Unmanned surface vessel energy-saving route planning method based on Biase-SVF-RRT*

Abstract

The invention relates to the field of USV global path planning, and discloses an unmanned surface vessel energy-saving path planning method based on Biase-SVF-RRT *. The method aims at solving the problems that an existing path planning algorithm is high in energy consumption, low in reverse flow area planning efficiency, insufficient in path smoothness and the like in a complex marine environment. The method comprises the following steps: integrating a navigation distance and an upstream function to establish a cost function, and realizing adaptive adjustment of a weight coefficient; adopting a target point direction to guide an extension node; periodically performing path evaluation based on a sliding window strategy, and dynamically updating the heuristic feasible expansion region according to an evaluation result; a soft constraint sampling mechanism based on path segment local discrimination; and smoothing the generated path, and outputting a global track with short distance, low energy consumption and high feasibility. According to the invention, the cruise task efficiency and the track precision of the USV in a wide sea area environment can be effectively improved, and the autonomous operation capability of the USV in strong countercurrent, strong wind current interference and complex obstacle areas is enhanced.

Description

Technical Field

[0001] This invention relates to the field of technology, and in particular to an energy-saving trajectory planning method for unmanned surface vessels based on Biased-SVF-RRT*. Background Technology

[0002] Unmanned surface vessels (USVs) have broad application prospects in fields such as environmental monitoring. Path planning, as a core supporting technology for achieving autonomous navigation capabilities, is directly related to mission completion efficiency and system operational safety. During actual navigation, the complex and ever-changing marine environment, especially ocean currents, affects the USV's energy consumption, navigation time, and course stability, posing significant challenges to path planning. Therefore, how to effectively model path planning and formulate trajectory strategies that balance energy efficiency, high efficiency, and safety has become a key issue in current USV intelligent navigation research.

[0003] Literature review reveals that current research on global path planning based on known environmental information primarily focuses on optimizing traditional indicators such as path length, obstacle avoidance safety, and trajectory smoothness, while neglecting dynamic marine environmental factors, especially the impact of ocean currents on USV navigation performance. Complex ocean currents not only alter the accuracy of USV dead reckoning but also affect their maneuverability and handling stability. Ignoring these factors during path planning can easily lead to track deviations, abnormally increased energy consumption, and even navigational safety risks. Therefore, USV path planning research urgently needs to incorporate dynamic flow field characteristics into the optimization system, establishing a multi-objective collaborative evaluation framework that comprehensively considers key indicators such as navigation distance, energy efficiency, turning costs, safety margin, and motion smoothness to ensure the feasibility, economy, and navigation safety of the path. This research direction is of great significance for improving the autonomy and mission reliability of USVs in complex marine environments. Summary of the Invention

[0004] The purpose of this invention is to solve the problems of high energy consumption, low planning efficiency in countercurrent areas, insufficient path smoothness, and neglect of wind field influence in existing path planning algorithms in complex marine environments.

[0005] To address the aforementioned issues, a biased sampling-guided vector field rapidly-exploring random tree optimization algorithm (Biased-SVF-RRT*) is proposed based on the SVF-RRT* algorithm. The technical solution is as follows:

[0006] An energy-saving trajectory planning method for unmanned surface vessels based on Biased-SVF-RRT* is implemented as follows:

[0007] Step 1: Obtain the starting point and target point of the unmanned surface vessel, and obtain obstacle information and ocean current information in the sea area where the unmanned surface vessel is located; set the maximum number of iterations, maximum upstream cost, sliding window period, and expansion step size.

[0008] Step 2: Perform random sampling within the global heuristic sampling interval to obtain sampling points.

[0009] Step 3: Traverse the set of nodes in the path tree, and take the node with the minimum comprehensive cost to reach the sampling point as the neighboring node, and determine the upstream cost from the neighboring node to the sampling point.

[0010] Step 4: Based on neighboring nodes, synthesize the combined ocean current direction, the direction of random sampling points, and the gravitational direction of the target point to obtain the combined direction; generate new nodes from neighboring nodes toward the combined direction according to a specified step size.

[0011] Step 5: Traverse the node set of the path tree. If a node in the node set of the path tree has no collision with the new node and its distance from the current tree is greater than the expansion step, then define the node as an efficient node; otherwise, define the node as an inefficient node. Get the number of inefficient nodes and the number of efficient nodes in the node set of the path tree; update the relevant parameters.

