Marine unmanned aerial vehicle path planning method

Abstract

The invention provides a maritime unmanned aerial vehicle path planning method. The method comprises the steps that 1, path planning task requirements, unmanned aerial vehicle parameters and maritime environment and constraint information are set; step 2, constructing a graph search space in the task sea area; 3, screening candidate edges in the graph search space according to the constraint information to obtain feasible sub-graphs; 4, constructing an edge-adding cost model, and calculating the total cost of each edge in the feasible subgraph; step 5, constructing a perceptual heuristic A * search scheme according to the total cost of each edge; 6, setting a safety corridor constraint, and executing two-stage path planning according to the steps 3 to 5 to obtain a coarse path and a fine path; step 7, performing B spline curvature limiting shaping on the fine path, and generating an executable track; and step 8, according to task sea area information updating, carrying out cost updating on edges in the feasible subgraph, and repeatedly executing the steps 5-8 to form a path planning closed loop.

Description

Technical Field

[0001] This invention relates to a method for unmanned aerial vehicle (UAV) path planning, and more particularly to a method for path planning of a maritime UAV. Background Technology

[0002] This section provides only background information relevant to this disclosure and is not necessarily prior art.

[0003] Unmanned aerial vehicles (UAVs) are increasingly used in near-shore inspections (such as offshore wind farms and oil and gas platforms), search and rescue and emergency delivery, maritime target surveillance and law enforcement evidence collection. These tasks rely heavily on UAV path planning. Maritime UAV path planning methods provide safe and efficient flight trajectories for autonomous navigation in complex marine environments, ensuring that UAVs can safely and efficiently complete their missions in these challenging and ever-changing environments. Compared to land or inland waters, the marine environment presents the following significant characteristics and challenges: 1) Uncertainty and strong spatiotemporal nonuniformity of sea state; 2) Maritime missions often require traversing large sea areas, demanding rapid convergence of path planning on large-scale maps and support for online replanning; 3) Multiple engineering constraints: safety constraints, time-dependent obstacles, etc.

[0004] Existing maritime UAV path planning methods fail to uniformly model sea conditions such as wind, waves, and currents, as well as constraints such as no-fly zones, dynamic vessels, communication links, power safety margins, and minimum turning radius / curvature. This results in algorithms primarily relying on geometric distances, easily leading to "paper shortest paths" that consume excessive power in headwinds, cross risk zones, or exceed curvature limits, lacking feasibility and safety. Furthermore, dynamic obstacles are often addressed through instantaneous detection and local heuristic collision avoidance, lacking uncertain spatiotemporal predictions based on vessel motion and forward-looking penalties, resulting in unstable safety margins. Additionally, in large-scale sea areas and high-frequency environmental updates, traditional A* single-layer full-map expansion and redundant calculations are severe, online replanning is time-consuming and converges slowly, and the optimization of targets is biased towards path length, ignoring engineering indicators such as energy consumption, risk, link reachability, and attitude / maneuver constraints, causing a disconnect between theoretical shortest paths and engineering usability.

[0005] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] Purpose of the invention: The technical problem to be solved by the present invention is to provide a method for path planning of marine unmanned aerial vehicles (UAVs) to address the shortcomings of the existing technology.

[0007] To address the aforementioned technical problems, this invention discloses a method for path planning of a maritime unmanned aerial vehicle (UAV), comprising the following steps:

[0008] Step 1: Define the path planning task requirements, UAV parameters, and marine environment and constraint information;

[0009] Step 2: Construct a graph search space within the mission area;

[0010] Step 3: Based on the constraint information, filter the candidate edges in the graph search space to obtain a feasible subgraph;

[0011] Step 4: Construct an edge-adding cost model and calculate the total cost of each edge in the feasible subgraph;

[0012] Step 5: Based on the total cost of each edge, construct a perceptual heuristic A* search scheme;

[0013] Step 6: Set safety corridor constraints and perform two-stage path planning according to steps 3 to 5 to obtain coarse path and fine path;

[0014] Step 7: Perform B-spline curvature shaping on the fine path and generate an executable trajectory;

[0015] Step 8: Update the cost of the edges in the feasible subgraph based on the updated task sea area information, and repeat steps 5 to 8 to form a closed loop of path planning.

[0016] Furthermore, the setting of path planning task requirements, UAV parameters, and marine environment and constraint information mentioned in step 1 includes:

[0017] Step 1-1: Define the path planning task requirements, including the three-dimensional coordinates of the task start and end points, the set of intermediate task points, and the allowed time window for task completion or arrival, as shown below:

[0018]

[0019] in, The three-dimensional coordinates representing the starting point of the task; The three-dimensional coordinates representing the endpoint of the task; Represents the set of intermediate task points; This indicates the allowed time window for task completion or arrival. These are preset parameters;

[0020] Steps 1-2: Set the drone parameters, including the minimum and maximum allowable flight speeds. and Minimum turning radius allowed for drones Maximum rate of ascent or descent Maximum permissible normal overload Total energy available for the mission and the energy safety margin required to be retained at the end of the mission ;

[0021] Steps 1-3: Define marine environmental and constraint information, including spatially and temporally related wind field vectors. Effective wave height field Ocean current velocity vector field Collection of static no-fly zones Dynamic obstacles and ship-occupied areas that change over time and the set of communication coverage areas that change over time .

