A path planning method for unmanned vessels in anchorage areas based on risk perception

Abstract

The present invention provides a risk-awareness-based unmanned vessel path planning method for an anchorage area, aiming to improve the safety and efficiency of unmanned vessels navigating in anchorage areas. The present invention first spatially divides and rasterizes the anchorage area using Voronoi polygons, introduces the ship domain and Gaussian influence function to perform risk assessment on each grid, and generates an initial path. The initial path is then smoothed twice in sequence based on the Bresenham line algorithm and the cubic B-spline algorithm, and the unmanned vessel autonomously navigates in the anchorage area according to the twice-smoothed path. The present invention ensures the safety of the path and avoids the influence of high-risk areas. At the same time, the path length is reduced while ensuring safety, thereby improving the path execution efficiency. In addition, the present invention reduces the number of turns by smoothing the path, thereby improving the operability and stability of the path.

Description

Technical Field

[0001] The present invention relates to the technical field of ship path planning, and more particularly to a risk perception-based unmanned ship path planning method for an anchorage area. Background Art

[0002] An unmanned surface vessel (USV) is a surface vehicle that can navigate autonomously and does not require crew operation. Unmanned vessels can reduce the risk of collision in complex navigation environments and undertake tasks such as maritime monitoring and material transportation. Generally, unmanned vessels are small in size and do not require manual operation, which significantly improves safety when crossing anchorage areas, increases operational efficiency, and reduces labor costs. Unmanned vessels can monitor the marine environment and the status of anchored ships in real time, effectively improving safety management efficiency. In addition, unmanned vessels can efficiently transport materials in various weather and sea conditions, and are particularly suitable for high-risk environments or areas that are not suitable for manual operation. In recent years, unmanned vessels have been increasingly used in fields such as marine monitoring, environmental protection, waterway patrols, and port operations.

[0003] Path planning, as a core technology in autonomous navigation for USVs, directly impacts navigation safety, efficiency, and energy consumption. In complex environments, path planning must not only consider the shortest path length but also effectively avoid obstacles, risk areas, and dynamic environmental factors. Currently, various path planning algorithms have been applied to different scenarios, including A* and Voronoi-based A* algorithms, Dijkstra's algorithm, artificial potential field (APF), rapidly expanding random tree (RRT), genetic algorithm (GA), and particle swarm optimization (PSO).

[0004] In anchorage areas, the high density of ships moored makes traditional path planning complicated and difficult to ensure both safety and efficiency. The close distance between ships increases the risk of collisions between unmanned vehicles (USVs). Although existing path planning algorithms have made significant progress in many fields, they still have some problems when applied to path planning in complex environments. Although the traditional A algorithm and Voronoi-based A algorithm can generate the shortest path or avoid obstacles, they fail to effectively consider path safety, and the generated paths are often not smooth enough, resulting in operational complexity and high energy consumption. Although the RRT algorithm has strong exploration capabilities, it lacks path smoothness and global optimization capabilities, and is prone to falling into local optimal solutions. Although the PSO algorithm performs well in global optimization, it has shortcomings in path safety and efficiency.

[0005] In recent years, the application of deep learning technology in path planning has been increasing, particularly in autonomous driving and drone navigation. Deep learning algorithms, trained on large amounts of historical data, can predict potential risks along a path and optimize path selection. These methods, typically based on network structures such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), are capable of performing path planning in complex environments. However, deep learning path planning methods typically require large amounts of training data and computing resources, and the model training time is long, resulting in poor real-time performance. Furthermore, deep learning methods are less effective in considering path safety in dynamic environments, particularly in handling unknown obstacles and risky areas.

[0006] Therefore, it is necessary to develop a path planning scheme that can ensure path safety while optimizing path smoothness and computational efficiency. Summary of the Invention

[0007] In order to overcome the defects of the above-mentioned existing technologies in that it is difficult to take into account path safety, path smoothness, computational efficiency and environmental adaptability when performing path planning in complex environments, the present invention provides a path planning method for unmanned vessels in anchorage areas based on risk perception. In combination with risk perception, the risk field is modeled using the Gaussian influence function. After path optimization, the path is smoothed using a dual-stage smoothing strategy (DPSS) to ensure the smoothness and feasibility of the path, and to identify and avoid risk areas while maintaining path planning efficiency.

[0008] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0009] A risk-aware path planning method for an unmanned vessel in an anchorage area includes the following steps:

[0010] S1: Obtain an environmental map of the anchorage area, use the coordinates of each anchored ship in the anchorage area as seed points to generate Voronoi polygons, divide the environmental map into a plurality of Voronoi polygons, and perform rasterization on the divided environmental map;

[0011] S2: Perform risk assessment on each grid inside each Voronoi polygon based on the Gaussian influence function, and perform initial path planning based on the risk assessment results to obtain the initial path;

[0012] S3: performing a first path smoothing on the initial path based on the Bresenham straight line algorithm to eliminate unnecessary turns and redundant nodes in the initial path, thereby obtaining a first smoothed path;

[0013] S4: performing a second path smoothing on the first smooth path based on a cubic B-spline algorithm to reduce the number of sharp turns in the first smooth path, thereby obtaining a second smooth path;

[0014] The unmanned vessel performs autonomous navigation in the anchorage area according to the second smooth path.

[0015] Preferably, in step S1, each of the Voronoi polygons is independently subjected to risk assessment;

[0016] The distance between the coordinates of each anchored ship and the boundary of its corresponding Voronoi polygon is half of the ship spacing; each anchored ship only affects the navigable waters within its corresponding Voronoi polygon.

[0017] Preferably, in step S1, the coordinates of the anchored ship are marked as (x0, y0), and the ship heading angle is recorded as θ, then the safety area around the anchored ship is an elliptical no-fly zone, the center point of the ellipse is (x0, y0), the semi-major axis is a, and the semi-minor axis is b, where a and b are the first and second preset parameters respectively.