[0012] Step 6: Perform obstacle avoidance detection on the connection between the neighboring nodes and the new node. If there is no collision, add the new node to the node set of the path tree and execute Step 7; if a collision occurs, update the iteration count and return to Step 2.

[0013] Step 7: Traverse all nodes in the node set of the path tree, select the node with the minimum comprehensive cost to reach the new node as the parent node of the new node, update the comprehensive cost from the parent node to the new node, and rewire based on the path connection between the parent node and the new node.

[0014] Step 8: If the number of iterations is less than the set maximum number of iterations, update the number of iterations and proceed to step 9; if the number of iterations is greater than or equal to the maximum number of iterations, proceed to step 11.

[0015] Step 9: Determine if the path tree has reached the target point. If it has, construct a new path from the starting point to the target point in the path tree and then proceed to Step 10.

[0016] If the target point is not reached, determine whether the number of iterations is a multiple of the sliding window period. If it is, determine whether to update the global heuristic sampling interval through dynamic window path detection; otherwise, if the number of iterations is not a multiple of the sliding window period, return to step 2.

[0017] Step 10: If a path from the starting point to the target point was not constructed in the previous iteration, obtain the sampling points directly through soft constraint heuristic sampling and return to step 3.

[0018] If a new path from the starting point to the target point has been constructed in the previous iteration, calculate the total cost of the new path.

[0019] If the total cost of the new path is less than the total cost of the original path, then update the global heuristic sampling interval Ξ. t+1 Then, for the new path, sampling points are obtained through soft constraint heuristic sampling, and the process returns to step 3.

[0020] If the total cost of the new path is greater than or equal to the total cost of the original path, then sampling points are obtained for the original path through soft constraint heuristic sampling, and the process returns to step 3.

[0021] Step 11: Output the current path, smooth the current path, and output the final path.

[0022] Furthermore, in step 1, it is also necessary to initialize the iteration count t = 1, initialize the node set V of the path tree to add the starting point of the unmanned surface vessel, at which point V contains only a unique node; initialize the global heuristic sampling interval Ξ1 = (-∞, +∞) and parameter λ1; and initialize the upstream coefficient ε0 = 1.

[0023] Furthermore, the method for calculating the comprehensive cost in step 3 includes:

[0024] The path tree's node set V contains nodes i that reach the sampling point p. rand The overall cost C total :

[0025] C total =q1C upstream (p i )+C dist (p i )

[0026] Where q1 and q2 are weight parameters, q2 = 1 - q1, C upstream (p i Let be the total upstream cost of the i-th node in the node set V of the path tree. C dist (p i Let be the sailing distance corresponding to the i-th node in the node set V of the path tree.

[0027] C(P j Let be the upstream cost from node j to node (j+1) in the node set V of the path tree.

[0028]

[0029] Among them, L j θ represents the distance from the j-th node to the (j+1)-th node in the node set V of the path tree; j Let f(s) be the angle between the current direction and the route direction at the j-th node in the node set V of the path tree, and f(s) be the flow velocity at a distance s from the j-th node.

[0030] Furthermore, the comprehensive direction described in step 4 The calculation methods include:

[0031]

[0032] in, To take into account the direction of ocean currents, For the direction of random sampling points, The direction of gravity at the target point;

[0033] c is the target gravitational bias function, determined by the upstream coefficient ε and the hyperbolic tangent function; m is... and inner product ω t As an intermediate variable,

[0034]

[0035] Furthermore, step 5 involves updating the relevant parameter λ. t+1 The method is as follows:

[0036]

[0037] Among them, N ineff N represents the number of inefficient nodes. eff For the number of efficient nodes, E s Given the efficiency factor E, the reference exploration efficiency factor is... s ∈(0,1), update the value of λ.

[0038] Furthermore, the dynamic window path detection method described in step 9 includes: denoting the Euclidean distance between the starting point and the target point of the unmanned surface vessel as D1, traversing the path tree node set V, and calculating the value of each node p in the path tree node set V. i The Euclidean distance to the target point; select the top-k path points with the smallest Euclidean distance to the target point and backtrack the path, calculate the comprehensive cost of each path, and select the path with the smallest comprehensive cost as the current optimal path. The distance between the end of this path and the target point is denoted as D2. Then update the upstream coefficient ε t+1 and global heuristic sampling interval Ξ t+1 And perform pruning.

