An unmanned ship autonomous navigation method and decision system based on real-time perception of laser radar

Abstract

The application discloses an unmanned ship autonomous navigation method and decision system based on laser radar real-time sensing, relates to the field of unmanned ship autonomous navigation, and comprises the following steps: collecting state information of an unmanned ship, terminal position information and two-dimensional laser radar data, and calculating position information of the terminal relative to the unmanned ship; fusing radar feature information extracted from the normalized two-dimensional laser radar data after normalizing the state information of the unmanned ship and the position information of the terminal relative to the unmanned ship; inputting the fused information into a trained unmanned ship autonomous navigation model, outputting expected yaw angles and expected speeds of the unmanned ship, combining a kinematics model and a dynamics model of the unmanned ship to calculate the propulsion speeds of each propeller; and judging whether the unmanned ship reaches the terminal, if yes, ending the navigation, and if not, continuing the navigation process. The application can realize autonomous learning navigation in unknown or dynamic environments, and has stronger environmental adaptability and decision optimization capability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of autonomous navigation of unmanned ships, in particular to an unmanned ship autonomous navigation method and decision system based on real-time perception of laser radar. BACKGROUND

[0002] With the increasing development of marine resources, the complexity and risk of maritime tasks have significantly increased, which poses unprecedented challenges to the autonomy and intelligence level of maritime unmanned systems. As a key component of maritime unmanned systems, unmanned ships have great potential in high-risk and high-difficulty tasks such as port cruising and environmental monitoring due to their high automation, excellent maneuverability and multi-functional characteristics. However, in complex marine environments, unmanned ships not only need to navigate to the preset target accurately, but also need to perceive the surrounding environment in real time and effectively avoid various marine obstacles to ensure the safety and efficiency of task execution. Therefore, it is of great significance to study an efficient and reliable autonomous navigation method to improve the application efficiency and practical deployment value of unmanned ships.

[0003] Currently, the autonomous navigation of unmanned ships mainly relies on path planning methods based on global information, such as ant colony optimization (ACO) and artificial potential field method (APF). Ant colony optimization simulates the cooperative behavior of ant colonies and finds the optimal path based on global environmental information, which is suitable for path planning in static or known environments. The artificial potential field method constructs a navigation potential field using the attractive force of the target point and the repulsive force of the obstacles, making the unmanned ship move in the direction of gradient descent. However, such methods usually rely on complete or partial known environmental maps and are difficult to adapt to dynamic and uncertain marine environments. For example, in complex water areas, environmental information may change at any time, making it difficult for global map-based planning methods to adjust the path in a timely manner, thereby affecting the navigation efficiency and safety.

[0004] Therefore, how to enable unmanned ships to navigate autonomously in unknown map information, achieve flexible and intelligent obstacle avoidance and target navigation, and improve environmental adaptability and task execution reliability is a technical problem that needs to be solved by those skilled in the art. SUMMARY

[0005] Therefore, the present application provides an unmanned ship autonomous navigation method and decision system based on real-time perception of laser radar, which solves the problems in the background art.

[0006] To achieve the above purpose, the present application provides the following technical solutions:

[0007] An unmanned ship autonomous navigation method based on real-time perception of laser radar, comprising the following steps:

[0008] S1: Collecting unmanned ship state information, end position information and two-dimensional laser radar data;

[0009] S2: Calculate the position information of the end point relative to the unmanned ship based on the unmanned ship state information and the end point position information;

[0010] S3: Normalize the two-dimensional laser radar data and extract features from the normalized two-dimensional laser radar data using a deep neural network to obtain radar feature information;

[0011] S4: Normalize the unmanned ship state information and the position information of the end point relative to the unmanned ship, and fuse with the radar feature information;

[0012] S5: Input the fused information into the trained unmanned ship autonomous navigation model to output the desired yaw angle and desired speed of the unmanned ship;

[0013] S6: Input the desired yaw angle and desired speed of the unmanned ship into the bottom controller of the unmanned ship, combine the kinematic model and dynamic model of the unmanned ship to calculate the propulsion speed of each propeller;

[0014] S7: Determine whether the unmanned ship has reached the end point, if so, the navigation ends, otherwise return to S1 to continue the navigation process.