[0018] Preferably, in step S2, performing risk assessment on each grid within each Voronoi polygon comprises the following steps:

[0019] The center coordinates of the grid are marked as (x, y). The risk value of the unmanned ship colliding with the anchored ship (x0, y0) at this grid location is calculated according to the following steps:

[0020] Using point (x, y) as the starting point and θ as the ship heading angle, the coordinates of the anchored ship (x0, y0) are used to rotate the ship heading coordinate system:

[0021] x rot =cos(θ)·(x-x0)+sin(θ)·(y-y0)

[0022] y rot =-sin(θ)·(x-x0)+cos(θ)·(y-y0)

[0023] Among them, (x rot ,y rot ) is the coordinate of the center point of the grid after rotation transformation;

[0024] Determine the center coordinates of the grid after rotation transformation (x rot ,y rot ) is located in the elliptical no-fly zone: If the point (x rot ,y rot )satisfy Then the point (x rot ,y rot ) In the elliptical no-fly zone, the distance d from the center of the grid to the boundary of the no-fly zone is set to 0; otherwise, the point (x rot,y rot ) is not in the elliptical no-fly zone, further calculate the point (x rot ,y rot ) to the boundary of the restricted area;

[0025] Based on the Gaussian influence function, the risk value D(n) of the grid is calculated according to the distance d:

[0026]

[0027] Where n is the identifier of the grid; σ is the parameter of the Gaussian influence function.

[0028] Preferably, when point (x rot ,y rot ) is not within the elliptical no-fly zone, calculate the angle θ of the point relative to the center of the elliptical no-fly zone point :

[0029] θ point =arctan2(y rot ,x rot )

[0030] According to the angle θ point , calculate the corresponding boundary point (x edge ,y edge ):

[0031] x edge =acos(θ point )

[0032] y edge =bsin(θ point )

[0033] Calculate the center coordinates of the grid after rotation transformation (x rot ,y rot ) to the boundary of the restricted area:

[0034]

[0035] Preferably, in step S2, initial path planning is performed based on any one of an A* algorithm, a genetic algorithm and a reinforcement learning algorithm according to the risk assessment result to obtain the initial path.

[0036] Preferably, in step S2, initial path planning is performed based on the A* algorithm and the risk assessment result, including the following steps:

[0037] The traditional A* algorithm is improved and expressed as:

[0038] f(n)=p(n)+h(n)

[0039]

[0040] Among them, f(n) is the total cost of the current node, p(n) represents the path cost from the starting point S to the current node n, h(n) represents the estimated minimum cost from the current node n to the target node G, (x G ,y G ) is the coordinate of the target node G; (x n ,y n ) is the coordinate of the current node n;

[0041] The improved A* algorithm is used to perform initial path planning in the rasterized environment map to obtain the initial path, which includes several discrete coordinate points P0, P1, .., P n .

[0042] Preferably, step S3 includes the following steps:

[0043] S31: Initialize the first smooth path P using the starting point P0 of the original path smoothed , expressed as: P smoothed ={P0};

[0044] S32: Take the starting point P0 as the starting point and the other points after the starting point in the original path as the end points, establish temporary connections using line segments, and use the Bresenham straight line algorithm to obtain the discrete point set P between the two end points of each line segment. line ;

[0045] S33: Calculate each discrete point set P line The risk value of each point in the filter is used to select the discrete point set P whose risk value of all internal points does not exceed the preset risk threshold. l ′ ine , and obtain P l ′ ine The corresponding line segments;

[0046] S34: In P l ′ ine In the corresponding line segments, add the endpoint of the longest line segment to the first smooth path P smoothed , update the first smooth path P smoothed ;

[0047] S35: The first smooth path P smoothed The newly added point is used as the starting point of the new line segment, and steps S32 to S34 are repeated to continuously update the first smooth path P smoothed, and the first smooth path P obtained by the last update smoothed Save and complete the first path smoothing.

[0048] Preferably, step S4 includes the following steps:

[0049] In the first smooth path P smoothed In the example, new control points are inserted at fixed distances to obtain the path P after inserting the control points. control , where P control ={P0, P1, ..., P m}, m is the number of path points after inserting the control point;

[0050] Using the path P after inserting the control points control Construct a third-order B-spline curve S(t):

[0051]

[0052] Among them, t∈[0,1] is the preset parameter; N i (t) is the basis function; P i P control The i-th point in ;

[0053] Generate path P according to the third-order B-spline curve S(t) b-spline , the path P b-spline as the second smooth path.

[0054] Preferably, the step S4 further includes:

[0055] Calculate the path P b-spline With the first smooth path P smoothed The absolute value of the total length difference is used to determine whether the absolute value of the difference is less than or equal to the preset change threshold. If so, the path P is b-spline As the second smooth path; otherwise, regenerate the path P b-spline , and make a new judgment.

[0056] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0057] The present invention provides a path planning method for an unmanned vessel in an anchorage area based on risk perception. The method comprises the following steps: first, obtaining an environmental map of the anchorage area, using the coordinates of each anchored vessel in the anchorage area as seed points to generate Voronoi polygons, dividing the environmental map into a plurality of Voronoi polygons, and performing rasterization processing on the divided environmental map; then, performing risk assessment on each grid within each Voronoi polygon based on a Gaussian influence function, and performing initial path planning based on the risk assessment results to obtain an initial path; then, performing a first path smoothing on the initial path based on a Bresenham straight line algorithm to eliminate unnecessary turns and redundant nodes in the initial path to obtain a first smoothed path; finally, performing a second path smoothing on the first smoothed path based on a cubic B-spline algorithm to reduce the number of sharp turns in the first smoothed path to obtain a second smoothed path; and the unmanned vessel autonomously navigates in the anchorage area according to the second smoothed path.

[0058] This paper proposes a risk-aware path optimization algorithm (denoted as RAPO algorithm), which has the following advantages over the existing technology:

[0059] 1) Ensure path security:

[0060] The primary benefit of this invention is ensuring the safety of the route. Through a risk assessment mechanism, the invention can effectively detect high-risk areas (such as obstacles, other ships, and dangerous waters) along the route. During the route generation process, a Gaussian influence function combined with a Voronoi diagram is used to perform a risk assessment on each point in the environment, ensuring that the route avoids high-risk areas. In practice, every point on the route undergoes a safety check, and the route is planned only within low-risk areas, effectively avoiding the potential safety flaws of traditional algorithms.

