A cooperative search method for surface targets by unmanned aerial vehicles and ships

Abstract

This invention belongs to the field of unmanned aerial vehicle (UAV) and surface vessel (SAV) cooperative search, specifically disclosing a method for cooperative search of surface targets by UAVs and SAVs, comprising: Step 1, constructing a multi-dimensional dynamic environment model, including dynamic environment modeling based on dynamic environment parameters; Step 2, constructing a heterogeneous platform kinematic model based on the multi-dimensional dynamic environment model constructed in Step 1; Step 3, constructing a distributed cooperative task allocation strategy based on the heterogeneous platform kinematic model constructed in Step 2; Step 4, predicting the target trajectory using an extended Kalman filter algorithm to obtain an optimal state estimate; Step 5, based on the optimal state estimate obtained in Step 4, performing collision-free path planning using the TangentBug real-time obstacle avoidance algorithm. The solution provided by this invention can solve the problems of existing technologies in adapting to dynamic environments, accurately predicting target trajectories, robustly allocating tasks, and safely avoiding obstacles in complex marine scenarios.

Description

Technical Field

[0001] This invention belongs to the field of unmanned aerial vehicle (UAV) ship cooperative search technology, and specifically relates to a method for cooperative search of UAV ship surface targets. Background Technology

[0002] With the increasing prominence of maritime strategic importance, utilizing unmanned and intelligent equipment to perform tasks such as maritime surveillance, target search, and patrol reconnaissance has become an important development direction for major maritime powers worldwide. Unmanned surface vessels (USVs) possess advantages such as long endurance, high payload capacity, and good stealth, making them suitable for prolonged close-range reconnaissance and response on the water surface. Unmanned aerial vehicles (UAVs), on the other hand, offer advantages such as high speed, wide range, and rapid response. Combining the high-altitude, wide-area reconnaissance capabilities of UAVs with the continuous surface tracking and response capabilities of UAVs to construct an integrated "air-sea" heterogeneous unmanned system can significantly improve the efficiency and robustness of maritime missions, and has broad application prospects.

[0003] In recent years, collaborative operations between UAVs and unmanned surface vessels (USVs) have become a research hotspot, with domestic and international scholars making significant progress in key technologies such as collaborative perception, task allocation, path planning, and collaborative control. In collaborative perception, algorithms such as adaptive weighted fusion have improved the detection and identification capabilities of weak targets on the water surface. Cross-domain collaboration and information fusion rely on core technologies such as tactical data links and Co-operation Programming (COP) to achieve information sharing and data fusion, generating real-time battlefield situation maps. Collaborative task allocation and decision-making utilize heuristic optimization and deep reinforcement learning to efficiently solve complex problems and enhance system robustness. Collaborative path planning employs algorithms such as Advanced Platform Path Facility (APF) and the Bug series to solve obstacle avoidance and collision avoidance and trajectory planning problems across multiple platforms while satisfying COLREGS rules and motion constraints, ensuring safe collaboration of heterogeneous unmanned systems. Despite the progress made in UAV-US collaborative technology, the following challenges remain: first, accurate nonlinear state estimation of maneuvering targets is impossible; second, robust and adaptive task allocation cannot be achieved in complex dynamic environments; and third, the collaborative obstacle avoidance and path planning problems of UAVs and USVs under heterogeneous motion characteristics cannot be solved. Summary of the Invention

[0004] In view of the aforementioned shortcomings of existing UAV-ship cooperative operations, the purpose of this invention is to propose a cooperative search method for UAV-ship surface targets.

[0005] To achieve the aforementioned objectives, the technical solution adopted by this invention is: a cooperative search method for surface targets by unmanned aerial vehicles (UAVs) and surface vessels, comprising:

[0006] Step 1: Construct a multi-dimensional dynamic environment model, including dynamic environment modeling based on dynamic environment parameters;

[0007] Step 2: Based on the multi-dimensional dynamic environment model constructed in Step 1, construct a heterogeneous platform kinematic model, which includes a UAV kinematic model, an unmanned surface vessel kinematic model, and a target kinematic model.

[0008] Step 3: Based on the UAV kinematic model, UAV kinematic model, and target kinematic model constructed in Step 2, construct a distributed collaborative task allocation strategy. The distributed collaborative task allocation strategy includes UAV strategy, UAV strategy, and collaborative decision-making mechanism.

[0009] Step 4: Use the extended Kalman filter algorithm to predict the target trajectory and obtain the optimal state estimate;

[0010] Step 5: Based on the optimal state estimate obtained in Step 4, the TangentBug real-time obstacle avoidance algorithm is used for path planning.

[0011] Furthermore, dynamic environmental parameters include wind field parameters and water flow velocity parameters. Dynamic environmental modeling includes wind field modeling, water flow velocity modeling, and the introduction of a dual-mode refresh mechanism, a grid indexing mechanism, and collision constraints to refresh obstacle positions.

[0012] The formula for wind field modeling is:

[0013]

[0014] in, Let be the wind speed at time t. The attenuation coefficient is... For the disturbance intensity, Standard Gaussian noise;

[0015] The formula for the dual-mode refresh mechanism is:

[0016]

[0017] in, , This refers to the last update time. The time of random fluctuation;

[0018] The grid index formula is:

[0019]

[0020] Where L and W represent the length and width of the scene;

[0021] The collision constraints are:

[0022]

[0023]

[0024] in For drone collection, For unmanned surface vessels, parameters Area with obstacles , These represent the distances between the drone, the unmanned surface vessel, and the obstacle, respectively. This indicates the minimum safe distance between a drone or unmanned surface vessel and an obstacle.

[0025] Furthermore, constructing the kinematic model of the drone includes:

[0026] Design an irregular polygonal closed patrol trajectory composed of multiple nodes, and define the track point matrix P;

[0027] The segmented trajectory of the UAV uses a straight-line interpolation model for smooth transition. The i-th segment of the trajectory (from P) i (x i ,y i ) to P i+1 (x i+1 ,y i+1 The parametric equation is:

[0028]

[0029] for Location at any given moment For the first Coordinates of the waypoints for Coordinates of the first waypoint For the first Segment trajectory start time; t i+1 Let be the end time of the i-th trajectory segment; s(t) is the time normalization parameter, ranging from... When t is 0, it is located at the th The coordinates of the waypoint are given when t=1. The coordinates of the waypoints; H is the flight altitude of the UAV;

[0030] Considering wind interference, the formula for dynamically adjusting the drone speed is:

[0031]

[0032] in Represents the initial velocity of the i-th segment of the trajectory. This indicates the drone's own acceleration. The time interval between two adjacent waypoints. Wind speed;

[0033] The speed constraints and minimum turning radius constraints for the UAV are as follows:

[0034]

[0035] in, This is the instantaneous speed modulus of the drone, and its value is less than the maximum speed. , It is the drone's turning angular velocity (radians per second). This is the turning radius of the drone, and its value is greater than the minimum turning radius. ;

[0036] Constructing the kinematic model of the unmanned surface vessel includes: patrol speed during patrol. Keeping constant, the parametric equations of the piecewise trajectory are:

[0037]

[0038] in , The coordinates of the current base point. , For the coordinates of the next base point, For dimensionless time parameters, , For the water flow , The directional disturbance velocity component, The heading angle pointing towards the target base point;

[0039] When the unmanned surface vessel (USV) is less than 800m from the target, it will detect the target and switch to pursuit mode. In pursuit mode, the pursuit trajectory parameter equation is:

[0040]

[0041] in Current position For the current course, The difference (range) between the updated target direction and the current heading. );

[0042] The pursuit speed is:

[0043]

[0044] in For patrol speed, maximum speed , To catch up with the start time, accelerate time. ;

[0045] The formula for the turning angle of an unmanned surface vessel is:

[0046]

[0047] in ;

[0048] Constructing the target kinematic model includes defining the position update formula as follows:

[0049]

[0050] in, yes Within a time Displacement increment in direction, yes Within a time The displacement increment in the direction;

[0051] Design a speed control algorithm and define acceleration. With deceleration They are respectively:

[0052]

[0053] in, The acceleration percentage parameter value is 0.1. The deceleration percentage parameter value is 0.15;

[0054] Design a heading control algorithm; the heading control formula is:

[0055]

[0056] in, This is the difference between the initial target direction and the current heading. This is the difference between the updated target direction and the current heading;

[0057] Orientation update rules for design goals:

[0058]

[0059] in, It is the initial orientation of the target. It is the target's current orientation. It is the target's updated orientation;

[0060] A simple obstacle avoidance strategy based on boundary distance is adopted to achieve target obstacle avoidance.

