase angle of the dock landing ship, The increasing direction is the direction of the dock landing ship, so the expected path It is approximately the same as the trajectory of the dock landing ship.

[0042] In the embodiment of the present disclosure, the mathematical description of the control target of the hovercraft is as follows:

[0043] In the formula, , , They are the expected path, expected heading and expected speed respectively; , , are the actual path, actual heading, and actual speed respectively; t is the current time, and T is the termination time. Since the lateral displacement of the hovercraft needs to be controlled during the recovery phase, the path tracking error here is Can be appropriately enlarged.

[0044] Figure 3 FIG. 4 shows a schematic diagram of the target ship and the curve in the north-east coordinate system in the LOS guidance method in an embodiment of the present disclosure, as shown in FIG. Figure 3 As shown, let the expected path be , the point on it is expressed in the north-east coordinate system as , , satisfy , the first and second order derivatives of curve O exist and are bounded. The hovercraft will move the point on the curve O is regarded as the target tracking point, NOE is the North East coordinate system, , The curve O is formed at the point Coordinate system of the SFLOS guidance method at .

[0045] In the embodiment of the present disclosure, the posture error vector is defined in the Serret-Frenet (SF) coordinate system: , the attitude error vector in the north-east coordinate system . Then according to the differential homeomorphism transformation, we can get the following relationship:

[0046] According to the properties of differential homeomorphism, if the error vector converges in the SF coordinate system, then the error vector converges in the N-E coordinate system. Therefore, we only need to design the control law to ensure that the error vector converges in the SF coordinate system. Let the transformation matrix in the formula be , then taking the derivative of the formula we can get the following relationship:

[0047] From the above formula, we can see that The derivative of is:

[0048] Substituting the kinematic relationship of the hovercraft into the formula, we can get:

[0049] By the same formula, we can know that The derivative of is:

[0050] From the above formula, we can know that the heading tracking error is The derivative of is:

[0051] Combining the above three equations, we can get the vector of the attitude error derivative of the hovercraft in the SFLOS guidance coordinate system:

[0052] The position error vector in the SFLOS guidance coordinate system is defined as , then the control objective description of path tracking is as follows: , then according to the above description, the following Lyapunov function can be constructed: , for the above Lyapunov function Derivative, and the vector of the attitude error derivative Substituting in, we get:

[0053] when When , the system has Lyapunov stability. To achieve this goal, we should consider Position tracking error in direction and Position tracking error in direction The convergence problem of Has the following form:

[0054] In the above formula, is the parameter to be designed and satisfies . Combining the above formulas, we can get:

[0055] Then for the virtual control Design a virtual control rate of the form:

[0056] The reason for using the inverse tangent function in the above formula is that its boundedness can avoid the position error Appears when too large Too large, resulting in excessive bow turning moment Definition The tracking error is , then according to the above formula we can get:

[0057] Substituting the formula into the equation, we get:

[0058] in, The expression is as follows:

[0059] When the system is stable , , When it converges to zero, it is easy to get , at this time , the Lyapunov stability of the system is proved.

[0060] The foresight distance parameter in the usual formula This section uses a variable to adapt to the change of navigation state during hovercraft path tracking. Design, that is, using the dynamic foresight distance parameters as shown below:

[0061] Among them, γ is an undetermined parameter, is the maximum value of the dynamic foresight distance parameter, is the minimum value of the dynamic foresight distance parameter. Using the foresight distance parameter in the above formula, we can Position tracking error in direction The size of the When larger, Closer to , the tracking speed is faster; and when When smaller, Closer to , the tracking speed is slower, thus effectively reducing overshoot.

[0062] In summary, the complete design of SFLOS guidance is as follows:

[0063] Among them, the above is the bow phase angle of the hovercraft, is the sideslip angle of the hovercraft, is the track angle of the hovercraft, is the rotation angle from the North East coordinate system to the SF coordinate system, , and are the heading tracking errors of the hovercraft in the SF coordinate system, where is the virtual control rate, is the foresight distance parameter in SFLOS guidance, For hovercraft Position tracking error in direction, For hovercraft The position tracking error in the direction, u, v, and U are the longitudinal velocity, lateral velocity, and combined velocity of the hovercraft, respectively.