[0061] 2) Reduce path length while ensuring safety:

[0062] While ensuring path safety, the present invention effectively reduces the total length of the path through an optimization algorithm. This reduction in path length not only improves path planning efficiency but also reduces energy consumption during operation. Compared with traditional algorithms, the present invention avoids lengthy detours by comprehensively considering path safety and optimization objectives, generating a more direct and shorter path.

[0063] 3) Improve path smoothness:

[0064] The present invention effectively improves the smoothness of the path by introducing the Bresenham straight line algorithm and the B-spline smoothing algorithm. Sharp turns and unnecessary deviations in the path are significantly reduced, and the generated path is smoother, facilitating smooth navigation operations for unmanned vehicles (USVs).

[0065] 4) Reduce computational complexity and improve computational efficiency:

[0066] During path planning and optimization, this invention utilizes the Bresenham algorithm, a computational algorithm based on integer operations, significantly reducing computational complexity and improving efficiency. This approach, particularly in large-scale environments, allows for rapid path calculation, risk assessment, and path optimization. Compared to traditional path planning algorithms, this invention achieves a better balance between efficiency and safety, adapting to path adjustment requirements in complex environments.

[0067] 5) Enhance the robustness and adaptability of path planning:

[0068] The path smoothing algorithm of the present invention has strong robustness and adaptability. In dynamic environments, especially in anchorage areas and complex marine environments, path optimization is crucial. By combining the Gaussian influence function with the B-spline smoothing algorithm, the path can be adjusted according to changes in the environment, ensuring that the path always remains safe and exhibits high stability and accuracy during path execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 This is a flow chart of a risk-aware unmanned vessel path planning method for an anchorage area provided in Example 1.

[0070] Figure 2 This is a schematic diagram of the elliptical ship field provided in Example 2.

[0071] Figure 3 This is a schematic diagram of the Gaussian influence function with σ=80 provided in Example 2.

[0072] Figure 4 This is a schematic diagram of risk distribution around a single anchored vessel provided in Example 2.

[0073] Figure 5 This is a schematic diagram of the path smoothing based on the Bresenham algorithm provided in Example 2.

[0074] Figure 6 This is a schematic diagram of the cubic B-spline path smoothing provided in Example 2.

[0075] Figure 7 This is a schematic diagram of the simulated ship position and Voronoi polygon division provided in Example 3.

[0076] Figure 8 This is a schematic diagram of the rasterization processing provided in Example 3.

[0077] Figure 9This is the risk distribution map of each anchored vessel provided in Example 3.

[0078] Figure 10 This is a diagram of the path planning result using the RAPO algorithm provided in Example 3 when the risk tolerance is 1.5. DETAILED DESCRIPTION

[0079] The accompanying drawings are for illustrative purposes only and are not to be construed as limiting this patent;

[0080] In order to better illustrate this embodiment, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product size;

[0081] It is understandable to those skilled in the art that some well-known structures and descriptions thereof may be omitted in the drawings.

[0082] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0083] Example 1

[0084] like Figure 1 As shown, this embodiment provides a method for planning a path for an unmanned vessel in an anchorage area based on risk perception, comprising the following steps:

[0085] S1: Obtain an environmental map of the anchorage area, use the coordinates of each anchored ship in the anchorage area as seed points to generate Voronoi polygons, divide the environmental map into a plurality of Voronoi polygons, and perform rasterization on the divided environmental map;

[0086] S2: Perform risk assessment on each grid inside each Voronoi polygon based on the Gaussian influence function, and perform initial path planning based on the risk assessment results to obtain the initial path;

[0087] S3: performing a first path smoothing on the initial path based on the Bresenham straight line algorithm to eliminate unnecessary turns and redundant nodes in the initial path, thereby obtaining a first smoothed path;

[0088] S4: performing a second path smoothing on the first smooth path based on a cubic B-spline algorithm to reduce the number of sharp turns in the first smooth path, thereby obtaining a second smooth path;

[0089] The unmanned vessel performs autonomous navigation in the anchorage area according to the second smooth path.

[0090] In the specific implementation process, the environmental map of the anchorage area is first obtained, the coordinates of each anchored ship in the anchorage area are used as seed points to generate Voronoi polygons, the environmental map is divided into several Voronoi polygons, and the divided environmental map is rasterized;

[0091] Then, risk assessment is performed on each grid inside each Voronoi polygon based on the Gaussian influence function, and initial path planning is performed based on the risk assessment results to obtain the initial path;

[0092] Then, the initial path is smoothed for the first time based on the Bresenham straight line algorithm to eliminate unnecessary turns and redundant nodes in the initial path and obtain the first smoothed path;

[0093] Finally, a second smoothing process is performed on the first smooth path based on the cubic B-spline algorithm to reduce the number of sharp turns in the first smooth path and obtain a second smooth path. The unmanned vessel autonomously navigates in the anchorage area according to the second smooth path.

[0094] This method first ensures the safety of the path and avoids the impact of high-risk areas. Second, while ensuring safety, it reduces the path length and improves the efficiency of path execution through a comprehensive optimization algorithm. Finally, it reduces the number of turns by smoothing the path, improving the operability and stability of the path. Compared with existing technologies, this method has significant improvements in the safety, optimization and computational efficiency of path planning, and is suitable for practical applications of autonomous navigation equipment such as USVs, and has important technical advantages.

