Hovercraft intelligent action planning and track control method based on reinforcement learning

Abstract

The invention discloses a hovercraft intelligent motion planning and track control method based on reinforcement learning. The method comprises the following steps: determining motion environment information of a full hovercraft; establishing a model of the full-hovering hovercraft, and obtaining a navigation speed and heading-transverse displacement decoupling control model; the expected speed, the expected heading position and the expected transverse position of the hovercraft are obtained; a hovercraft navigational speed controller based on intelligent self-adaptive linear active-disturbance-rejection control is designed, a hovercraft heading-transverse decoupling controller is designed based on an intelligent self-adaptive linear active-disturbance-rejection decoupling control algorithm, and it is guaranteed that the heading and transverse positions of the hovercraft are stable.

Description

Technical Field

[0001] The present invention relates to the field of ship intelligent decision-making and motion control, and in particular to a method for intelligent motion planning and track control of a hovercraft based on reinforcement learning. Background Art

[0002] Because of its amphibious and high-speed characteristics, the hovercraft has no underwater turning equipment and cannot turn freely, which greatly increases the difficulty of operating the hovercraft. In addition, its navigation stability is greatly affected by wind speed and direction, and it is easy to drift sideways during navigation, leading to safety accidents of the hovercraft. Ship intelligence can effectively solve the problems faced by ships in terms of safety, energy efficiency and cost, and is an inevitable trend in the development of ships. Realizing intelligent navigation based on manual operation rules and corresponding motion control methods for special operation requirements will greatly reduce the dangerous navigation state caused by manual operation and ensure the safe navigation of the hovercraft. Summary of the invention

[0003] In view of the above problems existing in the prior art, the present invention provides a method for intelligent motion planning and track control of a hovercraft based on reinforcement learning, comprising: Step 1: Determine the motion environment information of the full-lift air cushion vehicle; establish a full-lift air cushion vehicle model to obtain the heading, position, speed, and heel of the air cushion vehicle; Step 2: Using the heading, position, speed, and heel of the hovercraft obtained in step 1, the strong coupling problem between the heading control channel and the lateral control channel is simplified to obtain a speed control model and a heading-lateral displacement decoupling control model; Step 3: According to the speed control model in step 2, the bow-lateral displacement decoupling control model and the full-lift hovercraft model in step 1, the hovercraft navigation information is obtained, and the intelligent action of the hovercraft navigation process is planned by the deep deterministic policy gradient in the reinforcement learning algorithm to obtain the expected speed, expected bow and expected lateral position of the hovercraft; Step 4: According to the expected speed in step 3 and the speed control model in step 2, a hovercraft speed controller based on intelligent adaptive linear anti-disturbance control is designed, and the longitudinal speed is controlled by the thrust generated by the variable pitch air propeller to ensure the stability of the hovercraft speed; Step 5: According to the desired heading and lateral position in step 3 and the heading-lateral displacement decoupling control model in step 2, and based on the intelligent adaptive linear anti-disturbance decoupling control algorithm, the heading-lateral decoupling controller of the hovercraft is designed to ensure the stability of the heading and lateral position of the hovercraft.

[0004] The hovercraft model in step 1 includes a kinematic model and a dynamic model; The kinematic model is: ; In the formula, , , , They represent the north position, east position, heel angle and heading angle of the hovercraft in the northeast coordinate system, , , , They respectively represent the longitudinal velocity, lateral velocity, heel angular velocity and heading angular velocity of the hovercraft in the motion coordinate system.

[0005] The kinetic model described in step 1 is: ; In the formula, is the mass of the hovercraft, , , , They are the longitudinal force, transverse force, heel moment and bow moment of the hovercraft. , They are the moments of inertia of the hovercraft around the x and z axes respectively; , , , They respectively represent the longitudinal velocity, lateral velocity, heel angular velocity and heading angular velocity of the hovercraft in the motion coordinate system.

