Unmanned ship path tracking method, device and electronic equipment based on path gradient information

Abstract

The present application relates to the field of unmanned ship motion control technology, and specifically discloses an unmanned ship path tracking method, device and electronic equipment based on path gradient information, which constructs a motion mathematical model of the unmanned ship based on the unmanned ship's forward speed, angular velocity and lateral velocity, takes the position of a point on the axis of the unmanned ship as the control object, uses the expected path to establish a scalar field, and adjusts the angular velocity of the unmanned ship according to the error between the current position of the unmanned ship and the expected path and the gradient information of the scalar field, thereby reducing the error until it moves to the expected path. The method is not affected by the side slip problem, has good robustness, can make the position error exponential converge, and can maintain the same performance when tracking straight lines and curves.

Description

Technical Field

[0001] The present application relates to the technical field of unmanned ship motion control, and more specifically, to an unmanned ship path tracking method, device and electronic equipment based on path gradient information. Background Art

[0002] Traditional shrimp and crab farming usually requires manual management and monitoring, but there are problems such as high labor costs and low production efficiency. With the development of unmanned ship technology, people have begun to explore the application of unmanned ship technology in shrimp and crab farming to improve farming efficiency, reduce labor demand, reduce costs, and achieve more intelligent farming management. The motion control of unmanned ships is the most basic function to ensure the operation of unmanned shrimp and crab farming ships. In-depth research on motion control algorithms is needed to improve the navigation performance, safety and efficiency of unmanned ships and cope with complex environments and mission requirements.

[0003] At present, mainstream path tracking navigation controllers, such as the line of sight method and the tracking method, all use the idea of ​​geometric projection to solve the underactuation problem of unmanned ships, that is, converting the tracking error into the desired angle, and indirectly converge the tracking error to zero by controlling the angle deviation convergence. However, the geometric projection method is affected by the side slip problem, and will produce tracking deviations when the unmanned ship has a lateral speed, and the tracking effect is significantly different when tracking straight lines and curves, and its control performance has room for improvement.

[0004] Therefore, a method, device and electronic device for unmanned ship path tracking based on path gradient information are expected. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide an unmanned ship path tracking method, device and electronic device based on path gradient information, which constructs a mathematical model of the motion of the unmanned ship based on the forward speed, angular velocity and lateral velocity of the unmanned ship, takes the position of the point on the axis of the unmanned ship as the control object, and uses the expected path to establish a scalar field. The angular velocity of the unmanned ship is adjusted according to the error between the current position of the unmanned ship and the expected path and the gradient information of the scalar field, thereby reducing the error until it moves to the expected path. This method is not affected by the side slip problem, has good robustness, can make the position error exponential converge, and can maintain the same performance when tracking straight lines and curves.

[0006] Accordingly, according to one aspect of the present application, a method for tracking an unmanned ship path based on path gradient information is provided, which includes:

[0007] Step 1: Construct a mathematical model of the motion of the unmanned ship. The mathematical model is: ; ; ,in, Indicates the position of the unmanned ship in the world coordinate system, represents the angle of the unmanned ship, Indicates the speed of the unmanned ship in the forward direction, represents the angular velocity of the unmanned ship turning, represents the lateral speed of the unmanned ship, , and Indicates the position of the unmanned ship in the world coordinate system The time derivative of the unmanned ship and its position in the world coordinate system The time derivative of and the angle of the unmanned ship The time derivative of

[0008] Step 2: Describe the expected path of the unmanned ship in the form of an implicit function, the implicit function is in the form of ;

[0009] Step 3: Define a point on the axis of the unmanned ship ,in, ; , Indicates the position of the center of mass of the unmanned ship in the world coordinate system, for Relative to and calculate the Euclidean distance of the point Scalar value in a scalar field And the scalar value As this point The error from the expected path;

[0010] Step 4: Calculate the point The scalar field gradient of the point The scalar field gradient and the error are used to confirm the angular velocity of the unmanned ship's steering.

[0011] According to another aspect of the present application, a path tracking device for an unmanned ship based on path gradient information is provided, comprising:

[0012] The motion mathematical model building module is used to build the motion mathematical model of the unmanned ship. The motion mathematical model is ; ; ,in, Indicates the position of the unmanned ship in the world coordinate system, represents the angle of the unmanned ship, Indicates the speed of the unmanned ship in the forward direction, represents the angular velocity of the unmanned ship turning, represents the lateral speed of the unmanned ship, , and Indicates the position of the unmanned ship in the world coordinate system The time derivative of the unmanned ship and its position in the world coordinate system The time derivative of and the angle of the unmanned ship The time derivative of

[0013] The expected path description module is used to describe the expected path of the unmanned ship in the form of an implicit function, wherein the implicit function is in the form of ;

[0014] A path error calculation module is used to define a point on the axis of the unmanned ship. ,in, ; , Indicates the position of the center of mass of the unmanned ship in the world coordinate system, for Relative to and calculate the Euclidean distance of the point Scalar value in a scalar field And the scalar value As this point The error from the expected path;

[0015] Angular velocity confirmation module, used to calculate the point The scalar field gradient of the point The scalar field gradient and the error are used to confirm the angular velocity of the unmanned ship's steering.

