An overwater unmanned aerial vehicle path planning method and system fusing Dubins path curve and Jiangfeng energy consumption correction

Abstract

The application discloses a kind of fusion Dubins path curve and river wind energy consumption correction water unmanned aerial vehicle path planning method and system, comprising: obtaining the original space geographic coordinates of target point to be inspected and time window limit index;Dubins path state space based on physical turning radius constraint is constructed, and Dubins path state index table based on pose six-tuple is introduced to store high-concurrent trajectory data;Dynamic nonlinear electric quantity consumption penalty model is constructed by fusing along river real-time wind field environment vector, and the comprehensive energy consumption of each segmented path is calculated Value of generation;Time window urgency self-adaptive heuristic mixed genetic-ant colony path global collaborative solution is carried out, and global optimal inspection path sequence and control instruction are output;The application realizes the fast convergence solution of high-dimensional nonlinear mixed integer programming problem;Global optimal inspection path sequence and control instruction output can be directly used for unmanned aerial vehicle nest system task issuing, low-altitude maritime supervision scheduling and emergency response decision-making.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) traffic management technology, specifically relating to a method and system for path planning of waterborne UAVs that integrates Dubins trajectory curves and river wind energy consumption correction. Background Technology

[0002] Against the backdrop of the low-altitude economy boom and the evolution of next-generation smart maritime regulatory technologies, routine waterway inspections (such as dynamic evidence collection of illegal ship discharges, monitoring of damage to navigation marks, and emergency response to maritime accidents) using automated drone nests and industrial-grade drones have become a core means of enhancing the ability to perceive the entire waterway.

[0003] However, in actual high-risk industrial deployments and high-concurrency task scheduling on water, existing classical path planning methods (such as the traditional A* algorithm, conventional genetic algorithms with uncorrected dynamics, and mass-based Dijkstra's algorithm) have the following significant technical shortcomings:

[0004] 1. Ignoring Physical and Aerodynamic Constraints: Traditional path planning algorithms generally assume low-altitude aircraft as zero-dimensional simplified point masses, with path points rigidly connected by absolute geometric straight line segments. However, maritime inspection UAVs, in real high-speed flight, are subject to strong physical and dynamic constraints based on the minimum turning radius of the aircraft. When reaching a specific inspection target point or changing course, they cannot complete sharp, zigzag turns in three-dimensional space. Forcing the UAV to fly along a point-mass-like zigzag path will lead to autopilot overshoot, resulting in a high yaw rate, and even deviation from safe airspace, potentially increasing flight safety risks.

[0005] 2. Lack of energy consumption adaptation mechanism for changing wind fields on the water surface: Due to the strong driving force of the topographic funnel effect, the narrow inland waterway has significant and nonlinear local river wind disturbances. When the UAV is traveling along the river, the thrust output, air resistance, and battery power consumption when flying with the wind and against the wind show high asymmetry and nonlinear differences. Traditional algorithms only calculate energy consumption based on static Euclidean distance, which makes it impossible for the UAV to perceive the negative penalty of the wind field during the planning stage. When actually performing the return flight segment against the wind, the UAV often crashes into the river midway due to the surge in drag work and premature battery depletion.

[0006] 3. Instability and poor convergence timeliness in multidimensional and multi-constraint solutions: When the path planning system simultaneously accommodates nonlinear Dubins turning dynamics boundaries, piecewise infinitesimal vector integral energy consumption of spatiotemporally variable wind fields, and soft time window over-limit penalties with varying degrees of urgency, the optimization problem evolves into a typical high-dimensional nonlinear mixed-integer nonlinear programming (MINLP) combinatorial optimization problem. Conventional single heuristic algorithms have extremely narrow search spaces and converge very slowly, making them prone to getting trapped in local optima in the initial calculations, which makes it difficult to meet the real-time replanning requirements in smart maritime scenarios. Summary of the Invention

[0007] To address the shortcomings of the existing technologies, the present invention aims to provide a method and system for path planning of unmanned aerial vehicles (UAVs) that integrates Dubins trajectory curves and river wind energy consumption correction, in order to solve the problems of existing UAV path planning technologies that cannot adapt to dynamic radius, ignore wind field energy consumption penalties, and have poor timeliness in solving multiple constraints.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] The present invention provides a path planning method for a surface-to-water unmanned aerial vehicle (UAV) that integrates Dubins trajectory curves and river wind energy consumption correction, comprising the following steps:

[0010] 1) Obtain the original spatial geographic coordinates and time window constraints of the target point to be inspected;

[0011] 2) Construct a Dubins trajectory state space based on physical turning radius constraints, and introduce a Dubins trajectory state index table based on pose six-tuples to store high-concurrency trajectory data;

[0012] 3) Construct a dynamic nonlinear power consumption penalty model that integrates real-time wind field environmental vectors along the river, and calculate the comprehensive energy consumption value of each segment of the trajectory;

[0013] 4) Perform a global collaborative solution of hybrid genetic-ant colony path based on time window urgency adaptation, and output the globally optimal inspection trajectory sequence and control commands.

[0014] Further, step 1) specifically includes:

[0015] The riverside maritime support base was selected as the origin of the two-dimensional kilometer-scale projection. Obtain data distributed along the river through a maritime sensor network. Original spatial geographic latitude and longitude coordinate sequence of the target points to be inspected , And obtain the upper limit of the soft time window deadline specified for each inspection target point. , as a time window limitation indicator;

[0016] Based on the parametric geographic projection transformation algorithm, the latitude and longitude coordinates of each target point to be inspected are converted into Cartesian kilometer coordinates. As the original spatial geographic coordinates, this eliminates the computational latency caused by frequent calls to trigonometric functions in large-scale, high-concurrency computing due to the Earth's curvature. The specific conversion formula is as follows:

[0017] ;

[0018] ;

[0019] In the formula, The average radius of the Earth;

[0020] Based on the physical parameters of the UAV, a baseline performance constant is established: the cruising speed of the UAV is set as... The maximum usable battery capacity is The baseline rated output power under standard windless conditions is The minimum turning radius in aeronautical physics is .

[0021] Further, step 2) specifically includes:

[0022] Based on the drone's starting position and pose pose of the target point Calculate the length of a continuous smooth track that meets dynamic constraints. The specific steps are as follows:

[0023] Establish from the starting point Using a relative rectangular coordinate system as the reference, the radius of the turning center circle is set as the minimum turning radius. Calculate the initial pose respectively Left turn center coordinates and the center of the right turn coordinates ,as follows:

[0024] ;

[0025] ;

[0026] Calculate the pose of the target point Left turn center coordinates and the center of the right turn coordinates ,as follows:

[0027] ;

[0028] ;

[0029] Calculate the coordinate matrix of the tangent points of the inner and outer tangents of the two circles for each of the six topological patterns, the length of the straight line segment, the variable central angle corresponding to the first and second turning arcs; details are as follows:

[0030] Extract the corresponding starting center based on the specific topological branch of the current traversal. With the center of the target circle Among them, under the LSL and LSR branches, the initial center is... Mapped to the center of the left turn Under the RSR and RSL branches, the initial center of the circle Mapped to the center of the right turn Under the LSL and RSL branches, the target center Mapped to the center of the left turn Under the RSR and LSR branches, the target center Mapped to the center of the right turn Wherein, LSL represents left turn-straight-left turn, RSR represents right turn-straight-right turn, LSR represents left turn-straight-right turn, RSL represents right turn-straight-left turn, RLR represents right turn-left turn-right turn, and LRL represents left turn-right turn-left turn.