[0022] Furthermore, step 2, which involves constructing a graph search space within the mission area, specifically discretizing the continuous space into a finite number of nodes and candidate edges within the mission area, includes:

[0023] Step 2-1: Generate a set of nodes within the mission area. , means as follows:

[0024]

[0025] in, Represents discrete nodes The coordinates include the start point, end point, and intermediate spatial sampling points;

[0026] Step 2-2, for each node Candidate edges are generated under a predetermined set of heading angles and step size, as shown below:

[0027]

[0028] in, Indicates from node Pointing to node The directed edge, This represents the initial set of all candidate edges.

[0029] Furthermore, step 3, which involves filtering candidate edges in the graph search space, includes:

[0030] Step 3-1, process candidate edges The geometric trajectory is parameterized and represented as follows:

[0031]

[0032] in, Indicates along candidate edges Arc length is Spatial location; Indicate candidate edges The geometric length is calculated as follows:

[0033]

[0034] Step 3-2, using the reference speed Estimate the flight time of the edge , means as follows:

[0035]

[0036]

[0037] in, For the border Flight time estimation; To leave the node Time estimation; To reach the position on the trajectory The time of a moment;

[0038] Step 3-3, based on the static no-fly zone set Static no-fly zone constraints are implemented, as detailed below:

[0039] If there exists along the candidate edge Arc length is spatial location ,satisfy:

[0040]

[0041] Then determine the edge If the edge intersects with a static no-fly zone, delete that edge. ;

[0042] Steps 3-4: Based on the dynamic obstacles and ship occupancy zones that change over time. Dynamic obstacle constraints are applied, as follows:

[0043] If there exists along the candidate edge Arc length is spatial location ,satisfy:

[0044]

[0045] Then determine the edge If the border conflicts with a dynamic obstacle or a vessel-occupied area, delete that border. ;

[0046] Steps 3-5: Based on the time-varying set of communication coverage areas Communication constraints are implemented as follows:

[0047] Statistics along the border Not in communication coverage area Longest continuous time If the following conditions are met:

[0048]

[0049] If the edge does not satisfy the communication constraint, then delete the edge. ;in The maximum continuous duration of disconnection is the threshold.

[0050] Steps 3-6: Based on the total energy available for the task. and the energy safety margin required to be retained at the end of the mission Energy security constraints are implemented, as follows:

[0051] Let the current cumulative energy consumption be Estimate the edge The energy consumption is The conservative energy consumption estimate from the endpoint to the safe landing point is: If the following conditions are met:

[0052]

[0053] Therefore, it is determined that executing this edge would violate the energy safety margin, and the edge is deleted. ;

[0054] Steps 3-7: Based on the minimum allowable turning radius of the drone. Dynamic constraints are applied, as follows:

[0055] Estimate the equivalent turning radius and climb rate based on the edge connection relationship. If the following conditions are met:

[0056]

[0057] If the edge does not meet the minimum turning radius or climb rate limit, delete the edge. ,in, For this side Changes in heading angle before and after connection; Due to height variation; This is the equivalent turning radius;

[0058] Step 3-8: After deleting edges through steps 3-3 to 3-8, a feasible subgraph is obtained. , means as follows:

[0059]

[0060] in, The set of feasible edges that satisfy all constraints.

[0061] Furthermore, the construction of the edge-adding cost model described in step 4 includes:

[0062] Step 4-1: Calculate the distance cost based on the edge length. Calculate the time cost based on the side length and reference cruise speed. Specifically, it is expressed as follows:

[0063]

[0064]

[0065] in, The length of the side; The selected reference cruise speed;

[0066] Step 4-2: Calculate the energy consumption cost based on the power consumption at the sampling points. Specifically, it is expressed as follows:

[0067]

[0068]

[0069]

[0070] in, Indicates sampling point Wind speed and direction, Represents the unit vector of the edge direction; and Indicates sampling point Location and time; Indicates relative wind speed; This indicates the power consumption at that sampling point; , Represents the power model coefficients; Indicates the sampling arc length interval;

[0071] Step 4-3: Calculate the risk cost based on the comprehensive risk density at the sampling points. , means as follows:

[0072] Constructing the risk density function Then, approximate the edges using line integrals:

[0073]

[0074] in, This indicates the overall risk density at the sampling point;

[0075] Step 4-4: Calculate the curvature penalty cost based on the change in heading angle between adjacent edges, as shown below:

[0076]

[0077] in, Indicates the change in heading angle between adjacent sides; Indicates the maximum permissible curvature;

[0078] Steps 4-5, based on distance cost Time cost Energy consumption cost Space risk costs and curvature penalty cost Define the total cost of the edges. The details are as follows:

[0079]

[0080] in, The corresponding weights are the preset parameters in step 1-1. , means as follows:

[0081] Steps 4-6: Based on the total cost of the edges For any path Calculate the total cost , means as follows:

[0082]

[0083] in, Representing a path The total cost.

