An unmanned surface vehicle cluster obstacle avoidance method fusing artificial potential field method and dynamic window method

Abstract

This invention discloses an obstacle avoidance method for unmanned surface vessel (USV) swarms that integrates the artificial potential field method and the dynamic window method, belonging to the field of USV swarm obstacle avoidance. The method includes the following steps: dividing the USV swarm into a lead USV and several follower USVs, constructing a unified geographic coordinate system of USV-obstacle-target point; calculating the resultant force on the lead USV using an improved artificial potential field method, determining the motion trend direction of the USV swarm from the resultant force, and calculating a reference speed as a global path guide; solving for the optimal speed using an improved dynamic window method; fusing the reference speed and the optimal speed using adaptive weights to obtain the lead USV's fused speed, updating the lead USV's position with this fused speed; and considering the attractive forces between follower USVs and the lead USV, the repulsive forces between adjacent follower USVs, and formation constraints to complete the USV swarm obstacle avoidance. This invention balances path tracking accuracy, obstacle avoidance safety, and navigation efficiency, and possesses robust obstacle avoidance capabilities in complex environments.

Description

Technical Field

[0001] This invention relates to the field of obstacle avoidance in unmanned surface vessel (USV) swarms, and specifically to an obstacle avoidance method for USV swarms that integrates the artificial potential field method and the dynamic window method. Background Technology

[0002] Small unmanned surface vessels (USVs) are widely used in various fields such as marine and lake surveying and water quality monitoring due to their advantages of maneuverability, flexibility, and ease of operation. However, due to their inherent limitations, such as limited carrying capacity and weak ability to cope with complex environments, a single USV cannot meet the growing demand for marine and lake surveying. Compared with single-vessel operations, collaborative operations of USV swarms can effectively improve work efficiency, reduce costs, and alleviate the intensity of manual labor, making USV swarms a hot research area. Swarm operations usually require maintaining a certain formation, which brings technical challenges in areas such as collaborative control and path planning. Among these, swarm obstacle avoidance technology is one of the key technologies to ensure the efficient operation of USV swarms.

[0003] The artificial potential field method is an algorithm based on a virtual potential field. It controls obstacle avoidance and path planning of unmanned surface vessels (USVs) through potential field forces. It has received widespread attention and application due to its advantages such as simple model structure, low computational redundancy, and good real-time performance. However, the traditional artificial potential field method still has some shortcomings. The main issues are that when the obstacle is too close to the target point, the target may become unreachable; and when the repulsive and attractive forces in the potential field are equal in magnitude and opposite in direction, it is easy to get trapped in local minima.

[0004] Dynamic windowing, as an efficient local path planning algorithm, has been widely used in obstacle avoidance for mobile robots due to its excellent real-time performance and low computational cost. However, dynamic windowing is prone to getting trapped in local optima, which limits its planning efficiency and success rate. Summary of the Invention

[0005] Based on the above-mentioned technical problems, this invention proposes an obstacle avoidance method for unmanned surface vessel swarms that integrates the artificial potential field method and the dynamic window method.

[0006] The technical solution adopted in this invention is: An obstacle avoidance method for unmanned surface vessels (USVs) swarms that integrates the artificial potential field method and the dynamic window method includes the following steps: S1. Divide the unmanned surface vessel (USV) cluster into a lead vessel and several follower vessels. Both the lead vessel and the follower vessels are equipped with lidar and GNSS modules to acquire information and construct a unified geographic coordinate system of USV-obstacle-target point. S2. The improved artificial potential field method is used to calculate the resultant force on the pilot boat, and the motion trend direction of the unmanned surface vessel swarm is determined by the resultant force, and the reference speed of the pilot boat is determined. S3. Determine the optimal speed of the pilot boat using the improved dynamic window method; S4. The reference speed and the optimal speed of the pilot boat are fused by adaptive weights to obtain the fused speed of the pilot boat, and the position of the pilot boat is updated with the fused speed of the pilot boat. S5. Repeat S2-S4 to obtain the fusion speed of each following vessel and complete the obstacle avoidance of the unmanned vessel swarm.

[0007] The beneficial technical effects of the present invention are as follows: The unmanned surface vessel (USV) swarm obstacle avoidance method proposed in this invention, which integrates the improved artificial potential field method and the dynamic window method, has the following significant advantages through improved design of the artificial potential field-dynamic window fusion and downsampling Bezier curve path optimization: 1. Global balance of multi-objective optimization. Relying on a multi-source fusion two-dimensional speed evaluation mechanism, multi-dimensional indicators such as heading deviation and obstacle safety are normalized and fused to achieve multi-objective weighted balance optimization in speed space, taking into account path tracking accuracy, obstacle avoidance safety and navigation efficiency, and avoiding performance imbalance caused by over-optimization of a single indicator.