[0015] Optionally, in S1, the unmanned ship state information includes the coordinates (x u , y u ) of the unmanned ship, the yaw angle ψ u and the speed (u u , v u , r u ) of the unmanned ship; the end point position information is the coordinates (x t , y t ) of the end point; the two-dimensional laser radar data is represented as is the measurement data of the laser radar with an angular range of n s , which provides n L sets of angle-distance pairs for each scan, where n L represents the number of ranging beams of the radar, and the maximum detection distance of the laser radar is L max .

[0016] Optionally, in S2, the position information of the end point relative to the unmanned ship includes the distance d t between the unmanned ship and the end point and the azimuth angle ψ t of the end point relative to the unmanned ship, and the specific calculation formula is:

[0017]

[0018] The deviation between the yaw angle ψ u of the unmanned ship and the azimuth angle ψ t of the end point relative to the unmanned ship is the relative heading angle δψ for:

[0019] δ Ψ =Ψ u -Ψ t

[0020] In the formula: (x t y t (x) represents the coordinates of the endpoint. u y u () represents the coordinates of the unmanned vessel.

[0021] Optionally, in S3, the two-dimensional LiDAR data is normalized, specifically as follows:

[0022] Distance information is obtained by dividing by the maximum detection range L of the lidar. max Normalization is performed by dividing the angle information by the angular range n of the lidar. s Normalize.

[0023] Optionally, in S4, the unmanned vessel's state information and the destination's position information relative to the unmanned vessel are normalized, specifically as follows:

[0024] The speed of the unmanned vessel (u) u ,v u ,r u ) and bow roll angle ψ u Normalization is performed by dividing by the corresponding maximum value, and the coordinates (x) of the unmanned vessel are obtained. u ,y u The distance d between the unmanned vessel and the destination is normalized by dividing by the width W and length L of the map. t Normalized by dividing by the length of the map's diagonal, the relative heading angle δ ψ Normalization is achieved by dividing by 2π.

[0025] Optionally, in S5, the autonomous navigation model of the unmanned vessel is trained based on a dual-delay depth deterministic policy gradient framework, specifically including the following steps:

[0026] S51. Definition of State Space: The state space is the input to the autonomous navigation model of the unmanned vessel, containing all the information upon which decisions depend, and is defined as follows:

[0027] s = <o u o t o a o l >

[0028] In the formula: o u This represents the real-time status information of the unmanned vessel, including its location information (x). u ,y u ,ψ u) and speed information (u u ,v u ,r u );o t This represents the relative position information of the destination, specifically the distance d between the unmanned vessel and the destination. t and relative heading angle δ ψ ;o a This indicates the actions performed by the unmanned vessel at the previous time step. o l This represents the feature information of the lidar after dimensionality reduction processing;

[0029] S52. Action Space Definition: The action space is the output of the unmanned vessel's autonomous navigation model, compatible with the characteristics of the underlying controller, and is defined as follows:

[0030] a = d , ψ d >

[0031] In the formula: u d ψ represents the desired oscillation velocity. d Indicates the desired bow roll angle;

[0032] Considering the constraint characteristics of the underlying controller, the following constraints are imposed on the motion space:

[0033]

[0034] In the formula: u max ψ represents the maximum sway velocity of the unmanned vessel. max Indicates the maximum bow roll angle of the unmanned vessel;

[0035] S53. Set a reward mechanism: The reward function is a decision-driven mechanism that guides the unmanned vessel to perform the expected behavior pattern.

[0036] Optionally, in S53, the reward mechanism is set based on the state space and action space, specifically including:

[0037] 1) Target Point Reward Mechanism: Target point rewards are set to incentivize unmanned vessels to navigate to target points; target point rewards must meet the following conditions:

[0038]

[0039] In the formula: r safe The radius of the safe zone for the unmanned surface vessel (USV) is used to determine whether the USV has reached the target point or has collided with it; when the distance d between the USV and the destination... t Less than or equal to r safe When the unmanned vessel is deemed to have successfully reached the target point, a positive reward r is given. arrival ;

[0040] ​2) Collision Penalty Mechanism: Collision penalties are set to encourage unmanned vessels to avoid obstacles and maintain safe navigation; collision penalties must meet the following conditions:

[0041]

[0042] In the formula: d o This represents the distance between the unmanned surface vessel (USV) and the obstacle; when the distance d between the USV and the obstacle... o Less than or equal to the safe distance r safe When a collision is detected, a negative reward r is applied. collision ;

[0043] 3) Timeout Penalty Mechanism: A timeout penalty is set to prevent unmanned vessels from executing inefficient strategies; the timeout penalty must meet the following conditions:

[0044]

[0045] In the formula: t o This indicates the set time threshold; if the current time step t is greater than or equal to t o If the timeout occurs, a negative reward r will be applied. timeout ;

[0046] 4) Boundary Crossing Penalty Mechanism: A boundary crossing penalty is set to ensure that unmanned vessels navigate within designated areas; the boundary crossing penalty must meet the following conditions:

[0047]

[0048] In the formula: |x e |and|y e | represents the boundary range of the navigation area; when the current position of the unmanned vessel (x u y u When the boundary is exceeded, a negative reward r is applied. boundary ;

[0049] 5) Potential Field Reward Mechanism: A potential field reward is set to guide the unmanned surface vessel to actively avoid obstacles while approaching the finish line; the potential field reward meets the following conditions:

[0050]

[0051] R p =R att +R rep

[0052] In the formula: k att k represents the gravitational potential coefficient. rep Represents the repulsive potential field coefficient; the unmanned surface vessel receives a positive reward R from the gravitational field at each time step. att The negative reward R provided by the repulsive field rep ;

[0053] 6) Motion smoothing reward mechanism: A motion smoothing reward is set to suppress abrupt changes in the motion of the unmanned vessel between adjacent time steps; the motion smoothing reward satisfies the following conditions:

[0054]

[0055] In the formula: u d and ψ d These represent the actions taken at the current time step, u′ and u′ respectively. d and ψ′ d These represent the actions taken in the previous time step, σ and σ'. u and σ ψ Indicates parameter adjustment;

[0056] Based on the above reward and punishment mechanisms, the final reward function is defined as follows:

[0057] R=λ0R d +λ1R o +λ2R e +λ3R b +λ4R p +λ5R a

[0058] In the formula: λ0, λ1, λ2, λ3, λ4, and λ5 represent the corresponding weight coefficients.

[0059] Optionally, in S6, the unmanned surface vessel's underlying controller uses PID control, with the following expression:

[0060]

[0061] Where: K p K i and K d Let e(t) represent the proportional coefficient, integral coefficient, and derivative coefficient, respectively; and let w(t) and w(t) represent the input and output of the PID controller, respectively.

[0062]

[0063] w(t) is used to calculate the propulsion force on the port and starboard sides of the unmanned vessel, and then the overall propulsion force τ of the unmanned vessel is obtained:

[0064]

[0065] In the formula: T p and T s These represent the propulsion forces on the port and starboard sides, respectively, and B represents the center distance of the unmanned vessel's side hull.

[0066] The three-degree-of-freedom model of the unmanned vessel is as follows:

[0067]

[0068] In the formula: η=[x,y,ψ] T and v = [u, v, r] T Let J(η) represent the pose vector and velocity vector of the unmanned surface vessel (USV), respectively; J(η) represents the rotation matrix, used to transform variables in the geographic coordinate system to the fixed coordinate system of the hull; M represents the inertia matrix; C(v) represents the Coriolis-centrifugal force matrix of the USV; D(v) represents the damping matrix; G(η) represents the hydrostatic matrix; and τ represents the hydrostatic force matrix. wind and τ wave These represent the torque caused by wind and waves;

[0069] Based on the overall propulsion force τ of the unmanned vessel and the three-degree-of-freedom model of the unmanned vessel, the kinematic equations of the unmanned vessel are solved to achieve motion control.

[0070] An autonomous decision-making system for unmanned surface vessels (USVs) based on real-time lidar perception, executing any one of the above-described autonomous navigation methods for USVs based on real-time lidar perception, comprising:

[0071] The data acquisition module is used to collect unmanned vessel status information, destination location information, and two-dimensional lidar data;

[0072] The calculation module is used to calculate the position information of the destination relative to the unmanned vessel using the unmanned vessel's status information and the destination's position information;

[0073] The first data processing module is used to normalize the two-dimensional lidar data and use a deep neural network to extract features from the normalized two-dimensional lidar data to obtain radar feature information.

[0074] The second data processing module is used to normalize the unmanned vessel status information and the position information of the endpoint relative to the unmanned vessel, and fuse them with radar feature information.

[0075] The mapping module is used to input the fused information into the trained unmanned vessel autonomous navigation model and output the unmanned vessel's desired bow roll angle and desired speed.