[0031] 21) Calculate the turning-straight-turn trajectory pattern as follows:

[0032] Calculate the starting center With the center of the target circle Spatial geometric distance between and the azimuth and inclination angle of the line connecting the centers of the circles The formula is as follows:

[0033] ;

[0034] ;

[0035] In the formula, This represents the arctangent function in all quadrants, used to output the value in... The unique spatial azimuth within the interval;

[0036] External tangent pattern: Two circles have the same physical turning radius, and their common tangent segment is parallel to the line connecting the centers of the circles. The geometric length of the common tangent segment is... Equal to the spatial geometric distance between the two centers , It represents the geometric length of the common tangent segment between the two circles in the turn-straight-turn trajectory mode, and serves as the upper limit variable for the one-dimensional integral of the energy consumption cost of the straight segment.

[0037] For the LSL topology pattern branch, the angle of deviation of the external common tangent relative to the normal of the line connecting the centers is... The first tangent point is calculated. Second tangent point The planar kilometer coordinates are as follows:

[0038] ;

[0039] ;

[0040] ;

[0041] ;

[0042] For the RSR topology pattern branch, the angle of deviation of the external common tangent relative to the normal of the line connecting the centers is... The corresponding first tangent point is calculated. Second tangent point The planar kilometer coordinates are as follows:

[0043] ;

[0044] ;

[0045] ;

[0046] ;

[0047] Internal tangent pattern: There is a fixed geometric angle between the line connecting the centers of the two circles and the internal common tangent. The expression is: At this point, the geometric length of the internal common tangent segment... for:

[0048] ;

[0049] For the LSR topology pattern branch, the first tangent point corresponding to the LSR topology pattern branch is calculated. Second tangent point The Cartesian kilometer coordinates are as follows:

[0050] ;

[0051] ;

[0052] ;

[0053] ;

[0054] For the RSL topology pattern branch, calculate and obtain the first tangent point corresponding to the RSL topology pattern branch. Second tangent point The Cartesian kilometer coordinates are as follows:

[0055] ;

[0056] ;

[0057] ;

[0058] ;

[0059] Calculation of the variable central angle in the turn-straight-turn trajectory mode: Based on the kilometer coordinates of the tangent point plane, extract the instantaneous heading velocity vector angle of the UAV at the two tangent point positions. and By using the instantaneous heading velocity vector angle and with the starting heading angle and target heading angle By performing modulus difference calculations separately, the variable central angle radian value corresponding to the first segment of the turning arc under the current turning-straight-turn trajectory mode branch can be calculated. And the variable central angle radian value corresponding to the second segment of the turning arc. ;

[0060] 22) The turning-turning-turning trajectory pattern is calculated as follows:

[0061] In the turn-turn-turn trajectory mode, the geometric distance between the two circle centers satisfies the constraint boundary. This mode features a smooth, full-circle transition, excluding straight line segments and the geometric length of the common tangent segment. By introducing a radius of And the intermediate reverse transition center is externally tangent to both the initial center and the target center. Perform analysis;

[0062] The three centers form a circle in space with a length of [missing information]. Given an isosceles triangle, the isosceles base angles of the isosceles triangle are calculated to be... ;

[0063] For the RLR topology pattern branch, the starting center corresponds to the right turn center. The target circle's center corresponds to the right turn's center. The intermediate transition center is represented by the left turn center, and the planar kilometer coordinates of the intermediate transition center are calculated as follows:

[0064] ;

[0065] ;

[0066] Adjacent circular arcs exhibit a purely geometric external tangent topological relationship. The first point of tangency is constructed and solved using the midpoint equation. Second tangent point The planar kilometer coordinates are as follows:

[0067] ;

[0068] ;

[0069] ;

[0070] ;

[0071] For the LRL topology pattern branch, the starting center corresponds to the left turn center. The target circle's center corresponds to the left turn's center. The intermediate transition center is represented by the right turn center, and the planar kilometer coordinates of the intermediate transition center are calculated as follows:

[0072] ;

[0073] ;

[0074] Similarly, using the equation of the midpoint of the geometric tangency point, we can construct and solve for the first tangency point. Second tangent point The planar kilometer coordinates are as follows:

[0075] ;

[0076] ;

[0077] ;

[0078] ;

[0079] Calculation of variable central angle in turn-turn-turn trajectory mode: Calculate the geocentric turning angle swept from the starting point pose to the first tangent point position and from the second tangent point position to the target point pose under the turning radius constraint, and output the variable central angle radian value corresponding to the first turning arc. And the variable central angle radian value corresponding to the second segment of the turning arc. Based on the symmetry of the vertex angle of an isosceles triangle, the variable central angle in radians of the intermediate reverse transition arc segment is calculated. The calculation formula is as follows:

[0080] ;

[0081] In the formula, In the turn-turn-turn trajectory mode, the variable central angle radian value swept by the intermediate reverse transition turn arc segment is represented, which serves as the upper limit variable of the one-dimensional integral of the energy consumption cost of the second reverse turn segment.

[0082] Construct a unified trajectory operator that integrates topology switching indicator factors and calculates each topology mode branch. Total dynamic track length The expression is as follows:

[0083] ;

[0084] In the formula, It is represented as a set of six topological pattern branches; This is a binary indicator operator for topology categories, used to characterize the topology of the currently solved track. When, its value is 1, when When the time is right, its value is 0;

[0085] The minimum value of the integrated energy consumption calculation result corresponding to the six topology mode branches is extracted and used as the final Dubins track length under the connection of the starting point pose and the target point pose. The final Dubins track length is used as the core optimal metric in the Dubins track state space, thereby completing the construction and energy consumption calculation of the continuous and smooth track that meets the dynamic constraints. The function expression is as follows:

[0086] ;

[0087] A Dubins track state index table based on pose six-tuples is constructed. The Dubins track state index table adopts a high-concurrency key-value pair storage architecture with in-memory hashing. Its specific technical implementation and indexing mechanism are as follows: For the starting point pose and the target point pose in any iterative traversal, the kilometer rectangular coordinates and radian heading angles of the starting point pose and the target point pose are respectively normalized and truncated to retain preset valid values, and then normalized and combined into a unique six-tuple hash cache key, as shown in the following expression:

[0088] ;

[0089] The storage target value corresponding to the six-element hash cache key is defined as: the final Dubins track topology pattern code after geometric branching, the integral upper limit boundary of each turn / straight segment, and the final river wind correction comprehensive energy consumption cost under the current wind field environment. The mapping relationship is represented by the following quintuple state vector:

[0090] When the current track belongs to the CSC track mode, the state vector is represented as:

[0091] ;

[0092] When the current track belongs to the CCC track mode, the state vector is represented as:

[0093] ;

[0094] In the formula, These are the enumeration identifiers for the six topological patterns. This is the final integrated energy consumption result. During high-concurrency path retrieval scheduling, if the corresponding key is accurately matched in the memory retrieval unit using the six-element hash cache key, the geometric analysis and integration process is skipped, and the pre-stored state vector is directly output through the dynamic pointer. If no match is found, the geometric tangent point transformation unit is activated to perform a full-process solution, and the newly generated key-value pairs are dynamically written into the index table.

[0095] For any starting point pose and target point pose in any iterative traversal, the Cartesian kilometer coordinates and radian heading angles of the starting point pose and target point pose are respectively normalized and truncated to retain preset valid values, and then normalized and combined into a unique six-element hash cache key. ,as follows:

[0096] ;

[0097] During path retrieval scheduling, if an existing key matching the six-tuple hash cache key of the current iteration input is detected in the Dubins track state index table, a cache hit is determined, and the pre-stored five-tuple state vector corresponding to the key is directly output via memory read / write instructions; otherwise, if a matching key is not successfully retrieved in the Dubins track state index table, a cache miss is determined, and Dubins geometric analytical solution is performed. A complete geometric tangent point analytical solution process is executed for the current starting point pose and the target point pose, and the calculation results are stored in the Dubins track state index table and a corresponding mapping relationship is established.