[0006] The step 2 specifically includes: Step 2-1: Since the adaptive linear active disturbance rejection control is a control method that is not based on a model, the standard form of the hovercraft speed control model that can be directly obtained is: ; In the formula, is the longitudinal velocity of the hovercraft in the motion coordinate system, is the total disturbance of the system, To control the gain, To control the amount; Step 2-2: Based on the hovercraft model established in step 1, the bow control model and lateral displacement control model are obtained respectively: ; ; In the formula, is the bow aerodynamic moment, is the bow air momentum moment, is the bow hydrodynamic moment, is the air rudder turning moment, is the air propeller bow torque, Bow nozzle turning moment, is the lateral aerodynamic force, is the lateral air momentum force, is the lateral hydrodynamic force, is the lateral force of the air rudder, is the lateral thrust of the air propeller, is the lateral thrust of the bow nozzle, For hovercraft z The moment of inertia of the shaft, , represents the heel angle and heading angle of the hovercraft in the northeast coordinate system, m is the mass of the hovercraft, , , , y They respectively represent the longitudinal velocity, lateral velocity, bow angular velocity and lateral position of the hovercraft in the motion coordinate system.

[0007] Furthermore, in step 2, the heading control actuator and the lateral displacement control actuator are respectively an air rudder and a bow nozzle, and the heading control input is defined as Define the lateral displacement control input as the bow moment generated by the air rudder is the lateral force generated by the bow nozzle, which is expressed in mathematical form as follows: ;

[0008] In the formula, is the lateral rudder force coefficient, is the air density, is the incoming flow velocity on the air rudder, is the air rudder area, For vertical installation position, is the distance between the pressure center of the air rudder and the rudder axis, is the bow nozzle thrust, is the nozzle angle; The lateral force generated by the air rudder is defined as the interference of the heading control channel on the lateral control channel. τ uψ The bow turning moment generated by the bow nozzle is defined as the interference of the lateral control channel on the bow control channel τ uy , the mathematical form is as follows: ;

[0009] In the formula, This is the installation position of the bow nozzle.

[0010] The heading decoupling control model in step 2 is: ;

[0011] In the formula, , , , , , To calculate the derived base values ​​of the dimensionless hydrodynamic coefficients in the bow and transverse directions, is the longitudinal position of the hull acting on the hydrodynamic force or hydrodynamic moment, , are the heel angle and heading angle of the hovercraft in the northeast coordinate system, is the moment of inertia of the hovercraft around the z-axis, , are the lateral velocity and bow angular velocity of the hovercraft in the motion coordinate system, is the bow turning moment generated by the air rudder, τ uy is the interference of the lateral control channel on the heading control channel, is the heading control channel disturbance; The lateral displacement decoupling control model is: ;

[0012] In the formula, is the lateral control channel disturbance, u y is the lateral displacement control input, y, , respectively represent the lateral position, lateral velocity and bow angular velocity of the hovercraft in the motion coordinate system, m is the mass of the hovercraft, , , To calculate the derived base value of the lateral dimensionless hydrodynamic coefficient, It is the interference of the heading control channel on the lateral control channel.

[0013] The step 3 specifically includes: Step 3-1: Design the state space of all information that affects the hovercraft decision and the action space of all action sets of the hovercraft; Step 3-2: Design reward function: During the intelligent navigation of the hovercraft, the learning goal is to enable the hovercraft to approach the desired track; Step 3-3: Design termination conditions: Design the number of termination steps through experience and testing The training is terminated when the number of steps exceeds 2000.

[0014] The hovercraft speed controller in step 4 is specifically designed as follows: ;

[0015] In the formula, To control the amount, To control the gain, is the controller bandwidth, is the expected speed, , is the observed output of the speed observer.