[0016] According to another aspect of the present application, an electronic device is provided, comprising: a memory for storing instructions; a processor coupled to the memory, the processor being configured to execute the unmanned ship path tracking method based on path gradient information as described above based on the instructions stored in the memory.

[0017] Compared with the prior art, the unmanned ship path tracking method, device and electronic device provided by the present application based on path gradient information constructs a mathematical model of the motion of the unmanned ship based on the forward speed, angular velocity and lateral velocity of the unmanned ship, takes the position of the point on the axis of the unmanned ship as the control object, uses the expected path to establish a scalar field, and adjusts the angular velocity of the unmanned ship according to the error between the current position of the unmanned ship and the expected path and the gradient information of the scalar field, thereby reducing the error until it moves to the expected path. This method is not affected by the side slip problem, has good robustness, can make the position error exponential converge, and can maintain the same performance when tracking straight lines and curves. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 is a flow chart of an unmanned ship path tracking method based on path gradient information according to Embodiment 1 of the present application;

[0020] Figure 2 Schematic diagram of the scalar field formed by the expected path implicit function;

[0021] Figure 3 A curve diagram showing the motion trajectory of an unmanned ship and the expected path in the unmanned ship path tracking method of Example 1 of the present application in simulation verification;

[0022] Figure 4 A curve diagram of the position tracking error of the unmanned ship in the unmanned ship path tracking method of Example 1 of the present application in simulation verification;

[0023] Figure 5 is a block diagram of an unmanned ship path tracking device based on path gradient information according to an embodiment of the present application;

[0024] Figure 6 is a schematic diagram of the structure of an electronic device according to an embodiment of the present application;

[0025] Figure 7 is a flow chart of an unmanned ship path tracking method based on path gradient information according to Embodiment 2 of the present application;

[0026] Figure 8 This is a flowchart of step 4 in the unmanned ship path tracking method based on path gradient information according to Example 2 of the present application. DETAILED DESCRIPTION

[0027] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0028] Example 1

[0029] Figure 1 FIG. 1 is a flow chart of an unmanned ship path tracking method based on path gradient information according to Embodiment 1 of the present application. Figure 1As shown, the unmanned ship path tracking method based on path gradient information includes: step 1: constructing a mathematical model of the unmanned ship's motion; step 2: describing the expected path of the unmanned ship in the form of an implicit function; step 3: defining a point on the axis of the unmanned ship, and calculating the scalar value of the point in the scalar field and using the scalar value as the error between the point and the expected path; step 4: calculating the scalar field gradient of the point, and confirming the angular velocity of the unmanned ship's turning based on the scalar field gradient of the point and the error.

[0030] In the above-mentioned unmanned ship path tracking method based on path gradient information, the step 1: constructs a motion mathematical model of the unmanned ship. Specifically, the motion mathematical model is ; ; ,in, Indicates the position of the unmanned ship in the world coordinate system, represents the angle of the unmanned ship, Indicates the speed of the unmanned ship in the forward direction, represents the angular velocity of the unmanned ship turning, represents the lateral speed of the unmanned ship, , and Indicates the position of the unmanned ship in the world coordinate system The time derivative of the unmanned ship and its position in the world coordinate system The time derivative of and the angle of the unmanned ship The time derivative of .

[0031] This application describes the motion state of the unmanned ship in a two-dimensional plane by establishing a mathematical model, and associates the motion position of the unmanned ship with information such as angle and speed in mathematical form, so that the motion trajectory of the unmanned ship can be accurately controlled by controlling the speed parameters of the unmanned ship.

[0032] In the above-mentioned unmanned ship path tracking method based on path gradient information, the step 2: describing the expected path of the unmanned ship in the form of an implicit function. Wherein, the implicit function is in the form of The implicit function can describe various complex curves and straight lines. By changing the form of the function, different desired paths, such as straight lines, curves, etc., can be represented. For the implicit function, the coordinates of any point on the plane are Substituting this function, we get a value, so this function forms a scalar field, in which the set of points with a value of 0 is the desired path.

[0033] In one embodiment of the present application, the selected desired path is . The coordinates of any point on the plane Bring in the function The scalar field formed is Figure 2 As shown, the set of points with a value of 0 in this scalar field is the desired path.