[0008] 2. Robust obstacle avoidance capability in complex environments. By integrating the improved artificial potential field method and the improved dynamic window method, this invention retains the global path guidance of the artificial potential field method while leveraging the local trajectory prediction and forward simulation advantages of the dynamic window method. Furthermore, by combining adaptive improvements to both methods, it effectively solves the problems of local minima and target unreachability inherent in the traditional artificial potential field method, while also compensating for the lack of global guidance in the dynamic window method. This significantly improves the autonomous obstacle avoidance robustness of the unmanned surface vessel (USV) in complex dynamic environments. This invention uses the improved dynamic window method to kinematically correct the direction provided by the improved artificial potential field method, thereby effectively suppressing oscillations and escaping local minima.

[0009] 3. Improved Trajectory Smoothness and Tracking Stability. By using the downsampling Bezier curve trajectory smoothing method, discrete waypoints are fitted to a continuous and differentiable smooth reference trajectory, eliminating trajectory inflection points and abrupt motion changes at the source. This significantly reduces the control error of unmanned surface vessel path tracking, improves the stability and smoothness of trajectory tracking, and has high computational efficiency.

[0010] 4. Trajectory-Velocity Integrated Cooperative Control Performance. An integrated trajectory-velocity cooperative control system is constructed, fusing two-dimensional velocity optimization, obstacle avoidance constraints, and smooth trajectories generated by downsampling Bézier curves. This involves employing an improved artificial potential field method for obstacle avoidance constraints and an improved dynamic window method for two-dimensional velocity optimization. Combined with Bézier trajectory balancing, this solves the problem of separation between trajectory generation and velocity control in traditional methods, achieving closed-loop coordination of trajectory smoothing, velocity optimization, and obstacle avoidance safety, thus comprehensively improving the overall performance of autonomous navigation for unmanned surface vessels. Attached Figure Description

[0011] Figure 1This is a flowchart illustrating the obstacle avoidance method for unmanned surface vessels (USVs) swarms that integrates the artificial potential field method and the dynamic window method of the present invention. Figure 2 This is a schematic diagram illustrating the division of the obstacle's effective range in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the division of the target point's effective range in an embodiment of the present invention; Figure 4 This is a schematic diagram illustrating collision avoidance between unmanned surface vessels in an embodiment of the present invention; Figure 5 The diagram shows a comparison of the obstacle avoidance experimental results of the traditional artificial potential field method and the improved artificial potential field method; where (a) shows the obstacle avoidance experimental results of the traditional artificial potential field method, and (b) shows the obstacle avoidance experimental results of the improved artificial potential field method. Figure 6 A comparison of the results of the improved artificial potential field method and the method of the present invention for the oscillation problem is shown in the figure; where (a) shows the traditional artificial potential field method and (b) shows the fusion method of the present invention. Figure 7 A comparison of the results of the improved artificial potential field method and the method of the present invention for the local minimum problem is shown in the figure; where (a) shows the traditional artificial potential field method and (b) shows the fusion method of the present invention. Figure 8 The simulation comparison of obstacle avoidance of unmanned surface vessels (USVs) swarms using the improved artificial potential field method and the method of the present invention is shown in the figure; (a) shows the traditional artificial potential field method, and (b) shows the fusion method of the present invention. Detailed Implementation

[0012] This invention proposes an obstacle avoidance method for unmanned surface vessels (USVs) swarms that integrates the artificial potential field method and the dynamic window method. This method reconstructs the repulsive potential energy function, integrates an improved dynamic window method, proposes an adaptive potential field and dynamic window path planning method, and performs Bezier path optimization. This invention effectively solves the problems of target unreachability, local minima, and oscillations near obstacles inherent in the artificial potential field method, improving the smoothness and stability of path planning in USV obstacle avoidance.

[0013] An obstacle avoidance method for unmanned surface vessels (USVs) swarms that integrates the artificial potential field method and the dynamic window method includes the following steps: 1. Preparations before the experiment: The unmanned surface vessel (USV) swarm is numbered, consisting of a lead vessel and several follower vessels. The lidar, GNSS, IMU, and industrial control computer are powered on. Sensor time synchronization is calibrated, and the radar scanning resolution is set to 0.1°~1°.

[0014] 2. Environmental modeling preprocessing: The MID-360 lidar performs a 360° omnidirectional scan centered on the unmanned surface vessel (USV) to acquire obstacle distance, azimuth, and reflection intensity. It also acquires the current position, speed, heading angle, and roll rate of the lead and follower USVs within the USV swarm.

[0015] After point cloud filtering and denoising, a local raster map is generated, and the coordinates, size, speed, and heading of obstacles are extracted. A unified geographic coordinate system is established for the unmanned surface vessel, obstacles, and target points.

[0016] 3. Based on environmental modeling and grid maps, perform improved artificial potential field method to calculate the forces and reference velocity of the pilot boat: Calculate the gravitational force of the target point on the unmanned surface vessel. The direction is pointing towards the target point.

[0017] Divided into three zones (danger zone) Controllable area No impact zone ) and the target point in two sections (near range) Long distance Substitute the improved repulsive force function into the equation to calculate the repulsive force acting on the pilot boat. This repulsive force is an important component of the resultant force, and the final motion of the pilot boat is determined by the resultant force. The repulsive force acting on the pilot boat is a key component of the resultant force specifically responsible for obstacle avoidance and collision prevention.