[0016] The specific design of the air cushion craft bow-lateral decoupling controller in step 5 is as follows: Step 5-1: Simplify the heading decoupling control model in step 2: ;

[0017] In the formula, is the virtual control quantity of the heading control channel, ,in , , is the time-varying coefficient; Step 5-2: Simplify the lateral displacement decoupling control model in step 2: ;

[0018] In the formula, is the virtual control quantity of the lateral control channel, ,in , , is the time-varying coefficient; Step 5-3: Combine the control models in steps 5-1 and 5-2 into a matrix mode to obtain the hovercraft bow-lateral decoupling controller of the bow control channel and the lateral displacement control channel.

[0019] The heading-lateral decoupling controller of the air cushion craft for the heading control channel and the lateral displacement control channel is: ;

[0020] In the formula, is the expected heading error, is the desired heading, , , is the observation output of the heading observer, , , , , is the heading error feedback gain vector, is the virtual control quantity of the heading control channel, is the initial value of the virtual control variable of the heading controller, is the heading control quantity; ;

[0021] In the formula, is the expected lateral position error, , , is the observation output of the lateral position observer, , , , , is the lateral position error feedback gain vector, is the virtual control quantity of the lateral displacement control channel, is the initial value of the virtual control quantity of the lateral position controller, is the lateral position control quantity, is the expected lateral displacement.

[0022] Compared with the prior art, the beneficial effects achieved by the present invention are: 1) In this invention, a bow-lateral displacement decoupling control model based on the strong coupling between the simplified bow control channel and the lateral control channel is proposed, and the coupling relationship between the bow and lateral displacement is clarified, which provides a basis for the design of the subsequent decoupling controller; 2) In the present invention, a hovercraft intelligent motion planning design method based on DDPG algorithm is proposed. The deep neural network used in this method greatly enhances the feature extraction capability. The gradient strategy algorithm can effectively deal with the problem of continuous motion space of the hovercraft. The experience replay technology and the added target network can improve the training stability and convergence speed. The real-time expected navigation information of the hovercraft is obtained according to the operation requirements, and the motion planning is made; 3) In the present invention, an intelligent adaptive linear anti-disturbance control method based on BP neural network is proposed, which realizes the online adjustment of key parameters in the anti-disturbance control, greatly reduces the time for adjusting parameters, and enables the controller to observe disturbances more accurately and compensate for them, effectively avoiding the complexity and non-real-time disadvantages of offline training, improving the control accuracy and rapidity of the speed and bow-lateral decoupling controller, and improving the navigation safety of the hovercraft. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 Schematic diagram of the intelligent motion planning and track control method of the full-cushion hovercraft; Figure 2 Network structure diagram of deep deterministic policy gradient algorithm; Figure 3 Schematic diagram of intelligent adaptive linear anti-disturbance speed controller; Figure 4 Schematic diagram of the intelligent adaptive linear self-disturbance rejection heading-lateral decoupling controller. DETAILED DESCRIPTION

[0025] The present invention proposes a method for intelligent motion planning and track control of a hovercraft based on reinforcement learning, comprising: Step 1: Determine the motion environment information of the full-lift hovercraft, mainly including the navigation information of the hovercraft. Establish a full-lift hovercraft model to obtain the heading, position, speed, and heel of the hovercraft; the hovercraft model includes a kinematic model and a dynamic model; Step 1-1: The kinematic model of the hovercraft is: ;

[0026] In the formula, , , , They represent the north position, east position, heel angle and heading angle of the hovercraft in the northeast coordinate system, , , , They respectively represent the longitudinal velocity, lateral velocity, heel angular velocity and heading angular velocity of the hovercraft in the motion coordinate system.

[0027] Step 1-2: The dynamic model of the hovercraft is: ;

[0028] In the formula, is the mass of the hovercraft, , , , They are the longitudinal force, transverse force, heel moment and bow moment of the hovercraft. , The hovercraft circled x , z The moment of inertia of the shaft; , , , They respectively represent the longitudinal velocity, lateral velocity, heel angular velocity and heading angular velocity of the hovercraft in the motion coordinate system.