[0034] In the above-mentioned unmanned ship path tracking method based on path gradient information, the step 3: defines a point on the axis of the unmanned ship, and calculates the scalar value of the point in the scalar field and uses the scalar value as the error between the point and the expected path. Specifically, define a point on the axis of the unmanned ship : ; ,in, Indicates the position of the center of mass of the unmanned ship in the world coordinate system, for Relative to The Euclidean distance of is a small length compared to the length of the ship. Substituting into the implicit function, we get the point Scalar value in a scalar field , since the expected path is , therefore, when When it is not equal to 0, it means that the unmanned ship has deviated from the expected path. , indicating that the point is on one side of the expected path; , then on the other side, and The size of reflects the degree of deviation. Based on this, the point The scalar value of the scalar field is used as the error between the point and the expected path. The error can be used to measure the degree of deviation between the current position of the unmanned ship and the expected path, and then the angular velocity of the unmanned ship can be adjusted to reduce the error, so that the unmanned ship gradually approaches and stabilizes on the expected path.

[0035] In the above-mentioned unmanned ship path tracking method based on path gradient information, the step 4: calculates the scalar field gradient of the point, and based on the scalar field gradient of the point and the error, determines the angular velocity of the unmanned ship turning. It should be understood that the gradient reflects the direction and rate of the fastest change of the scalar field at the point. By calculating the scalar field At the point The gradient at , we can get the changing trend of the expected path near the current position of the unmanned ship. According to the gradient information, we can design the expected speed in the world coordinate system, so that When the expected speed points to the direction of gradient descent, When the expected speed points to the direction of the gradient rise, so that you can return to location.

[0036] Specifically, the design of the point The scalar field gradient and the error satisfy the following relationship: ,in, is a positive constant parameter, the size of which determines the speed of error convergence. In one embodiment of the present application, The expected velocity vector is obtained by solving the equation , the expected velocity vector is When , it points to the direction of gradient descent. When pointing to the direction of the gradient rise, the unmanned ship can return to the destination no matter where it is currently. The position of the unmanned ship is thus realized, thereby realizing the path tracking of the unmanned ship.

[0037] Further, the point The relationship between the scalar field gradient and the error is transformed into the hull coordinate system:

[0038]

[0039]

[0040]

[0041]

[0042] in, is the angular velocity of the unmanned ship turning. Measured by sensors such as the Global Positioning System, specifying the forward speed , solve the equation , you can get the desired angular velocity command In one embodiment of the present application, the forward speed is specified .

[0043] Since the path tracking task only requires the convergence of the path lateral error, which only occupies one degree of freedom, and the underactuated unmanned ship has two degrees of freedom, the path tracking problem can be converted into an optimization problem, and the expected forward speed is used as an optimization target. The relationship between the scalar field gradient and the error in the hull coordinate system is: As constraints, the optimization problem can be written as:

[0044]

[0045] st

[0046] Using this optimization problem as a framework, add speed control instructions to the optimization problem and The optimization goals , or add speed instructions and Inequality constraints , forming an optimization problem of the form:

[0047]

[0048] st

[0049]

[0050] The extra degree of freedom can be used to complete other mission objectives except the specified forward speed.

[0051] In one embodiment of the present application, the added optimization target is another point on the ship axis. Located on the desired path to avoid hitting banks in narrow farming areas, where ; ,Right now ,in

[0052]

[0053]

[0054]

[0055]

[0056] The constraint condition is set as the forward speed is no more than 0.6 m / s to ensure uniform feed distribution. The optimization problem is:

[0057]

[0058] st

[0059]

[0060] Solve the optimization problem and get the speed command and angular velocity command It can realize path tracking and complete the two tasks of preventing the unmanned shrimp and crab farming boat from hitting the embankment in the narrow farming area and ensuring the forward speed is not more than 0.6 m / s to ensure uniform feeding.

[0061] The embodiment is verified in the simulation platform, and the simulation results are as follows: Figure 3 to Figure 4 shown. Figure 3 The motion trajectory and expected path curve of the unmanned ship under the path tracking control method of the embodiment of the present application are shown in FIG. Figure 4 It is a position error curve diagram. It can be seen from the simulation results that under the path tracking control method of the embodiment of the present application, the position error index converges to 0, and the tracking performance in the straight section and the curved section remains consistent. The simulation results show the effectiveness of the path tracking control method described in the present application.