[0018] According to the unmanned surface vessel swarm system based on the pilot-follower model, the resultant force on the pilot vessel is: .

[0019] The motion trend direction of the unmanned surface vessel swarm is determined by the resultant force, and the reference speed of the navigator is output using an improved artificial potential field method. , serving as a global path guide.

[0020] ; In the formula, For the reference speed of the pilot boat, The maximum speed that the pilot boat can reach; Unit direction vector, , The combined force experienced by the pilot boat, | | is the modulus of the resultant force.

[0021] 4. Solve for the optimal speed of the pilot boat using the improved dynamic window method: To adapt and integrate with the improved artificial potential field method, a new velocity search space is constructed: ; An improved evaluation function is adopted: ; In the formula, Let be the linear velocity component of the unmanned surface vessel in the x-direction in the global coordinate system. Let be the linear velocity component of the unmanned surface vessel in the y-direction in the global coordinate system; It is the cost function for the change in heading angle. Let the target distance cost function be... It is a speed cost function. It is the obstacle collision cost function. , , d and d represent the weights of each part, respectively. This indicates normalization processing.

[0022] After traversing all feasible speed combinations and performing collision detection and trajectory evaluation, the optimal speed for the pilot boat is output. .

[0023] 5. Algorithm fusion: ; In the formula: The reference speed of the pilot boat is calculated based on the improved artificial potential field method; The optimal speed of the pilot boat is calculated based on the improved dynamic window method; For weight fusion.

[0024] According to fusion speed Update the position of the lead boat.

[0025] 6. Repeat the above steps to obtain the fusion speed of each following vessel and complete the obstacle avoidance of the unmanned vessel swarm.

[0026] The difference lies in the fact that the resultant force on each following vessel is calculated differently than that on the lead vessel.

[0027] The net force acting on a follower vessel in an unmanned surface vessel swarm is the attractive force exerted by the target point on the follower vessel. The attraction generated by the pilot boat The repulsive force generated by the pilot boat The repulsive force generated by the remaining follower boats and the repulsive force exerted by the obstacle on the following vessel. The resultant force of these five forces is as follows: ; In the formula, This represents the net force acting on a particular follower vessel. It is the vector sum of the repulsive forces exerted on the current following vessel by the other k following vessels, where k refers to the number of other following vessels in the unmanned surface vessel cluster besides the current following vessel.

[0028] Calculate the reference speed of the following vessel: ; In the formula, For the reference speed of the following boat, This is the maximum speed that the follower boat can achieve; 2 is the unit direction vector. , The net force acting on a particular follower vessel. It is the modulus of the resultant force.

[0029] The optimal speed of the following vessel is calculated using the same method as above, that is, the parameters of the lead vessel are replaced with those of the following vessel, and the above steps are repeated. That is, a new speed search space and an improved evaluation function are used to obtain the optimal speed of the following vessel. .

[0030] Repeat the above steps to perform velocity fusion, obtain the fused velocity of the following vessel, and update the position of the following vessel using the fused velocity of the following vessel.

[0031] In the above method, when no obstacle is detected, the position of the unmanned surface vessel (USV) is updated solely by calculating the attractive force generated by the target as the net force acting on the hull. When an obstacle is detected, a reference velocity is obtained through an improved artificial potential field method, an optimal velocity is generated through an improved dynamic window method, and then a fusion weight is considered to obtain a fusion velocity. The USV's position is updated using the fusion velocity, completing the fusion obstacle avoidance iteration.

[0032] Fusion weights The settings can be adjusted based on the distance between the unmanned surface vessel (USV) and obstacles. If it is in a controllable area, a balanced fusion method is used, and the fusion weight can be set to k=0.5. If it is in a dangerous area, the dynamic window method takes precedence, and the fusion weight can be set to k=0.2.

[0033] The above weight values ​​are typical examples. In practical applications, the fusion weight k can be dynamically adjusted according to the unmanned surface vessel's power characteristics, environmental complexity, and mission requirements to optimize path planning performance.

[0034] 7. Trajectory optimization based on downsampling Bézier curves: Taking the pilot boat as an example, the path generated by the fusion algorithm is smoothed: Obtain discrete path points of the fusion path According to step size Downsampling of path points yields sparse control points. Substituting these control points into the Bézier curve formula results in: ; Output a continuous, smooth final trajectory without inflection points.

[0035] The path optimization steps for the follower boat are the same as those for the lead boat. The only difference is that the lead boat is guided by the global target point, calculates the target gravity, and completes the global path planning from the starting point to the end point. The follower boat, on the other hand, uses the lead boat as a "virtual target," calculates the formation-maintaining gravity, avoids obstacles and other boats, and completes the local path planning to maintain the formation.

[0036] In the process of obstacle avoidance in the unmanned surface vessel (USV) swarm, the lead USV independently plans its path according to the fusion algorithm, and the following USVs repeat the above algorithm steps to ensure swarm collaborative obstacle avoidance and formation stability.