[0029] Step 2: Using the heading, position, speed, and heel of the hovercraft obtained in step 1, the strong coupling problem between the heading control channel and the lateral control channel is simplified to obtain a speed control model and a heading-lateral displacement decoupling control model; Specifically, the step 2 specifically includes: Step 2-1: Since the adaptive linear active disturbance rejection control is a control method that is not based on a model, the standard form of the hovercraft speed control model that can be directly obtained is: ;

[0030] In the formula, is the longitudinal velocity of the hovercraft in the motion coordinate system, is the total disturbance of the system, To control the gain, To control the amount.

[0031] Step 2-2: Based on the hovercraft model established in step 1, the bow control model and lateral displacement control model are obtained respectively: ;

[0032] ;

[0033] In the formula, is the bow aerodynamic moment, is the bow air momentum moment, is the bow hydrodynamic moment, is the air rudder turning moment, is the air propeller bow torque, Bow nozzle turning moment, is the lateral aerodynamic force, is the lateral air momentum force, is the lateral hydrodynamic force, is the lateral force of the air rudder, is the lateral thrust of the air propeller, is the lateral thrust of the bow nozzle, For hovercraft z The moment of inertia of the shaft, , represents the heel angle and heading angle of the hovercraft in the northeast coordinate system, m is the mass of the hovercraft, , , , y They respectively represent the longitudinal velocity, lateral velocity, bow angular velocity and lateral position of the hovercraft in the motion coordinate system.

[0034] Since the heading control actuator and the lateral displacement control actuator are the air rudder and the bow nozzle respectively, the heading control input is defined as Define the lateral displacement control input as the bow moment generated by the air rudder is the lateral force generated by the bow nozzle, which is expressed in mathematical form as follows: ;

[0035] In the formula, is the lateral rudder force coefficient, is the air density, is the incoming flow velocity on the air rudder, is the air rudder area, For vertical installation position, is the distance between the pressure center of the air rudder and the rudder axis, is the bow nozzle thrust, is the nozzle angle.

[0036] The air rudder will also generate lateral force, and the bow nozzle will also generate a bow torque, so it is necessary to consider the mutual disturbance between the two control channels; the lateral force generated by the air rudder is defined as the interference of the bow control channel on the lateral control channel τ uψ The bow turning moment generated by the bow nozzle is defined as the interference of the lateral control channel on the bow control channel τ uy , the mathematical form is as follows: ;

[0037] In the formula, This is the installation position of the bow nozzle.

[0038] Then the heading decoupling control model is: ;

[0039] In the formula, , , , , , To calculate the derived base values ​​of the dimensionless hydrodynamic coefficients in the bow and transverse directions, is the longitudinal position of the hull acting on the hydrodynamic force or hydrodynamic moment, , are the heel angle and heading angle of the hovercraft in the northeast coordinate system, is the moment of inertia of the hovercraft around the z-axis, , are the lateral velocity and bow angular velocity of the hovercraft in the motion coordinate system, is the bow turning moment generated by the air rudder, τ uy is the interference of the lateral control channel on the heading control channel, is the heading control channel disturbance; Then the lateral displacement decoupling control model is: ;

[0040] In the formula, is the lateral control channel disturbance, u y is the lateral displacement control input, y, , They represent the lateral position, lateral velocity and bow angular velocity of the hovercraft in the motion coordinate system, m is the mass of the hovercraft, , , To calculate the derived base value of the lateral dimensionless hydrodynamic coefficient, It is the interference of the heading control channel on the lateral control channel.