[0062] In summary, the unmanned ship path tracking method based on path gradient information according to the embodiment of the present application is explained, which constructs the mathematical model of the motion of the unmanned ship based on the forward speed, angular velocity and lateral velocity of the unmanned ship, takes the position of the point on the axis of the unmanned ship as the control object, and uses the expected path to establish a scalar field, and adjusts the angular velocity of the unmanned ship according to the error between the current position of the unmanned ship and the expected path and the gradient information of the scalar field, thereby reducing the error until it moves to the expected path. This method is not affected by the side slip problem, has good robustness, can make the position error exponential converge, and can maintain the same performance when tracking straight lines and curves.

[0063] Figure 5 FIG. 1 is a block diagram of an unmanned ship path tracking device based on path gradient information according to an embodiment of the present application. Figure 5 As shown, according to the unmanned ship path tracking device 100 based on path gradient information according to the embodiment of the present application, it includes: a motion mathematical model construction module 110, which is used to construct a motion mathematical model of the unmanned ship, and the motion mathematical model is ; ; ,in, Indicates the position of the unmanned ship in the world coordinate system, represents the angle of the unmanned ship, Indicates the speed of the unmanned ship in the forward direction, represents the angular velocity of the unmanned ship turning, represents the lateral speed of the unmanned ship, , and Indicates the position of the unmanned ship in the world coordinate system The time derivative of the unmanned ship and its position in the world coordinate system The time derivative of and the angle of the unmanned ship θ The expected path description module 120 is used to describe the expected path of the unmanned ship in the form of an implicit function, wherein the implicit function is ; Path error calculation module 130, used to define a point on the axis of the unmanned ship ,in, ; , Indicates the position of the center of mass of the unmanned ship in the world coordinate system, for Relative to and calculate the Euclidean distance of the point Scalar value in a scalar field And the scalar value As this point Angular velocity confirmation module 140 is used to calculate the error between the point The scalar field gradient of the point The scalar field gradient and the error are used to confirm the angular velocity of the unmanned ship's steering.

[0064] Here, those skilled in the art can understand that the specific operations of each step in the unmanned ship path tracking device based on path gradient information have been described in the above reference. Figures 1 to 4 The description of the unmanned ship path tracking method based on path gradient information has been introduced in detail, and therefore, its repeated description will be omitted.

[0065] Furthermore, the present application also provides an electronic device applied to an unmanned ship, and the electronic device can implement the path tracking method as described above. The structure of the electronic device is as follows: Figure 6 As shown, it includes a memory, a processor, and an input and output interface connected by a system bus.

[0066] The memory is used to store an operating system and a computer program. The computer program includes program instructions. When the program instructions are executed, the processor can execute an unmanned ship path tracking method based on path gradient information.

[0067] The processor is used to provide computing and control capabilities to support the operation of the electronic device. In an embodiment, the processor can be a single chip microcomputer, a programmable gate array, a digital signal processor, etc.

[0068] The input and output interfaces are used to communicate with other devices, including a timer interface output to the unmanned ship actuator, a serial communication interface for communicating with the global positioning system and the inertial navigation unit, and a network interface for communicating with the host computer.

[0069] All or part of the above technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling an electronic device to execute all or part of the steps of the unmanned ship path tracking method based on path gradient information described in this application.

[0070] Example 2

[0071] In particular, considering that only one point on the ship's axis is used as the control object, it may not be possible to fully reflect the posture changes of the entire ship. For example, when an unmanned ship is sailing in a complex environment, such as when affected by wind and waves or water currents, the ship may rotate or tilt, resulting in a single point error that is insufficient to accurately describe the overall position deviation of the ship. Therefore, in order to more accurately control the navigation trajectory of the unmanned ship, this application introduces a data processing technology based on deep learning on the basis of the above-mentioned scheme, using the position information of multiple points on the ship's axis as the control object, and intelligently optimizing the angular velocity adjustment strategy of the unmanned ship by comprehensively analyzing the errors between multiple positions on the ship's axis and the desired path and the gradient information of the scalar field.

[0072] Figure 7 FIG. 2 is a flow chart of an unmanned ship path tracking method based on path gradient information according to Embodiment 2 of the present application. Figure 7 As shown, the unmanned ship path tracking method based on path gradient information includes: Step 1: constructing a motion mathematical model of the unmanned ship, the motion mathematical model is ; ; ,in, Indicates the position of the unmanned ship in the world coordinate system, represents the angle of the unmanned ship, Indicates the speed of the unmanned ship in the forward direction, represents the angular velocity of the unmanned ship turning, represents the lateral speed of the unmanned ship, , and Indicates the position of the unmanned ship in the world coordinate system The time derivative of the unmanned ship and its position in the world coordinate system The time derivative of and the angle of the unmanned ship Step 2: Describe the expected path of the unmanned ship in the form of an implicit function, the implicit function is ; Step 3: Define the point set on the axis of the unmanned ship , and calculate each point in the point set The error between the expected path and the error point is calculated to obtain a set of errors. scalar field gradient to obtain a set of scalar field gradients; Step 4: Based on the set of errors and the set of scalar field gradients, determine the angular velocity of the unmanned ship turning.