[0064] Step 204, using DDPG and BP neural network to improve ADRC, and use it as the heading and speed controller of the hovercraft, combined with the SFLOS guidance algorithm to achieve the heading and speed control of the hovercraft.

[0065] In the disclosed embodiment, DDPG is used to improve the traditional ADRC to achieve control of speed and heading, wherein, based on the improvement of DDPG, a BP neural network is used to optimize the heading control, the optimized heading controller is used as the heading anti-disturbance control controller for controlling the hovercraft, the improved speed controller is used as the speed anti-disturbance control controller of the hovercraft, and a PID controller is used to control the lateral displacement of the hovercraft.

[0066] In the disclosed embodiment, the anti-disturbance controller consists of four parts: Tracking Differentiator (TD), Extended State Observer (ESO), Nonlinear State Error Feedback Law (NLSEF), and disturbance compensation device. The function of the TD link is to smooth the set expected value and extract its differential signal. The expression of the second-order tracking differentiator based on the discrete system is as follows:

[0067] Among them, Fu is the optimal result of the calculation, f a The specific expression of the function is:

[0068] Among them, the parameters With parameters The formula is:

[0069] Where k is the coefficient of the nonlinear feedback control rate, is the input signal, for The tracking signal, is the approximate differential signal of a The function parameter r is the speed factor, parameter h is the sampling time, and parameter is the filtering factor.

[0070] In the disclosed embodiment, the role of the ESO link is to estimate the state variables of the system and the total disturbance of the system. The self-disturbance rejection controller adopts the third-order ESO. The third-order ESO is an extended state observer of a second-order uncertain system. Since the controlled object is second-order, that is, there are two state variables, there are three state variables after expansion, and the corresponding ESO is a third-order ESO. Its state equation is as follows:

[0071] in, is the ESO adjustment gain, z1 to z n+1 x1 to x n+1 The observed value of , u is the control input, y is the control output, b is a constant, The function definition is as follows:

[0072] In the above formula, the specific value of a determines the function f a1The nonlinear shape of is a small positive number that prevents the function from oscillating at the origin. The size of the function f a1 The linearization interval of . From the above form, it can be seen that when the error is large, the function f a1 The gain is small, and the opposite is true when the error is small.

[0073] In the disclosed embodiment, the function of the NLSEF link is to obtain the state error by subtracting the output of the TD link from the output of the ESO link, and to perform a nonlinear combination to obtain the final control variable. Since the output of the TD link is the tracking signal of the input signal and the differential signal of the input signal, and the output of the ESO link is the system state variable after expansion, the state error of the system can be obtained by subtracting the output of the TD link from the output of the ESO, and then the state error is passed through the nonlinear function f a1 By combining them, the NLSEF link can be realized, and its specific form is as follows:

[0074] in, , 、a1、a2、 are adjustable parameters; v1(k) and v2(k) are the tracking signal and differential signal output by TD respectively; z1(k) and z2(k) are the estimated values ​​of the state variables by ESO.

[0075] In the disclosed embodiment, the above process regards the inaccuracy and coupling of the model as the internal disturbance of the system, and together with the external disturbance, it constitutes the total disturbance of the system, and then expands it into an additional state variable of the system, which is estimated by the ESO link together with other state variables. Therefore, ADRC realizes the transformation of the controlled object to linearization and determinism. Through dynamic compensation linearization, the control input u(t) required by the controlled object can be calculated from u0(t), and its specific form is as follows:

[0076] Among them, parameter b0 is the coefficient of dynamic compensation linearization and can be calculated according to the formula , the control input u(t) of the controller is used to obtain the control input required by the ESO. Generally, in engineering applications, b0 and b are taken as the same value.