[0039] Update upstream coefficient ε t+1 :

[0040]

[0041] Where, p near For neighboring nodes, C max For the maximum upstream cost;

[0042] Update the global heuristic sampling interval Ξ t+1 :

[0043] Ξ t+1 =[Φ min -k εt ΔΦ,Φ max +k εt ΔΦ]

[0044] Where, Φ t Φ is the set of all node flow values ​​in the path tree node set V in the current iteration. min and Φ max Φ t The minimum and maximum flow values, ΔΦ t For Φ max and Φ min The difference;

[0045] Furthermore, the soft-constraint heuristic sampling method described in step 10 includes:

[0046] Step 10.1: Determine the local path heuristic sampling space;

[0047] Divide the path into N equally spaced paths, resulting in N path points {p}. y1 ,p y2 ,...p yN}, based on each path point P ya Upstream cost and maximum upstream cost C max Calculate the local upstream coefficient ε of the path point. a And construct the corresponding heuristic sampling interval Ξ a .

[0048] Step 10.2: Place P ya To P y(a+1) The geometric midpoint coordinates c of the path between a Extract it and construct the KDTree spatial index structure.

[0049] Generate random sampling points p rand Based on KDTree, quickly determine and p rand The nearest c a Thus determining prand The path segment P belongs to ya .

[0050] Step 10.3: p rand Path segment P ya The corresponding heuristic sampling interval is Ξ a Directly retain the flow value belonging to the interval Ξ a sampling point p rand .

[0051] For heuristic intervals Ξ where the flow value does not belong to a local path segment a sampling point p rand If its stream function value Φ(p) rand ) is in the global heuristic sampling interval Ξ t+1 If it is inside, then according to probability P accept Retain;

[0052]

[0053] Where η is the steepness factor of the acceptance probability. Heuristic interval Ξ for local path segments a The central value of .

[0054] A computer device / apparatus / system includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

[0055] A computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implement the steps of the method according to any one of claims 1 to 7.

[0056] A computer program product comprising a computer program / instructions which, when executed by a processor, implement the steps of the method according to any one of claims 1 to 7.

[0057] The beneficial effects of this invention are as follows:

[0058] The present invention provides an energy-saving trajectory planning method for unmanned surface vessels based on Biased-SVF-RRT*, which can find an asymptotically optimal and efficient path for the USV that comprehensively considers distance, smoothness, and energy consumption when obstacle and ocean current information is known. The obtained path is characterized by short length, low energy consumption, high accuracy, and relatively smoothness, thereby efficiently completing the cruise mission, effectively saving resources, and optimizing work efficiency. Attached Figure Description

[0059] Figure 1 This is a flowchart of the present invention.

[0060] Figure 2 This is an environmental field modeling diagram for the present invention.

[0061] Figure 3 This is a schematic diagram of the periodic dynamic window path evaluation of the I-SVF-RRT* of the present invention.

[0062] Figure 4 This is a path planning diagram generated by the various algorithms of this invention. Detailed Implementation

[0063] The present invention will now be further described with reference to the accompanying drawings.

[0064] refer to Figure 1 An energy-saving trajectory planning method for unmanned surface vessels based on Biased-SVF-RRT* includes the following steps:

[0065] Step 1: Obtain the starting point of the unmanned surface vessel (USV), and acquire obstacle and current information for the sea area where the USV is located; set the maximum number of iterations t. max Maximum upstream cost C max The sliding window period T;

[0066] Initialize the iteration count t = 1, and add the starting point of the unmanned surface vessel to the node set V of the path tree. At this time, V contains only a unique node. Initialize the global heuristic sampling interval Ξ1 = (-∞, +∞) and the parameter λ1. Initialize the upstream coefficient ε0 = 1.

[0067] Step 2: In the global heuristic sampling interval Ξ t Random sampling is performed within the sample area to obtain sampling point p. rand ;

[0068] Step 3: Traverse the node set V of the path tree and find the sampling point p. rand The node with the lowest overall cost is selected as the neighboring node p. near Determine the neighboring node p near To sampling point p rand Upstream cost C upstream (p near );

[0069] The path tree's node set V contains nodes i that reach the sampling point p. rand The overall cost C total :

[0070] C total =q1C upstream (p i )+C dist (p i )

[0071] Where q1 and q2 are weight parameters, q2 = 1 - q1, C upstream (p i Let be the total upstream cost of the i-th node in the node set V of the path tree. C dist (p i Let be the sailing distance corresponding to the i-th node in the node set V of the path tree.