[0095] Example 2

[0096] This embodiment provides a method for planning a path for an unmanned vessel in an anchorage area based on risk perception, including the following steps:

[0097] S1: Obtain an environmental map of the anchorage area, use the coordinates of each anchored ship in the anchorage area as seed points to generate Voronoi polygons, divide the environmental map into a plurality of Voronoi polygons, and perform rasterization on the divided environmental map;

[0098] S2: Perform risk assessment on each grid inside each Voronoi polygon based on the Gaussian influence function, and perform initial path planning based on the risk assessment results to obtain the initial path;

[0099] S3: performing a first path smoothing on the initial path based on the Bresenham straight line algorithm to eliminate unnecessary turns and redundant nodes in the initial path, thereby obtaining a first smoothed path;

[0100] S4: performing a second path smoothing on the first smooth path based on a cubic B-spline algorithm to reduce the number of sharp turns in the first smooth path, thereby obtaining a second smooth path;

[0101] The unmanned vessel autonomously navigates in the anchorage area according to the second smooth path;

[0102] In step S1, each of the Voronoi polygons is independently subjected to risk assessment;

[0103] The distance between the coordinates of each anchored ship and the boundary of its corresponding Voronoi polygon is half the distance between ships; each anchored ship only affects the navigable waters within its corresponding Voronoi polygon;

[0104] In step S1, the coordinates of the anchored ship are marked as (x0, y0), and the ship heading angle is recorded as θ. Then, the safety area around the anchored ship is an elliptical no-navigation area, the center point of the ellipse is (x0, y0), the semi-major axis is a, and the semi-minor axis is b, where a and b are the first and second preset parameters respectively;

[0105] In step S2, risk assessment of each grid within each Voronoi polygon includes the following steps:

[0106] The center coordinates of the grid are marked as (x, y). The risk value of the unmanned ship colliding with the anchored ship (x0, y0) at this grid location is calculated according to the following steps:

[0107] Using point (x, y) as the starting point and θ as the ship heading angle, the coordinates of the anchored ship (x0, y0) are used to rotate the ship heading coordinate system:

[0108] x rot =cos(θ)·(x-x0)+sin(θ)·(y-y0)

[0109] y rot =-sin(θ)·(x-x0)+cos(θ)·(y-y0)

[0110] Among them, (x rot ,y rot ) is the coordinate of the center point of the grid after rotation transformation;

[0111] Determine the center coordinates of the grid after rotation transformation (x rot ,y rot ) is located in the elliptical no-fly zone: If the point (x rot ,y rot )satisfy Then the point (x rot ,y rot) In the elliptical no-fly zone, the distance d from the center of the grid to the boundary of the no-fly zone is set to 0; otherwise, the point (x rot ,y rot ) is not in the elliptical no-fly zone, further calculate the point (x rot ,y rot ) to the boundary of the restricted area;

[0112] Based on the Gaussian influence function, the risk value D(n) of the grid is calculated according to the distance d:

[0113]

[0114] Where n is the identifier of the grid; σ is the parameter of the Gaussian influence function;

[0115] When point (x rot ,y rot ) is not within the elliptical no-fly zone, calculate the angle θ of the point relative to the center of the elliptical no-fly zone point :

[0116] θ point =arctan2(y rot ,x rot )

[0117] According to the angle θ point , calculate the corresponding boundary point (x edge ,y edge ):

[0118] x edge =acos(θ point )

[0119] y edge =bsin(θ point )

[0120] Calculate the center coordinates of the grid after rotation transformation (x rot ,y rot ) to the boundary of the restricted area:

[0121]

[0122] In step S2, initial path planning is performed based on the risk assessment result based on any one of the A* algorithm, genetic algorithm, and reinforcement learning algorithm to obtain an initial path. In this embodiment, the initial path planning based on the A* algorithm includes the following steps:

[0123] The traditional A* algorithm is improved and expressed as:

[0124] f(n)=p(n)+h(n)

[0125]

[0126] Among them, f(n) is the total cost of the current node, p(n) represents the path cost from the starting point S to the current node n, h(n) represents the estimated minimum cost from the current node n to the target node G, (x G ,y G ) is the coordinate of the target node G; (x n ,y n ) is the coordinate of the current node n;

[0127] The improved A* algorithm is used to perform initial path planning in the rasterized environment map to obtain the initial path, which includes several discrete coordinate points P0, P1, .., P n ;

[0128] The step S3 comprises the following steps:

[0129] S31: Initialize the first smooth path P using the starting point P0 of the original path smoothed , expressed as: P smoothed ={P0};

[0130] S32: Take the starting point P0 as the starting point and the other points after the starting point in the original path as the end points, establish temporary connections using line segments, and use the Bresenham straight line algorithm to obtain the discrete point set P between the two end points of each line segment. line ;

[0131] S33: Calculate each discrete point set P line The risk value of each point in the filter is used to select the discrete point set P′ whose risk value of all internal points does not exceed the preset risk threshold. line , and obtain P′ line The corresponding line segments;

[0132] S34: At P′ line In the corresponding line segments, add the endpoint of the longest line segment to the first smooth path P smoothed , update the first smooth path P smoothed ;

[0133] S35: The first smooth path P smoothed The newly added point is used as the starting point of the new line segment, and steps S32 to S34 are repeated to continuously update the first smooth path P smoothed , and the first smooth path P obtained by the last update smoothed Save and complete the first path smoothing;

[0134] The step S4 comprises the following steps:

[0135] In the first smooth path P smoothed In the example, new control points are inserted at fixed distances to obtain the path P after inserting the control points. control , where P control ={P0, P1, ..., P m}, m is the number of path points after inserting the control point;

[0136] Using the path P after inserting the control points control Construct a third-order B-spline curve S(t):

[0137]

[0138] Among them, t∈[0,1] is the preset parameter; N i (t) is the basis function; P i P control The i-th point in ;

[0139] Generate path P according to the third-order B-spline curve S(t) b-spline , the path P b-spline As the second smooth path;

[0140] The step S4 further includes:

[0141] Calculate the path P b-spline With the first smooth path P smoothed The absolute value of the total length difference is used to determine whether the absolute value of the difference is less than or equal to the preset change threshold. If so, the path P is b-spline As the second smooth path; otherwise, regenerate the path P b-spline , and make a new judgment.

[0142] In the specific implementation process, the environmental map of the anchorage area is first obtained, the coordinates of each anchored ship in the anchorage area are used as seed points to generate Voronoi polygons, the environmental map is divided into several Voronoi polygons, and the divided environmental map is rasterized;

[0143] Then, risk assessment is performed on each grid inside each Voronoi polygon based on the Gaussian influence function. In this embodiment, risk assessment is performed independently on each Voronoi polygon.