[0077] In the disclosed embodiment, during the recovery process of the hovercraft, there are relatively strict requirements for the lateral displacement. Considering that the navigation speed of the hovercraft is relatively slow, in addition to the need to control the heading and speed, the lateral displacement of the hovercraft also needs to be accurately controlled. During the recovery process of the entire hovercraft, we can regard the movement of the dock landing ship as uniform linear motion. In addition, the heading angle of the hovercraft when it is about to enter the dock should be consistent with the heading angle of the dock landing ship to ensure the precise control of the ship alignment and lateral displacement during the recovery process. Therefore, a composite controller of ADRC and PID is used to achieve the heading, speed and lateral control of the hovercraft. Among them, ADRC is used to control the heading and speed of the hovercraft, while PID is used for lateral control. Because the environmental interference during the recovery process is relatively complex and will change during the control process, on the basis of using the self-disturbance rejection controller to control the heading and speed of the hovercraft, the deterministic policy gradient DDPG algorithm in reinforcement learning is used to optimize the design of the self-disturbance rejection controller. Since the influence of disturbance on speed and lateral displacement is small, this patent only optimizes the heading angle anti-disturbance controller. According to the experience of ADRC parameter setting, the unknown environmental disturbance has a great influence on the parameters of the NLSEF link in ADRC. and The impact is large, so this paper applies the DDPG algorithm to the parameter and Setting. Specifically: The speed anti-disturbance controller is designed as follows:

[0078] The heading active disturbance rejection controller is designed as follows:

[0079] The design of the lateral PID controller is as follows:

[0080] Figure 4 FIG. 4 shows a schematic diagram of the ADRC structure optimized by ANN in an embodiment of the present disclosure, such as Figure 4 As shown, obtaining the system information f is the key to improving the control ability of the ADRC controller. Among them, the external interference can be uncertain factors such as wave fluctuations, wind speed, wind direction, etc., and the controlled object is the hovercraft. The characteristics of ESO can be used with the help of the approximation ability of ANN to obtain system information, thereby improving the performance of the system. Specifically: configure the control parameters of the bow anti-disturbance controller improved based on the above-mentioned DDPG algorithm so that ADRC can control the hovercraft, obtain Z1, Z2, and Z3 in ESO and save them; use Z1 and Z2 as inputs and Z3 as the desired output, design the ANN network structure, and formulate the termination conditions for training, and use the BP algorithm to start training; after the training is completed, f ANNEmbedded into the controller for compensation. Therefore, the optimization results of the heading ADRC controller based on the neural network are as follows:

[0081] In the embodiment of the present disclosure, the reinforcement learning algorithm uses an Actor-Critic framework. Its Actor network is a deterministic policy network that directly outputs a specific action and is suitable for problems in continuous action space. The state received by the DDPG environment is the heading angle error. , differential of heading angle error , integral of heading angle error The differential of the control output u0 and the control output of NLSEF , control output integral ; The action output by the DDPG environment is the parameter of the NLSEF link at the current K moment , The value of and ; The reward function of the DDPG environment, that is, the critic function, is designed as follows:

[0082] Figure 5 The schematic diagram of the value neural network structure of the embodiment of the present disclosure is shown as follows: Figure 5 As shown in Figure 1, the critic neural network contains two inputs and one output. The input is the environment state received by the agent and the action output by the agent. The output is the Q value, i.e. The specific composition is shown in the following table: Table 2 Value neural network composition

[0083] Figure 6 FIG. 4 shows a schematic diagram of the action neural network structure in an embodiment of the present disclosure. Figure 6 As shown in the figure, the input of the action neural network is the environment state received by the agent, and the output is a two-dimensional vector corresponding to the NLSEF parameters of ADRC. and , its specific composition is shown in the following table: Table 3 Action neural network composition