[0098] Furthermore, step 3) specifically includes:

[0099] 31) The environmental wind resistance vector is obtained in real time by meteorological sensors deployed along the river for inspection. And extract wind speed and wind direction angle ;

[0100] 32) Extract the instantaneous heading angle of the UAV at the tangent of the straight segment or the integral element of the circular arc on the Dubins track. Acquire real-time environmental wind direction angles from a multi-source sensor network deployed around the navigable waterways along the Yangtze River. With ambient wind speed ; Calculate the spatial relative angle between the instantaneous flight heading angle and the ambient wind direction angle. The expression is as follows:

[0101] ;

[0102] 33) Construct a dynamic nonlinear power consumption penalty model that integrates real-time wind field environmental vectors along the river to quantitatively calculate the additional power loss of the amphibious UAV due to nonlinear changes in air resistance under non-uniform wind field conditions; the details are as follows:

[0103] Based on the spatial relative angle in step 32) Combining the principles of wind vector synthesis in aerodynamics, the environmental wind speed is... Decoupled projection is performed along the current direction of movement of the drone, and the axial wind speed components along the headwind and the wind direction experienced by the drone at the current integral element are calculated separately. With lateral wind speed component ,as follows:

[0104] ;

[0105] ;

[0106] Combined with the preset cruise relative speed of the waterborne drone in calm wind conditions The dynamic nonlinear actual air velocity after overcoming the air resistance of the wind field along the river was synthesized. The expression is as follows:

[0107] ;

[0108] Based on the nonlinear physical law in fluid mechanics that fluid resistance is proportional to the square of relative velocity, a nonlinear river wind energy consumption correction coefficient is constructed. The analytical equation, which serves as the core variable for the subsequent one-dimensional integral of energy consumption cost, is expressed as follows:

[0109] ;

[0110] In the formula, Forward drag penalty sensitivity coefficient along the instantaneous flight heading; This is the gain coefficient for power compensation due to crosswind drift caused by wind deflection. and It can be identified based on historical flight data; when hour, This manifests as a nonlinear gain penalty on energy consumption for the current flight segment under headwind conditions; when hour, This manifests as a negative but beneficial compensation for energy consumption costs under tailwind conditions;

[0111] 34) Based on the current optimal topology mode branch selected by the Dubins track state index table constructed in step 2), and the integral upper limit boundary corresponding to each segment, along the geometric trajectory composed of the tangent of the straight line segment or the arc segment of the Dubins track, establish a piecewise one-dimensional integral equation for the integral element in step 33), thereby calculating the final river wind correction comprehensive energy consumption value under the pose connection between the starting point and the target point. The expression is as follows:

[0112] ;

[0113] In the formula, The dynamic comprehensive energy consumption cost after correction by the environmental wind field is represented as the flight segment where the pose of the starting point and the target point are connected. The length of a continuous, smooth trajectory between the starting point and the target point that conforms to dynamic constraints; This is the baseline rated output power of the drone in a standard windless environment; This represents the spatial displacement distance of the current integral element.

[0114] Further, step 4) specifically includes:

[0115] 41) Dynamic time window urgency calculation: When the virtual optimization ant colony is at the starting point of the current flight segment... At that time, for the set of target points to be inspected. The target points to be inspected By reading the target points to be inspected The spatiotemporal constraints are used to calculate the corresponding soft time window urgency adaptive heuristic factor. ,as follows:

[0116] ;

[0117] In the formula, Indicates the inspection target point The specified soft time window deadline; This represents the total flight time consumed by an individual in the virtual optimization ant colony under the current path construction branch; The starting point of the flight segment Inspection target points The length of a continuous, smooth track that conforms to dynamic constraints; To prevent the denominator from being zero and to provide a small positive real number for nonlinear amplification of boundary urgency;

[0118] 42) Construction of initial trajectory sequence based on hybrid ant colony: based on the urgency heuristic factor of each target point And a heuristic factor that allows all other candidate points within the set of target points to be inspected. Perform normalized mapping and calculate the virtual optimization ant colony individuals from the starting point of the current flight segment. Move to the next target point to be inspected. Adaptive state transition probability The calculation expression is as follows:

[0119] ;

[0120] In the formula, This indicates the starting point of the current virtual optimization ant colony individual from the current flight segment. Transfer to the target point to be inspected The adaptive state transition probability value; This represents the set of allowed target points to be inspected that have not yet been visited by any individual member of the current virtual optimization ant colony; The starting point of the current flight segment and the target points to be inspected scalar of historical residual pheromone concentration along the connection path; This is a dynamic environmental energy consumption heuristic factor, whose value is the reciprocal of the overall energy consumption cost, i.e. ; These are the control weight indices for each heuristic factor, where β>0, γ>0, and δ>0. The starting point of the current flight segment With any candidate point within the set of allowed inspection target points scalar of historical residual pheromone concentration along the connection path; The starting point of the current flight segment With any candidate point Heuristic factors for dynamic environmental energy consumption under the connection path; For any candidate point The corresponding soft time window urgency adaptive heuristic factor;

[0121] Utilizing adaptive state transition probabilities Constructing a roulette wheel discrete probability selection operator: allowing the set of target points to be inspected. The adaptive state transition probability values ​​corresponding to all candidate points are sequentially accumulated by concatenating the first and last values, thereby achieving... The axis is divided into non-overlapping cumulative probability intervals corresponding to each candidate point; where, any i-th The cumulative probability interval boundary corresponding to each candidate point is represented as follows: ;

[0122] Generate a [something] through the computer kernel. Random numbers that are uniformly distributed within the interval ;

[0123] The random number The cumulative probability interval of each candidate point is compared with the target point to determine the next target point to be inspected. The point that falls within the cumulative probability interval is the next target point to be inspected. The mapping equation is determined to satisfy the following equation:

[0124] ;

[0125] Through the aforementioned roulette wheel selection operator with discrete probability, a cyclical iteration is performed until the set is reached. The set becomes empty, and the initial construction of a complete inspection topology track that is physically feasible for each individual virtual optimization ant colony is completed.

[0126] 43) Spatiotemporal state recursive update: Complete the next target point to be inspected. Once locked, the spatiotemporal state variables of the current virtual optimization ant colony individuals are immediately iteratively calculated; the total flight time consumed at the next moment is accumulated. Perform dynamic cumulative updates as follows:

[0127] ;

[0128] In the formula, For drones at the target points to be inspected The fixed dwell time required for evidence collection or hovering inspection; the locked target points to be inspected. From the target points to be inspected Remove from the list and add the points to be inspected. As the starting point of the new current flight segment Repeat steps 41) to 43) until the set is complete. It becomes an empty set, thus completing the initial construction of a complete inspection topology track with global spatiotemporal constraints.

[0129] 44) High-quality population gene injection: After all virtual optimization ant colony individuals have completed iterative traversal, a set of high-quality local track sequences with time window feasibility is output. This set of sequences is used as external high-quality path coding individuals and directly forced into the initial track coding population of the genetic algorithm to complete cross-algorithm solution space collaborative initialization and prior knowledge transfer.

[0130] 45) Post-genetic algorithm deep optimization: Utilizing a comprehensive fitness function that includes multi-dimensional physical boundaries and violation constraints. The candidate path coding individuals in the initial track coding population are iteratively evaluated using the following function:

[0131] ;

[0132] In the formula, The total river wind correction comprehensive energy consumption value for the entire inspection route; Route represents the set of all waterway segments included in the current route. For the soft time window exceeding the limit and timeout penalty term, Nodes represents the set of all inspection points; Target points to be inspected The moment; The weighting coefficient for penalties for late arrivals. The penalty coefficient for the boundary cliff that approaches infinity; A cliff-like boundary isolation penalty term is applied to the total charge exceeding the available battery capacity. The formula for the decision operator is as follows:

[0133] ;

[0134] The global iterative evolution is performed through selection, crossover, and mutation operators until the convergence termination condition is met, and the globally optimal inspection track sequence is output.