[0084] Furthermore, the construction of the perceptual heuristic A* search scheme described in step 5 includes:

[0085] Step 5-1, based on the maximum flight speed and considering the minimum turning radius Dubins shortest path length Construct a lower time bound that satisfies the minimum turning radius. , means as follows:

[0086]

[0087] in, For nodes To the finish line Time lower bound estimation; The Dubins shortest path length considering the minimum turning radius; Maximum flight speed;

[0088] Step 5-2, based on the minimum energy consumption coefficient per unit distance Construct a lower bound for energy consumption under ideal downwind conditions using the Euclidean distance from the current node to the destination. , means as follows:

[0089]

[0090] in, This represents the lower limit of energy consumption under ideal tailwind conditions. The minimum energy consumption coefficient per unit distance; This represents the Euclidean distance from the current node to the destination.

[0091] Step 5-3, based on the lower bound of the time required to satisfy the minimum turning radius. Lower bound of energy consumption under ideal tailwind conditions Constructing a comprehensive heuristic function , means as follows:

[0092]

[0093] in, These are non-negative weighting coefficients;

[0094] Step 5-4: Perform an A* search. During the search, calculate the minimum known cumulative cost from the starting point to the current node. and synthetic heuristic functions Define an evaluation function for each node. , means as follows:

[0095]

[0096] in, The minimum known cumulative cost from the starting point to the current node; For evaluation functions;

[0097] Priority queue sorting is performed based on the evaluation function;

[0098] Step 5-5: Iteratively select the evaluation function Expand the smallest node and proceed from the starting point to adjacent nodes. The current minimum cost Update adjacent nodes; where the distance from the starting point to adjacent nodes is... The current minimum cost The update method is as follows:

[0099]

[0100] in, From the starting point to the adjacent node The current minimum cost, From the starting point to the current node The current minimum cost is calculated based on the total cost of the path in steps 4-6. For the edge The overall cost, For assignment / update operations;

[0101] Steps 5-6, when the target node The search is complete when the target node is selected and expanded, or when the priority queue is empty; Begin by backtracking backwards using the recorded predecessor node relationships until the starting point. Extract the node sequence , means as follows:

[0102]

[0103] This node sequence This refers to the planning path that satisfies all hard constraints and minimizes the overall cost at the current resolution.

[0104] Furthermore, the two-stage path planning described in step 6 includes:

[0105] Step 6-1: Set the first search resolution, execute step 5, obtain the coarse path, and extract the key node sequence from the coarse path. ;

[0106] Step 6-2: Based on the sequence of key nodes, construct a security corridor, as shown below:

[0107]

[0108] in, As a key node Centered on, radius The neighborhood; This indicates the removal of restricted areas and dynamically occupied areas; This is a safe corridor area;

[0109] Step 6-3: Set the second search resolution and repeat steps 3 to 5 within the safe corridor to obtain a fine path.

[0110] Furthermore, step 7, which involves B-spline curvature shaping of the fine path, includes:

[0111] Nodes in a fine path Construct three control points Splines are represented as follows:

[0112]

[0113] in, For parameters The trajectory points below; For three spline basis functions;

[0114] Calculation parameters Trajectory points below curvature at The details are as follows:

[0115]

[0116] in, It is the first derivative; It is the second derivative; Represents the cross product of vectors; The vector magnitude; For parameters The curvature of the trajectory satisfies:

[0117] .

[0118] Furthermore, step 7, which involves generating an executable trajectory—that is, assigning velocity and time to nodes on the refined path after B-spline curvature shaping to generate a directly executable trajectory—includes:

[0119] Calculate the maximum permissible speed based on curvature and overload constraints. , means as follows:

[0120]

[0121] in, Let be the arc length along the trajectory; The curvature at that position; It is the acceleration due to gravity;

[0122] Calculate the actual flight speed based on the combined thrust limit and energy consumption budget. , means as follows:

[0123]

[0124] in, The upper limit of speed under thrust or power constraints; The upper limit of speed under energy budget constraints; This is the maximum permitted flight speed (speed limit) for the drone in this mission.

[0125] Parameterizing time is represented as follows:

[0126]

[0127] in, The arc length is The time of moment, Position of trajectory arc length The actual flight speed at that location.

[0128] The final executable trajectory is represented as follows:

[0129]

[0130] in, , , Position changing over time; For speed; This is the heading angle.

[0131] Furthermore, the cost update of the edges in the feasible subgraph described in step 8, i.e., when the marine environment and constraint information are updated, only the affected areas are incrementally updated, includes:

[0132] For each node Maintaining forward-looking costs , means as follows:

[0133]

[0134] in, For nodes The set of predecessor nodes; To reach a node from the predecessor via an edge The minimum expected cost; For nodes after environment update To the node The overall cost of the edge;

[0135] If the minimum known cumulative cost from the starting point to the current node is... and the cost of looking ahead If they are different, add the node to the priority queue and re-execute step 5.

[0136] Beneficial effects:

[0137] 1. This invention addresses the problem of "unexecutable / unsafe" paths under complex sea conditions and multiple constraints by uniformly introducing sea state energy consumption, risk distance, curvature / minimum turning radius, and communication / power safety constraints into the edge cost. It also performs spatiotemporal prediction and penalty for dynamic obstacles to generate executable paths that meet dynamic and safety requirements, thereby improving safety and reducing energy consumption.