[0076] The motion control module is used to input the desired bow angle and desired speed of the unmanned vessel into the unmanned vessel's underlying controller, and calculate the propulsion speed of each thruster by combining the kinematic and dynamic models of the unmanned vessel;

[0077] The judgment module is used to confirm whether the unmanned vessel has reached the destination. If it has reached the destination, the navigation ends; otherwise, the navigation process continues.

[0078] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses an autonomous navigation method and decision system for unmanned vessels based on real-time perception of lidar, which gets rid of the dependence on global environmental information and enables unmanned vessels to achieve autonomous navigation in unknown or dynamic environments, with stronger adaptability and robustness; secondly, by introducing deep reinforcement learning, unmanned vessels can autonomously learn efficient navigation strategies and achieve real-time path optimization in complex environments without relying on traditional heuristic rules. Attached Figure Description

[0079] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0080] Figure 1 A flowchart of the autonomous navigation method for unmanned vessels based on real-time lidar perception provided by the present invention;

[0081] Figure 2 This is a schematic diagram of the endpoint relative information conversion and radar scanning provided by the present invention;

[0082] Figure 3 The mathematical model of the three degrees of freedom of the unmanned vessel provided by this invention. Detailed Implementation

[0083] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0084] To enhance the autonomous navigation capability of unmanned surface vessels (USVs) in complex marine environments, this invention discloses an autonomous navigation method for USVs based on real-time lidar perception, such as... Figure 1 As shown, it includes the following steps:

[0085] S1: Collect unmanned vessel status information, destination location information, and two-dimensional lidar data;

[0086] Specifically, the unmanned surface vessel (USV) status information includes the USV's coordinates (x, y, y). u ,y u ), bow roll angle ψ u and the speed of the unmanned ship (u u v u ru The endpoint location information is the coordinates (x, y) of the endpoint. t y t Two-dimensional lidar data is represented as follows: The range of angles is n s The measurement data from the lidar provides n per scan. L A set of angular distance pairs, where n L The number of ranging beams represents the range of the lidar; the maximum detection range of the lidar is L. max .

[0087] S2: Calculate the position information of the destination relative to the unmanned vessel based on the unmanned vessel's state information and the destination's position information;

[0088] The location information of the endpoint relative to the unmanned vessel includes the distance d between the unmanned vessel and the endpoint. t And the azimuth angle ψ of the endpoint relative to the unmanned vessel t See Figure 2 The specific calculation formula is as follows:

[0089]

[0090] Unmanned vessel bow roll angle ψ u The azimuth angle ψ of the unmanned vessel relative to the endpoint t The deviation between them is the relative heading angle δ ψ for:

[0091] δ ψ =ψ u -ψ t

[0092] In the formula: (x t y t (x) represents the coordinates of the endpoint. u y u () represents the coordinates of the unmanned vessel.

[0093] S3: Normalize the two-dimensional lidar data and use a deep neural network to extract features from the normalized two-dimensional lidar data to obtain radar feature information, thereby achieving dimensionality reduction of high-dimensional radar data, preserving key environmental information, and improving computational efficiency.

[0094] In this embodiment, the distance information is obtained by dividing by the maximum detection range L of the lidar. max Normalization is performed by dividing the angle information by the angular range n of the lidar. s Normalization is performed to eliminate the influence of measurement scale.

[0095] S4: Normalize the unmanned vessel status information and the position information of the destination relative to the unmanned vessel, and fuse them with radar feature information;

[0096] In this embodiment, the speed of the unmanned vessel (x) u v u r u ) and bow roll angle ψ u Normalization is performed by dividing by the corresponding maximum value to ensure that the velocity and angle values ​​are normalized to a fixed range; the coordinates (x, y) of the unmanned vessel are normalized. u y u Normalization is performed by dividing by the map's width W and length L to ensure scale consistency of location features; the distance d between the unmanned vessel and the destination. t Normalized by dividing by the length of the map's diagonal, the relative heading angle δ ψ Normalization is achieved by dividing by 2π. The normalization operation aims to process the data to the range of [-1, 1], ensuring that the input data has a consistent numerical scale, thereby improving the training stability and convergence of the neural network model.