[0084] Figure 7 FIG. 4 shows a schematic diagram of the BP neural network structure in an embodiment of the present disclosure. Figure 7As shown in the figure, a neural network is a network structure composed of a large number of neurons connected to each other for information transmission. … is the input vector, … is the hidden layer output vector, … Output vector for the output layer, and are the weights from the input layer to the hidden layer and from the hidden layer to the output layer, respectively. is the expected value corresponding to the output layer, is the error value. BP neural network is a feedforward neural network that uses a negative gradient descent algorithm. This algorithm can effectively solve the XOR problem and other problems that a single-layer perceptron cannot solve. Its structure includes an input layer, one or more hidden layers, and an output layer. It has a strong approximation ability and is widely used in prediction, image processing, signal processing, pattern recognition, and knowledge processing. The main idea of ​​BP neural network is: given n input learning samples and their corresponding m output samples. Calculate the error e between the actual output time of the network and the target output, and use e to reversely adjust the weights so that the sum of the squares of the errors is minimized, thereby making the actual output closer and closer to the target output. Among them, the learning process of BP network mainly consists of four parts: (1) forward propagation of input patterns; (2) reverse propagation of output errors; (3) cyclic memory training; (4) identification of learning results. Among them, in the forward propagation process, the signal is transmitted from the input layer to the hidden layer, processed layer by layer, and then transmitted to the output layer. If the output of the output layer is not the expected value, the error is propagated in the opposite direction along the original propagation route, which is called the reverse propagation of the signal. The weights of each layer are modified to reduce the error, and then the signal starts its forward propagation process again. This cycle is repeated until the error is less than the expected value. The parameter settings of the DDPG algorithm are shown in the following table: Table 4 DDPG algorithm parameter settings

[0085] Step 205: When the difference between the control target and the expected value enters an acceptable error range, the hovercraft recovery guidance ends.

[0086] In the disclosed embodiment, during the recovery process of the hovercraft, it is necessary to observe the dock landing ship and obtain its real information, and the complex and changeable marine environmental interference has a great impact on the observer, so it is necessary to use Kalman filtering to remove the interference that occurs during observation in order to obtain accurate dock landing ship information. The recovery process of the hovercraft requires high accuracy. The Kalman filtering method can help eliminate environmental interference to a certain extent, provide accurate observation information, and thus improve the accuracy and safety of recovery. The specific form is:

[0087] Among them, B is the control output matrix, I is the identity matrix, P (k|k-1) is the covariance matrix of the state prediction at time k, which is used to represent the uncertainty of the estimate, X (k) represents the true state matrix of the system at time k, and A is the state transfer matrix. is the transposed matrix of matrix A, U(k) is the input of the system, Z(k) represents the true value of the system at time k, H is the measurement matrix, X(k|k) represents the updated state estimate at time k, X(k|k-1) represents the state prediction value at time k, X(k-1|k-1) represents the Kalman filter optimal estimate of the system at the previous moment, P(k|k) is the covariance of the optimal value estimated at time k, R represents the covariance of white noise, K(k) represents the Kalman gain, and Q represents the noise covariance.

[0088] Figure 8 A schematic diagram of a hovercraft path tracking device based on reinforcement learning in an embodiment of the present disclosure is shown. Figure 8 As shown, a hovercraft path tracking device based on reinforcement learning in an embodiment of the present disclosure includes: The construction unit 801 is used to obtain the initial position and motion information of the hovercraft and the dock landing ship, and determine the expected path of the hovercraft based on the initial position and motion information.

[0089] The construction unit 801 is also used for, wherein the initial position includes a first initial position of the hovercraft and a second initial position of the dock landing ship, and the initial motion information includes first initial motion information of the hovercraft and second initial motion information of the dock landing ship; constructing a northeast ground coordinate system and a hull coordinate system based on the first initial position and the second initial position, and constructing an expected path of the hovercraft in the northeast ground coordinate system based on the first initial motion information and the second initial motion information.

[0090] The path guidance unit 802 is used to calculate the target heading angle of the hovercraft on the expected path based on a first guidance algorithm.

[0091] The path guidance unit 802 is also used to construct a guidance coordinate system corresponding to the first guidance algorithm based on the second initial position, and define the attitude error vector of the north-east ground coordinate system and the guidance coordinate system respectively; construct a six-degree-of-freedom kinematic mathematical model of the hovercraft based on the north-east ground coordinate system and the hull coordinate system, and obtain a vector of the attitude error derivative of the hovercraft in the guidance coordinate system based on the six-degree-of-freedom kinematic mathematical model and the attitude error vector; define a position error vector based on the guidance coordinate system, construct a Lyapunov function based on the position error vector, set a dynamic foresight distance parameter based on the position error vector and the maximum and minimum values ​​of the foresight distance parameter of the hovercraft, and obtain the target heading angle based on the dynamic foresight distance parameter, the Lyapunov function and the vector of the attitude error derivative.