[0072] C(P j Let be the upstream cost from node j to node (j+1) in the node set V of the path tree.

[0073]

[0074] Among them, L j θ represents the distance from the j-th node to the (j+1)-th node in the node set V of the path tree; j Let f(s) be the angle between the direction of the ocean current and the direction of the route at the j-th node in the node set V of the path tree, and f(s) be the flow velocity at a distance s from the j-th node.

[0075] Step 4: Based on neighboring node p near The direction of the ocean currents will be considered. Random sampling point direction Gravitational direction relative to the target point Synthesis yields the comprehensive direction From neighboring node p near Towards a comprehensive direction Generate a new node p according to the specified step size. new ;

[0076]

[0077] in, m is and inner product ω t As an intermediate variable,

[0078] Step 5: Traverse the node set V of the path tree. If a node in V is related to the new node p... new If a node is a node with no collisions and its distance from the current tree is greater than the expansion step, then that node is defined as an efficient node; otherwise, it is defined as an inefficient node. The number N of inefficient nodes in the node set V of the path tree is then obtained. ineff and the number of efficient nodes N eff Update parameter λ t+1 ;

[0079]

[0080] in, Explore efficiency factor E s ∈(0,1);

[0081] Given a reference efficiency factor E s ∈(0,1), update the value of λ;

[0082] Step 6: For neighboring node p near With the new node p new The connection is used for obstacle avoidance detection. If there is no collision, the new node p is... new Add the node to the set V of the path tree and proceed to step 7; if a collision occurs, let t = t + 1, and then return to step 2;

[0083] Step 7: Traverse all nodes in the node set V of the path tree to reach the new node p. new The node with the lowest overall cost will be selected as the new node p. new parent node p min Update from parent node p min Reaching the new node p new The overall cost, based on the parent node p min and new node p new The path connections need to be rewired;

[0084] Step 8: If t < t max If t = t + 1, then proceed to step 9; if t ≥ t max Then proceed to step 11;

[0085] Step 9: Determine if the path tree has reached the target point. If it has, construct a new path from the starting point to the target point in the path tree and proceed to step 10.

[0086] If the target point is not reached, determine whether t is a multiple of the sliding window period T. If t is a multiple of T, determine whether to update the global heuristic sampling interval through dynamic window path detection; otherwise, if t is not a multiple of T, return to step 2.

[0087] Methods for dynamic window path detection:

[0088] Let D1 denote the Euclidean distance between the starting point and the target point of the unmanned surface vessel. Traverse the path tree node set V and calculate the value of each node p in the path tree node set V. i The Euclidean distance to the target point; select the top-k path points with the smallest Euclidean distance to the target point and backtrack the path, calculate the comprehensive cost of each path, and select the path with the smallest comprehensive cost as the current optimal path. The distance between the end of this path and the target point is denoted as D2. Then update the upstream coefficient ε t+1and global heuristic sampling interval Ξ t+1 And perform pruning;

[0089] Update upstream coefficient ε t+1 :

[0090]

[0091] Update the global heuristic sampling interval Ξ t+1 :

[0092] Ξ t+1 =[Φ min -k εt ΔΦ,Φ max +k εt ΔΦ]

[0093] Where, Φ t Φ is the set of all node flow values ​​in the path tree node set V in the current iteration. min and Φ max Φ t The minimum and maximum flow values, ΔΦ t For Φ max and Φ min The difference;

[0094] Step 10: If a path from the starting point to the target point was not constructed in the previous iteration, obtain the sampling point p directly through soft constraint heuristic sampling. rand Return to step 3;

[0095] If a new path from the starting point to the target point has been constructed in the previous iteration, calculate the total cost of the new path;

[0096] If the total cost of the new path is less than the total cost of the original path, then update the global heuristic sampling interval Ξ. t+1 Then, for the new path, sampling points p are obtained through soft constraint heuristic sampling. rand Return to step 3;

[0097] If the total cost of the new path is greater than or equal to the total cost of the original path, then sampling points p are obtained for the original path through soft-constraint heuristic sampling. rand Return to step 3;

[0098] Soft constraint heuristic sampling method:

[0099] Step 10.1: Determine the local path heuristic sampling space;

[0100] Divide the path into N equally spaced paths, resulting in N path points {p}. y1 ,p y2 ,...p yN}, based on each path point Pya Upstream cost and maximum upstream cost C max Calculate the local upstream coefficient ε of the path point. a And construct the corresponding heuristic sampling interval Ξ a ;