[0097] S5: Input the fused information into the trained unmanned vessel autonomous navigation model and output the unmanned vessel's desired bow roll angle and desired speed;

[0098] In this embodiment, the autonomous navigation model of the unmanned vessel is trained based on the dual-delay depth deterministic policy gradient (TD3) framework, specifically including the following steps:

[0099] S51. Definition of State Space: The state space is the input to the autonomous navigation model of the unmanned vessel, containing all the information upon which decisions depend, and is defined as follows:

[0100] s = <o u ,o t o a o l >

[0101] In the formula: o u This represents the real-time status information of the unmanned vessel, including its location information (x). u y u , ψ u ) and speed information (u u v u r u );o t This represents the relative position information of the destination, specifically the distance d between the unmanned vessel and the destination. t and relative heading angle δ ψ ;o a This indicates the actions performed by the unmanned vessel at the previous time step. o l This represents the feature information of the lidar after dimensionality reduction processing;

[0102] S52. Action Space Definition: The action space is the output of the unmanned vessel's autonomous navigation model, compatible with the characteristics of the underlying controller, and is defined as follows:

[0103] a = d ,ψ d >

[0104] In the formula: u d ψ represents the desired oscillation velocity. d Indicates the desired bow roll angle;

[0105] Considering the constraint characteristics of the underlying controller, the following constraints are imposed on the motion space:

[0106]

[0107] In the formula: u max The maximum pitch speed of the unmanned vessel is represented by ψ, which is set to 1. max The maximum bow roll angle of the unmanned vessel is represented by π.

[0108] S53. Setting a reward mechanism: The deep reinforcement learning model is used for autonomous navigation of unmanned vessels. The reward function is a decision-driven mechanism, which aims to guide the unmanned vessel to perform the expected behavior pattern.

[0109] The reward mechanism is set based on the state space and action space to optimize training effectiveness and ensure safety, specifically including:

[0110] 1) Target Point Reward Mechanism: Target point rewards are set to incentivize unmanned vessels to navigate to target points; target point rewards must meet the following conditions:

[0111]

[0112] In the formula: r safe The radius of the safe zone for the unmanned surface vessel (USV) is used to determine whether the USV has reached the target point or has collided with it; when the distance d between the USV and the destination... t Less than or equal to r safe When the unmanned vessel is deemed to have successfully reached the target point, a positive reward r is given. arrival ;

[0113] 2) Collision Penalty Mechanism: Collision penalties are set to encourage unmanned vessels to avoid obstacles and maintain safe navigation; collision penalties must meet the following conditions:

[0114]

[0115] In the formula: d o This represents the distance between the unmanned surface vessel (USV) and the obstacle; when the distance d between the USV and the obstacle... o Less than or equal to the safe distance r safe ​When a collision is detected, a negative reward r is applied. collision ;

[0116] 3) Timeout Penalty Mechanism: A timeout penalty is set to prevent unmanned vessels from executing inefficient strategies; the timeout penalty must meet the following conditions:

[0117]

[0118] In the formula: t o This indicates the set time threshold; if the current time step t is greater than or equal to t o If the timeout occurs, a negative reward r will be applied. timeout ;

[0119] 4) Boundary Crossing Penalty Mechanism: A boundary crossing penalty is established to ensure that unmanned vessels navigate within designated areas and prevent aimless navigation; the boundary crossing penalty must meet the following conditions:

[0120]

[0121] In the formula: |x e |and|y e | represents the boundary range of the navigation area; when the current position of the unmanned vessel (x u y u When the boundary is exceeded, a negative reward r is applied. boundary ;

[0122] 5) Potential Field Reward Mechanism: A potential field reward is set to guide the unmanned surface vessel to actively avoid obstacles while approaching the destination, thereby improving navigation safety and efficiency; the potential field reward meets the following conditions:

[0123]

[0124] R p =R att +R rep

[0125] In the formula: k att k represents the gravitational potential coefficient. rep Represents the repulsive potential field coefficient; the unmanned surface vessel receives a positive reward R from the gravitational field at each time step. att The negative reward R provided by the repulsive field rep ;

[0126] 6) Motion Smoothing Reward Mechanism: A motion smoothing reward is set to suppress abrupt changes in the unmanned vessel's motion between adjacent time steps, ensuring the stability of the output motion; the motion smoothing reward satisfies the following conditions:

[0127]

[0128] In the formula: ud and ψ d These represent the actions taken at the current time step, u′ and u′ respectively. d and ψ′ d These represent the actions taken in the previous time step, σ and σ'. u and σ ψ Indicates parameter adjustment;

[0129] Based on the above reward and punishment mechanisms, the final reward function is defined as follows:

[0130] R=λ0R d +λ1R o +λ2R e +λ3R b +λ4R p +λ5R a

[0131] In the formula: λ0, λ1, λ2, λ3, λ4, and λ5 represent the corresponding weight coefficients.