[0061] Furthermore, the drone strategy includes: when a target appears in a specific direction and within a specific area of ​​the drone's speed, the drone switches to tracking mode, and its speed direction points towards the target's location. If multiple targets exist within the detection area, the drone's speed direction points towards the center point of the multiple targets. The formula for calculating the target center point is:

[0062]

[0063] in, The coordinates of the UAV target point, The number of targets detected. These are the x and y coordinates of the i-th target, respectively.

[0064] The formula for calculating the heading angle of the drone at this time is:

[0065]

[0066] in For the drone's heading angle, The current coordinates of the drone. The heading angle is the center point of the n detected targets;

[0067] The unmanned surface vessel (USV) strategy includes: when the USV detects or is dispatched to a target, it switches to pursuit mode, with the following turning angle constraints:

[0068]

[0069] in, This represents the change in the unmanned surface vessel's turning angle. The speed of the unmanned surface vessel;

[0070] When multiple boats are pursuing each other in coordination, the principle of distance priority and load balancing should be adopted to select one unmanned boat for pursuit.

[0071] The construction of a collaborative decision-making mechanism includes: constructing real-time state matrices for UAVs and unmanned surface vessels (USVs) respectively. The real-time state matrices for UAVs and USVs are as follows:

[0072]

[0073] in, For platform location, For platform speed, Let i represent the platform task status, i = 1, 2, 3, ...;

[0074] The parameters of the target are obtained hierarchically, and the target set is defined as follows:

[0075] ;

[0076] in, Let represent the target's number and its position coordinates at time t, respectively. These represent the target's velocity and heading angle at the current time t, respectively.

[0077] Furthermore, in addition to using the extended Kalman algorithm for trajectory prediction, the method also includes: establishing a hybrid motion model of the target based on the target's motion characteristics, wherein the hybrid motion model includes a linear motion model and a circular motion model;

[0078] Establishing a hybrid motion model of the target includes defining the target state vector:

[0079]

[0080] The linear motion model is as follows:

[0081]

[0082] in, The sampling time interval, for Always the goal is The position of the axis for The velocity component of the target at any given moment. for acceleration component, for Location at any given moment for The velocity component at any given moment;

[0083] The circular motion model is as follows:

[0084]

[0085] in, and These are the position increments of the circular arc motion in the x and y directions, respectively;

[0086] The extended Kalman filter algorithm includes initialization, prediction, and update phases;

[0087] Initialization includes setting the initial state vector and covariance matrix, and defining the process noise covariance matrix and observation matrix:

[0088] The process noise covariance matrix is:

[0089] ;

[0090] The observation matrix is:

[0091] ;

[0092] The prediction phase is used to determine the current state prediction and the corresponding uncertainty quantification based on the hybrid motion model and the optimal state estimate of the previous time step, including state prediction and covariance prediction.

[0093] The update phase is used to correct the state prediction values ​​obtained in the prediction phase based on the observation data at the current time, thereby obtaining the optimal state estimate and the corresponding uncertainty quantification, including Kalman gain calculation, state update and covariance update.

[0094] Furthermore, state prediction is used to derive the state prediction value at the current moment through the hybrid motion model, and the expression is:

[0095]

[0096] in, For the first The optimal state estimate after the update at each time step. The state transition matrix is ​​the objective function, used to describe the transition from state to state 1. Time to the The evolutionary relationship at any moment, Then it is the first The predicted state value at time;

[0097] Covariance prediction is used to quantify the uncertainty of state prediction values, and its expression is:

[0098]

[0099] In the formula, It is the first The covariance matrix of the optimal state estimate at time step 1. State transition matrix transpose, The process noise covariance matrix is... That is, the first The covariance matrix of the state prediction at time step.

[0100] Furthermore, the Kalman gain is used to determine the weighting of the predicted and observed results in the final state estimation, expressed as:

[0101]

[0102] in, For the observation noise matrix transpose, To observe the noise covariance matrix, That is, the first Kalman gain at time step;

[0103] The state update obtains the optimal state estimate for the current time step by combining observed data and Kalman gain to correct the state prediction value in the prediction stage. The expression is:

[0104]

[0105] In the formula, For the first The observed value at time, For the derived theoretical observation values, To observe the residuals, Then it is the first The optimal state estimate after correction at time step;

[0106] Covariance update is used to quantify the uncertainty of the optimal state estimate after the state update, and its expression is:

[0107]

[0108] in, It is the identity matrix. It is the first The covariance matrix of the optimal state estimate at time step.

[0109] Furthermore, step 5 includes: constructing a basic geometric model;

[0110] The path planning logic is set based on the constructed basic geometric model, and path planning is performed according to the path planning logic. The path planning logic includes straight line movement logic, line segment and circle intersection detection logic, tangent validity determination logic, boundary movement logic, and auxiliary determination logic.

[0111] Furthermore, constructing the basic geometric model includes:

[0112] Calculate location and distance:

[0113]

[0114] This is the current location of the unmanned surface vessel. For the target location, Let P and Q be the Euclidean distance between them.

[0115] Description of obstacles and collision condition determination: Describe obstacles as circles and use their center coordinates. and radius To describe;

[0116] The conditions for a collision to occur are:

[0117] in, For the unmanned surface vessel to reach the center of the obstacle The distance; The radius of the obstacle;

[0118] Calculate the current position of the unmanned surface vessel Tangent to the circumference of the obstacle: Calculate the direction angle from the current position P of the unmanned surface vessel to the center C of the obstacle. ;

[0119] Calculate the angle between the tangent and line segment PC based on geometric relationships. :

[0120]

[0121] Let r be the distance from the unmanned surface vessel to the center of the obstacle, and r be the radius of the obstacle.

[0122] The direction angles of the two tangents relative to line segment PC and They are respectively:

[0123]

[0124] Calculate the two tangent points and Coordinates; Point of tangency arrive The distance between the points is The coordinates are:

[0125]

[0126] Similarly, tangent point The coordinates are:

[0127] .

[0128] The aforementioned method for cooperative search of surface targets by unmanned aerial vehicles and surface vessels can achieve the following beneficial effects:

[0129] This technology addresses the challenges of adapting to dynamic environments, accurately predicting target trajectories, robustly allocating tasks, and safely avoiding obstacles in complex marine scenarios. Ultimately, it enables efficient and accurate searching of dynamic surface targets in unfamiliar sea areas, providing reliable technical solutions for practical scenarios such as marine environmental monitoring, maritime search and rescue, maritime patrol, and anti-smuggling.

[0130] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0131] Figure 1 This is a flowchart illustrating a method for collaborative search of surface targets by unmanned aerial vehicles (UAVs) according to the present invention. Detailed Implementation

[0132] The features and exemplary embodiments of various aspects of the present invention will now be described in detail. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention may be practiced without requiring some of these specific details. The following description of embodiments is merely intended to provide a better understanding of the invention by illustrating examples of the invention.