[0144] The ship domain is a concept used to represent the safe area around a ship. Generally, it is defined as a two-dimensional area around a ship that other ships should avoid entering to prevent collision. The size and shape of the ship domain can vary depending on the size, speed and navigation environment of the ship. Figure 2As shown in Figure 1, the ship's area is usually quantified by boundary radii in four directions: front (bow), rear (stern), left (port), and right (starboard). These boundary radii are usually expressed in multiples of the ship's length (L). By establishing a no-fly zone around the ship, collision accidents can be effectively prevented.

[0145] Reasonable establishment of a no-fly zone can significantly reduce the risk of collision, improve navigation efficiency, and enhance overall safety. In this embodiment, an ellipse is directly used as the restricted area. Using an ellipse as the restricted area around an anchored ship has significant advantages. The major axis of the elliptical restricted area is aligned with the longitudinal axis of the ship, providing a larger front and rear safety distance, and the minor axis provides a lateral safety distance to prevent other ships from approaching the side of the anchored ship, thereby reducing the risk of collision. In this embodiment, the elliptical restricted area is set based on the captain's navigation experience. For example, if the length of the ship is L, the semi-major axis of the major axis is 1.2L; if the width of the ship is W, the semi-minor axis of the minor axis is 2W, and the grid risk value within the restricted area is set to infinity. In this embodiment, the coordinates of the anchored ship are marked as (x0, y0), and the ship's heading angle is recorded as θ. Then the safety area around the anchored ship is an elliptical restricted area, the center point of the ellipse is (x0, y0), the semi-major axis is a, and the semi-minor axis is b.

[0146] When an unmanned vessel (USV) navigates in an anchorage area, anchored vessels pose a certain risk to the USV. This risk can be characterized by a Gaussian function. The Gaussian influence function is a variant of the Gaussian function and is mainly used to describe the exponential influence of a quantity with distance or time. Its form is shown in the following formula:

[0147]

[0148] In the Gaussian influence function, the value of f(x) ranges from (0,1), which is consistent with the typical risk value range; the Gaussian influence function is used to describe the ship domain and assess risk with high reliability and effectiveness. In risk assessment, the Gaussian influence function represents the attenuation of risk with distance, providing an intuitive and computationally simple model for path planning and obstacle avoidance; its smoothness and symmetry ensure the continuity and uniformity of risk distribution, making it particularly suitable for representing high-risk areas near moored ships, where the risk gradually decreases with increasing distance; in this embodiment, the map is divided into Voronoi polygons, and the distance from each ship to the boundary of the Voronoi polygon is half the distance between ships. The risk brought by each moored ship is limited to its assigned Voronoi polygon; for example, a Gaussian influence function with a parameter of σ = 80 can be used to depict the risk value, as shown in Figure 3 As shown;

[0149] The Gaussian influence function ensures that risk decreases with increasing distance, naturally simulating the risk posed by anchored ships to the surrounding environment. By calculating the distance d from a point to the boundary of the restricted area and applying the Gaussian influence function, an accurate risk assessment can be provided for path planning, thereby enhancing navigation safety and the effectiveness of path selection. In this embodiment, the map is converted into a grid, where d is the distance from the center of the grid to the boundary of the restricted area, and is calculated as follows:

[0150] The center coordinates of the grid are marked as (x, y). The risk value of the unmanned ship colliding with the anchored ship (x0, y0) at this grid location is calculated according to the following steps:

[0151] Using point (x, y) as the starting point and θ as the ship heading angle, the coordinates of the anchored ship (x0, y0) are used to rotate the ship heading coordinate system:

[0152] x rot =cos(θ)·(x-x0)+sin(θ)·(y-y0)

[0153] y rot =-sin(θ)·(x-x0)+cos(θ)·(y-y0)

[0154] Among them, (x rot ,y rot ) is the coordinate of the center point of the grid after rotation transformation;

[0155] Determine the center coordinates of the grid after rotation transformation (x rot ,y rot ) is located in the elliptical no-fly zone: If the point (x rot ,y rot )satisfy Then the point (x rot ,y rot ) In the elliptical no-fly zone, the distance d from the center of the grid to the boundary of the no-fly zone is set to 0; otherwise, the point (x rot ,y rot ) is not within the elliptical no-fly zone, calculate the angle θ of this point relative to the center of the elliptical no-fly zone point :

[0156] θ point =arctan2(y rot ,x rot )

[0157] According to the angle θ point , calculate the corresponding boundary point (x edge ,y edge ):

[0158] xedge =acos(θ point )

[0159] y edge =bsin(θ point )

[0160] Calculate the center coordinates of the grid after rotation transformation (x rot ,y rot ) to the boundary of the restricted area:

[0161]

[0162] Then, based on the Gaussian influence function, the risk value D(n) of the grid is calculated according to the distance d:

[0163]

[0164] Where n is the identifier of the grid; σ is the parameter of the Gaussian influence function;

[0165] Each grid outside the restricted navigation area has a risk value, which is set in the range of [1,2). When d>0, the point is outside the restricted navigation area, and the risk gradually decreases with the increase of distance; when d=0, the point is on the boundary, and the risk value is set to infinity (∞); when d<0, the point is inside the restricted navigation area, and the risk value is also set to infinity (∞). The risk caused by a single ship to its surrounding environment is displayed on the grid map, as shown in the figure below. Figure 4 As shown, the risk decreases gradually from the inside to the outside;

[0166] Initial path planning is then performed based on the risk assessment results to obtain an initial path. The traditional A* algorithm uses only path length as a heuristic function, which often results in the planned path being close to obstacles and unable to effectively guide the unmanned vehicle (USV) to navigate safely and smoothly. To address this issue, the RAPO algorithm combines the ship domain and the Gaussian influence function to determine the risk area created by anchored ships for other vessels. The risk function is incorporated as part of the RAPO path planning evaluation function.