[0101] Step 10.2: Place P ya To P y(a+1) The geometric midpoint coordinates c of the path between a Extract it and construct the KDTree spatial index structure;

[0102] Generate random sampling points p rand Based on KDTree, quickly determine and p rand The nearest c a Thus determining p rand The path segment P belongs to ya ;

[0103] Step 10.3: p rand Path segment P ya The corresponding heuristic sampling interval is Ξ a Directly retain the flow value belonging to the interval Ξ a sampling point p rand ;

[0104] For heuristic intervals Ξ where the flow value does not belong to a local path segment a sampling point p rand If its stream function value Φ(p) rand ) is in the global heuristic sampling interval Ξ t+1 If it is inside, then according to probability P accept Retain;

[0105]

[0106] Where η is the steepness factor of the acceptance probability. Heuristic interval Ξ for local path segments a The central value;

[0107] Step 11: Output the current path, smooth the current path, and output the final path.

[0108] Example:

[0109] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams illustrating only the basic structure of the invention. Therefore, they only show the components relevant to the invention and are intended to explain rather than limit the invention. The specific process is as follows:

[0110] refer to Figure 1 and 2This invention constructs an environmental model based on the direction, magnitude, and layout of ocean currents in a real sea area. Based on this model, it employs the I-SVF-RRT* algorithm (Improved Stream-based Vector Field Rapidly-exploring Random Tree Star) for global path planning for USVs. First, a real sea area environmental model is established. The target point's influence is incorporated into the selection of the path node's expansion direction, considering not only the entire path's upstream and downstream current conditions but also local upstream and downstream current conditions. This reduces the negative impact of the flow field on the node's growth direction when the starting point or target point is in a downstream environment, thus accelerating the finding of the initial path. Second, before finding the first feasible path, a sliding window path detection is activated every set period T iterations. This addresses the limitation of the SVF-RRT* algorithm (Rapidly-exploring Random Tree Star based on Stream-based Vector Field) in finding the initial path being equivalent to the VF-RRT* algorithm (Rapidly-exploring Random Tree Star based on Vector Field), as well as the poor heuristic interval contraction effect in downstream environments. After finding the initial path, a soft-constraint sampling mechanism based on local discrimination of path segments is employed. This not only considers the matching degree between each path segment and the flow field but also introduces a rapid spatial location matching technique to determine the segment affiliation of sampling points, and uses this as the basis for heuristic discrimination and flexible sampling control. Furthermore, weighting coefficients are set in the cost function to comprehensively consider route distance and energy consumption, and a cubic quasi-uniform B-spline curve is introduced to smooth the output path to meet the continuity requirements of USV navigation.

[0111] Input: Ocean current velocity field at a depth of 0.5m in a certain region. Starting Point X init Target point X goal Free region χ free Obstacles χ OBS And the tree containing the starting point: Tree = (V, E);

[0112] Output: a low-energy, safe, and smooth path from the starting point to the destination;

[0113] Case Study:

[0114] refer to Figure 2 The present invention will be further explained using the global path planning problem in a certain coastal environment as an example. Figure 2 To select a complex ocean current environment in a certain region, 'a' represents the ocean current field environment map, 'b' represents the environment map of current field values, and the I-SVF-RRT* algorithm parameters are as follows:

[0115] The expansion step size is set to 0.2, the dynamic window path evaluation period is 100, the reference exploration efficiency factor is 0.5, the initial energy cost weight q1 = 0.8, the initial distance weight ratio q2 = 0.2, the maximum number of iterations is 3000, the initial comprehensive cost is ∞, the initial upstream coefficient is 0.5, the initial number of inefficient nodes is 0, the initial number of efficient nodes is 0, the initial inefficient node rate is 0.5, the ocean current field intensity weight is 0.5, the number of path segments is 5, and the number of candidate paths for sliding window path evaluation is 5.