[0037] In the improved artificial potential field method, the reference speed calculation methods for the pilot and follower vessels are identical except for the resultant force when calculating the resultant force vector; in the improved dynamic window method for solving the optimal speed, the optimal speed calculation methods for the pilot and follower vessels are the same.

[0038] The present invention will now be described in more detail with reference to specific embodiments.

[0039] Example 1: like Figure 1 As shown, an obstacle avoidance method for unmanned surface vessels (USVs) swarms that integrates the artificial potential field method and the dynamic window method includes the following steps: S1: Environmental modeling preprocessing; S11: Radar data acquisition; The MID-360 lidar deployed on the bow of the unmanned surface vessel (USV) performs a 360° omnidirectional scan with an angular resolution of 0.1° to 1°, centered on the USV, and collects point cloud data within a 360° range around the USV. The point cloud data includes at least obstacle distance information, azimuth information, and reflection intensity information.

[0040] S12: Point cloud data filtering; The Kalman filter algorithm is used to denoise and fuse the point cloud data collected by the MID-360 lidar, removing sea surface clutter, surge interference and sensor noise points to generate a clean environmental point cloud dataset.

[0041] S13: Raster map construction; Based on the preprocessed point cloud data, a local environmental grid map of the unmanned surface vessel is constructed. At the same time, feature information of static obstacles (reefs, buoys, docks) and dynamic obstacles (other ships, floating objects, neighboring vessels in the cluster) is extracted. The feature information includes at least the obstacle's position coordinates, size parameters, movement speed, and heading angle.

[0042] S14: Construction of the kinematic model of the unmanned surface vessel; The unmanned surface vessel's (USV) position, speed, heading angle, and yaw rate are obtained through the GNSS module, and a unified geographic coordinate system is constructed for the USV, obstacles, and target points.

[0043] S2: The improved artificial potential field method is used to calculate the resultant force on the pilot boat, and the motion trend direction of the unmanned surface vessel swarm is determined by the resultant force, and the reference speed of the pilot boat is determined. S21: Optimize the repulsive potential field function; In this invention, the unmanned surface vessel moves in space, and the gravitational potential field generated by the target point always acts as a force throughout the process, while the repulsive potential field generated by obstacles in the path only appears when the unmanned surface vessel approaches the obstacle within a certain range.

[0044] By improving the repulsive potential field function of the traditional artificial potential field method, the range of action of the obstacle is divided into three parts, and the range of action of the target point is divided into two parts. Different repulsive force adjustment coefficients are added to the repulsive force within the range of action of different obstacles and the target point. Near the target point, the repulsive force adjustment coefficient is set to be related to the distance to the target point, so that the repulsive force at the target point is 0, and the net potential force of the unmanned surface vessel at the target point is always minimized globally.

[0045] like Figure 2 As shown, define a The area of ​​effect between the unmanned surface vessel and obstacles is divided into danger zones (using...). (representation), controllable area (using) (represented), unaffected area (using) express).

[0046] Taking the pilot boat as a reference, in the formula: The coordinates of the pilot boat's current position in the global coordinate system. The range of influence of the obstacle. Distance to the collision hazard zone of the obstacle. The distance between the pilot boat and the obstacle is the Euclidean distance.

[0047] like Figure 3 As shown, an L is defined as the range of action between the pilot boat and the obstacle. The range of action between the pilot boat and the target point is divided into a near range (using...). (representation), long distance (using) express).

[0048] L is the threshold value of the effective range between the pilot boat and the target point. These are the position coordinates of the target point in the global coordinate system. The distance between the pilot boat and the target point is expressed in Euclidean form.

[0049] The repulsive potential field function acting on the pilot boat is modified as follows: ; In the formula, Let be the repulsive potential field function acting on the pilot boat. , This is a weighted index representing the rate of change of the intensity of the repulsive potential field with respect to the distance between the pilot vessel and the target point. is the coefficient of the repulsive potential field.

[0050] The repulsive force experienced by the pilot boat is: ; In the formula, The repulsive force experienced by the pilot boat. The gradient operator represents the gradient with respect to the potential field function. Find the gradient. The basic obstacle avoidance repulsion force component, which is related to the distance to the obstacle, is used to drive the unmanned surface vessel away from the obstacle; The modified repulsive force component, which is related to the distance to the target point, is used to solve the target inaccessibility problem of the traditional artificial potential field method, ensuring that the unmanned surface vessel can successfully reach the target point.

[0051] (1) When it is within the vicinity of the target point, i.e. ; When it is within a controllable area, that is hour; ; When in a danger zone, i.e. hour; ; in, The direction is from the obstacle towards the pilot boat. The direction is from the pilot boat towards the target point. As the pilot boat gradually approaches the target point, the repulsive potential energy will continuously decrease. When the pilot boat reaches the target point, the repulsive force drops to 0, at which point the total repulsive potential energy experienced by the pilot boat reaches its global minimum. By selecting an appropriate repulsive force adjustment coefficient, the problem of target unreachability can be effectively solved.