[0133] This invention provides a method for searching surface targets on unmanned aerial vehicles (UAVs) that integrates multi-dimensional dynamic environment modeling, precise kinematic control of heterogeneous unmanned platforms, hierarchical intelligent prediction, and distributed collaborative decision-making. By deeply combining the high-altitude wide-area detection advantages of UAVs with the continuous surface tracking capabilities of UAVs, it solves the problems of existing technologies in adapting to dynamic environments, accurately predicting target trajectories, robustly allocating tasks, and safely avoiding obstacles in complex marine scenarios. Ultimately, it achieves efficient and accurate searching for dynamic surface targets in unfamiliar sea areas, providing a reliable technical solution for practical scenarios such as marine environmental monitoring, maritime search and rescue, maritime patrol, and anti-smuggling. The specific solution is as follows:

[0134] Please see Figure 1 This embodiment of a method for cooperative search of surface targets by unmanned aerial vehicles and surface vessels includes:

[0135] Step 1: Construct a multi-dimensional dynamic environment model, which specifically includes dynamic environment modeling and obstacle refresh.

[0136] Specifically, the wind field parameters follow a Markov process, and the wind speed is dynamically updated through an attenuation coefficient and Gaussian noise, as shown in the formula:

[0137]

[0138] in The attenuation coefficient is... For the disturbance intensity, To simulate the irregular fluctuations of a real stroke, a perturbation that conforms to a normal distribution is randomly added each time the wind speed change is calculated, using standard Gaussian noise.

[0139] The flow parameters are based on the average flow velocity, with random superposition of velocity disturbances ranging from -0.3 to 0.3 m / s. The impact of radian flow direction shift on the unmanned surface vessel (USV) is calculated through velocity component synthesis. The specific calculation formula is as follows:

[0140] (1) Formula for random disturbance of water flow velocity

[0141]

[0142] in For the final water flow velocity, For average speed, The value represents the random velocity disturbance, and its range is... m / s

[0143] (2) Formula for random disturbance of water flow direction

[0144]

[0145] in The final direction of water flow. The average direction of water flow The value represents the directional random perturbation, with a range of [value missing]. rad.

[0146] To simulate the dynamic displacement characteristics of obstacles in a real ocean, a dual-mode refresh mechanism combining periodicity and randomness is adopted. The periodic refresh uses a base period of 1800 seconds, superimposed with random fluctuations ranging from 0 to 300 seconds, as shown in the formula:

[0147]

[0148] in , This is the last update time.

[0149] Simultaneously, using the grid index formula:

[0150]

[0151] L and W represent the length and width of the scene, ensuring that obstacles are evenly distributed throughout the scene.

[0152] To avoid obstacles overlapping with the unmanned platform, the collision constraint formula is set as follows:

[0153]

[0154]

[0155] in For drone collection, For unmanned surface vessels, parameters Area with obstacles , These represent the distances between the drone, the unmanned surface vessel, and the obstacle, respectively. This refers to the minimum safe distance between the equipment and the obstacle.

[0156] In addition, an active refresh function is designed. By manually clicking the O key, the system can immediately recalculate dynamic environmental parameters such as obstacle position, water flow speed, and wind direction and transmit them to the UAV boat decision system in real time to deal with sudden scenarios during equipment operation and avoid environmental information lag.

[0157] Step 2: Based on the multi-dimensional dynamic environment model constructed in Step 1, construct a heterogeneous platform kinematic model. The heterogeneous platform kinematic model includes the UAV kinematic model, the unmanned surface vessel kinematic model, and the target kinematic model.

[0158] Specifically, to match the characteristics of heterogeneous platforms with the complex marine environment, this embodiment constructs kinematic models of UAVs, unmanned surface vessels, and targets. By accurately depicting the motion laws and their coupling relationship with the environment, it provides reliable dynamic support for collaborative search.

[0159] In this embodiment, the construction of the UAV kinematic model includes:

[0160] Design an irregular polygonal closed patrol trajectory composed of multiple nodes. In this embodiment, taking a 5*5 nautical mile activity range as an example, the waypoint matrix is ​​defined as follows:

[0161] (1.4)

[0162] Two virtual base points are added at the midpoints of the top and bottom sides of the fixed trajectory square. These virtual base points move at 1.1 to 1.2 times the drone's cruising speed, moving independently along the auxiliary path and always positioned 500-800m ahead of the drone at key turning points. Leading guidance is used to correct turning timing. Segmented trajectories use a linear interpolation model to achieve smooth transitions. The i-th segment of the trajectory (from P...) i (x i ,y i ) to P i+1 (x i+1 ,y i+1 The parametric equation is:

[0163] (1.5)

[0164] in, for Location at any given moment For the first Coordinates of the waypoints for Coordinates of the first waypoint For the first Segment trajectory start time; t i+1 Let be the end time of the i-th trajectory segment; s(t) is the time normalization parameter, ranging from... When t is 0, it is located at the th The coordinates of the waypoints, when t=1, are The coordinates of the first waypoint; H is the flight altitude of the UAV.

[0165] Considering wind interference, the formula for dynamically adjusting the drone speed is:

[0166] (1.6)

[0167] in This represents the initial velocity (or velocity upon reaching the waypoint) of the i-th trajectory. This indicates the drone's own acceleration.

[0168] The time interval between two adjacent waypoints. The speed of environmental impact (wind speed).

[0169] The speed constraints and minimum turning radius constraints for the UAV are as follows:

[0170] (1.7)

[0171] in, This is the instantaneous speed modulus of the drone, and its value is less than the maximum speed. , It is the drone's turning angular velocity (radians per second). This is the turning radius of the drone, and its value is greater than the minimum turning radius. .

[0172] In constructing the kinematic model of the unmanned aerial vehicle (UAV), to accurately characterize its motion state and dynamic changes in a complex ocean wind field environment, a state vector is defined:

[0173] (1.8)

[0174] in Represents three-dimensional position. The horizontal velocity component is... The velocity is in the vertical direction. These are the two components of the heading angle.

[0175] At the same time, a state transition equation is constructed to characterize the state from time t. arrive Changes:

[0176]

[0177] in, It is a moment The system state vector, It is a moment The state transition matrix, It is a moment The control input vector, It is a moment The control input matrix, It is a moment The process noise vector, It is a moment The process noise matrix, and That is the moment The new state vector.

[0178] This state transition equation integrates three types of factors: self-motion, wind field disturbance, and control input. The state transition matrix is ​​as follows:

[0179]

[0180] in, For the simulation time step, Represented as an n-dimensional identity matrix, since position and velocity are three-dimensional and heading angle is two-dimensional, the first two rows are three-dimensional identity matrices and the last row is a two-dimensional identity matrix.

[0181] Substitute the state vector Equation (1.8) yields the UAV's position update at time k+1:

[0182] (1.11)

[0183] Drone speed update at time k+1:

[0184] (1.12)

[0185] The heading angle of the drone at time k+1 is updated as follows:

[0186] (1.13)

[0187] The wind field disturbance matrix can quantify the disturbance of the wind field on the horizontal position and velocity, which is consistent with the horizontal wind resistance characteristics of multi-rotor UAVs.

[0188] (1.14)

[0189] The control input matrix is ​​controlled by horizontal acceleration commands. , and heading angle control commands To achieve precise control of speed and direction, the control input matrix W is:

[0190] (1.15)

[0191] Once the drone detects a target, it switches from regular patrol mode to target tracking mode. At this time, the target point is updated in real time to the target's current location, and the speed is dynamically adjusted. and heading angle To achieve continuous tracking, the wind field interference matrix during the process By introducing random wind fluctuations, the control input matrix W adjusts the control commands according to the target position to ensure that the UAV can stably track the target under complex wind fields.