[0041] Step 3: According to the speed control model in step 2, the bow-lateral decoupling control model and the full-lift hovercraft model in step 1, the hovercraft navigation information is obtained. The Deep Deterministic Policy Gradient (DDPG) in the reinforcement learning algorithm is used to design the intelligent action planning method for the hovercraft navigation process. The algorithm network structure is implemented using the actor-critic algorithm. The specific structure is as follows: Figure 2 As shown, it mainly includes the design or selection of state space, action space, reward function, network structure, and termination condition to obtain the desired speed, desired heading, and desired lateral position of the hovercraft; The step 3 specifically includes: Step 3-1: Design the state space of all information that affects the decision-making of the hovercraft and the action space of all action sets of the hovercraft: In the scenario of the present invention, the required state information mainly includes the attitude of the hovercraft given for special operation requirements, which is represented by vector S as follows: ;

[0042] In the formula, the elements in vector S are x', u', y', v' , '、 r' , '、 p' They are the longitudinal position, longitudinal speed, lateral position, lateral speed, heading angle, heading angular velocity, heel angle and heel angular velocity of the hovercraft under special operation requirements, and they are all continuous variables.

[0043] During the operation of the hovercraft, the action should be designed as the expected speed, expected lateral displacement and expected heading angle of the hovercraft. Therefore, three action information are selected to form the action space, which is expressed by vector A as follows: ;

[0044] In the formula, 、 、 are the expected speed, expected lateral displacement and expected heading angle of the hovercraft respectively.

[0045] Step 3-2: Design reward function: During the intelligent navigation of the hovercraft, the learning goal is to enable the hovercraft to approach the desired track.

[0046] Apply a positive excitation close to the desired position, defined as follows: ; In the formula, is the relative longitudinal distance between the hovercraft and the desired position at the current moment, is a reward coefficient greater than 0, is the relative lateral distance between the hovercraft and the desired position at the current moment, is a reward coefficient greater than 0, r x is the vertical distance reward function, r y is the lateral distance reward function.

[0047] A positive excitation is applied to approach the desired heading direction, which is defined as follows: ;

[0048] In the formula, is the relative heading angle between the hovercraft and the desired heading at the current moment, is a reward coefficient greater than 0, is the heading reward function.

[0049] For the expected speed term, define its reward function: ;

[0050] In the formula, , , are the relative longitudinal velocity, relative lateral velocity and relative heading angular velocity of the hovercraft at the current moment and the desired velocity, , , is a reward coefficient greater than 0, r u is the longitudinal velocity reward function, r v is the lateral velocity reward function, r r is the heading angular velocity reward function.

[0051] The hovercraft should also be kept stable during navigation. Therefore, positive incentives are applied to stable motion, and the more stable it is, the higher the reward value. The variables related to stability are the heel angle and heel angular velocity of the hovercraft, and the corresponding reward function is defined as: ;

[0052] ;

[0053] In the formula, , are the heel angle and heel angular velocity of the hovercraft at the current moment, , are the reward coefficients greater than 0, is the heel angle reward function, is the heel angular velocity reward function.

[0054] In summary, we can get the total reward function r for: ;

[0055] Step 3-3: Set the termination condition: Set the number of termination steps through experience and testing The training is terminated when the number of steps exceeds 2000.

[0056] Step 4: According to the desired speed in step 3 and the speed control model in step 2, design the hovercraft speed controller based on intelligent adaptive linear anti-disturbance control. The specific structure of the controller is as follows: Figure 3 As shown, the longitudinal speed is controlled by the thrust generated by the variable pitch air propeller to ensure the stability of the hovercraft speed; Linear ADRC omits the tracking differentiator and introduces the concept of bandwidth by linearizing and simplifying the nonlinear part of traditional ADRC, greatly reducing the number of parameters that need to be adjusted, and focusing the configuration on the linear extended state observer (ESO) and linear state error feedback (LSEF). The speed controller is rewritten as an extended system as follows: ;

[0057] In the formula, x 1 , x 2 is in an expansion state; u is the longitudinal velocity of the hovercraft in the motion coordinate system, is the total disturbance of the system, To control the gain, To control the amount; Then the linear expansion observer is designed as: ;

[0058] In the formula, , is the velocity error feedback gain vector, , is the observed output of the speed observer, To control the gain, To control the amount; The corresponding hovercraft speed controller is: ;

[0059] In the formula, is the controller bandwidth, The system input is the expected speed, , is the observed output of the speed observer, To control the gain, To control the amount.