[0092] The optimization unit 803 is used to use the target heading angle as an input of a reinforcement learning network and an anti-disturbance controller to generate a first control parameter, and train the first control parameter based on a first neural network to generate a target control parameter.

[0093] The optimization unit 803 is also used to control the operation of the anti-disturbance controller based on the first control parameter, and obtain the observation value output when the anti-disturbance controller is running; the observation value includes the input signal feedback observation value and the compensation coefficient feedback observation value, and the input signal feedback observation value is input into the first neural network for training to obtain observation value optimization data; the observation value optimization data is combined with the compensation coefficient feedback observation value to obtain the control optimization coefficient, and the first control parameter is optimized based on the control optimization coefficient to obtain the target control parameter.

[0094] The control unit 804 is used to control the hovercraft to move along the desired path based on the target control parameter.

[0095] The control unit 804 is also used to input the target control parameter into the anti-disturbance controller to control the bow direction of the unmanned boat when it moves along the expected path; input the first control parameter into the anti-disturbance controller to control the speed of the unmanned boat when it moves along the expected path; and control the lateral displacement of the unmanned boat when it moves along the expected path based on the first control system.

[0096] In an exemplary embodiment, the construction unit 801, the path guidance unit 802, the optimization unit 803, the control unit 804, etc. can be implemented by one or more central processing units (CPU), graphics processing units (GPU), application specific integrated circuits (ASIC), DSP, programmable logic devices (PLD), complex programmable logic devices (CPLD), field programmable gate arrays (FPGA), general-purpose processors, controllers, microcontrollers (MCU), microprocessors, or other electronic components.

[0097] Regarding the device in the above embodiment, the specific manner in which each module and unit performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0098] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device and a readable storage medium.

[0099] Fig. 9 A schematic block diagram of an example electronic device 900 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0100] like Fig. 9As shown, the device 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the device 900 can also be stored. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0101] A number of components in the device 900 are connected to the I / O interface 905, including: an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0102] The computing unit 901 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 901 performs the various methods and processes described above, such as a hovercraft path tracking method based on reinforcement learning. For example, in some embodiments, a hovercraft path tracking method based on reinforcement learning may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the hovercraft path tracking method based on reinforcement learning described above may be performed. Alternatively, in other embodiments, the computing unit 901 may be configured to execute a hovercraft path tracking method based on reinforcement learning in any other appropriate manner (for example, by means of firmware).

[0103] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0104] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0105] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0106] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0107] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0108] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0109] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0110] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present disclosure, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0111] The above is only a specific embodiment of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present disclosure, which should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A hovercraft path tracking method based on reinforcement learning, characterized in that: The method comprises: Acquire initial positions and motion information of the hovercraft and the dock landing ship, and determine a desired path of the hovercraft based on the initial positions and motion information; Calculating a target heading angle of the hovercraft on the desired path based on a first guidance algorithm; Using the target heading angle as an input of a reinforcement learning network and an anti-disturbance controller to generate a first control parameter, and training the first control parameter based on a first neural network to generate a target control parameter; Controlling the hovercraft to move along the desired path based on the target control parameter; The step of training the first control parameter based on the first neural network to generate the target control parameter includes: Controlling the operation of the anti-disturbance controller based on the first control parameter, and obtaining an observed value outputted when the anti-disturbance controller is in operation; The observation values ​​include input signal feedback observation values ​​and compensation coefficient feedback observation values, and the input signal feedback observation values ​​are input into the first neural network for training to obtain observation value optimization data; The observation value optimization data is combined with the compensation coefficient feedback observation value to obtain a control optimization coefficient, and the first control parameter is optimized based on the control optimization coefficient to obtain the target control parameter.

2. The method according to claim 1, characterized in that: The determining of the expected path of the hovercraft based on the initial position and motion information comprises: The initial position includes a first initial position of the hovercraft and a second initial position of the dock landing ship, and the motion information includes first initial motion information of the hovercraft and second initial motion information of the dock landing ship; A north-east ground coordinate system and a hull coordinate system are constructed based on the first initial position and the second initial position, and an expected path of the hovercraft in the north-east ground coordinate system is constructed based on the first initial motion information and the second initial motion information.