[0073] It should be understood that when the unmanned ship is actually sailing, based on the unmanned ship motion model and the scalar field definition of the expected path, different position points on the ship axis will behave differently in the scalar field, that is, the magnitude and direction of the error and scalar field gradient will be different. Therefore, the present application selects multiple position points along the ship axis (for example, at equal intervals) Select For each point, according to its position (in, ； ， Indicates the position of the center of mass of the unmanned ship in the world coordinate system, for Relative to ), calculate the Euclidean distance of the point Scalar value in a scalar field As this point The error from the expected path , forming a set of errors. At the same time, for each point Calculate its scalar field gradient, that is , and we get the set of scalar field gradients.

[0074] Figure 8 FIG. 4 is a flow chart of step 4 in the unmanned ship path tracking method based on path gradient information according to Embodiment 2 of the present application. Figure 8 As shown, step 4 includes: S41, merging the set of errors and the set of scalar field gradients to obtain a set of {errors, scalar field gradients}; S42, performing low-dimensional embedding coding based on a fully connected layer on the set of {errors, scalar field gradients} to obtain a set of error-joint scalar field gradient low-dimensional embedding coding vectors; S43, performing feature dynamic compensation aggregation on the set of error-joint scalar field gradient low-dimensional embedding coding vectors to obtain an error-joint scalar field gradient global significant aggregation coding feature vector; S44, inputting the error-joint scalar field gradient global significant aggregation coding feature vector into an angular velocity estimation module based on a decoder to obtain the angular velocity of the unmanned ship steering.

[0075] Specifically, in step S41, the set of errors and the set of scalar field gradients are combined to obtain a set of {errors, scalar field gradients}. Here, by making a one-to-one correspondence between the error of each point and the scalar field gradient, a set of {errors, scalar field gradients} is formed, so as to integrate the error information and scalar field gradient information of each point, and fully display the relationship between different positions on the axis of the unmanned ship and the expected path, thereby avoiding the limitations of single-point information and more accurately describing the overall position deviation of the hull.

[0076] Specifically, in step S42, the set of {error, scalar field gradient} is subjected to low-dimensional embedding coding based on a fully connected layer to obtain a set of error-scalar field gradient low-dimensional embedding coding vectors. It should be understood that the error represents the degree of deviation of the point from the expected path. , indicating that the point is on one side of the expected path; , then on the other side, the scalar field gradient reflects the direction and rate of the fastest change of the scalar field at that point. Among them, the positive or negative error determines the way the gradient direction is used, that is, When , the expected speed is in the direction of gradient descent, making the unmanned ship close to the expected path; When , the unmanned boat is guided to regress along the direction of gradient ascent. Therefore, in order to effectively learn the relationship between the error and scalar field gradient of each point, the present application adopts a fully connected layer based on deep learning to associate and encode {error, scalar field gradient} of each point. It should be known that the fully connected layer is the basic structure of the neural network, and its internal neurons are connected to all elements of the input data, so that the complex relationship between the elements in the input data can be captured. In this application, a fully connected layer is used to map the error and scalar field gradient information of each point to a low-dimensional space. While retaining the key information of the original error and scalar field gradient, the nonlinear transformation of the fully connected layer can capture the complex relationship between the error and the scalar field gradient, thereby extracting the joint feature representation of the two, that is, the error combined with the scalar field gradient low-dimensional embedded coding vector, so as to more comprehensively reflect the comprehensive deviation between each position point on the axis of the unmanned ship and the expected path.

[0077] Specifically, in step S43, the set of the error-joint scalar field gradient low-dimensional embedded coding vectors is subjected to feature dynamic compensation aggregation to obtain the error-joint scalar field gradient global significant aggregated coding feature vector. It should be understood that after the above processing, the joint feature representation between the error information and the scalar field gradient information of each position point is effectively captured. In order to further achieve the accurate description and control of the overall position deviation of the unmanned ship, it is necessary to perform information aggregation processing on the set of the error-joint scalar field gradient low-dimensional embedded coding vectors. In particular, considering that different position points have different influences on the overall path planning of the unmanned ship, that is, the error and scalar field gradient information of certain key position points contribute more to the path adjustment of the unmanned ship, and this difference in importance will change with time and environmental changes. Therefore, in order to more accurately guide the navigation of the unmanned ship, the present application proposes a feature dynamic compensation aggregation method, which finds the hub features of the set of low-dimensional embedded coding vectors of the error joint scalar field gradient, and dynamically adjusts the information weights of different position points based on the hub features, thereby strengthening the information of key position points and weakening the information of non-key position points, so as to obtain a more comprehensive and representative error joint scalar field gradient global significant aggregation coding feature representation, and more accurately describe the relationship between the unmanned ship as a whole and the expected path.