[0060] Step 5: According to the desired heading and lateral position in step 3 and the heading-lateral displacement decoupling control model in step 2, the heading-lateral decoupling controller of the hovercraft is designed based on the intelligent adaptive linear anti-disturbance decoupling control algorithm. The specific structure of the controller is as follows: Figure 4 As shown, the bow and lateral positions of the hovercraft are guaranteed to be stable, thereby improving the navigation safety of the hovercraft.

[0061] The step 5 specifically includes: Step 5-1: Simplify the heading decoupling control model in step 2: ;

[0062] In the formula, is the virtual control quantity of the heading control channel, ,in , , is the time-varying coefficient; Step 5-2: Simplify the lateral displacement decoupling control model in step 2: ;

[0063] In the formula, is the virtual control quantity of the lateral control channel, ,in , , is the time-varying coefficient; Step 5-3: Combine the control models in steps 5-1 and 5-2 into a matrix model and use the heading acceleration and lateral acceleration To express: ;

[0064] In the formula, represents the system disturbance and system state uncertainty, is the virtual control quantity matrix; The specific expression is: , ;

[0065] Combining the heading control input and lateral displacement control input in step 2, the coupling system gain matrix is ​​obtained: for: ;

[0066] Its inverse is the conversion matrix between the virtual control amount and the actual control amount. In actual situations, we may encounter In the case of irreversibility, we can find an approximate reversible matrix instead. For vertical installation position, is the distance between the pressure center of the air rudder and the rudder axis, This is the installation position of the bow nozzle.

[0067] According to the virtual control amount of the heading control channel The heading-lateral decoupling controller of the hovercraft with the heading control channel is obtained as: ;

[0068] In the formula, is the expected heading error, is the desired heading, , , is the observation output of the heading observer, , , , , is the heading error feedback gain vector, is the virtual control quantity of the heading control channel, is the initial value of the virtual control variable of the heading controller, is the heading control quantity; According to the virtual control The air cushion craft bow-lateral decoupling controller of the lateral control channel is obtained as: ;

[0069] In the formula, is the expected lateral position error, , , is the observation output of the lateral position observer, , , , , is the lateral position error feedback gain vector, is the virtual control quantity of the lateral displacement control channel, is the initial value of the virtual control quantity of the lateral position controller, is the lateral position control quantity, is the expected lateral displacement.

[0070] The intelligent motion planning method for hovercraft based on DDPG algorithm proposed in the present invention can quickly provide the hovercraft with accurate expected relative speed, expected relative heading and expected relative lateral position, and provide a basis for the design of the lower-level controller; the intelligent adaptive linear anti-disturbance control method proposed in the present invention designs the hovercraft speed controller and the heading-lateral decoupling controller by combining BP neural network and anti-disturbance control, and the designed controller can realize online adjustment of controller parameters, which greatly reduces the adjustment time of parameters, and can also observe disturbances more accurately and compensate for them, so as to meet the control requirements under special operation navigation; the intelligent navigation method proposed in the present invention realizes autonomous track tracking of hovercraft through the mutual cooperation between the motion planning layer and the control layer, which increases the navigation safety of the hovercraft.

[0071] In this specification, the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the embodiments described later, the description is relatively simple, and the relevant parts can be referred to the partial description of the previous embodiments.