[0116] refer to Figure 3 and 4 The I-SVF-RRT* algorithm's dynamic window periodic path evaluation results determine the current optimal path and update parameters, reducing the number of iterations required to find the initial path. A comparison with the path planned by the VF-RRT* algorithm shows that VF-RRT* has low efficiency in full-plane expansion. By restricting the sampling interval based on the flow domain, the problem of the SVF-RRT* algorithm's insignificant sampling interval reduction effect in complex environments or countercurrent fields is effectively solved, improving the algorithm's optimization and convergence efficiency, and thus improving pathfinding efficiency. Although the overall tree structure shows a tendency to cluster towards the target point, the tree distribution density and expansion direction differ in different flow regions (such as the central downstream corridor and the upper countercurrent zone). This differentiation trend reflects the role of the local discrimination mechanism of the path segment: the algorithm automatically adjusts the sampling area according to the flow function distribution of the current path segment, thus enabling faster path convergence in downstream sections and retaining more exploration flexibility in countercurrent sections. Secondly, during the expansion of new nodes, not only is the global upstream / downstream current situation evaluated based on the upstream coefficient, but the local upstream / downstream current situation is also evaluated based on the angle between the navigation direction and the ocean current direction of the local segment. This jointly realizes an adaptive guidance expansion mechanism for the target point direction, solving the problem of low pathfinding success rate in upstream environments. Finally, a smooth path satisfying the USV navigation curvature is obtained through cubic uniform B-spline smoothing, improving the feasibility of path planning.

[0117] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for energy-saving trajectory planning of unmanned surface vessels based on Biased-SVF-RRT*, characterized in that: The specific implementation process is as follows: Step 1: Obtain the starting point and target point of the unmanned surface vessel, and obtain obstacle information and ocean current information in the sea area where the unmanned surface vessel is located; set the maximum number of iterations, maximum upstream cost, sliding window period, and expansion step size; Step 2: Perform random sampling within the global heuristic sampling interval to obtain sampling points; Step 3: Traverse the set of nodes in the path tree, and take the node with the minimum comprehensive cost to reach the sampling point as the neighboring node, and determine the upstream cost from the neighboring node to the sampling point. Step 4: Based on neighboring nodes, synthesize the combined ocean current direction, the direction of random sampling points, and the gravitational direction of the target point to obtain the combined direction; Generate new nodes from neighboring nodes in the synthesis direction according to a specified step size; Step 5: Traverse the node set of the path tree. If a node in the node set of the path tree has no collision with the new node and its distance from the current tree is greater than the expansion step, then define the node as an efficient node; otherwise, define the node as an inefficient node. Obtain the number of inefficient nodes and the number of efficient nodes in the node set of the path tree. Update the relevant parameters; Step 6: Perform obstacle avoidance detection on the connection between the neighboring nodes and the new node. If there is no collision, add the new node to the node set of the path tree and execute Step 7; if a collision occurs, update the iteration count and return to Step 2. Step 7: Traverse all nodes in the node set of the path tree, take the node with the minimum comprehensive cost to reach the new node as the parent node of the new node, update the comprehensive cost from the parent node to the new node, and rewire based on the path connection between the parent node and the new node. Step 8: If the number of iterations is less than the set maximum number of iterations, update the number of iterations and proceed to step 9; if the number of iterations is greater than or equal to the maximum number of iterations, proceed to step 11. Step 9: Determine if the path tree has reached the target point. If it has, construct a new path from the starting point to the target point in the path tree and then proceed to Step 10. If the target point is not reached, determine whether the number of iterations is a multiple of the sliding window period. If it is a multiple of the sliding window period, determine whether to update the global heuristic sampling interval through dynamic window path detection. Otherwise, if the number of iterations is not a multiple of the sliding window period, return to step 2; Step 10: If a path from the starting point to the target point was not constructed in the previous iteration, obtain the sampling points directly through soft constraint heuristic sampling and return to step 3; If a new path from the starting point to the target point has been constructed in the previous iteration, calculate the total cost of the new path; If the total cost of the new path is less than the total cost of the original path, then update the global heuristic sampling interval Ξ. t+1 Then, for the new path, sampling points are obtained through soft constraint heuristic sampling, and the process returns to step 3; If the total cost of the new path is greater than or equal to the total cost of the original path, then for the original path, sampling points are obtained through soft constraint heuristic sampling, and the process returns to step 3. Step 11: Output the current path, smooth the current path, and output the final path.

2. The energy-saving trajectory planning method for unmanned surface vessels based on Biased-SVF-RRT* as described in claim 1, characterized in that: In step 1, it is also necessary to initialize the iteration count t=1, initialize the node set V of the path tree to add the starting point of the unmanned surface vessel, at which time there is only one node in V; initialize the global heuristic sampling interval Ξ1=(-∞,+∞) and parameter λ1; initialize the upstream coefficient ε0=1.