[0192] In this embodiment, the construction of the unmanned surface vessel kinematic model includes:

[0193] In this embodiment, the unmanned surface vessel adopts a rectangular closed patrol trajectory based on four fixed base points, and the base point coordinate matrix R is defined as:

[0194] (1.16)

[0195] Each base point moves in a clockwise direction in a cyclical manner, with adjacent base points switching in sequence. The unmanned surface vessel (USV) tracks the assigned base points to achieve formation patrols, thereby ensuring regular coverage of the target sea area.

[0196] Patrol speed in patrol mode Keep constant, specifically In two-dimensional planar motion, the unmanned surface vessel's own velocity component is: Superimposed water flow velocity components Then, the actual velocity vector is synthesized as follows:

[0197] (1.17)

[0198] In this state, the motion of the unmanned surface vessel between adjacent base points adopts a linear interpolation model, and its piecewise trajectory parametric equations are as follows:

[0199] (1.18)

[0200] in , The coordinates of the current base point. , For the coordinates of the next base point, A dimensionless time parameter describing the progress of motion (range: ), , For the water flow , The directional disturbance velocity component, The heading angle is the angle pointing towards the target base point.

[0201] When the unmanned surface vessel (USV) is less than 800 meters from the target, it will detect the target and switch to pursuit mode, with the target position dynamically updated.

[0202] In the initial stage of detection (detection time less than 10 seconds), the target's real-time position is:

[0203]

[0204] In the later stages of detection (detection duration greater than 10 seconds), the predicted interception point is:

[0205]

[0206] In pursuit mode, if an unmanned surface vessel (USV) detects a target that is not being pursued by other USVs, it will automatically initiate pursuit. The pursuit trajectory parameter equation is as follows:

[0207] (1.19)

[0208] in This is the current location of the unmanned surface vessel. This is the current heading of the unmanned surface vessel. The difference (range) between the updated target direction and the current heading of the unmanned surface vessel. ).

[0209] During pursuit, the speed gradually increases from patrol speed to maximum speed, using the following formula:

[0210] (1.20)

[0211] The maximum speed , To catch up with the start time, accelerate time. .

[0212] The turning angle of an unmanned surface vessel (USV) is limited by the minimum turning radius, as shown in the formula:

[0213]

[0214] in The coefficient 2 is a safety redundancy to prevent excessively fast turns.

[0215] To accurately characterize the dynamic changes in the motion state of the unmanned surface vessel, its state vector is defined as follows:

[0216]

[0217] in and The coordinates of the unmanned surface vessel's horizontal and vertical positions are given. The heading angle of the unmanned surface vessel; The speed of the unmanned surface vessel. For status indicators, Indicates patrol status. Indicates a pursuit state.

[0218] State transitions are achieved through state transition equations:

[0219]

[0220] The state transfer characteristics of the unmanned surface vessel's motion are described using a state transition matrix:

[0221]

[0222] Control input matrix Includes adjustments to integrated speed and steering, and interference items. The corrections for water flow disturbances and obstacle avoidance in the TangentBug algorithm are as follows:

[0223] Unmanned surface vessel control input matrix

[0224] The control input matrix is ​​a 5×2 dimensional matrix (matching state vector). (dimensions)

[0225] (1.24)

[0226] In the state transition equation of the unmanned surface vessel, the input... , For speed increments, This is the heading angle increment.

[0227] Unmanned surface vessel interference items

[0228] (1.25)

[0229] in, For water flow velocity, In terms of water flow direction, For time step. The heading angle deviation is caused by the TangentBug obstacle avoidance algorithm. The flow resistance coefficient is indicated by the status symbol. Unaffected by environmental factors (determined solely by pattern logic).

[0230] When the target is captured (collision radius ≤ 100m) or lost, the unmanned surface vessel's status is indicated. Automatically switch to 0, that is, switch from pursuit mode back to patrol mode.

[0231] In this embodiment, constructing the target kinematic model includes: firstly, kinematic calculations are performed, using a two-direction decomposition method for uniformly accelerated linear motion, assuming the target is in... The magnitude and direction of the velocity remain constant over time. Using trigonometric functions of motion, its position update formula is:

[0232] (1.26)

[0233] in, yes Within a time Displacement increment in direction, yes Within a time The displacement increment in the direction. Constructing the target kinematic model specifically includes:

[0234] Designing a speed control algorithm: The speed control algorithm design requires setting the target's acceleration and deceleration parameters to ensure smooth and natural acceleration, rapid deceleration in emergencies, and quick-response braking. The acceleration and deceleration formulas are as follows:

[0235] (1.27)

[0236] in, The acceleration percentage parameter value is 0.1; The deceleration percentage parameter is 0.15, and the braking intensity is 1.5 times the acceleration.

[0237] Establishing a heading control algorithm: In terms of heading algorithm control, the first step is to calculate and normalize the angle difference. The heading difference is normalized to ensure it falls within the range of... Within the interval, to ensure that the shortest rotation path is always chosen when turning, the specific formula is as follows:

[0238] (1.28)

[0239] in, The difference between the initial target direction and the current heading. This is the difference between the updated target direction and the current heading of the unmanned surface vessel.

[0240] A smooth transition algorithm is established: This algorithm handles target orientation updates, enabling bounded orientation control. If the difference between the current orientation and the target orientation is less than or equal to the maximum possible turning angle within a time step, it is directly set as the target orientation; otherwise, the target orientation is rotated at the maximum turning rate. The maximum turning angle within the time step is... ,in It is the maximum steering angular velocity. The time step and orientation update rule are as follows:

[0241] (1.29)

[0242] in, It is the direction of the target. It is the current orientation. This is the updated orientation.

[0243] A state transition algorithm is established: To avoid repetitive mechanical behavior and simulate the uncertainty of real-world decision-making, a random duration is generated for each state using a uniform distribution. This duration is a uniformly distributed real number randomly selected between a given minimum and maximum value, as shown in the formula:

[0244] (1.30)

[0245] in It is the duration of the target's state. This represents the minimum duration of the state. This represents the maximum duration of the state. To conform to the interval A uniformly distributed random number, which can be used to... and Random numbers are generated between these values ​​to simulate the uncertainty of real-world decision-making.

[0246] Obstacle avoidance for achieving the target: Since there are boundaries to the motion, a simple reflection obstacle avoidance strategy based on boundary distance is adopted to avoid boundary collisions and prevent the target from entering obstacles or leaving the scene area.

[0247] The target will determine whether the next position is valid. When invalidity is detected, it indicates that the target is near the boundary of an obstacle or scene. At this time, obstacle avoidance mode is triggered, and the speed decreases by 20%.

[0248] Let the current position of the target be The next position is The boundary is The obstacle is Then when At that time, among them The radius of the equivalent circumcircle of the obstacle is given, and the target switches to obstacle avoidance mode.

[0249] The nearest boundary is determined by boundary distance calculation, and trigonometric functions are used to constrain it based on the relative relationship between the current direction of movement and the risk source. Determine the escape direction and reduce speed by linear scaling to minimize the risk of collision.

[0250] The specific calculation steps are as follows:

[0251] (1) Nearest boundary determination:

[0252] Define the boundary range of the rectangle: , The target's current location is

[0253] (1.31)

[0254] in, / The boundaries are respectively at Minimum and maximum values ​​in the y-axis direction. / These represent the vertical distances from the target to the left / right and top / bottom boundaries, respectively. The minimum distance from the target to the four boundaries.

[0255] (2) Calculate the escape direction after constraints

[0256] (1.32)

[0257] in, The central coordinates of the risk source The directional angle of the target pointing towards the source of risk. The initial escape direction angle, Let the target's current direction of motion be the angle. The maximum allowable steering angle for a single movement of the target. This is a truncation function that restricts the input value to a certain range. Within the interval, This is the final escape direction angle after the steering angle constraint.