[0052] When the pilot boat is in a danger zone, obstacle avoidance should take precedence over reaching the target point. In this situation, the direction of the resultant force should be more inclined towards obstacle avoidance so that the pilot boat can better evade obstacles. Because the direction of the resultant force is determined by… To determine, when in a danger zone, It should be greater than when the pilot boat is within close range of the target point. ,Right now .

[0053] That is, through The value of to To control, and thus control the direction of the resultant force. When hour, and Follow For changes of the same order, the resultant force maintains its tendency to point towards the target point while moving away from the obstacle, avoiding deviation from the overall path due to excessive repulsive force. As the pilot boat approached the target point ( (reduce) Will be more It decays faster. The dominance of the force will become stronger and stronger, thus ensuring that in dangerous areas, the main role of the combined force is to push the pilot boat away from the obstacle, rather than being affected by the corrective force of the target direction.

[0054] (2) When at a distance from the target point, i.e. ; When it is within a controllable area, that is hour; ; When in a danger zone, i.e. hour; ; in, The direction is from the obstacle towards the pilot boat. When the pilot boat is in a danger zone, in order to better avoid the obstacle, the repulsion adjustment coefficient needs to be increased compared to the controllable zone. At that time, it should meet the following requirements. .

[0055] when hour, = Dangerous areas and controllable areas If they are the same size, the repulsive force will not increase. When hour, Will be more Much larger, such a dangerous area It will be significantly larger than the controllable area. This achieves the purpose of increasing repulsive force and forcing obstacle avoidance.

[0056] S22: Obstacle avoidance of unmanned surface vessels swarms based on an improved artificial potential field method; The unmanned surface vessel (USV) swarm formation model used in this invention employs a pilot-follower model. The pilot vessel determines the overall direction and speed of the swarm, while the follower vessels adjust their own routes by calculating the distance and angle between themselves and the pilot vessel.

[0057] According to the unmanned surface vessel swarm system based on the pilot-follower model, the resultant force on the pilot vessel is... It can be represented as: ; in, The attraction of the target point to the pilot boat, The repulsive force experienced by the pilot boat.

[0058] The improvement of the artificial potential field method in this invention mainly lies in its repulsive potential field function, while the attractive force uses the gravitational potential field function and gravity from the traditional artificial potential field method. Existing methods can be used for calculation.

[0059] The lead boat will affect the other follower boats, but it is not affected by the other follower boats. That is, the resultant force on the lead boat does not take the follower boats into account, while the follower boats need to take into account the forces between them and the lead boat, as well as the forces between the lead boats, in order to maintain formation and other functions.

[0060] The direction of motion of the unmanned surface vessel swarm is determined by the combined forces of the lead vessels, and the reference velocity of the lead vessels is output using an improved artificial potential field method. .

[0061] ; In the formula, For the reference speed of the pilot boat, The maximum speed that the pilot boat can reach; Unit direction vector, , The combined force experienced by the pilot boat, | | is the magnitude (size) of the resultant force.

[0062] S3: Solve for the optimal speed of the pilot boat using an improved dynamic window method; S31: Optimize the evaluation function; The dynamic window method is mainly used to avoid local obstacles in static or dynamic environments and generate local paths for unmanned surface vessels (USVs). The core of the dynamic window method lies in searching for candidate velocities in the velocity space and selecting the optimal velocity using an evaluation function, thereby formulating the local motion plan for the USV.

[0063] To better integrate with the improved artificial potential field method, the improved velocity search space of this invention is as follows: ; In the formula, Let be the linear velocity component of the unmanned surface vessel in the x-direction in the global coordinate system. Let be the linear velocity component of the unmanned surface vessel in the y-direction in the global coordinate system; This is the maximum speed that the unmanned surface vessel can achieve.

[0064] The improved evaluation function is: ; In the formula, Let be the linear velocity component of the unmanned surface vessel in the x-direction in the global coordinate system. Let be the linear velocity component of the unmanned surface vessel in the y-direction in the global coordinate system; It is the cost function for the change in heading angle. Let the target distance cost function be... It is a speed cost function. It is the obstacle collision cost function. , , d and d represent the weights of each part, respectively. This indicates normalization processing, which involves normalizing the four cost functions. Specifically, the velocity cost function and the target distance cost function remain consistent with the original evaluation function.

[0065] S32: Improved cost function; This invention introduces a new heading angle change cost function for the dynamic window method, which effectively penalizes large turning behaviors by calculating the minimum angle difference between the current heading angle and the predicted heading angle. A significant improvement is made to the obstacle collision cost function, replacing the traditional binary judgment with a distance-based continuous function. The new obstacle collision cost function increases linearly as the distance to the obstacle decreases, providing more refined gradient information for obstacle avoidance decisions, thus enabling earlier and smoother obstacle avoidance.

[0066] The improved heading angle change cost function is: ; in, It is the heading angle for predicting speed. It is the heading angle at the current speed.