3. The energy-saving trajectory planning method for unmanned surface vessels based on Biased-SVF-RRT* as described in claim 2, characterized in that, The method for calculating the overall cost in step 3 includes: The path tree's node set V contains nodes i that reach the sampling point p. rand The overall cost C total ; C total =q1C upstream (p i )+C dist (p i ) Where q1 and q2 are weight parameters, q2 = 1 - q1, C upstream (p i Let be the total upstream cost of the i-th node in the node set V of the path tree. C dist (p i Let be the sailing distance corresponding to the i-th node in the node set V of the path tree. C(P j Let be the upstream cost from node j to node (j+1) in the node set V of the path tree. Among them, L j θ represents the distance from the j-th node to the (j+1)-th node in the node set V of the path tree; j Let f(s) be the angle between the current direction and the route direction at the j-th node in the node set V of the path tree, and f(s) be the flow velocity at a distance s from the j-th node.

4. The energy-saving trajectory planning method for unmanned surface vessels based on Biased-SVF-RRT* as described in claim 2, characterized in that, The comprehensive direction described in step 4 The calculation methods include: in, To take into account the direction of ocean currents, For the direction of random sampling points, The direction of gravity at the target point; c is the target gravitational bias function, determined by the upstream coefficient ε and the hyperbolic tangent function; m is... and inner product ω t As an intermediate variable, 5. The energy-saving trajectory planning method for unmanned surface vessels based on Biased-SVF-RRT* as described in claim 2, characterized in that, Step 5 describes updating the relevant parameter λ t+1 The method is as follows: Where, N ineff N represents the number of inefficient nodes. eff For the number of efficient nodes, E s Given the efficiency factor E, the reference exploration efficiency factor is... s ∈(0,1), update the value of λ.

6. The energy-saving trajectory planning method for unmanned surface vessels based on Biased-SVF-RRT* as described in claim 3, characterized in that, The dynamic window path detection method described in step 9 includes: denoting the Euclidean distance between the starting point and the target point of the unmanned surface vessel as D1, traversing the path tree node set V, and calculating the value of each node p in the path tree node set V. i The Euclidean distance to the target point; select the top-k path points with the smallest Euclidean distance to the target point and backtrack the path, calculate the comprehensive cost of each path, and select the path with the smallest comprehensive cost as the current optimal path. The distance between the end of this path and the target point is denoted as D2. Then update the upstream coefficient ε t+1 and global heuristic sampling interval Ξ t+1 And perform pruning; Update upstream coefficient ε t+1 : Where, p near For neighboring nodes, C max For the maximum upstream cost; Update the global heuristic sampling interval Ξ t+1 : X t+1 =[Φ min -k εt DF,F max +k εt [D.F.] Where, Φ t Φ is the set of all node flow values ​​in the path tree node set V in the current iteration. min and Φ max Φ t The minimum and maximum flow values, ΔΦ t For Φ max and Φ min The difference; 7. The energy-saving trajectory planning method for unmanned surface vessels based on Biased-SVF-RRT* as described in claim 6, characterized in that, The soft-constraint heuristic sampling method described in step 10 includes: Step 10.1: Determine the local path heuristic sampling space; Divide the path into N equally spaced paths, resulting in N path points {p}. y1 ,p y2 ,...p yN }, based on each path point P ya Upstream cost and maximum upstream cost C max Calculate the local upstream coefficient ε of the path point. a And construct the corresponding heuristic sampling interval Ξ a ; Step 10.2: Place P ya To P y(a+1) The geometric midpoint coordinates c of the path between a Extract it and construct the KDTree spatial index structure; Generate random sampling points p rand Based on KDTree, quickly determine and p rand The nearest c a Thus determining p rand The path segment P belongs to ya ; Step 10.3: p rand Path segment P ya The corresponding heuristic sampling interval is Ξ a Directly retain the flow value belonging to the interval Ξ a sampling point p rand ; For heuristic intervals Ξ where the flow value does not belong to a local path segment a sampling point p rand If its stream function value Φ(p) rand ) is in the global heuristic sampling interval Ξ t+1 If it is inside, then according to probability P accept Retain; Where η is the steepness factor of the acceptance probability. Heuristic interval Ξ for local path segments a The central value of .

8. A computer device / equipment / system, comprising a memory, a processor, and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that: When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 7.

10. A computer program product comprising a computer program / instructions, characterized in that: When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1 to 7.

Images