[0258] (3) Perform linear scaling to reduce deceleration

[0259] (1.33)

[0260] in, For the target's safe speed in obstacle avoidance mode, The maximum permissible speed of movement for the target. The real-time distance from the target to the risk source. The equivalent circumcircle radius of the obstacle Preset safe distance threshold.

[0261] Boundary conditions: hour (Stop moving); hour , restore maximum speed.

[0262] Step 3: Based on the UAV kinematic model, UAV kinematic model, and target kinematic model constructed in Step 2, construct a distributed collaborative task allocation strategy. The distributed collaborative task allocation strategy includes UAV strategy, UAV strategy, and collaborative decision-making mechanism.

[0263] Specifically, the core of the distributed collaborative task allocation strategy in this embodiment revolves around the collaborative operation of UAVs and unmanned surface vessels. UAVs, with their high-altitude perspective and rapid mobility, undertake tasks such as wide-area detection, information relay, and area patrol, thereby achieving wide-area target search and multi-platform information interaction. Unmanned surface vessels, relying on their water surface endurance and precise control characteristics, undertake tasks such as close-range tracking and intelligent obstacle avoidance, thereby completing continuous target tracking and obstacle avoidance operations in complex aquatic environments.

[0264] In this embodiment, the UAV achieves target detection and area control through the logic of "target discovery - tracking / patrol switching". The specific UAV strategy includes:

[0265] Target detection and tracking trigger: When a target appears within a semi-circular area with a radius of 3 kilometers and a 60-degree angle symmetrical to the drone's velocity direction, the drone switches to tracking mode, and its velocity direction points towards the target's location. If multiple targets exist within the detection area, the drone's velocity direction points towards the center point of all targets. The formula for calculating the target center point is:

[0266] (1.34)

[0267] in, The coordinates of the UAV target point, The number of targets detected. These are the x and y coordinates of the i-th target, respectively.

[0268] The formula for calculating the heading angle of the drone at this time is:

[0269] (1.35)

[0270] in, For the drone's heading angle, The current coordinates of the drone. The number of detected targets is determined by the UAV's heading angle pointing towards the center point of n targets.

[0271] In addition, when the target is false, the drone will detect the target for 10 seconds and then the target will disappear. If there are no other targets in the drone's detection area at this time, the drone will return to patrol mode. If all targets within the drone's detection range are within the detection range of the unmanned surface vessel or other drones, the drone will also return to patrol mode.

[0272] In this embodiment, the unmanned surface vessel (USV) uses "target detection / deployment - pursuit state switching - intelligent allocation" as its core logic to achieve accurate target tracking and task load balancing. The specific USV strategy includes:

[0273] 1. Pursuit Mode Switching: When the unmanned surface vessel (USV) detects or is dispatched to a target, it switches to pursuit mode. The USV's turning angle constraint is based on the minimum turning radius. (Unit: meter), must meet the following requirements:

[0274]

[0275] in, This represents the change in the unmanned surface vessel's turning angle. The speed of the unmanned surface vessel.

[0276] 2. Division of labor in multi-boat coordinated pursuit

[0277] Let the set of unmanned surface vessels participating in the collaboration be . The allocation rule adopts "distance priority + load balancing":

[0278] Distance priority: Assign the target to the unmanned surface vessel closest to it, i.e. ,in Let v be the distance between the unmanned surface vessel v and the target t.

[0279] Load balancing: If multiple targets are detected at the same time, calculate the distance of each target to the four unmanned surface vessels (USVs) and sort them. If the distance of a USV to all targets is in the bottom 50% of the sorted distances, then the USV will not be assigned a target to pursue and will continue patrolling to avoid excessive aggregation of USVs at one end of the scene.

[0280] In this embodiment, the collaborative decision-making mechanism achieves global optimization for multi-platform collaboration through global state management, hierarchical acquisition of target parameters, and dynamic allocation rules. Specifically, establishing the collaborative decision-making mechanism includes:

[0281] First, a real-time state matrix is ​​constructed for the unmanned aerial vehicles (UAVs) and unmanned surface vessels (USVs) to dynamically monitor the position, speed, and mission status of each platform. The matrix form is as follows:

[0282] (1.36)

[0283] in, For platform location, For platform speed, In this embodiment, the number of drones is 2, so the real-time status matrix of the drones has two rows. The number of unmanned surface vessels (USVs) is 4, so the real-time status matrix of the USVs has 4 rows. Of course, in other embodiments, the number of drones and USVs can be selected according to the actual situation.

[0284] The target parameters are obtained hierarchically, and the target set T is defined as follows:

[0285]

[0286] in, These represent the target's identity (number) and location information (location coordinates at time t), respectively. These represent the target's velocity and heading angle at the current time t, respectively; only position information is acquired within the first 10 seconds after the target is detected; if the target is confirmed to be real after 10 seconds, its velocity is then acquired. Heading angle And other basic information.

[0287] Finally, following the dynamic allocation rules, if the target information is updated, the allocation of the target to the unmanned surface vessel will also be updated synchronously. Through the logic of "distance priority + load balancing", the target allocation is always kept in the global optimal state.

[0288] Step 4: Use the extended Kalman filter algorithm to predict the target trajectory and obtain the optimal state estimate.

[0289] Specifically, in unmanned aerial vehicle (UAV) and ship-assisted target pursuit missions, targets are typically characterized by high maneuverability and complex movement patterns. Traditional straight-line prediction methods struggle to accurately estimate the target's future position, leading to low pursuit efficiency. This system employs the EKF algorithm to predict the target trajectory.

[0290] The Extended Kalman Filter (EKF) is a state estimation algorithm suitable for nonlinear systems. By locally linearizing the nonlinear model, it enables real-time prediction and updating of the state of dynamic systems, and is particularly suitable for dynamic estimation of target trajectories.

[0291] Specifically, the core idea of ​​the Extended Kalman Filter (EKF) algorithm is to achieve state estimation through an iterative "prediction update" process: predicting the state at the next moment based on the system's motion model, and then correcting the prediction results by combining observation data, thereby reducing noise interference and improving the accuracy of state estimation. For target motion systems with nonlinear characteristics, EKF approximates the nonlinear model as a linear model through a first-order Taylor expansion, allowing it to be computed using the framework of the Kalman filter.

[0292] In this embodiment, establishing the target motion model includes:

[0293] To describe the motion state of the target, the state vector is defined as:

[0294]

[0295] in, The target's position coordinates, For velocity components, For acceleration components, For heading angle, ω is the angular velocity.

[0296] Based on the motion characteristics of maritime targets, a hybrid motion model supporting linear and circular motion is established;

[0297] The linear motion model is (when hour):

[0298] (1.37)

[0299] in, This represents the sampling time interval. for Always the goal is The position of the axis for The velocity component of the target at any given moment. for acceleration component, for The position of the next moment. for The velocity component at time; where the velocity update is determined by the product of acceleration and time, while acceleration, heading angle and angular velocity remain constant in linear motion.

[0300] The state transition equation corresponding to the linear motion of the target is:

[0301]

[0302] The corresponding circular motion model can be obtained (when...) hour):

[0303] (1.38)

[0304] in, and These represent the position increments in the x and y directions of the circular motion, respectively, and are related to the angular velocity, velocity components, and time interval. The heading angle update is determined by the product of the angular velocity and time, and the final position is obtained by superimposing the initial position and the position increments.

[0305] The state transition equation corresponding to the target's circular motion is:

[0306]

[0307] The implementation steps of the extended Kalman filter algorithm are as follows:

[0308] EKF follows a closed-loop process of "initialization-prediction-update". By iteratively executing this process, it achieves continuous optimization estimation of the target state.

[0309] 1. Initialization

[0310] Set the initial state vector and the initial covariance matrix Define the process noise covariance Q and the observation matrix.