[0135] Constructing a population selection operator based on the roulette wheel mechanism: for individuals encoding any path in the current generation of the genetic population. Based on the comprehensive fitness function value The discrete selection probability of it being inherited by the next generation of temporary population is calculated as follows:

[0136] ;

[0137] In the formula, Indicates the number of the current genetic population. The probability value of the selected track evolution adaptation of each candidate path encoding individual; This represents the preset global population size of the genetic algorithm; the probability value of adaptive selection is used based on the trajectory evolution. Construct a roulette wheel evolutionary selection operator by generating a... Uniformly distributed random numbers within the interval This allows for the identification of candidate path-encoding individuals for the next generation of iterative evolution. The roulette wheel evolution selection operator satisfies the following mapping equation:

[0138] ;

[0139] By repeatedly executing the above roulette evolution selection operator mapping equation The selection of the track population was completed in one step, and an intermediate evolutionary population was constructed.

[0140] Specifically, the intermediate evolutionary population is subjected to global cyclic evolution by introducing scene-adaptive crossover and adaptive mutation operators. The specific techniques include:

[0141] Two candidate path coding individuals from the intermediate evolutionary population are randomly selected, and the order of some of their target points to be inspected is swapped to achieve track topology crossover; at the same time, the order of visits to inspection points in a single candidate path coding individual is randomly changed to achieve track topology variation.

[0142] After each crossover and mutation operation, the newly generated candidate path coding individual is fed back to steps 2) and 3) in real time. The total length of the continuous smooth track mapped by the candidate path coding individual that meets the dynamic constraints is recalculated, and the dynamic comprehensive energy consumption cost and time window penalty corrected by the environmental wind field are dynamically updated until the preset convergence termination condition is met, and the globally optimal inspection track sequence is output.

[0143] Finally, the expected flight heading angles corresponding to each segmented infinitesimal trajectory in the globally optimal inspection trajectory sequence are extracted. and composite airspeed , as a dynamic control command for tracking the flight path of unmanned aerial vehicles (UAVs) on the sea.

[0144] This invention also provides a path planning system for a seaplane unmanned aerial vehicle (UAV) that integrates Dubins trajectory curves and river wind energy consumption correction, comprising:

[0145] The data acquisition unit is used to acquire the original spatial geographic coordinates and time window limit indicators of the target point to be inspected.

[0146] The first building unit is used to construct the Dubins trajectory state space based on physical turning radius constraints, and introduces a Dubins trajectory state index table based on pose six-tuples to store high-concurrency trajectory data;

[0147] The second building unit is used to construct a dynamic nonlinear power consumption penalty model that integrates real-time wind field environmental vectors along the river, and to calculate the comprehensive energy consumption cost of each segment of the flight path.

[0148] The instruction generation unit is used to perform global collaborative solution of hybrid genetic-ant colony path based on time window urgency adaptation, and output the globally optimal inspection trajectory sequence and control instructions.

[0149] The beneficial effects of this invention are:

[0150] This invention integrates Dubins trajectory curves, river wind energy consumption correction, and time window urgency adaptation to make the path planning results of the amphibious UAV more consistent with real aviation physical constraints and complex meteorological environments, while meeting the needs of real-time path planning and safety requirements. The method of this invention can perform parallel Dubins six-topology branch solution, nonlinear drag coefficient quantification, and piecewise vector integration for different starting point-target point pose combinations and river wind field environments, thereby obtaining a series of physically feasible and energy-efficient candidate trajectories.

[0151] This invention combines a Dubins trajectory state index table based on pose six-tuples with a hybrid genetic-ant colony dual-layer heuristic search to achieve faster convergence in solving high-dimensional nonlinear mixed integer programming problems. The output globally optimal inspection trajectory sequence and control commands can be directly used for UAV nest system task issuance, low-altitude maritime regulatory scheduling, and emergency response decision-making. Through multi-dimensional spatiotemporal constraint tracing and dual penalty function isolation, the specific causes of energy consumption exceeding limits or time-disruption imbalances can be clearly identified, supporting accurate trajectory replanning and power warning, and providing a basis for airworthiness safety verification and closed-loop adjustment of the flight control system. Attached Figure Description

[0152] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0153] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present invention.

[0154] Reference Figure 1 As shown, the present invention provides a method for path planning of a waterborne unmanned aerial vehicle (UAV) that integrates Dubins trajectory curves and river wind energy consumption correction. The steps are as follows:

[0155] 1) Obtain the original spatial geographic coordinates and time window constraints of the target point to be inspected; specifically including:

[0156] The riverside maritime support base was selected as the origin of the two-dimensional kilometer-scale projection. Obtain data distributed along the river through a maritime sensor network. Original spatial geographic latitude and longitude coordinate sequence of the target points to be inspected , And obtain the upper limit of the soft time window deadline specified for each inspection target point. , as a time window limitation indicator;

[0157] Based on the parametric geographic projection transformation algorithm, the latitude and longitude coordinates of each target point to be inspected are converted into Cartesian kilometer coordinates. As the original spatial geographic coordinates, this eliminates the computational latency caused by frequent calls to trigonometric functions in large-scale, high-concurrency computing due to the Earth's curvature. The specific conversion formula is as follows:

[0158] ;

[0159] ;

[0160] In the formula, The value is the average radius of the Earth; in this embodiment, it is taken as 6371.393 km.

[0161] Based on the physical parameters of the UAV, a baseline performance constant is established: the cruising speed of the UAV is set as... The maximum usable battery capacity is The baseline rated output power under standard windless conditions is The minimum turning radius in aeronautical physics is .

[0162] 2) Construct a Dubins trajectory state space based on physical turning radius constraints, and introduce a Dubins trajectory state index table based on pose six-tuples to store high-concurrency trajectory data; specifically including:

[0163] Based on the drone's starting position and pose pose of the target point Calculate the length of a continuous smooth track that meets dynamic constraints. The specific steps are as follows:

[0164] Establish from the starting point Using a relative rectangular coordinate system as the reference, the radius of the turning center circle is set as the minimum turning radius. Calculate the initial pose respectively Left turn center coordinates and the center of the right turn coordinates ,as follows:

[0165] ;

[0166] ;

[0167] Calculate the pose of the target point Left turn center coordinates and the center of the right turn coordinates ,as follows:

[0168] ;

[0169] ;

[0170] Calculate the coordinate matrix of the tangent points of the inner and outer tangents of the two circles for each of the six topological patterns, the length of the straight line segment, the variable central angle corresponding to the first and second turning arcs; details are as follows:

[0171] Extract the corresponding starting center based on the specific topological branch of the current traversal. With the center of the target circle Among them, under the LSL and LSR branches, the initial center is... Mapped to the center of the left turn Under the RSR and RSL branches, the initial center of the circle Mapped to the center of the right turn Under the LSL and RSL branches, the target center Mapped to the center of the left turn Under the RSR and LSR branches, the target center Mapped to the center of the right turn Wherein, LSL represents left turn-straight-left turn, RSR represents right turn-straight-right turn, LSR represents left turn-straight-right turn, RSL represents right turn-straight-left turn, RLR represents right turn-left turn-right turn, and LRL represents left turn-right turn-left turn.

[0172] 21) Calculate the turn-straight-turn (CSC) trajectory pattern (including LSL, RSR, LSR, and RSL topology pattern branches), as follows:

[0173] Calculate the starting center With the center of the target circle Spatial geometric distance between and the azimuth and inclination angle of the line connecting the centers of the circles The formula is as follows:

[0174] ;

[0175] ;

[0176] In the formula, This represents the arctangent function in all quadrants, used to output the value in... The unique spatial azimuth within the interval;

[0177] For external tangent patterns (LSL and RSR topological branches): the two circles have the same physical turning radius, their common tangent segment is parallel to the line connecting the centers of the circles, and the geometric length of the common tangent segment is... Equal to the spatial geometric distance between the two centers , It represents the geometric length of the common tangent segment between the two circles in the turn-straight-turn trajectory mode, and serves as the upper limit variable for the one-dimensional integral of the energy consumption cost of the straight segment.