[0067] The improved obstacle collision cost function is: ; in, Indicates the distance to the nearest obstacle. This indicates the safe distance from obstacles, consistent with the safe distance in the artificial potential field method.

[0068] Speed ​​cost function: ; In the formula, v is the linear velocity of the currently sampled unmanned surface vessel; that is, the greater the velocity, the higher the score for this item.

[0069] Target distance cost function: ; In the formula, For the reason The simulated trajectory end position These are the coordinates of the target point.

[0070] This invention improves the heading angle change cost function to calculate the minimum angle difference between the current heading angle and the predicted heading angle, effectively penalizing large turning behavior; and improves the collision cost function by changing the traditional binary judgment to a distance-based continuous function. The new collision cost function increases linearly as the distance to the obstacle decreases, providing more delicate gradient information for obstacle avoidance decision-making, thereby enabling earlier and smoother obstacle avoidance.

[0071] The dynamic window method calculates the cost function values ​​for each candidate velocity pair from all feasible velocity pairs generated within the velocity window, and substitutes these values ​​into the evaluation function to calculate the overall score. The velocity with the highest overall score is the optimal velocity command that is both safe and efficient in moving towards the target, and allows for smooth turning.

[0072] S33. Generate all feasible velocity pairs in the velocity search space. For each candidate velocity pair among all feasible velocity pairs, calculate all cost function values ​​and substitute them into the evaluation function to calculate the comprehensive score. Select the velocity pair with the highest comprehensive score as the optimal speed of the pilot boat.

[0073] The dynamic window method calculates all cost function values ​​for each candidate speed pair from all feasible speed pairs generated in the speed window, substitutes them into the evaluation function to calculate the comprehensive score, and then the speed with the highest comprehensive score is the optimal speed command that is both safe and efficient in moving towards the target and has a smooth turning.

[0074] S4. The reference speed and the optimal speed of the pilot boat are fused by adaptive weights to obtain the fused speed of the pilot boat, and the position of the pilot boat is updated with the fused speed of the pilot boat.

[0075] The formula for calculating the fusion speed of the pilot boat is as follows: ; In the formula: The reference speed of the pilot boat is calculated based on the improved artificial potential field method; The optimal speed of the pilot boat is calculated based on the improved dynamic window method; For fusion weights; According to fusion speed Update the position of the lead boat.

[0076] Reference speed It generates a speed command biased towards the global path based on the resultant force direction of the target's attraction and the obstacle's repulsion; optimal speed. The feasible speed is calculated using the dynamic window method. It is the optimal solution among a set of safe and executable local obstacle avoidance speed commands generated under the current motion constraints of the unmanned surface vessel.

[0077] To determine the fusion weights, a balanced fusion method is used if the area is controllable, with a weight k=0.5; if the area is dangerous, a dynamic window method is used, with a weight k=0.2. Of course, in practical applications, the fusion weight k can be adjusted according to the unmanned surface vessel's dynamic characteristics, environmental complexity, and mission requirements.

[0078] S5. Repeat S2-S4 to obtain the fusion speed of each following vessel and complete the obstacle avoidance of the unmanned vessel swarm.

[0079] The net force acting on a follower vessel in an unmanned surface vessel swarm is the attractive force exerted by the target point on the follower vessel. The attraction generated by the pilot boat The repulsive force generated by the pilot boat The repulsive force generated by the remaining follower boats and the repulsive force exerted by the obstacle on the following vessel. The resultant force of these five forces is as follows: ; In the formula, This represents the net force acting on a particular follower vessel. It is the vector sum of the repulsive forces exerted on the current following vessel by the other k following vessels, where k refers to the number of other following vessels in the unmanned surface vessel cluster besides the current following vessel.

[0080] The five forces acting on the following vessel can be obtained using conventional methods, such as a simplified artificial potential field force model. This model only uses the basic ideas of attraction and repulsion, eliminating the need to construct complex potential field functions like those used on the lead vessel. For example, the repulsive force exerted by obstacles on the following vessel can be calculated directly based on the distance using the traditional artificial potential field repulsion formula.

[0081] Furthermore, calculate the reference speed of the following vessel: ; In the formula, For the reference speed of the following boat, This is the maximum speed that the follower boat can achieve; Unit direction vector, , The net force acting on a particular follower vessel. It is the modulus of the resultant force.

[0082] Repeat S3. When the unmanned surface vessel is a follower vessel, the follower vessel parameters are used to obtain the optimal speed of the follower vessel.

[0083] Repeat step S4 to perform velocity fusion, obtain the fused velocity of the following vessel, and update the position of the following vessel based on the fused velocity of the following vessel.

[0084] Example 2: Based on Example 1, a repulsive force adjustment coefficient is further determined, which includes a weighting exponent and a repulsive potential field coefficient. The specific steps are as follows: (1) Set initial parameters, including the initial repulsive potential field coefficient. Initial weight index , .