[0311] The process noise covariance matrix is:

[0312]

[0313] The observation matrix is:

[0314]

[0315] The observation matrix extracts directly measurable position components from the multi-dimensional state of the system through linear mapping, transforming the high-dimensional predicted state into predicted observations of the same dimension as the measured values. This enables effective comparison between model predictions and sensor measurements, providing a prerequisite for subsequent state correction.

[0316] 2. Forecasting Phase

[0317] The prediction phase is the fundamental step in the extended Kalman filter (EKF) to achieve state estimation. Its core is to predict the state at the current moment and quantify the corresponding uncertainty based on the system motion model and the optimal state estimate at the previous moment. Specifically, it includes two key steps: state prediction and covariance prediction.

[0318] State prediction is the deduction of the state prediction value at the current moment through a motion model, and its mathematical expression is:

[0319]

[0320] in, Indicates the first The optimal state estimate after the update at each time step. The state transition matrix is ​​as described above. , , used to describe the state from the th Time to the The evolutionary relationship at any moment, Then it is the first The predicted state value at time.

[0321] Covariance prediction is used to quantify the uncertainty of state prediction values, and its calculation formula is as follows:

[0322]

[0323] In the formula, It is the first The covariance matrix of the optimal state estimate at time step 1. State transition matrix transpose, Let be the process noise covariance matrix, representing the inherent error of the motion model itself. That is, the first Covariance moments of state prediction at time step.

[0324] In summary, by calculating the above two formulas, the prediction from the historical state to the current state is completed, and the uncertainty of this prediction process is simultaneously quantified, thus completing the initial input for the extended Kalman filter iterative optimization.

[0325] 3. Update Phase

[0326] The update phase is a key step in the extended Kalman filter (EKF) to achieve state estimation. Its core is to correct the state prediction value obtained in the prediction phase based on the observation data at the current time to obtain a better state estimate and corresponding uncertainty quantification. Specifically, it includes three key steps: Kalman gain calculation, state update and covariance update.

[0327] Kalman gain calculation is used to determine the weighting of "predicted results" and "observed results" in the final state estimation, and its mathematical expression is as follows:

[0328]

[0329] in, Observation matrix transpose, To observe the noise covariance matrix, which is used to quantify the deviation between the sensor's measured values ​​and the physical true values. That is, the first Kalman gain at time step.

[0330] The state update, by combining observed data and Kalman gain, corrects the state prediction value during the prediction phase to obtain the optimal state estimate for the current time step. The calculation formula is as follows:

[0331]

[0332] In the formula, For the first The observed value at time, For the derived theoretical observation values, To observe the residuals, Then it is the first The optimal state estimate after correction at time step.

[0333] Covariance update is used to quantify the uncertainty of the optimal state estimate after the state update, and its mathematical expression is:

[0334]

[0335] in, It is the identity matrix. It is the first The covariance matrix of the optimal state estimate at time step 1 is numerically compared to... This usually decreases, reflecting the role of observational data in improving the accuracy of state estimation.

[0336] In summary, the update phase obtains the optimal state estimate and corresponding uncertainty quantification at the current moment through the calculation of the above three formulas, thus completing one iterative optimization process of the extended Kalman filter.

[0337] This embodiment, based on the extended Kalman filter algorithm, performs three-layer processing (long-range, medium-range, and short-range) according to the distance between the unmanned surface vessel and the target, as detailed below:

[0338] At long distances, the pursuit time is long due to the large distance between the unmanned surface vessel (USV) and the target. Therefore, target trajectory prediction algorithms based on extended Kalman filters will prolong the prediction time. To shorten the pursuit time, we predict the target's distance and position at a longer time before pursuing it. Furthermore, at greater distances, the prediction yields higher returns, requiring only a rough prediction of the approximate location and distance.

[0339] At medium range, the distance between the unmanned surface vessel (USV) and the target decreases significantly, and the pursuit time is relatively shortened. Therefore, a balance must be struck between efficiency and accuracy. To address this, an extended Kalman filter algorithm is used to predict the target's distance over a relatively equal time frame, thus enabling pursuit.

[0340] In close-range scenarios, the unmanned surface vessel (USV) is already close to the target's core detection range. At this point, "tracking accuracy" becomes the core requirement. Therefore, it is necessary to shorten the predicted target movement time and distance to achieve real-time and accurate target locking. Specifically, trajectory prediction will adopt a "high-frequency, short-cycle" update mode, relying on the extended Kalman filter algorithm to quickly predict the target's position in the near future (e.g., within a few seconds) until the target handling conditions are met (e.g., distance ≤ 100m collision radius). This will enable accurate tracking and handling of the target, effectively avoiding near-range "misses" caused by the cumulative error of long-distance predictions.

[0341] Step 5: Based on the optimal state estimate obtained in Step 4, the TangentBug real-time obstacle avoidance algorithm is used for path planning.

[0342] Specifically, the TangentBug real-time obstacle avoidance algorithm is an obstacle avoidance algorithm based on local environment perception. By switching between two modes, "straight-line approach to the target" and "circumventing along the obstacle boundary", it achieves collision-free path planning in dynamic environments and has the characteristics of low computational complexity and strong real-time performance.

[0343] The core idea of ​​TangentBug's real-time obstacle avoidance algorithm is to acquire the position and range information of obstacles through sensors, and prioritize moving towards the target along a straight line. When an obstacle is detected on the path, it switches to obstacle avoidance mode, detouring along the tangent direction of the obstacle until an obstacle-free path towards the target is obtained again. Its core principle is to calculate the obstacle avoidance tangent point and movement direction through geometric relationships, ensuring a safe and efficient path.

[0344] In this embodiment, the TangentBug real-time obstacle avoidance algorithm is used for path planning, including: basic geometric calculations and establishing path planning logic.

[0345] Specifically, basic geometric calculations include:

[0346] 1. Location and Distance Calculation

[0347] For any two points in a two-dimensional plane, such as the current position of an unmanned surface vessel... and target location The Euclidean distance between them Calculated using the following formula:

[0348]

[0349] At the same time, from point Point of view Direction angle It can be achieved through the arctangent function The result is obtained, and its range is within The formula is as follows:

[0350]

[0351] 2. Obstacle Description and Collision Detection

[0352] This invention simplifies environmental obstacles into circular models for processing. A circular obstacle can be represented by its center coordinates. and radius Full description.

[0353] To determine whether an unmanned surface vessel (USV) will collide with an obstacle, the distance from the USV to the center of the obstacle can be compared. distance With the radius of the obstacle The size is used to determine the collision. The condition for determining a collision is:

[0354]

[0355] 3. Tangent calculation

[0356] When the unmanned surface vessel is outside the obstacle (i.e. Furthermore, when it is necessary to bypass the obstacle, the distance from the current position of the unmanned surface vessel must be calculated. Tangents to the circumference of the obstacle. These tangents are the key path guides for entering the "Boundary Bypass" mode.

[0357] The calculation process is as follows:

[0358] (1) Calculate the azimuth angle from the unmanned surface vessel's position P to the center of the obstacle C. .

[0359] (2) Calculate the angle between the tangent and line segment PC based on geometric relationships. This angle can be obtained through trigonometric relationships, where... Let r be the distance from the unmanned surface vessel to the center of the obstacle, and r be the radius of the obstacle.

[0360]

[0361] Therefore, the direction angles of the two tangents relative to PC and They are respectively:

[0362]

[0363] (3) Based on the above angles and distances, the two tangent points can be calculated. and The coordinates of the point of tangency. arrive The distance between the points is Therefore, its coordinates are:

[0364]

[0365] Similarly, tangent point The coordinates are:

[0366]

[0367] Path planning logic includes:

[0368] 1. Linear movement logic

[0369] In "target-oriented" mode, the unmanned surface vessel (USV) will move in the direction pointing towards the target from its current position. Assuming the USV's step size is *s*, its position at the next moment... The coordinate update formula is as follows:

[0370]

[0371] in, This represents the direction angle from the current position to the target.