[0072] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for intelligent motion planning and track control of a hovercraft based on reinforcement learning, characterized in that: include: Step 1: Determine the motion environment information of the full-lift hovercraft; Establish a full-lift hovercraft model and obtain the hovercraft's heading, position, speed, and heel; Step 2: Using the heading, position, speed, and heel of the hovercraft obtained in step 1, the strong coupling problem between the heading control channel and the lateral control channel is simplified to obtain a speed control model and a heading-lateral displacement decoupling control model; Step 3: According to the speed control model in step 2, the bow-lateral displacement decoupling control model and the full-lift hovercraft model in step 1, the hovercraft navigation information is obtained, and the intelligent action of the hovercraft navigation process is planned by the deep deterministic policy gradient in the reinforcement learning algorithm to obtain the expected speed, expected bow and expected lateral position of the hovercraft; Step 4: According to the expected speed in step 3 and the speed control model in step 2, a hovercraft speed controller based on intelligent adaptive linear anti-disturbance control is designed, and the longitudinal speed is controlled by the thrust generated by the variable pitch air propeller to ensure the stability of the hovercraft speed; Step 5: According to the desired heading and lateral position in step 3 and the heading-lateral displacement decoupling control model in step 2, and based on the intelligent adaptive linear anti-disturbance decoupling control algorithm, the heading-lateral decoupling controller of the hovercraft is designed to ensure the stability of the heading and lateral position of the hovercraft.

2. The method for intelligent motion planning and track control of a hovercraft based on reinforcement learning according to claim 1, characterized in that: The hovercraft model in step 1 includes a kinematic model and a dynamic model; The kinematic model is: ； In the formula, , , , They represent the north position, east position, heel angle and heading angle of the hovercraft in the northeast coordinate system, , , , They respectively represent the longitudinal velocity, lateral velocity, heel angular velocity and heading angular velocity of the hovercraft in the motion coordinate system.

3. The method for intelligent motion planning and track control of a hovercraft based on reinforcement learning according to claim 2, characterized in that: The kinetic model described in step 1 is: ； In the formula, is the mass of the hovercraft, , , , They are the longitudinal force, transverse force, heel moment and bow moment of the hovercraft. , They are the moments of inertia of the hovercraft around the x and z axes respectively; , , , They respectively represent the longitudinal velocity, lateral velocity, heel angular velocity and heading angular velocity of the hovercraft in the motion coordinate system.

4. The method for intelligent motion planning and track control of a hovercraft based on reinforcement learning according to claim 1, characterized in that: The step 2 specifically includes: Step 2-1: Since the adaptive linear active disturbance rejection control is a control method that is not based on a model, the standard form of the hovercraft speed control model that can be directly obtained is: ； In the formula, is the longitudinal velocity of the hovercraft in the motion coordinate system, is the total disturbance of the system, To control the gain, To control the amount; Step 2-2: Based on the hovercraft model established in step 1, the bow control model and lateral displacement control model are obtained respectively: ； ； In the formula, is the bow aerodynamic moment, is the bow air momentum moment, is the bow hydrodynamic moment, is the air rudder turning moment, is the air propeller bow torque, Bow nozzle turning moment, is the lateral aerodynamic force, is the lateral air momentum force, is the lateral hydrodynamic force, is the lateral force of the air rudder, is the lateral thrust of the air propeller, is the lateral thrust of the bow nozzle, For hovercraft z The moment of inertia of the shaft, , represents the heel angle and heading angle of the hovercraft in the northeast coordinate system, m is the mass of the hovercraft, , , , y They respectively represent the longitudinal velocity, lateral velocity, bow angular velocity and lateral position of the hovercraft in the motion coordinate system.

5. The method for intelligent motion planning and track control of a hovercraft based on reinforcement learning according to claim 4, characterized in that: In step 2, the heading control actuator and the lateral displacement control actuator are the air rudder and the bow nozzle respectively. The heading control input is defined as Define the lateral displacement control input as the bow moment generated by the air rudder is the lateral force generated by the bow nozzle, which is expressed in mathematical form as follows: ； In the formula, is the lateral rudder force coefficient, is the air density, is the incoming flow velocity on the air rudder, is the air rudder area, For vertical installation position, is the distance between the pressure center of the air rudder and the rudder axis, is the bow nozzle thrust, is the nozzle angle; The lateral force generated by the air rudder is defined as the interference of the heading control channel on the lateral control channel. τ uψ The bow turning moment generated by the bow nozzle is defined as the interference of the lateral control channel on the bow control channel τ uy , the mathematical form is as follows: ； In the formula, This is the installation position of the bow nozzle.