[0167] The improved A* algorithm is expressed as:

[0168] f(n)=p(n)+h(n)

[0169]

[0170] Among them, f(n) is the total cost of the current node, p(n) represents the path cost from the starting point S to the current node n, h(n) represents the estimated minimum cost from the current node n to the target node G, (x G ,y G ) is the coordinate of the target node G; (xn ,y n ) is the coordinate of the current node n;

[0171] The improved A* algorithm is used to perform initial path planning in the rasterized environment map to obtain the initial path, which includes several discrete coordinate points P0, P1, .., P n ;

[0172] The initial path generated by the improved A* algorithm, although combined with risk assessment, may contain many redundant nodes and turns. In order to optimize this path, this method uses the Bresenham straight line algorithm to check the connection between each pair of adjacent nodes. The schematic diagram of the Bresenham straight line path smoothing is shown as follows Figure 5 As shown in Figure 2; if the risk values ​​of all intermediate nodes between the current node and the remote node are within an acceptable range (not exceeding the set threshold), these nodes can be directly connected, thus skipping the redundant nodes in the middle;

[0173] The overall idea of ​​the first smoothing is to fix the starting point, continuously traverse new end points, find the end point with the risk value that meets the requirements and is the farthest, then use this farthest end point as the new starting point, repeat the traversal, and finally skip the redundant nodes in the middle to generate the first smooth path. Specifically:

[0174] S31: Initialize the first smooth path P using the starting point P0 of the original path smoothed , expressed as: P smoothed ={P0};

[0175] S32: Take the starting point P0 as the starting point and the other points after the starting point in the original path as the end points, establish temporary connections using line segments, and use the Bresenham straight line algorithm to obtain the discrete point set P between the two end points of each line segment. line ;

[0176] S33: Calculate each discrete point set P line The risk value of each point in the filter is used to select the discrete point set P′ whose risk value of all internal points does not exceed the preset risk threshold. line , and obtain P′ line The corresponding line segments;

[0177] S34: At P′ line In the corresponding line segments, add the endpoint of the longest line segment to the first smooth path P smoothed , update the first smooth path P smoothed ;

[0178] S35: The first smooth path P smoothedThe newly added point is used as the starting point of the new line segment, and steps S32 to S34 are repeated to continuously update the first smooth path P smoothed , and the first smooth path P obtained by the last update smoothed Save and complete the first path smoothing;

[0179] After the initial smoothing by the Bresenham algorithm, although redundant nodes and sharp turns have been partially reduced, significant angle variations may still exist. These variations will result in larger turning angles, thereby increasing the energy consumption and operational difficulty of the USV in actual operation. To further optimize the smoothness and continuity of the path, this embodiment introduces a path smoothing method based on cubic B-splines in the second stage of DPSS. The mathematical definition of the B-spline curve is shown in the following formula:

[0180]

[0181] In the above formula, C(u) represents a point on the curve, P i is the i-th control point, N i，k (u) is the B-spline basis function, where k = 3 represents a cubic B-spline;

[0182] The recursive definition of the cubic B-spline basis function is as follows:

[0183] For the zero-order B-spline basis function, the formula is as follows:

[0184]

[0185] For higher order B-spline basis functions, as shown in the following formula:

[0186]

[0187] In order to generate a smooth B-spline curve, this embodiment uses uniformly distributed knot vectors. If there are n+1 control points, the knot vector is usually defined as:

[0188] u={u0,v1,...,u k -1,u k ,...,u n ,u n+1 ,...,u n+k}

[0189] These evenly distributed knot vectors ensure smooth transitions between control points in the B-spline curve;

[0190] Specifically, in this embodiment, on the first smooth path P smoothed In the example, new control points are inserted at fixed distances to obtain the path P after inserting the control points. control, where P control ={P0, P1, ..., P m}, m is the number of path points after inserting the control point;

[0191] Using the path P after inserting the control points control Construct a third-order B-spline curve S(t):

[0192]

[0193] Among them, t∈[0,1] is the preset parameter; N i (t) is the basis function; P i P control The i-th point in ;

[0194] Generate path P according to the third-order B-spline curve S(t) b-spline ;

[0195] In order to verify the path smoothing effect, it is necessary to compare the path length, number of turns and curvature before and after smoothing; the number of turns of the path should be significantly reduced, the curvature should be kept within a reasonable range; the change in path length should be small, so as to ensure the smoothness of the optimized path; this embodiment calculates the path P b-spline With the first smooth path P smoothed The absolute value of the total length difference is used to determine whether the calculated value is less than or equal to the preset change threshold. If so, the path P is b-spline As the second smooth path; otherwise, regenerate the path P b-spline , make a new judgment;

[0196] Path P b-spline The total length is calculated according to the following formula:

[0197]

[0198] Finally, the path P that meets the conditions b-spline As the second smooth path, Figure 6 As shown, the unmanned vessel autonomously navigates according to the second smooth path in the anchorage area;

[0199] This method first ensures the safety of the path and avoids the impact of high-risk areas. Second, while ensuring safety, it reduces the path length and improves the efficiency of path execution through a comprehensive optimization algorithm. Finally, it reduces the number of turns by smoothing the path, improving the operability and stability of the path. Compared with existing technologies, this method has significant improvements in path planning safety, optimization, and computational efficiency, making it suitable for practical applications of autonomous navigation equipment such as USVs and has important technical advantages.

[0200] It is worth mentioning that the path smoothing and optimization algorithm used in this embodiment can also be combined with other types of path planning and optimization algorithms to replace or enhance them to adapt to different application requirements. The following are several possible alternatives:

[0201] 1) Genetic Algorithm-Based Path Optimization: During path planning, if a more global path search and optimization is required, a genetic algorithm (GA) can be used instead of the traditional A* algorithm. By simulating natural selection and genetic operations, GAs can effectively find optimal paths in complex environments and are particularly suitable for multi-objective optimization and nonlinear problems. Although GAs have a high computational complexity, they are still highly applicable in certain applications requiring greater precision.

[0202] 2) Path Planning Based on Reinforcement Learning: In dynamic environments, if real-time path adjustments are required, reinforcement learning (RL) can be considered as an alternative to path planning. RL can continuously optimize path selection based on environmental feedback and is particularly suitable for autonomous vehicles, drones, and path planning in dynamic environments with complex obstacles. In this case, the optimization of path safety and efficiency will be automatically adjusted by the learning algorithm to adapt to complex and dynamically changing environments.