[0178] For the LSL topology pattern branch, the angle of deviation of the external common tangent relative to the normal of the line connecting the centers is... The first tangent point is calculated. Second tangent point The planar kilometer coordinates are as follows:

[0179] ;

[0180] ;

[0181] ;

[0182] ;

[0183] For the RSR topology pattern branch, the angle of deviation of the external common tangent relative to the normal of the line connecting the centers is... The corresponding first tangent point is calculated. Second tangent point The planar kilometer coordinates are as follows:

[0184] ;

[0185] ;

[0186] ;

[0187] ;

[0188] For the internal tangent pattern (LSR and RSL topology pattern branches): there is a fixed geometric angle between the line connecting the two centers and the internal tangent. The expression is: At this point, the geometric length of the internal common tangent segment... for:

[0189] ;

[0190] For the LSR topology pattern branch, the first tangent point corresponding to the LSR topology pattern branch is calculated. Second tangent point The Cartesian kilometer coordinates are as follows:

[0191] ;

[0192] ;

[0193] ;

[0194] ;

[0195] For the RSL topology pattern branch, calculate and obtain the first tangent point corresponding to the RSL topology pattern branch. Second tangent point The Cartesian kilometer coordinates are as follows:

[0196] ;

[0197] ;

[0198] ;

[0199] ;

[0200] Calculation of the variable central angle in the turn-straight-turn trajectory mode: Based on the kilometer coordinates of the tangent point plane, extract the instantaneous heading velocity vector angle of the UAV at the two tangent point positions. and By using the instantaneous heading velocity vector angle and with the starting heading angle and target heading angle By performing modulus difference calculations separately, the variable central angle radian value corresponding to the first segment of the turning arc under the current turning-straight-turn trajectory mode branch can be calculated. And the variable central angle radian value corresponding to the second segment of the turning arc. ;

[0201] 22) The turn-turn-turn (CCC) trajectory pattern (including RLR and LRL topology pattern branches) is calculated as follows:

[0202] In the turn-turn-turn trajectory mode, the geometric distance between the two circle centers satisfies the constraint boundary. This mode features a smooth, full-circle transition, excluding straight line segments and the geometric length of the common tangent segment. By introducing a radius of And the intermediate reverse transition center is externally tangent to both the initial center and the target center. Perform analysis;

[0203] The three centers form a circle in space with a length of [missing information]. Given an isosceles triangle, the isosceles base angles of the isosceles triangle are calculated to be... ;

[0204] For the RLR topology pattern branch, the starting center corresponds to the right turn center. The target circle's center corresponds to the right turn's center. The intermediate transition center is represented by the left turn center, and the planar kilometer coordinates of the intermediate transition center are calculated as follows:

[0205] ;

[0206] ;

[0207] Adjacent circular arcs exhibit a purely geometric external tangent topological relationship. The first point of tangency is constructed and solved using the midpoint equation. Second tangent point The planar kilometer coordinates are as follows:

[0208] ;

[0209] ;

[0210] ;

[0211] ;

[0212] For the LRL topology pattern branch, the starting center corresponds to the left turn center. The target circle's center corresponds to the left turn's center. The intermediate transition center is represented by the right turn center, and the planar kilometer coordinates of the intermediate transition center are calculated as follows:

[0213] ;

[0214] ;

[0215] Similarly, using the equation of the midpoint of the geometric tangency point, we can construct and solve for the first tangency point. Second tangent point The planar kilometer coordinates are as follows:

[0216] ;

[0217] ;

[0218] ;

[0219] ;

[0220] Calculation of variable central angle in turn-turn-turn trajectory mode: Calculate the geocentric turning angle swept from the starting point pose to the first tangent point position and from the second tangent point position to the target point pose under the turning radius constraint, and output the variable central angle radian value corresponding to the first turning arc. And the variable central angle radian value corresponding to the second segment of the turning arc. Based on the symmetry of the vertex angle of an isosceles triangle, the variable central angle in radians of the intermediate reverse transition arc segment is calculated. The calculation formula is as follows:

[0221] ;

[0222] In the formula, In the turn-turn-turn trajectory mode, the variable central angle radian value swept by the intermediate reverse transition turn arc segment is represented, which serves as the upper limit variable of the one-dimensional integral of the energy consumption cost of the second reverse turn segment.

[0223] Construct a unified trajectory operator that integrates topology switching indicator factors and calculates each topology mode branch. Total dynamic track length The expression is as follows:

[0224] ;

[0225] In the formula, It is represented as a set of six topological pattern branches; This is a binary indicator operator for topology categories, used to characterize the topology of the currently solved track. When, its value is 1, when When the time is right, its value is 0;

[0226] The minimum value (min) of the integrated energy consumption calculation results corresponding to the six topology mode branches is extracted and used as the final Dubins track length under the connection of the starting point pose and the target point pose. The final Dubins track length is used as the core optimal metric in the Dubins track state space, thereby completing the construction and energy consumption calculation of the continuous and smooth track that meets the dynamic constraints. The function expression is as follows:

[0227] ;

[0228] A Dubins Trajectory StateIndex Table based on pose six-tuples is constructed. This table employs a high-concurrency key-value storage architecture with in-memory hashing. The specific technical implementation and indexing mechanism are as follows: For the starting point pose and target point pose in any iterative traversal, the kilometer Cartesian coordinates and radian heading angles of the starting point pose and target point pose are respectively normalized and truncated to retain preset valid values, and then normalized and combined into a unique six-tuple hash cache key, as shown in the following expression:

[0229] ;

[0230] The storage target value (Value) corresponding to the six-element hash cache key is defined as: the final Dubins track topology pattern code after geometric branching, the integral upper limit boundary of each turn / straight segment, and the final river wind correction comprehensive energy consumption value corresponding to the current wind field environment. The mapping relationship is represented by the following quintuple state vector:

[0231] When the current track belongs to the CSC track mode, the state vector is represented as:

[0232] ;

[0233] When the current track belongs to the CCC track mode, the state vector is represented as:

[0234] ;

[0235] In the formula, These are the enumeration identifiers for the six topological patterns. This is the final integrated energy consumption result. During high-concurrency path retrieval scheduling, if the corresponding key is accurately matched in the memory retrieval unit using the six-element hash cache key, the geometric analysis and integration process is skipped, and the pre-stored state vector is directly output through the dynamic pointer. If no match is found, the geometric tangent point transformation unit is activated to perform a full-process solution, and the newly generated key-value pairs are dynamically written into the index table.

[0236] For any starting point pose and target point pose in any iterative traversal, the Cartesian kilometer coordinates and radian heading angles of the starting point pose and target point pose are respectively normalized and truncated to retain preset valid values, and then normalized and combined into a unique six-element hash cache key. ,as follows:

[0237] ;

[0238] During path retrieval scheduling, if an existing key matching the six-tuple hash cache key of the current iteration input is detected in the Dubins track state index table, a cache hit is determined, and the pre-stored five-tuple state vector corresponding to the key is directly output via memory read / write instructions; otherwise, if a matching key is not successfully retrieved in the Dubins track state index table, a cache miss is determined, and Dubins geometric analytical solution is performed. A complete geometric tangent point analytical solution process is executed for the current starting point pose and the target point pose, and the calculation results are stored in the Dubins track state index table and a corresponding mapping relationship is established.