[0138] 2. This invention addresses the problem of insufficient planning efficiency and stability in large-scale sea areas and online tasks by adopting a hybrid framework of "layered mapping + local grid acceleration + search symmetry reduction (bidirectional) + incremental update". Without changing the optimality / suboptimality guarantee, it reduces invalid expansion and redundant calculations, and supports rapid convergence under high-frequency environmental updates. Attached Figure Description

[0139] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, and the advantages of the present invention in the above and / or other aspects will become clearer.

[0140] Figure 1 This is a schematic diagram of the overall process of the present invention.

[0141] Figure 2 This is a schematic diagram illustrating the simulation planning results analysis in the example. Detailed Implementation

[0142] This invention aims to propose a path planning method for maritime unmanned aerial vehicles (UAVs) based on an improved A* algorithm. By embedding multiple constraints such as sea state energy consumption, risk distance, curvature / minimum turning radius, and communication and power security into edge costs, and combining this with spatiotemporal prediction and penalties for dynamic obstacles, a path that meets dynamic and safety requirements, is directly executable, and energy-efficient is generated. Simultaneously, a hybrid framework of "hierarchical mapping, local mesh acceleration, bidirectional search symmetry reduction, and incremental updates" is employed to significantly reduce ineffective expansion and redundant computation without compromising the optimality / suboptimality of A*, achieving rapid convergence and stable online replanning in large-scale sea areas. Ultimately, this results in comprehensive improvements in mission completion rate, energy consumption, risk exposure, link availability, and maneuver smoothness.

[0143] This invention addresses two core problems in maritime UAV path planning under complex sea conditions and multiple engineering constraints: "paper optimal but unexecutable" and low efficiency of online replanning in large-scale sea areas. It conducts research on "model-algorithm integration." Firstly, at the model level, sea state elements such as wind field, wave height, and ocean currents, along with engineering constraints such as static no-fly zones, dynamic ship occupancy, communication coverage, power safety margin, and minimum turning radius / curvature, are unified onto the same path cost line. The path is viewed as a series of feasible edges connected together, with the cost of each edge consisting of length / time, energy consumption driven by apparent wind, risk density from sea conditions and occupancy, and curvature penalty. A few-point line integral approximation ensures that "edge costs are additive, and path costs are additive," thus providing a provable convergence basis for heuristic search. Simultaneously, hard constraint judgments (no-fly zones, links, energy, time windows) are applied at the edge level, ensuring that the search is performed only on subgraphs that are "engineering-executable." Secondly, at the algorithm level, two key improvements are introduced using A* as the framework: First, a heuristic for sea state perception is designed, which combines the geometric / Dubins lower bound with the energy / time lower bound under the assumption of "tailwind and unobstructed" conditions, ensuring or providing an acceptable level of flexibility. Under the premise of suboptimal bounds, the guidance is significantly improved; secondly, the fixed neighborhood is replaced with motion primitives that satisfy the turning radius and climb rate, so that the expansion direction naturally conforms to the platform dynamics, the original result is smoother, and the back-end shaping cost is smaller. In order to match the large-scale and highly time-varying characteristics of maritime missions, this invention adopts a two-stage solution from coarse to fine: first, the skeleton and portal are quickly calculated on the coarse resolution map, and then the skeleton-induced safe corridor is recalculated at fine resolution, with jump point compression and bidirectional expansion to reduce the search diameter; the sea state and ship flow updates are updated by block cost caching and LPA* / D* Lite incremental repair, only the affected area is recalculated, and stable online replanning is achieved under high frequency updates. The final output polyline path is B-spline curvature-limited shaping under corridor constraints, and time retiming is performed in combination with apparent wind and thrust / overload constraints to obtain a directly executable track.

[0144] The technical solution of this invention is as follows Figure 1 As shown, the specific steps are as follows:

[0145] I. Task and Environmental Information Acquisition

[0146] First, the system acquires mission requirements, UAV platform parameters, and marine environment and constraint information, and converts them into inputs that the planning algorithm can directly use.

[0147] 1. Task information can be represented as:

[0148]

[0149] In the formula, The three-dimensional coordinates representing the starting point of the task; The three-dimensional coordinates representing the endpoint of the task; Represents the set of intermediate task points; Indicates the allowed time window for task completion or arrival; This represents the weighting coefficients of components such as distance, time, energy consumption, safety risk, and curvature penalty in the path cost.

[0150] 2. The key performance parameters of the UAV platform meet the following requirements:

[0151] , , , , ,

[0152] In the formula, , These are the minimum and maximum permissible flight speeds for drones, respectively. It refers to the flight speed of the drone; The minimum allowable turning radius for a drone. It is the turning half-radius of the drone; For maximum climb / descent rate, It is the drone's climb / descent rate; For the maximum permissible normal overload, For the normal overload of the drone; Total energy available for the mission; The required energy safety margin to be maintained at the end of the mission. For the power of drones.

[0153] 3. Marine environmental and constraint information modeling is as follows:

[0154] , , , , ,

[0155] In the formula, A wind field vector representing spatial location and time relationship; Indicates the effective wave height field; Represents the ocean current velocity vector field; Represents the set of static no-fly zones; This represents a dynamic set of obstacles and vessel-occupied areas that change over time. This represents the set of communication coverage areas that change over time.