[0203] The path planning and optimization algorithm of this embodiment is not only applicable to unmanned vessels (USVs), but can also be flexibly applied to other types of automated navigation platforms, such as:

[0204] Autonomous vehicles: This algorithm can be used for path planning in autonomous vehicles, ensuring they avoid obstacles and maintain a smooth path. By optimizing the vehicle path, it reduces sharp turns and ineffective paths, thereby improving driving efficiency and ride comfort.

[0205] Unmanned aerial vehicles (UAVs): For UAVs, especially in complex urban or post-disaster environments, the path planning method of this embodiment can effectively avoid obstacles in flight (such as buildings, power lines, and flocks of birds), optimize the flight path while ensuring safety, and improve flight stability and endurance.

[0206] The path planning method of this embodiment can also be used in other fields, such as:

[0207] Disaster rescue and emergency response: In disaster rescue or emergency response scenarios, the path optimization algorithm of this embodiment can help emergency vehicles, robots, or drones quickly plan the optimal path;

[0208] Maritime logistics and shipping management: The path planning solution of this embodiment can also be applied to maritime logistics and shipping management, providing ships with efficient and safe navigation routes, avoiding high-risk waters, and improving shipping efficiency and safety.

[0209] In some applications, path planning must not only focus on safety and path length, but may also need to consider other optimization objectives, such as:

[0210] Minimizing energy consumption: In unmanned ships or autonomous vehicles, the path planning goal can be expanded to minimize energy consumption. By optimizing the curves and smoothness of the path, sudden acceleration and braking can be reduced, thus extending the energy life.

[0211] Obstacle avoidance optimization: In complex environments, the goal of path planning may be to optimize obstacle avoidance. That is, in addition to ensuring path safety, it is also necessary to consider how to efficiently avoid sudden obstacles (such as other ships, pedestrians, animals, etc.) in a dynamic environment to ensure obstacle avoidance effectiveness.

[0212] Example 3

[0213] This embodiment provides a simulation experiment for verifying the effectiveness of the risk-awareness-based unmanned vessel path planning method for an anchorage area described in Example 2.

[0214] The simulation experiment was performed on a computer equipped with a Microsoft Windows 11 operating system, an Intel i5 3.10GHz twelve-core CPU, and 16GB of memory. To verify the rationality and efficiency of the RAPO algorithm proposed in Example 2, the simulation was performed on a 2D static grid map using PyCharm as the development environment.

[0215] In the specific implementation process, referring to the anchorage and anchored ships in a certain water area, the size of the simulated anchorage is set to 5.5km×4.8km, including 60 anchored ships, and the length of each ship is between 90 and 150 meters; the distance between ships is set to 500 to 750 meters; the heading of each ship is evenly distributed between 135° and 165°; the specific positions of the anchored ships are determined by setting their position coordinates; the length of 13 ships is between 90 and 110 meters; the length of 24 ships is between 110 and 130 meters; the length of 23 ships is between 130 and 150 meters; using the coordinates of these anchored ships, the anchorage area is divided by Voronoi polygons to prepare for the risk of introducing anchored ships; through Voronoi polygon division, the distribution of ships in the defined anchorage area is simulated as follows Figure 7 As shown;

[0216] When processing environmental maps, grid-based maps are the most commonly used representation and processing form because they can effectively convey spatial information and support the application of various algorithms; considering that the normal length of an unmanned vehicle (USV) is about 10 meters, the grid size is set to 30 meters × 30 meters; Figure 8 The raster maps used in risk assessments are shown;

[0217] Before the simulation experiment, a risk value was assigned to each grid cell according to the Gaussian influence function. Each anchored ship formed a risk area, and the anchored ship was used as a seed point for the Voronoi polygon. Each polygon was a risk assessment unit. It was assumed that each anchored ship only affected the navigable waters within its corresponding Voronoi polygon. The distance between anchored ships ranged from 500 to 750 meters, and the shortest distance from an anchored ship to the boundary of its Voronoi polygon was approximately 250 meters. Considering that the main influence range of the Gaussian distribution is concentrated in [-3σ, +3σ], the corresponding actual risk range is 250 meters. Therefore, 3σ was set to 250, resulting in σ≈80. Therefore, the parameter σ in the Gaussian influence function was set to 80.

[0218] After converting the map into a raster, each grid is assigned a risk value. For the no-fly zone, the risk value is set to infinite; the grid in this area is displayed in white; the risk value of the risk area ranges from 1 to 2; Figure 9 The risk distribution diagram of anchored ships is shown, where "Start" is the starting point and "Goal" is the end point;

[0219] The RAPO algorithm combines risk assessment and DPSS. To verify that the RAPO algorithm can be applied to path planning in anchorage areas, simulation experiments were conducted. Figure 10 The figure shows the path planning results of the RAPO algorithm when the path risk value is 1.5. The three solid lines represent the initial path generated by the risk-modified A* algorithm in RAPO, the path smoothed by the Bresenham algorithm in the first stage, and the final path smoothed by cubic B-spline in the second stage.

[0220] Path planning results show that the RAPO algorithm significantly improves path length and number of turns by combining it with DPSS, and performs well under different path risk tolerances. Table 1 shows the path length, number of turns, and maximum turning angle results of the RAPO algorithm when the path risk tolerances are 1.2, 1.5, and 1.8.

[0221] Table 1 Comparison of path length, number of turns and maximum turning angle before and after smoothing under different risk tolerances

[0222]

[0223] Compared with the original paths generated by risk assessment in the RAPO algorithm, the lengths of the smoothed paths were reduced by 7.13%, 7.60% and 7.70% respectively; the number of turns was reduced by 81.13%, 90.57% and 94.34% respectively, while the maximum turning angles were reduced by 17.78%, 13.33% and 11.11% respectively; in the comparison of smoothed paths under different risk tolerances, the path length when the risk tolerance was 1.5 was reduced by 4.9% compared with that when the risk tolerance was 1.2, and the number of turns was reduced by 57.14%; the path length when the risk tolerance was 1.8 was reduced by 0.51% compared with that when the risk tolerance was 1.5, and the number of turns was reduced by 50%; the path length when the risk tolerance was 1.8 was reduced by 0.61% compared with that when the risk tolerance was 1.2, and the number of turns was reduced by 70%; with the increase of risk tolerance, the generated path length continued to shorten and the number of turns decreased, thereby reducing the operating difficulty and energy consumption of the USV and ensuring the economy of the path.