[0372] 2. Line segment and circle intersection detection logic

[0373] To determine whether the straight path from the current position P of the unmanned surface vessel to the target position Q will collide with a circular obstacle C, it is necessary to perform an intersection detection between the line segment and the circle.

[0374] Specifically, this includes: (1) constructing the equation of a straight line.

[0375] From point and A defined straight line can be represented using the general formula. It represents the coefficient. They are respectively:

[0376]

[0377] (2) Calculate the distance from the center of the circle to the line.

[0378] Center of circular obstacle The formula for calculating the distance D to the above line is:

[0379]

[0380] (3) Determine whether the foot of the perpendicular is on the line segment.

[0381] To determine whether an obstacle truly blocks the path between points P and Q, it's necessary to check if the foot of the perpendicular from the center of the circle to the line falls on line segment PQ. Let the foot of the perpendicular be F; then the position of F can be determined using the parameter t. The parameter t is calculated as follows, where... Indicates that the foot of the perpendicular F lies on line segment PQ:

[0382]

[0383] Comprehensive judgment: if and only if and At this point, line segment PQ intersects with the circular obstacle C, meaning the straight path is not feasible.

[0384] 3. Tangent validity determination logic

[0385] When an obstacle is detected blocking the path, the algorithm calculates two tangent lines. To select the optimal obstacle avoidance direction, it needs to determine which tangent line is more "oriented" towards the target. The core of tangent line validity determination is comparing the deviation between the direction from the current position to the tangent point and the direction towards the target.

[0386] Specifically, for each tangent point T, calculate its direction angle. relative to the target direction angle The minimum included angle, if the minimum included angle is less than If the tangent line is true, then the tangent line is considered valid. Its mathematical expression is:

[0387]

[0388] in, It is the direction angle from the current position P to the tangent point T. It is the direction angle from the current position P to the target Q.

[0389] 4. Boundary Movement Logic

[0390] In "borderline detour" mode, the unmanned surface vessel (USV) needs to move along the edge of the obstacle. To ensure safety, the USV maintains a small safety distance from the obstacle. At this point, the trajectory of the unmanned surface vessel can be approximated as a circle centered at the obstacle center C, with a radius of... The arc. Its coordinate update formula is as follows:

[0391]

[0392] in, The angle of the current position relative to the center of the obstacle. For angle increment ( (Calculated from step size s and radius r). These are the coordinates of the next position after moving along the arc.

[0393] 5. Auxiliary Decision Logic

[0394] (1) Leaving the boundary conditions

[0395] While circling the boundary, the unmanned surface vessel (USV) needs to continuously determine whether it has rounded a sufficient angle to re-attempt towards the target. The key criterion for this judgment is the distance from its current position P to the target Q. Is it less than the distance from the collision point H to the target Q when entering obstacle avoidance mode? Its formula is expressed as:

[0396]

[0397] When the conditions are met, the unmanned surface vessel has bypassed the protrusions of the obstacle, and at this point, it can switch back to the "approaching target" mode.

[0398] (2) Recent obstacle screening

[0399] When multiple obstacles exist within the detection range, to determine the priority obstacle to avoid, it is necessary to filter out the nearest obstacle that poses the greatest threat to the current path. The selection of the nearest obstacle is achieved by calculating the distance from the unmanned surface vessel (USV) to the surface of each obstacle. For circular obstacles, this distance... Subtract the radius of the obstacle from the distance from the unmanned surface vessel to the center of the obstacle:

[0400]

[0401] TangentBug real-time obstacle avoidance algorithm is preferred. The obstacle with the smallest value is the nearest obstacle that needs to be dealt with first.

[0402] Specifically, the TangentBug real-time obstacle avoidance algorithm process is as follows:

[0403] 1. Initialization: When the system starts, it acquires the initial position P and target position Q of the unmanned surface vessel, and sets the movement step size s and safety distance. Parameters such as these.

[0404] 2. Environmental perception: The unmanned surface vessel uses sensors to detect the surrounding environment in real time and obtain information on the location and size of obstacles.

[0405] 3. Path determination: Perform "line segment and circle intersection detection" to determine whether the straight path from the current position P to the target Q is unobstructed.

[0406] 4. Mode Selection and Execution:

[0407] a. If the path is clear: Enter "Towards Target" mode and calculate the next position according to the "Straight Line Movement Formula". And move.

[0408] b. If the path is blocked: enter the "detour along the boundary" mode.

[0409] i. Calculate the two tangent lines from the current position to the nearest obstacle, and select a valid tangent line using the "tangent validity determination formula".

[0410] ii. Control the unmanned surface vessel to move along the selected tangential direction, or enter the movement mode along the obstacle boundary, and update the position according to the "movement along the boundary formula".

[0411] 5. Obstacle avoidance process monitoring: During the obstacle avoidance process, continuously use the "departure boundary condition formula" to determine whether it is possible to get away from the obstacle.

[0412] 6. Mode switching and looping: Once the "leaving the boundary condition" is met, the algorithm considers that it has bypassed the obstacle and then switches back to the "approaching the target" mode.

[0413] 7. Termination condition: Repeat the above steps until the unmanned surface vessel reaches the target location or receives a stop command.

[0414] The above description is merely a preferred embodiment of the present invention. Any simple modifications, equivalent changes, and alterations made by those skilled in the art to the above embodiments without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for cooperative search of surface targets by unmanned aerial vehicles and surface vessels, characterized in that, include: Step 1: Construct a multi-dimensional dynamic environment model, including dynamic environment modeling based on dynamic environment parameters; Step 2: Based on the multi-dimensional dynamic environment model constructed in Step 1, construct a heterogeneous platform kinematic model. The heterogeneous platform kinematic model includes the UAV kinematic model, the unmanned surface vessel kinematic model, and the target kinematic model. Step 3: Based on the UAV kinematic model, UAV kinematic model, and target kinematic model constructed in Step 2, construct a distributed collaborative task allocation strategy. The distributed collaborative task allocation strategy includes UAV strategy, UAV strategy, and collaborative decision-making mechanism. Step 4: Use the extended Kalman filter algorithm to predict the target trajectory and obtain the optimal state estimate; Step 5: Based on the optimal state estimate obtained in Step 4, the TangentBug real-time obstacle avoidance algorithm is used for collision-free path planning. The drone strategy includes: when a target appears in a specific direction and within a specific area of ​​the drone's speed, the drone switches to tracking mode, and its speed direction points towards the target's location. If multiple targets exist within the detection area, the drone's speed direction points towards the center point of the multiple targets. The formula for calculating the target center point is: ； in, The coordinates of the UAV target point, The number of targets detected. These are the x and y coordinates of the i-th target, respectively. The formula for calculating the heading angle of the drone at this time is: ； in For the drone's heading angle, The current coordinates of the drone. The heading angle is the center point of the n detected targets; The unmanned surface vessel (USV) strategy includes: when the USV detects or is dispatched to a target, it switches to pursuit mode, with the following turning angle constraints: ； in, This represents the change in the unmanned surface vessel's turning angle. R represents the speed of the unmanned surface vessel. min This is the minimum turning radius of the unmanned surface vessel; When multiple boats are pursuing each other in coordination, the principle of distance priority and load balancing should be adopted to select one unmanned boat for pursuit. Building a collaborative decision-making mechanism includes: constructing real-time state matrices for unmanned aerial vehicles (UAVs) and unmanned surface vessels (USVs); The parameters of the target are obtained hierarchically, and the target set is defined as follows: ； in, Let represent the target's number and its position coordinates at time t, respectively. These represent the target's velocity and heading angle at the current time t, respectively.