[0085] (2) In a scenario where only the target's gravitational field exists, observe whether the pilot boat can successfully reach the target point: if it cannot reach the target point and oscillates near the target point, it indicates that the repulsive force decays too slowly when approaching the target point, and needs to be increased. If there is no repulsive force during the arrival process and the path is entirely dominated by gravity, it indicates that the overall repulsive force is too weak and needs to be increased. .

[0086] (3) Observe the obstacle avoidance behavior of the pilot boat in the presence of obstacles: If the pilot boat gets too close to the obstacle and there is a risk of collision, it means the repulsive force is too weak and needs to be increased. Or adjust ; If the pilot boat deviates significantly from the overall path or takes excessive detours, it indicates that the repulsive force is too strong and needs to be reduced. Or adjust , .

[0087] (4) Compare different parameter combinations and select the set of repulsive potential field coefficients with the best overall performance. and weight index , This serves as the final value, thus determining the repulsion adjustment coefficient.

[0088] Example 3: Based on Example 2, a repulsion rule was added to avoid collisions between unmanned surface vessels (USVs). The rule was designed as follows: each USV establishes a radius of... The collision area and radius are The collision hazard zone can have a radius of [missing information]. The annular region is considered as a collision buffer zone, such as Figure 4 As shown.

[0089] As the unmanned surface vessels approach each other, when the hulls Located on the hull Outside the collision buffer zone, there is no repulsive force between the two; when the hull... Located in the hull Within the collision buffer zone, the repulsive force increases linearly with decreasing distance; when the hull... Located in the hull When the collision hazard zone is reached, the repulsive force increases rapidly in an exponential relationship. The specific design expression for the repulsive force is as follows: ; in, The repulsive force coefficient controls the intensity of the repulsive force. This represents the position vector of hull i in the global coordinate system. This represents the position vector of hull j in the global coordinate system; It is a piecewise adjustment function, the core function of which is to dynamically control the magnitude and variation of the repulsive force based on the actual distance between the two boats; It is an effective boundary for inter-boat repulsion force artificially set in the artificial potential field method. If the boat length is M, it is generally set to be greater than 6M. The distance between the collision buffer zone and the danger zone between boats is generally taken as 3-6 meters. Let be the Euclidean distance between hull i and hull j.

[0090] function The expression is: ; in, This is the radius of the safe zone for the unmanned surface vessel; anything smaller than this zone is considered a collision. It is the radius of the repulsive force between the boats. Let be the actual distance between hull i and hull j. Assume that in the cluster, the distance between hull i and hull j is... Unmanned surface vessels that communicate with each other have One, then the hull The repulsive force generated by the other boats in the group is The resultant force of the three forces is: ; This formula represents the total inter-vessel repulsion force acting on hull i, which is the single-vessel repulsion force exerted on hull i by m neighboring vessels interacting with it within the unmanned surface vessel cluster. The vector superposition result, in its physical sense, integrates the local collision avoidance actions among multiple vessels into a unified resultant force, which serves as a constraint for the motion control of the unmanned surface vessel, guiding the vessel to simultaneously avoid all neighboring vessels and achieving safe and coordinated navigation of the swarm.

[0091] That is, the total repulsive force is the vector sum of the repulsive forces of m individual boats, and its modulus ranges from [0, m× When no adjacent vessels are within the effective range, the resultant force is 0; when all m adjacent vessels are in the danger zone and their directions are completely aligned, the resultant force reaches its theoretical maximum value of m× Its role is to participate in the adjustment of the following vessel's motion attitude and navigation control, avoid collisions within the formation, and maintain the desired formation of the group.

[0092] In layman's terms, the combined force of the lead boat determines the swarm's course, and all adjustments made by the following boats must revolve around this direction. The combined force of the lead boat also determines the overall speed of the unmanned surface vessel (USV) swarm; that is, the speed adjustments of the following boats are based on the lead boat's speed. The combined force of the following boats includes: the target's gravitational pull, the lead boat's gravitational / repulsive pull, the repulsive pull of the other boats, and the repulsive pull of obstacles. Among these, the gravitational / repulsive pull of the lead boat on the following boats is directly determined by the lead boat's position and motion. Maintaining formation and adjusting motion are essentially responses to changes in position, course, and speed determined by the combined force of the lead boat.

[0093] Example 4: Based on Example 3, Bézier curves are further used to perform post-processing optimization of the planned path, which is described in detail below: S41: Adaptive Potential Field and Dynamic Window Path Planning Algorithm; The improved artificial potential field method solves the goal unreachability problem, but it still suffers from local minima and oscillations, especially near obstacles. The dynamic window method dynamically generates velocity windows, combining robot kinematic constraints and environmental obstacle information to select the optimal velocity pair in real time, avoiding getting trapped in local minima. Near obstacles, the trajectory prediction and cost evaluation mechanisms of the dynamic window method can suppress the high-frequency oscillations of the potential field method, generating smoother obstacle avoidance paths and effectively mitigating the local minima and oscillation problems of the artificial potential field method. Therefore, an adaptive potential field and dynamic window path planning algorithm is proposed. When encountering obstacles, a hybrid potential field dynamic window path planning algorithm is used for path planning; when there are no obstacles, the artificial potential field method is used.