[0156] II. Spatial Discretization and Candidate Edge Construction

[0157] Within the mission area, the continuous space is discretized into a finite number of nodes and candidate edges to construct a graph search space.

[0158] 1. Generate a set of nodes within the planned sea area:

[0159]

[0160] In the formula, Represents the coordinates of discrete nodes, including the start point, end point, and intermediate spatial sampling points.

[0161] 2. For each node Candidate edges are generated under a predetermined set of heading angles and step size:

[0162]

[0163] In the formula, Indicates from node Pointing to node The directed edges; This represents the initial set of all candidate edges.

[0164] III. Boundary-level filtering based on hard constraints

[0165] Before the search, each candidate edge is subject to hard constraints, and only edges that are feasible for the project are retained to form a feasible subgraph.

[0166] 1. For candidate edges Parameterize the geometric trajectory:

[0167]

[0168] In the formula, Indicates along the edge Arc length is Spatial location; Representing an edge Geometric length.

[0169] 2. Use reference speed Estimate the flight time of the edge:

[0170]

[0171]

[0172] In the formula, For the border Flight time estimation; To leave the node Time estimation; To reach the position on the trajectory The time of a moment.

[0173] 3. Static no-fly zone constraints: If they exist make

[0174]

[0175] Then determine the edge If the edge intersects with a static no-fly zone, that edge is deleted. In the formula, This is a set of static no-fly zones.

[0176] 4. Dynamic obstacle constraints: If they exist make

[0177]

[0178] Then determine the edge If an edge conflicts with a dynamic obstacle or a vessel-occupied area, it is deleted. In the formula, For a moment A set of dynamic obstacle regions.

[0179] 5. Communication constraints: Statistics along the edge Not in communication coverage area Longest continuous time ,like

[0180]

[0181] If the edge does not satisfy the communication constraint, it is deleted. In the formula, This refers to the duration of continuous communication interruption. This is the maximum allowed continuous disconnection time by the system.

[0182] 6. Energy security constraints: Assume the current cumulative energy consumption is... The estimated energy consumption of this side is The conservative energy consumption estimate from the endpoint to the safe landing point is: ,like

[0183]

[0184] If executing this edge would violate the energy safety margin, then the edge is deleted. In the formula, Energy already consumed; For flight edge The projected energy consumption; This is a conservative energy consumption estimate for the subsequent return / mission completion.

[0185] 7. Dynamic Constraints: Estimate the equivalent turning radius and climb rate based on the edge connection relationship.

[0186]

[0187] If the edge does not meet the minimum turning radius or climb rate limit, then the edge is deleted. In the formula, This refers to the change in heading angle before and after the edge is joined; Due to height variation; This is the equivalent turning radius.

[0188] After applying the above constraints, the feasible subgraph can be obtained:

[0189]

[0190] In the formula, The set of feasible edges that satisfy all hard constraints.

[0191] IV. Construction of Edge-Addable Cost Model

[0192] Based on the feasible subgraph, a unified additive cost is constructed for each edge, providing an objective function for heuristic search.

[0193] 1. The total edge cost is defined as:

[0194]

[0195] In the formula, For the edge The overall cost; As a cost of distance; For the cost of time; For the cost of energy consumption; For the cost of space risks; The cost of curvature penalty; For the corresponding weights.

[0196] 2. Distance and Time Costs

[0197]

[0198]

[0199] In the formula, The length of the side; The selected reference cruise speed.

[0200] 3. Energy consumption cost

[0201]

[0202]

[0203]

[0204] In the formula, Represents the unit vector of the edge direction; , Indicates the location and time of the sampling point; Indicates relative wind speed; This indicates the power consumption at that point; , Represents the power model coefficients calibrated by the platform; This indicates the sampling arc length interval.

[0205] 4. Risk Cost: Constructing the Risk Density Function Then, approximate the edges using line integrals:

[0206]

[0207] In the formula, This represents the overall risk density at the sampling point, including a weighted average of factors such as wave height, nearby obstacles, and ship traffic density. Indicates the arc length step size.

[0208] 5. Curvature penalty cost: This item is used to suppress sharp turns and improve trajectory smoothness.

[0209]

[0210] In the formula, Indicates the change in heading angle between adjacent sides; Indicates the maximum permissible curvature;

[0211] The costs mentioned above are all calculated on an edge-by-edge basis, for any path Cost calculation is as follows:

[0212]

[0213] In the formula, Representing a path The total cost.

[0214] V. Perceptual Heuristic A* Search

[0215] 1. Construct a lower bound for the time that satisfies the minimum turning radius:

[0216]

[0217] In the formula, For nodes Lower bound estimate of the time to the endpoint; The Dubins shortest path length considering the minimum turning radius; This is the maximum flight speed.

[0218] 2. Constructing the lower bound of energy consumption under ideal downwind conditions:

[0219]

[0220] In the formula, This is an estimate of the lower bound of energy consumption. The minimum energy consumption coefficient per unit distance; This represents the Euclidean distance from the current node to the destination.

[0221] 3. Comprehensive Heuristic Functions

[0222]

[0223] In the formula, These are non-negative weighting coefficients used to adjust the influence of time and energy consumption lower bounds; their selection ensures... A heuristic that is acceptable or has bounded suboptimality.