6. The method for intelligent motion planning and track control of a hovercraft based on reinforcement learning according to claim 5, characterized in that: The heading decoupling control model in step 2 is: ； In the formula , , , , , To calculate the derived base values ​​of the dimensionless hydrodynamic coefficients in the bow and transverse directions, is the longitudinal position of the hull acting on the hydrodynamic force or hydrodynamic moment, , are the heel angle and heading angle of the hovercraft in the northeast coordinate system, is the moment of inertia of the hovercraft around the z-axis, , are the lateral velocity and bow angular velocity of the hovercraft in the motion coordinate system, is the bow moment generated by the air rudder, τ uy is the interference of the lateral control channel on the heading control channel, is the heading control channel disturbance; The lateral displacement decoupling control model is: ； In the formula, is the lateral control channel disturbance, u y is the lateral displacement control input, y, , They represent the lateral position, lateral velocity and bow angular velocity of the hovercraft in the motion coordinate system, m is the mass of the hovercraft, , , To calculate the derived base value of the lateral dimensionless hydrodynamic coefficient, It is the interference of the heading control channel on the lateral control channel.

7. The method for intelligent motion planning and track control of a hovercraft based on reinforcement learning according to claim 1, characterized in that: The step 3 specifically includes: Step 3-1: Design the state space of all information that affects the hovercraft decision and the action space of all action sets of the hovercraft; Step 3-2: Design reward function: During the intelligent navigation of the hovercraft, the learning goal is to enable the hovercraft to approach the desired track; Step 3-3: Design termination conditions: Design the number of termination steps through experience and testing The training is terminated when the number of steps exceeds 2000.

8. The method for intelligent motion planning and track control of a hovercraft based on reinforcement learning according to claim 1, characterized in that: The hovercraft speed controller in step 4 is specifically designed as follows: ； In the formula, To control the amount, To control the gain, is the controller bandwidth, is the expected speed, , is the observed output of the speed observer.

9. The method for intelligent motion planning and track control of a hovercraft based on reinforcement learning according to claim 6, characterized in that: The specific design of the air cushion craft bow-lateral decoupling controller in step 5 is as follows: Step 5-1: Simplify the heading decoupling control model in step 2: ； In the formula, is the virtual control quantity of the heading control channel, ,in , , is the time-varying coefficient; Step 5-2: Simplify the lateral displacement decoupling control model in step 2: ； In the formula, is the virtual control quantity of the lateral control channel, ,in , , is the time-varying coefficient; Step 5-3: Combine the control models in steps 5-1 and 5-2 into a matrix mode to obtain the hovercraft bow-lateral decoupling controller of the bow control channel and the lateral displacement control channel.

10. The method for intelligent motion planning and track control of a hovercraft based on reinforcement learning according to claim 9, characterized in that: The heading-lateral decoupling controller of the air cushion craft for the heading control channel and the lateral displacement control channel is: ； In the formula, is the expected heading error, is the desired heading, , , is the observation output of the heading observer, , , , , is the heading error feedback gain vector, is the virtual control quantity of the heading control channel, is the initial value of the virtual control variable of the heading controller, is the heading control quantity; ； In the formula, is the expected lateral position error, , , is the observation output of the lateral position observer, , , , , is the lateral position error feedback gain vector, is the virtual control quantity of the lateral displacement control channel, is the initial value of the virtual control quantity of the lateral position controller, is the lateral position control quantity, is the expected lateral displacement.

Images