[0132] S6: Input the desired bow angle and desired speed of the unmanned vessel into the unmanned vessel's underlying controller, and calculate the propulsion speed of each thruster by combining the kinematic and dynamic models of the unmanned vessel, so as to achieve precise control of the unmanned vessel;

[0133] In this embodiment, the unmanned surface vessel's underlying controller uses a PID controller, with the following expression:

[0134]

[0135] Where: K p K i and K d Let e(t) represent the proportional coefficient, integral coefficient, and derivative coefficient, respectively; and let w(t) and w(t) represent the input and output of the PID controller, respectively.

[0136]

[0137] w(t) is used to calculate the propulsion force on the port and starboard sides of the unmanned vessel, and then the overall propulsion force τ of the unmanned vessel is obtained:

[0138]

[0139] In the formula: T p and T s These represent the propulsion forces on the port and starboard sides, respectively, and B represents the center distance of the unmanned vessel's side hull.

[0140] Reference Figure 3 The three-degree-of-freedom model of the unmanned vessel is as follows:

[0141]

[0142] In the formula: η = [x, y, ψ] T and v = [u, v, r] T Let J(η) represent the pose vector and velocity vector of the unmanned surface vessel (USV), respectively; J(η) represents the rotation matrix, used to transform variables in the geographic coordinate system to the fixed coordinate system of the hull; M represents the inertia matrix; C(v) represents the Coriolis-centrifugal force matrix of the USV; D(v) represents the damping matrix; G(η) represents the hydrostatic matrix; and τ represents the hydrostatic force matrix. wind and τ wave These represent the disturbance torque caused by wind and waves, respectively. Based on the overall propulsion force τ of the unmanned vessel and the three-degree-of-freedom model of the unmanned vessel, the kinematic equations of the unmanned vessel are solved to achieve motion control.

[0143] S7: Determine if the unmanned vessel has reached the destination. If it has, navigation ends; otherwise, return to S1 to continue the navigation process.

[0144] and Figure 1 Corresponding to the method described above, this embodiment of the invention also provides an autonomous decision-making system for unmanned vessels based on real-time lidar perception, used for... Figure 1 The specific implementation of the method, as provided in this embodiment of the invention, is an autonomous decision-making system for unmanned vessels based on real-time lidar perception, which can be applied to computer terminals or various mobile devices, specifically including:

[0145] The data acquisition module is used to collect unmanned vessel status information, destination location information, and two-dimensional lidar data;

[0146] The calculation module is used to calculate the position information of the destination relative to the unmanned vessel using the unmanned vessel's status information and the destination's position information;

[0147] The first data processing module is used to normalize the two-dimensional lidar data and use a deep neural network to extract features from the normalized two-dimensional lidar data to obtain radar feature information.

[0148] The second data processing module is used to normalize the unmanned vessel status information and the position information of the endpoint relative to the unmanned vessel, and fuse them with radar feature information.

[0149] The mapping module is used to input the fused information into the trained unmanned vessel autonomous navigation model and output the unmanned vessel's desired bow roll angle and desired speed.

[0150] The motion control module is used to input the desired bow angle and desired speed of the unmanned vessel into the unmanned vessel's underlying controller, and calculate the propulsion speed of each thruster by combining the kinematic and dynamic models of the unmanned vessel;

[0151] The judgment module is used to confirm whether the unmanned vessel has reached the destination. If it has reached the destination, the navigation ends; otherwise, the navigation process continues.

[0152] Compared to traditional methods such as ant colony optimization and artificial potential field methods that rely on global information, this embodiment employs an autonomous navigation method based on deep reinforcement learning, enabling the unmanned surface vessel (USV) to navigate autonomously even without map information. By perceiving the environment in real time and optimizing decisions using a deep reinforcement learning model, the USV can dynamically adjust its trajectory in complex and uncertain marine environments, achieving flexible and intelligent obstacle avoidance and target navigation. Furthermore, to further enhance environmental perception capabilities, this embodiment uses a two-dimensional lidar to acquire high-precision point cloud data, allowing the USV to make autonomous decisions based on local perception information even without global information support, thereby improving its environmental adaptability and mission execution reliability.