2. The method for cooperative search of surface targets by unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, Dynamic environmental parameters include wind field parameters and water flow velocity parameters. Dynamic environmental modeling includes wind field modeling, water flow velocity modeling, and the introduction of a dual-mode refresh mechanism, a grid indexing mechanism, and collision constraints to refresh obstacle positions. The formula for wind field modeling is: ； in, Let be the wind speed at time t. The attenuation coefficient is... For the disturbance intensity, Standard Gaussian noise; The formula for the dual-mode refresh mechanism is: ； in, , This refers to the last update time. The time of random fluctuation; The grid index formula is: ； Where L and W represent the length and width of the scene. ; The collision constraints are: ； ； in For drone collection, For unmanned surface vessels, parameters The area is an obstacle zone. , These represent the distances between the drone, the unmanned surface vessel, and the obstacle, respectively. This indicates the minimum safe distance between a drone or unmanned surface vessel and an obstacle.

3. The method for cooperative search of surface targets by unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, Constructing a UAV kinematic model includes: Design an irregular polygonal closed patrol trajectory composed of multiple nodes, and define the track point matrix P; The segmented trajectory of the UAV uses a straight-line interpolation model for smooth transition. The parametric equation of the i-th segment of the trajectory is: ； Where t represents the current time, for Location at any given moment For the first Coordinates of the waypoints for Coordinates of the first waypoint For the first Segment trajectory start time; t i+1 Let be the end time of the i-th trajectory segment; s(t) is the time normalization parameter, ranging from... When t is 0, it is the first... The coordinates of the waypoint are given when t=1. The coordinates of the waypoints; H is the flight altitude of the UAV; Considering wind interference, the formula for dynamically adjusting the drone speed is: ； in Represents the initial velocity of the i-th segment of the trajectory. This indicates the drone's own acceleration. The time interval between two adjacent waypoints. Wind speed; The speed constraints and minimum turning radius constraints for the UAV are as follows: ； in, This is the instantaneous speed modulus of the drone, and its value is less than the maximum speed. , It is the turning angular velocity of the drone. This is the turning radius of the drone, and its value is greater than the minimum turning radius. ; Constructing the kinematic model of the unmanned surface vessel includes: patrol speed during patrol. Keeping constant, the parametric equations of the piecewise trajectory are: ； in , The coordinates of the current base point. , For the coordinates of the next base point, For dimensionless time parameters, , The water flow is respectively in , The directional disturbance velocity component, The heading angle pointing towards the target base point; When the unmanned surface vessel (USV) is less than 800m from the target, it will detect the target and switch to pursuit mode. In pursuit mode, the pursuit trajectory parameter equation is: ； in Current position For the current course, This is the difference between the updated target direction and the current heading; The pursuit speed is: ； in For patrol speed, maximum speed , To catch up with the start time, accelerate time. ; The formula for the turning angle of an unmanned surface vessel is: ； in ; Constructing the target kinematic model includes: defining the position update formula: ； in, yes Within a time Displacement increment in direction, yes Within a time The displacement increment in the direction; Design a speed control algorithm and define acceleration. With deceleration They are respectively: ； in, The acceleration percentage parameter value is 0.

1. The deceleration percentage parameter value is 0.15; Design a heading control algorithm; the heading control formula is: ； in, This is the difference between the initial target direction and the current heading. This is the difference between the updated target direction and the current heading; Orientation update rules for design goals: ； in, It is the initial orientation of the target. It is the target's current orientation. It is the target's updated orientation; A simple obstacle avoidance strategy based on boundary distance is adopted to achieve target obstacle avoidance.

4. The method for cooperative search of surface targets by unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, Before performing the extended Kalman filter algorithm, the method further includes: establishing a hybrid motion model of the target based on the target's motion characteristics, wherein the hybrid motion model includes a linear motion model and a circular motion model; Establishing a hybrid motion model of the target includes defining the target state vector: ； The linear motion model is as follows: ； in, The sampling time interval, for Always the goal is The position of the axis for The velocity component of the target at any given moment. for acceleration component, for Location at any given moment for The velocity component at any given moment; The circular motion model is as follows: ； in, and These represent the position increments in the x and y directions, respectively, for the circular motion.

5. The method for cooperative search of surface targets by unmanned aerial vehicles and surface vessels according to claim 1, characterized in that, The extended Kalman filter algorithm includes initialization, prediction, and update phases; Initialization includes setting the initial state vector and covariance matrix, and defining the process noise covariance matrix and observation matrix; The prediction phase is used to determine the current state prediction and the corresponding uncertainty quantification based on the hybrid motion model and the optimal state estimate of the previous time step, including state prediction and covariance prediction. The update phase is used to correct the state prediction values ​​obtained in the prediction phase based on the observation data at the current time, thereby obtaining the optimal state estimate and the corresponding uncertainty quantification, including Kalman gain calculation, state update and covariance update.

6. The method for cooperative search of surface targets by unmanned aerial vehicles and surface vessels according to claim 5, characterized in that, State prediction is used to derive the state prediction value at the current moment through a hybrid motion model. The expression is: ； in, For the first The optimal state estimate after the update at each time step. The state transition matrix is ​​the objective function, used to describe the transition from state to state 1. Time to the The evolutionary relationship at any moment, Then it is the first The predicted state value at time; Covariance prediction is used to quantify the uncertainty of state prediction values, and its expression is: ； In the formula, It is the first The covariance matrix of the optimal state estimate at time step 1. State transition matrix transpose, The process noise covariance matrix is... That is, the first The covariance matrix of the time-state prediction.

7. The method for cooperative search of surface targets by unmanned aerial vehicles and surface vessels according to claim 5, characterized in that, Kalman gain is used to determine the weighting of the predicted and observed results in the final state estimation, and its expression is: ； in, For the observation noise matrix transpose, To observe the noise covariance matrix, That is, the first Kalman gain at time step; The state update obtains the optimal state estimate for the current time step by combining observed data and Kalman gain to correct the state prediction value in the prediction stage. The expression is: ； In the formula, For the first The observed value at time, For the derived theoretical observation values, To observe the residuals, Then it is the first The optimal state estimate after correction at time step; Covariance update is used to quantify the uncertainty of the optimal state estimate after the state update, and its expression is: ； in, It is the identity matrix. It is the first The covariance matrix of the optimal state estimate at time step.

8. The method for cooperative search of surface targets by unmanned aerial vehicles and surface vessels according to claim 1, characterized in that, Step 5 includes: constructing the basic geometric model; The path planning logic is set based on the constructed basic geometric model, and path planning is performed according to the path planning logic. The path planning logic includes straight line movement logic, line segment and circle intersection detection logic, tangent validity determination logic, boundary movement logic, and auxiliary determination logic.

9. A method for cooperative search of surface targets by unmanned aerial vehicles (UAVs) according to claim 8, characterized in that, Constructing the basic geometric model includes: Calculate location and distance: ； This is the current location of the unmanned surface vessel. For the target location, Let P and Q be the Euclidean distance between them. Description of obstacles and collision condition determination: Describe obstacles as circles, using their center coordinates. and radius To describe; The conditions for a collision to occur are: ; in, For the unmanned surface vessel to reach the center of the obstacle The distance; The radius of the obstacle; Calculate the current position of the unmanned surface vessel Tangent to the circumference of the obstacle: Calculate the direction angle from the current position P of the unmanned surface vessel to the center C of the obstacle. ; Based on geometric relationships, the angle between the tangent and line segment PC can be calculated using trigonometric functions. : ； Let r be the distance from the unmanned surface vessel to the center of the obstacle, and r be the radius of the obstacle. The direction angles of the two tangents relative to PC and They are respectively: ； According to the angle , and tangent point , arrive The distance between the two points is used to calculate the two tangent points. and The coordinates.

Images