[0078] In a specific example of the present application, the step S43 includes: first, extracting the hub feature of the set of the error joint scalar field gradient low-dimensional embedding coding vectors to obtain the error joint scalar field gradient hub feature coding vector, which is expressed as:

[0079]

[0080] in, represents the hub feature extraction network, represents the set of low-dimensional embedding coding vectors of the error joint scalar field gradient, , , and They represent the first, second, and third in the set of low-dimensional embedded coding vectors of the error joint scalar field gradient respectively. and The error is combined with the scalar field gradient low-dimensional embedding encoding vector, is the number of low-dimensional embedding coding vectors of the error joint scalar field gradient, and Respectively represent the weight parameter matrix and bias term of the hub feature extraction network, represents the hub feature relevance score transformation vector of the hub feature extraction network, Indicates the The corresponding hub feature relevance scoring factor, represents matrix multiplication, represents the normalized exponential function, Indicates the The corresponding normalized hub feature relevance score factor, Represents the error joint scalar field gradient pivot feature encoding vector.

[0081] That is, firstly, a hub feature extraction network is constructed to perform hub feature correlation analysis on each error joint scalar field gradient low-dimensional embedding coding vector to learn the key position information that affects the unmanned ship's navigation path tracking, and the key position features of the set of the error joint scalar field gradient low-dimensional embedding coding vectors are extracted and the global information is integrated, thereby comprehensively considering the multi-point position deviation information on the ship's axis, extracting the core hub features that affect the overall path tracking of the unmanned ship, and obtaining the error joint scalar field gradient hub feature coding vector.

[0082] Next, the complementary information of each error joint scalar field gradient low-dimensional embedded coding vector relative to the error joint scalar field gradient hub feature coding vector in the set of the error joint scalar field gradient low-dimensional embedded coding vector is extracted to obtain a set of error joint scalar field gradient position node-hub feature complementary information embedded coding vectors, which is expressed as follows:

[0083]

[0084] in, represents the Sigmoid activation function, represents the normalized error joint scalar field gradient low-dimensional embedding encoding vector, represents the normalized error joint scalar field gradient pivot feature encoding vector, and represents different weight matrices, represents point convolutional coding, represents the error joint scalar field gradient-pivot feature differential encoding vector, represents the difference operation, To take the absolute value, Indicates the The corresponding error joint scalar field gradient position node-hub feature complementary information is embedded in the encoding vector.

[0085] Here, considering that only using the error joint scalar field gradient hub feature coding vector as the basis for unmanned ship navigation control may ignore some useful information contained in non-hub features. Therefore, in order to avoid the loss of useful information that may be caused in the process of hub feature extraction, the present application further integrates and associates the useful information in each error joint scalar field gradient low-dimensional embedded coding vector with the error joint scalar field gradient hub feature coding vector as the core, so as to capture the subtle information differences between all position points, thereby achieving a more refined feature description of the overall navigation state of the unmanned ship. Specifically, first, by calculating the complementary information of each error joint scalar field gradient low-dimensional embedded coding vector relative to the error joint scalar field gradient hub feature coding vector, the information difference and complementarity of each error joint scalar field gradient low-dimensional embedded coding vector relative to the hub feature is captured, and a set of error joint scalar field gradient position node-hub feature complementary information embedded coding vectors is generated, so as to restore the feature components of each error joint scalar field gradient low-dimensional embedded coding vector that are not considered and absorbed in the hub feature extraction process, and reveal the complementarity and difference between the features of each position point.

[0086] Secondly, each error joint scalar field gradient position node-hub feature complementary information embedding coding vector in the set of the error joint scalar field gradient position node-hub feature complementary information embedding coding vector is respectively marked with complementary information significance to obtain a set of error joint scalar field gradient complementary information attention weights, which is expressed by the formula:

[0087]

[0088]

[0089] in, express The corresponding significant identification factor is and They represent the weight parameter matrix and bias term of the complementary information saliency identification module, represents the complementary information saliency score conversion vector of the complementary information saliency identification module, represents the exponential function operation with e as the base, express The corresponding error joint scalar field gradient complementary information attention weight.

[0090] That is, the attention mechanism is further used to evaluate the importance of each error joint scalar field gradient position node-hub feature complementary information embedded coding vector, so as to dynamically adjust the weight distribution of each error joint scalar field gradient position node-hub feature complementary information embedded coding vector in the information aggregation process, so that more critical and decisive information can obtain greater weight in the final unmanned ship navigation control decision, thereby realizing the fine characterization and differentiated processing of the deviations of each position point of the unmanned ship, and improving the accuracy and adaptability of the unmanned ship path tracking.