[0224] The same or similar reference numerals correspond to the same or similar components;

[0225] The terms used in the drawings to describe positional relationships are for illustrative purposes only and should not be construed as limiting this patent;

[0226] Obviously, the above embodiments of the present invention are merely examples for the purpose of clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the claims of the present invention.

Claims

1. A risk-aware unmanned vessel path planning method for anchorage areas, characterized in that: The following steps are involved: S1: Obtain an environmental map of the anchorage area, use the coordinates of each anchored ship in the anchorage area as seed points to generate Voronoi polygons, divide the environmental map into a plurality of Voronoi polygons, and perform rasterization on the divided environmental map; S2: Risk assessment of each grid within each Voronoi polygon is performed based on the Gaussian influence function, including the following steps: The center point of the grid is marked as (x, y). The risk value of the unmanned ship colliding with the anchored ship (x0, y0) at this grid position is calculated according to the following steps: Using the point (x, y) as the starting point and θ as the ship heading angle, the coordinates of the anchored ship (x0, y0) are used to rotate the ship heading coordinate system: in,( , ) is the coordinate of the center point of the grid after rotation transformation; Determine the center coordinates of the grid after rotation transformation ( , ) is located in the elliptical no-fly zone: If point ( , )satisfy , then point ( , ) In the elliptical no-fly zone, the distance d from the center of the grid to the boundary of the no-fly zone is set to 0; otherwise, the point ( , ) is not within the elliptical no-fly zone, further calculation of point ( , ) the distance d to the boundary of the restricted area; Based on the Gaussian influence function, the risk value of the grid is calculated according to the distance d. : Where n represents the identifier of the grid; σ represents the parameters of the Gaussian influence function; And perform initial path planning based on the risk assessment results to obtain the initial path; S3: performing a first path smoothing on the initial path based on the Bresenham straight line algorithm to eliminate unnecessary turns and redundant nodes in the initial path, thereby obtaining a first smoothed path; S4: performing a second path smoothing on the first smooth path based on a cubic B-spline algorithm to reduce the number of sharp turns in the first smooth path, thereby obtaining a second smooth path; The unmanned vessel performs autonomous navigation in the anchorage area according to the second smooth path.

2. The method for planning an unmanned vessel path in an anchorage area based on risk perception according to claim 1 is characterized in that: In step S1, each of the Voronoi polygons is independently subjected to risk assessment; The distance between the coordinates of each anchored ship and the boundary of its corresponding Voronoi polygon is half of the ship spacing; each anchored ship only affects the navigable waters within its corresponding Voronoi polygon.

3. The method for planning an unmanned vessel path in an anchorage area based on risk perception according to claim 1, characterized in that: In step S1, the coordinates of the anchored ship are marked as (x0, y0), and the ship heading angle is recorded as θ. Then, the safety area around the anchored ship is an elliptical no-navigation area, the center point of the ellipse is (x0, y0), the semi-major axis is a, and the semi-minor axis is b, where a and b are the first and second preset parameters respectively.

4. The method for planning an unmanned vessel path in an anchorage area based on risk perception according to claim 3 is characterized in that: When the point ( , ) is not within the elliptical no-fly zone, calculate the angle of the point relative to the center of the elliptical no-fly zone : According to the angle , calculate the corresponding boundary points on the elliptical no-fly zone ( , ): Calculate the center point coordinates of the grid after rotation transformation ( , ) to the boundary of the restricted area: .

5. The method for planning an unmanned vessel path in an anchorage area based on risk perception according to claim 4 is characterized in that: In step S2, initial path planning is performed based on any one of the A* algorithm, genetic algorithm and reinforcement learning algorithm according to the risk assessment result to obtain the initial path.

6. The method for planning an unmanned vessel path in an anchorage area based on risk perception according to claim 5 is characterized in that: In step S2, initial path planning is performed based on the A* algorithm and the risk assessment results, including the following steps: The traditional A* algorithm is improved and expressed as: ， in, is the total cost of the current node, Indicates starting point S The path cost to the current node n, represents the estimated minimum cost from the current node n to the target node G, ( x G , y G ) is the target node G The coordinates of , ) is the coordinate of the current node n; The improved A* algorithm is used to perform initial path planning in the rasterized environment map to obtain the initial path, which includes several discrete coordinate points. .

7. A risk-aware unmanned vessel path planning method for anchorage areas according to claim 1 or 6, characterized in that: The step S3 comprises the following steps: S31: Using the starting point of the original path Initialize the first smooth path , expressed as: ; S32: Set the starting point As the starting point, other points after the starting point in the original path are used as the end point, and temporary connections are established using line segments. The Bresenham straight line algorithm is used to obtain the discrete point set between the two endpoints of each line segment. ; S33: Calculate each discrete point set The risk value of each point in the filter is used to filter out the discrete point set whose risk value of all internal points does not exceed the preset risk threshold. , and obtain The corresponding line segments; S34: In Among the corresponding line segments, add the longest line segment endpoint to the first smooth path , update the first smooth path ; S35: The first smooth path The newly added point is used as the starting point of the new line segment, and steps S32 to S34 are repeated to continuously update the first smooth path. , and the first smooth path obtained by the last update Save and complete the first path smoothing.

8. The method for planning an unmanned vessel path in an anchorage area based on risk perception according to claim 7 is characterized in that: The step S4 comprises the following steps: In the first smooth path Insert new control points at fixed distances and obtain the path after inserting the control points. ,in, , is the number of path points after inserting the control point; Using the path after inserting the control points Constructing a third-order B-spline curve : Among them, t∈[0,1] is the preset parameter; is the basis function; for The i-th point in ; According to the third-order B-spline curve Generate Path , the path as the second smooth path.

9. The method for planning an unmanned vessel path in an anchorage area based on risk perception according to claim 8, characterized in that: The step S4 further includes: Calculate the path With the first smooth path The absolute value of the total length difference is used to determine whether the absolute value of the difference is less than or equal to the preset change threshold. If so, the path as the second smooth path; otherwise, regenerate the path , and make a new judgment.

Images