3. The method according to claim 2, characterized in that The calculating the target heading angle of the hovercraft on the expected path based on the first guidance algorithm includes: constructing a guidance coordinate system corresponding to the first guidance algorithm based on the second initial position, and defining attitude error vectors of the north-east ground coordinate system and the guidance coordinate system respectively; Constructing a six-degree-of-freedom kinematic mathematical model of the hovercraft based on the north-east coordinate system and the hull coordinate system, and obtaining a vector of the attitude error derivative of the hovercraft in the guidance coordinate system based on the six-degree-of-freedom kinematic mathematical model and the attitude error vector; defining a position error vector based on the guidance coordinate system, and constructing a Lyapunov function based on the position error vector; A dynamic foresight distance parameter is set based on the position error vector and the maximum and minimum values ​​of the foresight distance parameter of the hovercraft, and the target heading angle is obtained based on the dynamic foresight distance parameter, the Lyapunov function and the vector of the attitude error derivative.

4. The method according to claim 1, characterized in that: The controlling the hovercraft to move along the desired path based on the target control parameter comprises: Inputting the target control parameter into the anti-disturbance controller to control the heading of the hovercraft when it moves along the desired path; Inputting the first control parameter into the anti-disturbance controller to control the speed of the hovercraft when it moves along the expected path; The lateral displacement of the hovercraft when moving along the desired path is controlled based on the first control system.

5. A hovercraft path tracking device based on reinforcement learning, characterized in that: The device comprises: A construction unit is used to obtain initial positions and motion information of the hovercraft and the dock landing ship, and determine a desired path of the hovercraft based on the initial positions and motion information; A path guidance unit, configured to calculate a target heading angle of the hovercraft on the desired path based on a first guidance algorithm; an optimization unit, configured to use the target heading angle as an input of a reinforcement learning network and an anti-disturbance controller to generate a first control parameter, and train the first control parameter based on a first neural network to generate a target control parameter; a control unit, configured to control the hovercraft to move along the desired path based on the target control parameter; The optimization unit is also used to control the operation of the anti-disturbance controller based on the first control parameter, and obtain the observation value output when the anti-disturbance controller is running; the observation value includes an input signal feedback observation value and a compensation coefficient feedback observation value, and the input signal feedback observation value is input into the first neural network for training to obtain observation value optimization data; the observation value optimization data is combined with the compensation coefficient feedback observation value to obtain a control optimization coefficient, and the first control parameter is optimized based on the control optimization coefficient to obtain the target control parameter.

6. The device according to claim 5, characterized in that The construction unit is further configured to: the initial position includes a first initial position of the air cushion vessel and a second initial position of the dock landing ship; and the motion information includes first initial motion information of the air cushion vessel and second initial motion information of the dock landing ship; Constructing a north-east ground coordinate system and a hull coordinate system based on the first initial position and the second initial position, and constructing an expected path of the hovercraft in the north-east ground coordinate system based on the first initial motion information and the second initial motion information; The path guidance unit is further used to construct a guidance coordinate system corresponding to the first guidance algorithm based on the second initial position, and define attitude error vectors of the north-east ground coordinate system and the guidance coordinate system respectively; A six-degree-of-freedom kinematic mathematical model of the hovercraft is constructed based on the north-east coordinate system and the hull coordinate system, and a vector of the attitude error derivative of the hovercraft in the guidance coordinate system is obtained based on the six-degree-of-freedom kinematic mathematical model and the attitude error vector; a position error vector is defined based on the guidance coordinate system, a Lyapunov function is constructed based on the position error vector, a dynamic foresight distance parameter is set based on the position error vector and the maximum and minimum values ​​of the foresight distance parameter of the hovercraft, and the target heading angle is obtained based on the dynamic foresight distance parameter, the Lyapunov function and the vector of the attitude error derivative.

7. The device according to claim 5, characterized in that The control unit is also used to input the target control parameter into the anti-disturbance controller to control the bow direction of the hovercraft when it moves along the expected path; input the first control parameter into the anti-disturbance controller to control the speed of the hovercraft when it moves along the expected path; and control the lateral displacement of the hovercraft when it moves along the expected path based on the first control system.

8. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 4.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to make a computer execute the method according to any one of claims 1-4.

Images