[0224] 4. During the A* search process, define the following for each node:

[0225]

[0226] In the formula, The minimum known cumulative cost from the starting point to the node; This is the evaluation function used for priority queue sorting.

[0227] Algorithm Iteration Selection Expand the smallest node and update its neighboring nodes according to the following formula:

[0228]

[0229] in, From the starting point to the adjacent node The current minimum cost, For the edge The overall cost, This is for assignment / update operations.

[0230] When the target node The search terminates when a node is selected and expanded, or when the priority queue is empty. At this point, the search continues from the target node. Begin by backtracking backwards using the recorded predecessor node relationships until the starting point. Extract the node sequence The sequence This refers to the planning path that satisfies all hard constraints and minimizes the overall cost at the current resolution.

[0231] VI. Two-Phase Planning and Safe Corridor Constraints

[0232] 1. Coarse-grained global planning and key node extraction: Set a large spatial sampling step size. (i.e., coarse resolution), run step five at coarse resolution to obtain the coarse path. Extract all discrete nodes contained in the coarse path and construct a sequence of key nodes in order. .

[0233] 2. Construct a safe corridor:

[0234]

[0235] In the formula, As a key node Centered on, radius The neighborhood; A safe passageway area after the no-fly zone and dynamic obstacles have been cleared.

[0236] 3. Refined replanning within the corridor: setting smaller spatial sampling steps. (i.e., fine resolution, satisfying) Only in the safe corridor The graph search space is reconstructed within the covered area, and steps three through five are repeated to obtain a high-precision fine path. .

[0237] seven, Spline curvature shaping

[0238] 1. Using polyline path nodes Construct three control points Spline:

[0239]

[0240] In the formula, For parameters The trajectory points below; For three Spline basis functions.

[0241] 2. Curvature is calculated as follows:

[0242]

[0243] In the formula, It is the first derivative; It is the second derivative; Represents the cross product of vectors; The vector magnitude; For parameters Curvature of the trajectory.

[0244] Requires all parameters satisfy:

[0245]

[0246] 8. Time Resetting and Executable Track Generation

[0247] Based on a trajectory that satisfies spatial and curvature constraints, velocity and time are allocated to generate a directly executable trajectory.

[0248] 1. Curvature and overload constraints determine the maximum permissible speed:

[0249]

[0250] In the formula, Let be the arc length along the trajectory; The curvature at that position; It is the acceleration due to gravity; This is the upper speed limit obtained from the overload limit.

[0251] 2. The combined thrust limitation and energy consumption budget are obtained , Determine the actual flight speed:

[0252]

[0253] In the formula, This represents the upper limit of speed under thrust / power constraints. This represents the upper limit of speed under energy budget constraints.

[0254] 3. Time parameterization

[0255]

[0256] In the formula, The arc length is The time of a moment;

[0257] 4. The final output is an executable trajectory:

[0258]

[0259] In the formula, , , Position changing over time; For speed; This is the heading angle.

[0260] IX. Incremental Online Replanning

[0261] When updates to sea state or ship flow information lead to changes in environmental and constraint information, incremental updates are performed only on the affected areas to avoid recalculating the entire map. For each node in the search graph... Maintaining a forward-looking cost :

[0262]

[0263] In the formula, For nodes The set of predecessor nodes; For nodes after environment update arrive The latest cost of the edge; This represents the destination node calculated based on the current state of neighboring nodes. The potential minimum cost.

[0264] By comparing the known costs of the nodes and the cost of looking ahead Identify nodes with inconsistent states (i.e.) (Nodes). Only these inconsistent nodes affected by environmental changes are added to a priority queue, and the search logic in step 5 is re-executed, thereby quickly completing local path repair and replanning.

[0265] Example:

[0266] like Figure 2 As shown, in order to verify the effectiveness of the proposed maritime unmanned aerial vehicle (UAV) path planning method, in The experiment was conducted in a simulated sea area. The mission was set so that the UAV would start from a shore-based origin (indicated by Start in the diagram, coordinates...). Fly to the destination of the mission at sea (as shown by Goal in the image, coordinates) .

[0267] The simulation environment includes the following key elements:

[0268] 1. Environmental disturbance field: A persistent northwesterly wind field exists in the sea area (indicated by the gray arrow in the figure), with a wind speed set at... .

[0269] 2. Constraints: There is a static no-fly zone (circular area in the figure) on the straight path, and there is a communication blind zone in the sea area to the east (comm blind zone in the figure, shaded area).

[0270] 3. Dynamic obstacle: There is a moving ship near the waterway (Dynamic Ship in the figure, shown by a square).

[0271] Figure 2 Although the straight path shown by the thin dashed line has the shortest geometric distance, it is directly eliminated by this method in the "edge-level filtering based on hard constraints" step because it crosses a static no-fly zone.

[0272] The thick solid line in the figure shows the final optimized path generated by the method of this invention. Compared with detouring to the north, the path planned by the algorithm is significantly shifted to the south. This is because this invention introduces energy consumption costs when constructing the edge-adding cost model. By detouring south, drones can more effectively utilize the tailwind provided by the northwest wind, thereby significantly reducing flight energy consumption.