[0091] Then, based on the set of attention weights of the error joint scalar field gradient complementary information, the set of error joint scalar field gradient position node-hub feature complementary information embedded coding vectors is modulated to obtain a set of significant modulated error joint scalar field gradient position node-hub feature complementary information embedded coding vectors, which is expressed by the formula:

[0092]

[0093] in, , , and Respectively , , and The corresponding error joint scalar field gradient position node-hub feature complementary information embedding encoding vector, , , and Respectively , , and The corresponding error joint scalar field gradient complementary information attention weight, A set of encoding vectors that represent the complementary information of the node-hub features embedded in the significant modulation error joint scalar field gradient position.

[0094] In this way, through the selective enhancement or suppression of features, it is possible to strengthen key information and weaken non-key information, thereby reducing the interference of noise and redundant information on the navigation control decision of the unmanned ship while retaining useful information, and obtaining a set of significant modulation error joint scalar field gradient position node-hub feature complementary information embedded coding vectors.

[0095] Finally, the error joint scalar field gradient hub feature encoding vector and the significant modulation error joint scalar field gradient position node-hub feature complementary information embedding encoding vector are fused to obtain the error joint scalar field gradient global significant aggregation encoding feature vector, which is expressed as:

[0096]

[0097] in, represents a cascade function, Denotes the error joint scalar field gradient global significant aggregate encoding feature vector.

[0098] That is, through the feature cascade method, the set of the significant modulation error joint scalar field gradient position node-hub feature complementary information embedded coding vectors is fused with the error joint scalar field gradient hub feature coding vector, so as to comprehensively consider the overall path deviation of the unmanned ship and the local detail features of each position point, and generate an error joint scalar field gradient aggregation coding representation vector, thereby improving the accuracy and comprehensiveness of the feature expression, and providing a more reliable and accurate basis for subsequent angular velocity adjustment and desired path tracking.

[0099] Specifically, in step S44, the error combined with the scalar field gradient global significant aggregation coding feature vector is input into the angular velocity estimation module based on the decoder to obtain the angular velocity of the unmanned ship turning. It should be understood that the decoder is a structure corresponding to the encoder in deep learning, and its function is to decode the low-dimensional coding features into outputs with actual physical meanings. In the present application, a multi-layer neural network structure is used inside the decoder, and the key feature information reflecting the path deviation of the unmanned ship is gradually parsed and extracted by transferring and transforming the error combined with the scalar field gradient global significant aggregation coding feature vector layer by layer, and based on this, the error combined with the scalar field gradient global significant aggregation coding feature vector is mapped to the angular velocity numerical space, and the angular velocity value required for the unmanned ship to turn is calculated by the regression algorithm at the output layer of the decoder, so as to reflect the degree of angle adjustment required for the unmanned ship to track the desired path. In this way, the accurate control and adjustment of the navigation state of the unmanned ship can be achieved, and the accuracy and stability of the unmanned ship path tracking can be improved.

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for unmanned ship path tracking based on path gradient information, characterized in that: include: Step 1: Construct a mathematical model of the motion of the unmanned ship. The mathematical model is: ; ; ,in, Indicates the position of the unmanned ship in the world coordinate system, represents the angle of the unmanned ship, Indicates the speed of the unmanned ship in the forward direction, represents the angular velocity of the unmanned ship turning, represents the lateral speed of the unmanned ship, , and Indicates the position of the unmanned ship in the world coordinate system The time derivative of the unmanned ship and its position in the world coordinate system The time derivative of and the angle of the unmanned ship θ The time derivative of Step 2: Describe the expected path of the unmanned ship in the form of an implicit function, the implicit function is in the form of ; Step 3: Define a point on the axis of the unmanned ship ,in, ; , Indicates the position of the center of mass of the unmanned ship in the world coordinate system, for Relative to and calculate the Euclidean distance of the point Scalar value in a scalar field And the scalar value As this point The error from the expected path; Step 4: Calculate the point The scalar field gradient of the point The scalar field gradient and the error are used to determine the angular velocity of the unmanned ship turning; The scalar field gradient is , the point The scalar field gradient and the error satisfy the following relationship: in, is the expected velocity vector, is a constant parameter; Transform the relationship to the ship coordinate system: in, is the angular velocity of the unmanned ship turning.