[0239] 3) Construct a dynamic nonlinear energy consumption penalty model that integrates real-time wind field environmental vectors along the river, and calculate the comprehensive energy consumption value of each segment of the flight path; specifically including:

[0240] 31) The environmental wind resistance vector is obtained in real time by meteorological sensors deployed along the river for inspection. And extract wind speed and wind direction angle ;

[0241] 32) Extract the instantaneous heading angle of the UAV at the tangent of the straight segment or the integral element of the circular arc on the Dubins track. Acquire real-time environmental wind direction angles from a multi-source sensor network deployed around the navigable waterways along the Yangtze River. With ambient wind speed ; Calculate the spatial relative angle between the instantaneous flight heading angle and the ambient wind direction angle. The expression is as follows:

[0242] ;

[0243] 33) Construct a dynamic nonlinear power consumption penalty model that integrates real-time wind field environmental vectors along the river to quantitatively calculate the additional power loss of the amphibious UAV due to nonlinear changes in air resistance under non-uniform wind field conditions; the details are as follows:

[0244] Based on the spatial relative angle in step 32) Combining the principles of wind vector synthesis in aerodynamics, the environmental wind speed is... Decoupled projection is performed along the current direction of movement of the drone, and the axial wind speed components along the headwind and the wind direction experienced by the drone at the current integral element are calculated separately. With lateral wind speed component ,as follows:

[0245] ;

[0246] ;

[0247] Combined with the preset cruise relative speed of the waterborne drone in calm wind conditions The dynamic nonlinear actual air velocity after overcoming the air resistance of the wind field along the river was synthesized. The expression is as follows:

[0248] ;

[0249] Based on the nonlinear physical law in fluid mechanics that fluid resistance is proportional to the square of relative velocity, a nonlinear river wind energy consumption correction coefficient is constructed. The analytical equation, which serves as the core variable for the subsequent one-dimensional integral of energy consumption cost, is expressed as follows:

[0250] ;

[0251] In the formula, Forward drag penalty sensitivity coefficient along the instantaneous flight heading; This is the gain coefficient for power compensation due to crosswind drift caused by wind deflection. and It can be identified based on historical flight data; when hour, This manifests as a nonlinear gain penalty on energy consumption for the current flight segment under headwind conditions; when hour, This manifests as a negative but beneficial compensation for energy consumption costs under tailwind conditions;

[0252] 34) Based on the current optimal topology mode branch selected by the Dubins track state index table constructed in step 2), and the integral upper limit boundary corresponding to each segment, along the geometric trajectory composed of the tangent of the straight line segment or the arc segment of the Dubins track, establish a piecewise one-dimensional integral equation for the integral element in step 33), thereby calculating the final river wind correction comprehensive energy consumption value under the pose connection between the starting point and the target point. The expression is as follows:

[0253] ;

[0254] In the formula, The dynamic comprehensive energy consumption cost after correction by the environmental wind field is represented as the flight segment where the pose of the starting point and the target point are connected. The length of a continuous, smooth trajectory between the starting point and the target point that conforms to dynamic constraints; This is the baseline rated output power of the drone in a standard windless environment; This represents the spatial displacement distance of the current integral element.

[0255] 4) Perform a global collaborative solution for hybrid genetic-ant colony pathfinding inspired by time window urgency adaptation, and output the globally optimal inspection trajectory sequence and control commands; specifically including:

[0256] 41) Dynamic time window urgency calculation: When the virtual optimization ant colony is at the starting point of the current flight segment... At that time, for the set of target points to be inspected. The target points to be inspected By reading the target points to be inspected The spatiotemporal constraints are used to calculate the corresponding soft time window urgency adaptive heuristic factor. ,as follows:

[0257] ;

[0258] In the formula, Indicates the inspection target point The specified soft time window deadline; This represents the total flight time consumed by an individual in the virtual optimization ant colony under the current path construction branch; The starting point of the flight segment Inspection target points The length of a continuous, smooth track that conforms to dynamic constraints; To prevent the denominator from being zero and to provide a small positive real number for nonlinear amplification of boundary urgency;

[0259] 42) Construction of initial trajectory sequence based on hybrid ant colony: based on the urgency heuristic factor of each target point And a heuristic factor that allows all other candidate points within the set of target points to be inspected. Perform normalized mapping and calculate the virtual optimization ant colony individuals from the starting point of the current flight segment. Move to the next target point to be inspected. Adaptive state transition probability The calculation expression is as follows:

[0260] ;

[0261] In the formula, This indicates the starting point of the current virtual optimization ant colony individual from the current flight segment. Transfer to the target point to be inspected The adaptive state transition probability value; This represents the set of allowed target points to be inspected that have not yet been visited by any individual member of the current virtual optimization ant colony; The starting point of the current flight segment and the target points to be inspected scalar of historical residual pheromone concentration along the connection path; This is a dynamic environmental energy consumption heuristic factor, whose value is the reciprocal of the overall energy consumption cost, i.e. ; These are the control weight indices for each heuristic factor, where β>0, γ>0, and δ>0. The starting point of the current flight segment With any candidate point within the set of allowed inspection target points scalar of historical residual pheromone concentration along the connection path; The starting point of the current flight segment With any candidate point Heuristic factors for dynamic environmental energy consumption under the connection path; For any candidate point The corresponding soft time window urgency adaptive heuristic factor;

[0262] Utilizing adaptive state transition probabilities Constructing a roulette wheel discrete probability selection operator: allowing the set of target points to be inspected. The adaptive state transition probability values ​​corresponding to all candidate points are sequentially accumulated by concatenating the first and last values, thereby achieving... The axis is divided into non-overlapping cumulative probability intervals corresponding to each candidate point; where, any i-th The cumulative probability interval boundary corresponding to each candidate point is represented as follows: ;

[0263] Generate a [something] through the computer kernel. Random numbers that are uniformly distributed within the interval ;

[0264] The random number The cumulative probability interval of each candidate point is compared with the target point to determine the next target point to be inspected. The point that falls within the cumulative probability interval is the next target point to be inspected. The mapping equation is determined to satisfy the following equation:

[0265] ;

[0266] Through the aforementioned roulette wheel selection operator with discrete probability, a cyclical iteration is performed until the set is reached. The set becomes empty, and the initial construction of a complete inspection topology track that is physically feasible for each individual virtual optimization ant colony is completed.

[0267] 43) Spatiotemporal state recursive update: Complete the next target point to be inspected. Once locked, the spatiotemporal state variables of the current virtual optimization ant colony individuals are immediately iteratively calculated; the total flight time consumed at the next moment is accumulated. Perform dynamic cumulative updates as follows:

[0268] ;

[0269] In the formula, For drones at the target points to be inspected The fixed dwell time required for evidence collection or hovering inspection; the locked target points to be inspected. From the target points to be inspected Remove from the list and add the points to be inspected. As the starting point of the new current flight segment Repeat steps 41) to 43) until the set is complete. It becomes an empty set, thus completing the initial construction of a complete inspection topology track with global spatiotemporal constraints.

[0270] 44) High-quality population gene injection: After all virtual optimization ant colony individuals have completed iterative traversal, a set of high-quality local track sequences with time window feasibility is output. This set of sequences is used as external high-quality path coding individuals and directly forced into the initial track coding population of the genetic algorithm to complete cross-algorithm solution space collaborative initialization and prior knowledge transfer.

[0271] 45) Post-genetic algorithm deep optimization: Utilizing a comprehensive fitness function that includes multi-dimensional physical boundaries and violation constraints. The candidate path coding individuals in the initial track coding population are iteratively evaluated using the following function:

[0272] ;

[0273] In the formula, The total river wind correction comprehensive energy consumption value for the entire inspection route; Route represents the set of all waterway segments included in the current route. For the soft time window exceeding the limit and timeout penalty term, Nodes represents the set of all inspection points; Target points to be inspected The moment; The weighting coefficient for penalties for late arrivals. The penalty coefficient for the boundary cliff that approaches infinity; A cliff-like boundary isolation penalty term is applied to the total charge exceeding the available battery capacity. The formula for the decision operator is as follows:

[0274] ;

[0275] The global iterative evolution is performed through selection, crossover, and mutation operators until the convergence termination condition is met, and the globally optimal inspection track sequence is output.