[0273] Furthermore, the path is not a simple polyline connection between key nodes (hollow dots in the diagram), but rather undergoes B-spline curvature shaping to generate a path that satisfies the minimum turning radius of the UAV. A smooth flight path. Simultaneously, during the detailed planning phase, through spatiotemporal prediction of dynamic obstacles, the path was successfully established. The area avoided the navigation zone of moving vessels and ensured that the continuous loss of contact time when crossing communication blind spots did not exceed the safety threshold.

[0274] The final results show that this method can generate an executable track that balances safety (obstacle avoidance, communication constraints) and economy (energy saving by utilizing wind farms) under complex sea conditions and multiple constraints.

[0275] In its specific implementation, this application provides a computer storage medium and a corresponding data processing unit. The computer storage medium is capable of storing a computer program, which, when executed by the data processing unit, can run the invention's content regarding a path planning method for a maritime unmanned aerial vehicle (UAV) and some or all of the steps in various embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0276] Those skilled in the art will clearly understand that the technical solutions in the embodiments of the present invention can be implemented using computer programs and their corresponding general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of computer programs, i.e., software products. These computer program software products can be stored in a storage medium and include several instructions to cause a device containing a data processing unit (which may be a personal computer, server, microcontroller, MCU, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.

[0277] This invention provides a method and approach for path planning of unmanned aerial vehicles (UAVs) at sea. Many methods and approaches exist for implementing this technical solution; the above description is merely a preferred embodiment. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.

Claims

1. A method for path planning of a marine unmanned vehicle, the method comprising: The method comprises the following steps: Step 1, setting path planning task requirements, unmanned aerial vehicle parameters and sea environment and constraint information; Step 2, constructing a graph search space in the task sea area; Step 3, screening candidate edges in the graph search space according to the constraint information to obtain a feasible subgraph; Step 4, constructing an addable edge cost model to calculate the total cost of each edge in the feasible subgraph; Step 5, constructing a perception heuristic A* search scheme according to the total cost of each edge; Step 6, setting a safety corridor constraint, and executing two-stage path planning according to steps 3 to 5 to obtain a coarse path and a fine path; Step 7, performing B-spline limited curvature shaping on the fine path, and generating an executable flight path; Step 8, updating the cost of the edges in the feasible subgraph according to the task sea area information, and repeating steps 5 to 8 to form a path planning closed loop.

2. The path planning method for the offshore unmanned vehicle according to claim 1, wherein, The setting of path planning task requirements, unmanned aerial vehicle parameters and sea environment and constraint information in step 1 comprises: Step 1-1, setting path planning task requirements, including three-dimensional coordinates of task start and end points, a set of intermediate task points, and an allowed time window for task completion or arrival, represented as follows: ； wherein, a three-dimensional coordinate representing a task start point; a three-dimensional coordinate representing a task end point; a set of intermediate task points; a permitted time window for task completion or arrival, is a preset parameter; Step 1-2, setting the drone parameters, including the minimum and maximum flight speed allowed by the drone and the minimum turning radius allowed by the drone the maximum climb or descent rate the maximum normal overload allowed the total energy available for the mission and the energy safety margin required to be reserved at the end of the mission ; Step 1-3, setting maritime environment and constraint information, including spatial position and time dependent wind field vector , significant wave height field , ocean current velocity vector field , static no-fly zone set , dynamic obstacle and vessel occupancy zone set changing over time , and communication coverage area set changing over time .

3. The method of claim 2, wherein, The construction of a graph search space in the task sea area in step 2, i.e. discretizing the continuous space into finite nodes and candidate edges in the task sea area, comprises: Step 2-1, generating a node set in the task sea area is represented as follows: ； wherein, representing coordinates of discrete nodes including start, end and intermediate spatial sampling points; Step 2-2, for each node Generating candidate edges at a set of predetermined heading angles and step size is represented as follows: ； wherein, represents a directed edge from node pointing to node , represents an initial set of all candidate edges.

4. The method of claim 3, wherein, The screening of candidate edges in the graph search space in step 3 comprises: Step 3-1, parameterizing the geometric trajectory of the candidate edge is parameterized as follows: ； wherein, represents the spatial position of the candidate edge along the candidate edge with an arc length of represents the geometric length of the candidate edge . Step 3-2, using a reference speed Estimating the time of flight of the edge ; Step 3-3, Static No Fly Zone Set The static no fly zone constraint is performed as follows: if there exists a spatial position along the candidate edge with an arc length of satisfying: ； then determine edge Intersect with static no-fly zone, delete edge ; Step 3-4, dynamic obstacle according to time-varying dynamic obstacle and ship occupation area set , dynamic obstacle constraint is performed, specifically as follows: if there exists a spatial position along the candidate edge with an arc length of satisfying: ； then determine edge in conflict with dynamic obstacle or ship occupied area, delete the edge ; Step 3-5 - Based on the set of communication coverage areas over time The communication constraints are performed as follows: Statistics along the edge Not in the communication coverage area The longest continuous time If the following conditions are met: ； If the edge does not satisfy the communication constraint, the edge is deleted ; wherein is a maximum continuous outage duration threshold value; Step 3 - 6, total energy available to the task and the energy safety margin required to be reserved at the end of the task , the energy safety constraint is performed as follows: Let the current accumulated energy consumption be , the estimated energy consumption of the edge be , the conservative energy consumption from the edge end point to the safe landing point be , if ； If the execution of the edge will break the energy safety margin, the edge is deleted ; Step 3-7, according to the minimum turning radius allowed by the drone Dynamics constraints are performed, as follows: Estimating the equivalent turning radius and climb rate according to the edge connection relationship, and if the following conditions are met: ； if the edge does not satisfy the minimum turn radius or climb rate limit, delete the edge wherein, is the edge change in heading before and after the link; is the change in altitude; is the equivalent turn radius; Step 3-8, after the edge deletion by step 3-3 to step 3-8, the feasible subgraph is obtained is represented as follows: ； wherein, is the set of feasible edges that satisfy all constraints.