[0153] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0154] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An autonomous navigation method for unmanned vessels based on real-time lidar perception, characterized in that, Includes the following steps: S1: Collect unmanned vessel status information, destination location information, and two-dimensional lidar data; S2: Calculate the position information of the destination relative to the unmanned vessel based on the unmanned vessel's state information and the destination's position information; S3: Normalize the two-dimensional lidar data and use a deep neural network to extract features from the normalized two-dimensional lidar data to obtain radar feature information. S4: Normalize the unmanned vessel status information and the position information of the destination relative to the unmanned vessel, and fuse them with radar feature information; S5: Input the fused information into the trained unmanned vessel autonomous navigation model and output the unmanned vessel's desired bow roll angle and desired speed; S6: Input the desired bow roll angle and desired velocity of the unmanned ship into the unmanned ship's underlying controller, and calculate the propulsion speed of each thruster by combining the kinematic and dynamic models of the unmanned ship; S7: Determine whether the unmanned vessel has reached the destination. If it has, navigation ends; otherwise, return to S1 to continue the navigation process. In S5, the autonomous navigation model of the unmanned vessel is trained based on a dual-delay depth deterministic policy gradient framework. Includes the following steps: S51. Definition of State Space: The state space is the input to the autonomous navigation model of the unmanned vessel, containing all the information upon which decisions depend, and is defined as follows: In the formula: This indicates the real-time status information of the unmanned vessel, including its location information. and speed information ; This indicates the relative position information of the destination, i.e., the distance between the unmanned vessel and the destination. and relative heading angle ; This indicates the actions performed by the unmanned vessel at the previous time step. ; This represents the feature information of the lidar after dimensionality reduction processing. S52. Action Space Definition: The action space is the output of the unmanned vessel's autonomous navigation model, compatible with the characteristics of the underlying controller, and is defined as follows: In the formula: Indicates the desired oscillation speed. Indicates the desired bow roll angle; Considering the constraint characteristics of the underlying controller, the following constraints are imposed on the motion space: In the formula: This indicates the maximum pitch speed of the unmanned vessel. Indicates the maximum bow roll angle of the unmanned vessel; S53. Set a reward mechanism: The reward function is a decision-driven mechanism that guides the unmanned vessel to perform the expected behavior pattern.

2. The autonomous navigation method for unmanned vessels based on real-time lidar perception according to claim 1, characterized in that, In S1, the unmanned surface vessel's status information includes its coordinates. Bow roll angle and the speed of unmanned ships The endpoint location information is the coordinates of the endpoint. Two-dimensional lidar data is represented as The range of angles is The measurement data from the lidar provides [data] for each scan. Group of angular distance pairs, where, The number of ranging beams represents the range of the radar; the maximum detection range of the lidar is... .

3. The autonomous navigation method for unmanned vessels based on real-time lidar perception according to claim 2, characterized in that, In S2, the position information of the endpoint relative to the unmanned vessel includes the distance between the unmanned vessel and the endpoint. and the azimuth of the endpoint relative to the unmanned vessel The specific calculation formula is as follows: Unmanned boat bow roll Azimuth angle of the unmanned vessel relative to the destination The deviation between them is the relative heading angle. for: In the formula: The coordinates of the endpoint The coordinates are for the unmanned vessel.

4. The autonomous navigation method for unmanned vessels based on real-time lidar perception according to claim 1, characterized in that, In S3, the two-dimensional lidar data is normalized, specifically by dividing the distance information by the lidar's maximum detection range. Normalization is performed by dividing the angle information by the angular range of the lidar. Normalize.

5. The autonomous navigation method for unmanned vessels based on real-time lidar perception according to claim 3, characterized in that, In S4, the unmanned vessel's state information and the endpoint's position information relative to the unmanned vessel are normalized. Specifically, this includes the unmanned vessel's speed. and bow angle The coordinates of the unmanned surface vessel are normalized by dividing by the corresponding maximum value. By dividing by the width of the map W and length L Normalization is performed on the distance between the unmanned vessel and the destination. Normalized by dividing by the length of the map's diagonal, relative heading angle By dividing Normalize.