[0276] Constructing a population selection operator based on the roulette wheel mechanism: for individuals encoding any path in the current generation of the genetic population. Based on the comprehensive fitness function value The discrete selection probability of it being inherited by the next generation of temporary population is calculated as follows:

[0277] ;

[0278] In the formula, Indicates the number of the current genetic population. The probability value of the selected track evolution adaptation of each candidate path encoding individual; This represents the preset global population size of the genetic algorithm; the probability value of adaptive selection is used based on the trajectory evolution. Construct a roulette wheel evolutionary selection operator by generating a... Uniformly distributed random numbers within the interval This allows for the identification of candidate path-encoding individuals for the next generation of iterative evolution. The roulette wheel evolution selection operator satisfies the following mapping equation:

[0279] ;

[0280] By repeatedly executing the above roulette evolution selection operator mapping equation The selection of the track population was completed in one step, and an intermediate evolutionary population was constructed.

[0281] Specifically, the intermediate evolutionary population is subjected to global cyclic evolution by introducing scene-adaptive crossover and adaptive mutation operators. The specific techniques include:

[0282] Two candidate path coding individuals from the intermediate evolutionary population are randomly selected, and the order of some of their target points to be inspected is swapped to achieve track topology crossover; at the same time, the order of visits to inspection points in a single candidate path coding individual is randomly changed to achieve track topology variation.

[0283] After each crossover and mutation operation, the newly generated candidate path coding individual is fed back to steps 2) and 3) in real time. The total length of the continuous smooth track mapped by the candidate path coding individual that meets the dynamic constraints is recalculated, and the dynamic comprehensive energy consumption cost and time window penalty corrected by the environmental wind field are dynamically updated until the preset convergence termination condition is met, and the globally optimal inspection track sequence is output.

[0284] Finally, the expected flight heading angles corresponding to each segmented infinitesimal trajectory in the globally optimal inspection trajectory sequence are extracted. and composite airspeed , as a dynamic control command for tracking the flight path of unmanned aerial vehicles (UAVs) on the sea.

[0285] This invention also provides a path planning system for a seaplane unmanned aerial vehicle (UAV) that integrates Dubins trajectory curves and river wind energy consumption correction, comprising:

[0286] The data acquisition unit is used to acquire the original spatial geographic coordinates and time window limit indicators of the target point to be inspected.

[0287] The first building unit is used to construct the Dubins trajectory state space based on physical turning radius constraints, and introduces a Dubins trajectory state index table based on pose six-tuples to store high-concurrency trajectory data;

[0288] The second building unit is used to construct a dynamic nonlinear power consumption penalty model that integrates real-time wind field environmental vectors along the river, and to calculate the comprehensive energy consumption cost of each segment of the flight path.

[0289] The instruction generation unit is used to perform global collaborative solution of hybrid genetic-ant colony path based on time window urgency adaptation, and output the globally optimal inspection trajectory sequence and control instructions.

[0290] This invention has many specific applications. The above description is only a preferred embodiment of this invention. It should be noted that for those skilled in the art, several improvements can be made without departing from the principle of this invention, and these improvements should also be considered within the scope of protection of this invention.

Claims

1. A path planning method for a surface-to-water unmanned aerial vehicle (UAV) that integrates Dubins trajectory curves and river wind energy consumption correction, comprising the following steps: 1) Obtain the original spatial geographic coordinates and time window constraints of the target point to be inspected; 2) Construct a Dubins trajectory state space based on physical turning radius constraints, and introduce a Dubins trajectory state index table based on pose six-tuples to store high-concurrency trajectory data; 3) Construct a dynamic nonlinear power consumption penalty model that integrates real-time wind field environmental vectors along the river, and calculate the comprehensive energy consumption value of each segment of the trajectory; 4) Perform a global collaborative solution of hybrid genetic-ant colony path based on time window urgency adaptation, and output the globally optimal inspection trajectory sequence and control commands.

2. The method for path planning of a seaplane unmanned aerial vehicle (UAV) that integrates Dubins trajectory curves and river wind energy consumption correction as described in claim 1, is characterized in that... Step 1) specifically includes: The riverside maritime support bases were selected as the origin for a two-dimensional, kilometer-scale projection, and the distribution along the river was acquired through a maritime sensor network. The original spatial geographic latitude and longitude coordinate sequence of each target point to be inspected is obtained, and the upper limit of the soft time window deadline for each target point is used as the time window limit indicator. Based on the parametric geographic projection transformation algorithm, the latitude and longitude coordinates of each target point to be inspected are converted into plane rectangular kilometer coordinates as the original spatial geographic coordinates; Based on the physical parameters of the UAV, benchmark performance constants are established, including: the UAV's cruise speed, maximum available battery capacity, benchmark rated output power in a standard windless environment, and minimum turning radius in aeronautical physical conditions.

3. The method for path planning of a seaplane unmanned aerial vehicle (UAV) that integrates Dubins trajectory curves and river wind energy consumption correction according to claim 2, is characterized in that... Step 2) specifically includes: The length of a continuous and smooth trajectory that meets dynamic constraints is calculated based on the initial pose and target pose of the UAV. The specific steps are as follows: Establish from the starting point Using a relative rectangular coordinate system as the reference, the radius of the turning center circle is set as the minimum turning radius, and the coordinates of the left turning center and the right turning center are calculated respectively under the starting point pose; Calculate the coordinates of the left-turn center and the right-turn center under the target point pose; Calculate the coordinate matrix of the tangent points of the inner and outer tangents of the two circles for each of the six topological patterns, the length of the straight line segment, the variable central angle corresponding to the first and second turning arcs; details are as follows: Based on the specific topological branch of the current traversal, extract the corresponding starting center and target center. Specifically, under the LSL and LSR branches, the starting center is mapped to the left turn center; under the RSR and RSL branches, the starting center is mapped to the right turn center; under the LSL and RSL branches, the target center is mapped to the left turn center; under the RSR and LSR branches, the target center is mapped to the right turn center. Here, LSL represents left turn-straight-left turn, RSR represents right turn-straight-right turn, LSR represents left turn-straight-right turn, RSL represents right turn-straight-left turn, RLR represents right turn-left turn-right turn, and LRL represents left turn-right turn-left turn. 21) Calculate the turning-straight-turn trajectory pattern; 22) Calculate the turn-turn-turn trajectory pattern; Construct a unified trajectory operator that integrates topology switching indicator factors and calculates each topology mode branch. Total dynamic track length ; The minimum value of the integrated energy consumption calculation result corresponding to the six topology mode branches is extracted and used as the final Dubins track length under the connection of the starting point pose and the target point pose. The final Dubins track length is used as the core optimal metric representation value in the Dubins track state space, thereby completing the construction and energy consumption calculation of the continuous smooth track that meets the dynamic constraints. A Dubins track state index table based on pose six-tuples is constructed. The Dubins track state index table adopts a memory hash-level high-concurrency key-value pair storage architecture. Its specific technical implementation and indexing mechanism are as follows: For the starting point pose and the target point pose in any iterative traversal, the kilometer rectangular coordinates and radian heading angles of the starting point pose and the target point pose are respectively normalized and truncated to retain preset valid values, and then normalized and combined into a unique six-tuple hash cache key. The target storage value corresponding to the six-ary hash cache key is defined as: the final Dubins track topology pattern code after geometric branch filtering, the upper limit boundary of integration for each turn / straight segment, and the final river wind correction comprehensive energy consumption value corresponding to the current wind field environment; during high-concurrency path retrieval scheduling, if the corresponding key is accurately hit in the memory retrieval unit through the six-ary hash cache key, the geometric analysis and integration calculation process is directly bypassed, and the pre-stored state vector is directly output through the dynamic pointer. If no match is found, the geometric tangent point transformation unit is activated to perform a full-process solution, and the newly generated key-value pairs are dynamically written into the index table. For any starting point pose and target point pose in any iterative traversal, the Cartesian kilometer coordinates and radian heading angles of the starting point pose and target point pose are respectively normalized and truncated to retain preset valid values, and then normalized and combined into a unique six-element hash cache key. ; During path retrieval scheduling, if an existing key matching the six-tuple hash cache key of the current iteration input is detected in the Dubins track state index table, a cache hit is determined, and the pre-stored five-tuple state vector corresponding to the key is directly output via memory read / write instructions; otherwise, if a matching key is not successfully retrieved in the Dubins track state index table, a cache miss is determined, and Dubins geometric analytical solution is performed. A complete geometric tangent point analytical solution process is executed for the current starting point pose and the target point pose, and the calculation results are stored in the Dubins track state index table and a corresponding mapping relationship is established.