5. The method of claim 4, wherein, The construction of an addable edge cost model in step 4 comprises: Step 4-1, calculate distance cost according to edge length ; calculate time cost according to edge length and reference cruise speed ; Step 4-2, calculate the energy cost according to the power consumption of the sampling point ; Step 4-3, calculate risk cost according to the integrated risk density at the sampling point ; Step 4-4, calculate the curvature penalty cost according to the heading angle change between adjacent edges ; Steps 4-5, based on distance cost Time cost Energy consumption cost Space risk costs and curvature penalty cost Define the total cost of the edges. The details are as follows: ； wherein, is a corresponding weight, i.e. a preset parameter in step 1-1 is represented as follows: ； Step 4-6, according to the total cost of the synthesis of the edges For any path Calculate the total cost .

6. The method of claim 5, wherein, The construction of a perception heuristic A* search scheme in step 5 comprises: Step 5-1, based on the maximum flight speed and considering the minimum turning radius Dubins shortest path length Construct a lower time bound that satisfies the minimum turning radius. ; Step 5-2, according to the minimum energy consumption coefficient of unit distance and the Euclidean distance from the current node to the end point, construct the lower bound of energy consumption under ideal wind conditions ; Step 5 - 3, lower bound on time to meet minimum turning radius lower bound on energy consumption under ideal downwind conditions , constructing a comprehensive heuristic function ; Step 5 - 4, A* search is performed, during which the minimum known cumulative cost from the start to the current node is updated and the composite heuristic function an evaluation function is defined for each node ; Prior queue sorting according to the evaluation function; Step 5-5, iteratively select evaluation function the minimum node expansion and update the adjacent node according to the current minimum cost from the starting point to the adjacent node ; wherein the update method of the current minimum cost from the starting point to the adjacent node is as follows: , the adjacent node is updated; wherein the update method of the current minimum cost from the starting point to the adjacent node is as follows: , the adjacent node is updated; wherein the update method of the current minimum cost from the starting point to the adjacent node is as follows: , the adjacent node is updated; wherein the update method of the current minimum cost from the starting point to the adjacent node is as follows: ； wherein, is the current minimum cost from the start to the adjacent node , is the current minimum cost from the start to the current node , calculated from the total cost of the path in step 4-6, is the combined cost of the edge , is the assignment update operation; Step 5-6, when the target node is selected and expanded, or the priority queue is empty, the search is completed; starting from the target node , the recorded predecessor node relationship is reversed to trace back to the starting point , and the node sequence is extracted; The node sequence That is, the planning path with the minimum integrated cost that satisfies all hard constraints at the current resolution.

7. The method of claim 6, wherein, The execution of two-stage path planning in step 6 comprises: Step 6-1, set the first search resolution, perform step 5, obtain a coarse path, and extract a key node sequence in the coarse path ; Step 6-2, constructing a safety corridor according to the key node sequence, specifically represented as follows: ； wherein, is a key node centered, radius neighborhood; represents a removal forbidden zone and a dynamic occupancy zone; is a safety corridor region; Step 6-3, setting a second search resolution, and re-executing steps 3 to 5 in the safety corridor to obtain a fine path.

8. The method of claim 7, wherein, The B-spline limited curvature shaping of the fine path in step 7 comprises: The nodes in the fine path Construct a cubic spline with the control points as knots, represented as follows: ； wherein is a parameter under the trajectory point; is a cubic spline basis function; Computing parameters under the trajectory point at the curvature .

9. The method of claim 8, wherein, The generation of an executable flight path in step 7, i.e. assigning speed and time to the nodes on the fine path after B-spline limited curvature shaping to generate a directly executable flight path, comprises: According to the curvature and overload constraints, the maximum allowable speed is calculated ; Calculate actual flight speed according to integrated thrust limit and energy budget ; Parameterizing the time, represented as follows: ； wherein, is the arc length of the time of the moment, is the actual flight speed at the arc length position of the trajectory; Finally, outputting an executable flight path, represented as follows: ； wherein , , is the position as a function of time; is the velocity; is the heading angle.

10. The method of claim 9, wherein, The cost updating of the edges in the feasible subgraph in step 8, i.e. when the sea environment and constraint information are updated, only the affected area is incrementally updated, comprises: for each node maintain look-ahead cost ; If the minimum known cumulative cost from the start to the current node is different from the look-ahead cost then the node is added to the priority queue and step 5 is re-executed.

Images