6. The autonomous navigation method for unmanned vessels based on real-time lidar perception according to claim 1, characterized in that, In S53, the reward mechanism is set based on the state space and action space, specifically including: 1) Target Point Reward Mechanism: Target point rewards are set to incentivize unmanned vessels to navigate to target points; target point rewards must meet the following conditions: In the formula: This represents the safe zone radius of the unmanned surface vessel (USV), used to determine whether the USV has reached the target point or has collided with it; the distance between the USV and the destination... Less than or equal to When the unmanned vessel is deemed to have successfully reached the target point, a positive reward is given. ; 2) Collision Penalty Mechanism: Collision penalties are set to encourage unmanned vessels to avoid obstacles and maintain safe navigation; collision penalties must meet the following conditions: In the formula: This indicates the distance between the unmanned surface vessel (USV) and an obstacle; when the distance between the USV and the obstacle... Less than or equal to the safe distance When a collision occurs, a negative reward is applied. ; 3) Timeout Penalty Mechanism: A timeout penalty is set to prevent unmanned vessels from executing inefficient strategies; the timeout penalty must meet the following conditions: In the formula: This indicates the set time threshold; if the current time step t Greater than or equal to If the timeout occurs, a negative reward will be applied. ; 4) Boundary Crossing Penalty Mechanism: A boundary crossing penalty is set to ensure that unmanned vessels navigate within designated areas; the boundary crossing penalty must meet the following conditions: In the formula: and These represent the boundary range of the navigation area; and the current position of the unmanned vessel. Apply a negative reward when the boundary is exceeded. ; 5) Potential Field Reward Mechanism: A potential field reward is set to guide the unmanned surface vessel to actively avoid obstacles while approaching the finish line; the potential field reward meets the following conditions: In the formula: Represents the gravitational potential field coefficient. This represents the repulsive potential field coefficient; the unmanned surface vessel receives a positive reward from the gravitational field at each time step. and the negative rewards provided by the repulsive field ; 6) Motion smoothing reward mechanism: A motion smoothing reward is set to suppress abrupt changes in the actions of the unmanned vessel between adjacent time steps; the motion smoothing reward satisfies the following conditions: In the formula: and These represent the actions taken at the current time step. and These represent the actions taken in the previous time step. and Indicates parameter adjustment; Combining the above reward and penalty mechanisms, the final reward function is defined as follows: In the formula: , , , , , These represent the corresponding weighting coefficients.

7. The autonomous navigation method for unmanned vessels based on real-time lidar perception according to claim 1, characterized in that, In S6, the unmanned surface vessel's underlying controller uses PID, with the following expression: In the formula: , and These represent the proportional coefficient, integral coefficient, and differential coefficient, respectively. and Let the input and output of the PID be represented as follows: The propulsion force on the port and starboard sides of the unmanned vessel is obtained through differential drive calculations, and thus the overall propulsion force of the unmanned vessel is obtained. : In the formula: and These represent the propulsion forces on the port and starboard sides, respectively. B Indicates the center distance of the side hull of the unmanned vessel; The three-degree-of-freedom model of the unmanned vessel is as follows: In the formula: and These represent the pose vector and velocity vector of the unmanned surface vessel, respectively. This represents a rotation matrix used to transform variables in a geographic coordinate system to a fixed coordinate system of the ship's hull. M Represents the inertia matrix. Represents the Coriolis-centrifugal force matrix of the unmanned vessel. Represents the damping matrix. Represents the hydrostatic matrix. and These represent the torque caused by wind and waves; and the overall propulsion force of the unmanned vessel. The kinematic equations of the unmanned vessel are solved using a three-degree-of-freedom model to achieve motion control.

8. An autonomous decision-making system for unmanned vessels based on real-time lidar perception, characterized in that, The method for autonomous navigation of unmanned vessels based on real-time lidar perception as described in any one of claims 1-7 includes: The data acquisition module is used to collect unmanned vessel status information, destination location information, and two-dimensional lidar data; The calculation module is used to calculate the position information of the destination relative to the unmanned vessel using the unmanned vessel's status information and the destination's position information; The first data processing module is used to normalize the two-dimensional lidar data and use a deep neural network to extract features from the normalized two-dimensional lidar data to obtain radar feature information. The second data processing module is used to normalize the unmanned vessel status information and the position information of the endpoint relative to the unmanned vessel, and fuse them with radar feature information. The mapping module is used to input the fused information into the trained unmanned vessel autonomous navigation model and output the unmanned vessel's desired bow roll angle and desired speed. The motion control module is used to input the desired bow angle and desired speed of the unmanned vessel into the unmanned vessel's underlying controller, and calculate the propulsion speed of each thruster by combining the kinematic and dynamic models of the unmanned vessel; The judgment module is used to confirm whether the unmanned vessel has reached the destination. If it has reached the destination, the navigation ends; otherwise, the navigation process continues.

Images