4. The method for path planning of a seaplane unmanned aerial vehicle (UAV) that integrates Dubins trajectory curves and river wind energy consumption correction according to claim 3, is characterized in that... Step 21) specifically includes: Calculate the spatial geometric distance between the starting center and the target center, as well as the azimuth angle of the line connecting the centers; External tangent mode: The two circles have the same physical turning radius, and their common tangent segment is parallel to the line connecting the centers of the circles. The geometric length of the common tangent segment is equal to the spatial geometric distance between the centers of the two circles. It represents the geometric length of the common tangent segment between the two circles in the turning-straight-turn trajectory mode, and it serves as the upper limit variable of the one-dimensional integral of the energy consumption cost of the straight segment. For the LSL topology pattern branch, the angle of deviation of the external common tangent relative to the normal of the line connecting the centers is... The planar kilometer coordinates of the first and second tangent points are calculated. For the RSR topology pattern branch, the angle of deviation of the external common tangent relative to the normal of the line connecting the centers is... The corresponding planar kilometer coordinates of the first and second tangent points are calculated respectively; Internal tangent pattern: There is a fixed geometric angle between the line connecting the centers of the two circles and the internal tangent; For the LSR topology pattern branch, calculate the Cartesian kilometer coordinates of the first and second tangent points corresponding to the LSR topology pattern branch; For the RSL topology pattern branch, calculate the Cartesian kilometer coordinates of the first and second tangent points corresponding to the RSL topology pattern branch; Calculation of the variable central angle in the turn-straight-turn trajectory mode: Based on the kilometer coordinates of the tangent point plane, the instantaneous heading velocity vector angles at the two tangent point positions of the UAV are extracted. By performing modulo difference calculations on the instantaneous heading velocity vector angles and the initial heading angle and target heading angle respectively, the variable central angle radian value corresponding to the first segment of the turn arc and the variable central angle radian value corresponding to the second segment of the turn arc under the current turn-straight-turn trajectory mode branch are calculated.

5. The method for path planning of a seaplane unmanned aerial vehicle (UAV) that integrates Dubins trajectory curves and river wind energy consumption correction according to claim 4, is characterized in that... Step 22) specifically includes: In the turn-turn-turn trajectory mode, the transition is a smooth full circle, without straight segments, and the geometric length of the common tangent segment is 0; by introducing a radius of... Furthermore, the analysis is performed using the intermediate reverse transition center, which is externally tangent to both the initial and target centers. The minimum turning radius in aeronautical physics; The three centers form a circle in space with a length of [missing information]. Given an isosceles triangle, calculate the isosceles base angles of the isosceles triangle; For the RLR topology mode branch, the starting circle center corresponds to the right turn circle center, the target circle center corresponds to the right turn circle center, and the intermediate transition circle center is represented as the left turn circle center. The planar kilometer coordinates of the intermediate transition circle center are calculated. The relationship between adjacent circular arcs is purely geometric and externally tangent. The plane kilometer coordinates of the first and second tangent points are constructed and solved using the midpoint equation. For the LRL topology mode branch, the starting circle center corresponds to the left turn circle center, the target circle center corresponds to the left turn circle center, and the intermediate transition circle center is represented as the right turn circle center. The planar kilometer coordinates of the intermediate transition circle center are calculated. Similarly, using the equation of the midpoint of the geometric tangency point, we can construct and solve the planar kilometer coordinates of the first and second tangency points; Calculation of variable central angle in turn-turn-turn trajectory mode: Calculate the geocentric turning angle swept from the starting point pose to the first tangent point position and from the second tangent point position to the target point pose under the turning radius constraint. Output the variable central angle radian value corresponding to the first turning arc and the second turning arc. Based on the symmetry of the vertex angle of the isosceles triangle, calculate the variable central angle radian value of the intermediate reverse transition arc segment, which serves as the upper limit variable of the one-dimensional integral of the energy consumption cost of the second reverse turn.

6. The method for path planning of a seaplane unmanned aerial vehicle (UAV) that integrates Dubins trajectory curves and river wind energy consumption correction according to claim 5, is characterized in that... Step 3) specifically includes: 31) The environmental wind resistance vector is obtained in real time by meteorological sensors deployed along the river for inspection, and the wind speed and wind direction angle are extracted. 32) Extract the instantaneous flight heading angle of the UAV at the tangent of the straight segment of the Dubins track or the integral element of the circular arc, and obtain the real-time environmental wind direction angle and environmental wind speed sent by the multi-source sensor network deployed around the navigation channel along the river; calculate the spatial relative angle between the instantaneous flight heading angle and the environmental wind direction angle. 33) Construct a dynamic nonlinear power consumption penalty model that integrates real-time wind field environmental vectors along the river to quantitatively calculate the additional power loss of the amphibious UAV due to nonlinear changes in air resistance under non-uniform wind field conditions; the details are as follows: Based on the spatial relative angle in step 32), and combined with the principle of wind vector synthesis in aerodynamics, the environmental wind speed is decoupled and projected onto the current movement direction of the UAV, and the axial wind speed components and lateral wind speed components experienced by the UAV at the current integral element are calculated respectively. By combining the preset cruise relative speed of the waterborne UAV in calm wind conditions, the dynamic nonlinear actual air speed after overcoming the air resistance of the wind field along the river is synthesized. Based on the nonlinear physical law in fluid mechanics that fluid resistance is proportional to the square of relative velocity, an analytical equation for the nonlinear river wind energy consumption correction coefficient is constructed, which serves as the core variable integrand for the subsequent one-dimensional integral of energy consumption cost. 34) Based on the current optimal topology mode branch selected by the Dubins track state index table constructed in step 2) and the integral upper limit boundary corresponding to each segment, along the geometric trajectory composed of the tangent of the straight line segment or the arc segment of the Dubins track, establish a piecewise one-dimensional integral equation for the integral element in step 33) to calculate the final comprehensive energy consumption value of the river wind correction under the pose connection between the starting point and the target point.

7. The method for path planning of a seaplane unmanned aerial vehicle (UAV) that integrates Dubins trajectory curves and river wind energy consumption correction according to claim 6, is characterized in that... Step 4) specifically includes: 41) Calculation of urgency within a dynamic time window; 42) Construction of initial flight path sequences based on hybrid ant colonies; 43) Spatiotemporal state recursive update 44) Injection of genes from high-quality populations; 45) Post-genetic algorithm deep optimization: global cyclic evolution is performed through selection, crossover and mutation operators until the convergence termination condition is met, and the globally optimal inspection trajectory sequence is output; the expected flight heading angle and synthetic airspeed corresponding to each segment of the global optimal inspection trajectory sequence are extracted as dynamic control commands for tracking the trajectory of the UAV.

8. A path planning system for a surface-to-water unmanned aerial vehicle (UAV) that integrates Dubins trajectory curves and river wind energy consumption correction, characterized in that, include: The data acquisition unit is used to acquire the original spatial geographic coordinates and time window limit indicators of the target point to be inspected. The first building unit is used to construct the Dubins trajectory state space based on physical turning radius constraints, and introduces a Dubins trajectory state index table based on pose six-tuples to store high-concurrency trajectory data; The second building unit is used to construct a dynamic nonlinear power consumption penalty model that integrates real-time wind field environmental vectors along the river, and to calculate the comprehensive energy consumption cost of each segment of the flight path. The instruction generation unit is used to perform global collaborative solution of hybrid genetic-ant colony path based on time window urgency adaptation, and output the globally optimal inspection trajectory sequence and control instructions.

Images