2. A method for unmanned ship path tracking based on path gradient information, characterized in that: include: Step 1: Construct a mathematical model of the motion of the unmanned ship. The mathematical model is: ; ; ,in, Indicates the position of the unmanned ship in the world coordinate system, represents the angle of the unmanned ship, Indicates the speed of the unmanned ship in the forward direction, represents the angular velocity of the unmanned ship turning, represents the lateral speed of the unmanned ship, , and Indicates the position of the unmanned ship in the world coordinate system The time derivative of the unmanned ship and its position in the world coordinate system The time derivative of and the angle of the unmanned ship θ The time derivative of Step 2: Describe the expected path of the unmanned ship in the form of an implicit function, the implicit function is in the form of ; Step 3: Define the point set on the axis of the unmanned ship , and calculate each point in the point set The error between the expected path and the error point is calculated to obtain a set of errors. The scalar field gradient of the point set is The error between the expected path and ; Step 4: Determine the angular velocity of the unmanned ship turning based on the set of errors and the set of scalar field gradients, including: Combining the set of errors and the set of scalar field gradients to obtain a set of {errors, scalar field gradients}; Performing low-dimensional embedding coding based on a fully connected layer on the set of {error, scalar field gradient} to obtain a set of error-joint scalar field gradient low-dimensional embedding coding vectors; Performing feature dynamic compensation aggregation on the set of the error-joint scalar field gradient low-dimensional embedded coding vectors to obtain an error-joint scalar field gradient global significant aggregation coding feature vector; The error combined with the scalar field gradient global significant aggregation encoding feature vector is input into a decoder-based angular velocity estimation module to obtain the angular velocity of the unmanned ship steering.

3. The unmanned ship path tracking method based on path gradient information according to claim 2, characterized in that: The set of the error-joint scalar field gradient low-dimensional embedded coding vectors is subjected to feature dynamic compensation aggregation to obtain an error-joint scalar field gradient global significant aggregation coding feature vector, including: Extracting the pivot features of the set of error joint scalar field gradient low-dimensional embedding coding vectors to obtain an error joint scalar field gradient pivot feature coding vector; Extracting complementary information of each error joint scalar field gradient low-dimensional embedded coding vector in the set of the error joint scalar field gradient low-dimensional embedded coding vector relative to the error joint scalar field gradient hub feature coding vector to obtain a set of error joint scalar field gradient position node-hub feature complementary information embedded coding vectors; Performing complementary information significance marking on each error joint scalar field gradient position node-hub feature complementary information embedding coding vector in the set of error joint scalar field gradient position node-hub feature complementary information embedding coding vectors to obtain a set of error joint scalar field gradient complementary information attention weights; Based on the set of attention weights of the error joint scalar field gradient complementary information, the set of error joint scalar field gradient position node-hub feature complementary information embedded coding vectors is subjected to attention modulation to obtain a set of significant modulated error joint scalar field gradient position node-hub feature complementary information embedded coding vectors; The error joint scalar field gradient hub feature coding vector and the set of the significant modulation error joint scalar field gradient position node-hub feature complementary information embedding coding vectors are fused to obtain the error joint scalar field gradient global significant aggregation coding feature vector.

4. An unmanned ship path tracking device based on path gradient information, which can be used to execute the unmanned ship path tracking method based on path gradient information according to claim 1, characterized in that: include: The motion mathematical model building module is used to build the motion mathematical model of the unmanned ship. The motion mathematical model is ; ; ,in, Indicates the position of the unmanned ship in the world coordinate system, represents the angle of the unmanned ship, Indicates the speed of the unmanned ship in the forward direction, represents the angular velocity of the unmanned ship turning, represents the lateral speed of the unmanned ship, , and Indicates the position of the unmanned ship in the world coordinate system The time derivative of the unmanned ship and its position in the world coordinate system The time derivative of and the angle of the unmanned ship The time derivative of The expected path description module is used to describe the expected path of the unmanned ship in the form of an implicit function, wherein the implicit function is in the form of ; A path error calculation module is used to define a point on the axis of the unmanned ship. ,in, ; , Indicates the position of the center of mass of the unmanned ship in the world coordinate system, for Relative to and calculate the Euclidean distance of the point Scalar value in a scalar field And the scalar value As this point The error from the expected path; Angular velocity confirmation module, used to calculate the point The scalar field gradient of the point The scalar field gradient and the error are used to determine the angular velocity of the unmanned ship turning; The scalar field gradient is , the point The scalar field gradient and the error satisfy the following relationship: in, is the expected velocity vector, is a constant parameter; Transform the relationship to the ship coordinate system: in, is the angular velocity of the unmanned ship turning.

5. An electronic device, characterized in that: include: A memory for storing instructions; A processor is coupled to the memory, and the processor is configured to execute the unmanned ship path tracking method based on path gradient information according to claim 1 based on the instructions stored in the memory.

Images