[0094] To optimize the path planning performance of unmanned surface vessels (USVs), an adaptive adjustment fusion strategy based on environmental risk is implemented. This involves improving the fusion weight coefficients of the artificial potential field method and the dynamic window method to achieve intelligent switching of algorithmic advantages. In controllable areas, a balanced fusion mode (K=0.5) is adopted, fully leveraging the global guidance of the artificial potential field method and the motion feasibility advantage of the dynamic window method. The artificial potential field method generates gravitational and repulsive fields based on the distribution of the target point and obstacles, providing macroscopic path guidance for the USV; the dynamic window method samples and optimizes the velocity space, outputting appropriate velocity combinations. The two work together to maintain effective approach to the target point while generating smooth and feasible trajectories, significantly improving path planning quality. When entering hazardous areas, the system automatically increases the weight of the dynamic window method (K=0.2) to prioritize obstacle avoidance safety. In this case, the algorithm primarily uses the dynamic window method, relying on its forward simulation and collision detection capabilities to generate emergency avoidance strategies in real time. This adaptive fusion mechanism based on environmental risk assessment ensures that the USV maintains the optimal decision-making mode in different scenarios, balancing path optimality and navigation safety.

[0095] Regarding path smoothness, the algorithm effectively suppresses the high-frequency oscillation phenomenon near obstacles in the artificial potential field method through forward simulation using the dynamic window method and trajectory optimization mechanism. The fusion weight adjustment mechanism further enhances this advantage, ensuring the stability of obstacle avoidance actions by using the dynamic window method as the main approach in dangerous areas, and balancing the advantages of both methods in controllable areas, making the heading angle change of the planned path smoother.

[0096] In addressing local minima, the adaptive potential field and dynamic window path planning algorithms achieve significant breakthroughs through trajectory prediction and cost function design using the dynamic window method. When the artificial potential field method gets stuck in a local minimum due to resultant force equilibrium, the dynamic window method can discover a better escape path through forward simulation, and its multi-objective cost function ensures that a feasible solution can always be found. This complementary mechanism enables the algorithm to autonomously escape potential field traps and maintain continuous progress towards the target point, significantly improving the success rate of path planning in complex environments.

[0097] S42: Optimization of Bézier curves; To further improve the smoothness and traceability of paths generated by the adaptive potential field and dynamic window path planning algorithms, Bézier curves are introduced for post-processing optimization of the planned paths. Bézier curves are parametric curves that effectively eliminate sharp inflection points during the generation of discrete points in path planning, producing continuous and smooth curves, thus resulting in a fusion path planning algorithm based on Bézier curve optimization.

[0098] Given a set of control points The formula for the Bézier curve is shown below.

[0099] ; in, Let be a point on the Bézier curve. For time, As control points, Let be the combination number, denoted as: ; By downsampling the control points, we obtain: ; in, The sampling interval is... These are the control points after downsampling.

[0100] The final smoothed trajectory is shown in the following formula: .

[0101] Based on the above embodiments, the innovation of this invention is mainly reflected in the following aspects: 1. An obstacle avoidance control framework that integrates the artificial potential field method and the dynamic window method; To address the problems of local minima and dynamic obstacle avoidance lag in traditional artificial potential field methods, as well as insufficient global guidance in dynamic window methods, this invention proposes an improved obstacle avoidance control framework that integrates artificial potential field methods and dynamic window methods. The artificial potential field method provides global path guidance, while the dynamic window method enables trajectory prediction and forward simulation in the local velocity space. The attraction / repulsion cost of the potential field method is integrated into the velocity evaluation system of the dynamic window method, achieving synergistic optimization of global path guidance and local dynamic obstacle avoidance. This overcomes the inherent defects of single methods and improves the autonomous obstacle avoidance and path tracking performance of unmanned surface vessels in complex environments.

[0102] 2. Multi-source fusion two-dimensional velocity evaluation mechanism; To address the multi-objective optimization requirements of unmanned surface vessels in the two-dimensional velocity space, a multi-source fusion two-dimensional velocity evaluation mechanism is constructed: integrating multi-dimensional indicators such as heading deviation cost and obstacle safety cost, a comprehensive two-dimensional velocity evaluation function is built to achieve multi-objective weighted fusion in the velocity space, providing a unified evaluation standard for velocity optimization, and taking into account both path tracking accuracy and obstacle avoidance safety performance.

[0103] 3. A trajectory smoothing method based on downsampling Bézier curves; To address the issues of discontinuous and inflection-point-progressed fitted trajectories from discrete waypoints, a downsampling Bézier curve trajectory smoothing method is proposed. First, the original pathpoints are sparsified by step-size downsampling. Then, based on the nth-degree Bézier curve formula, and with the Bernstein basis function weighted by binomial coefficients as the core, the downsampling pathpoints are fitted into a continuous and differentiable smooth reference trajectory. This achieves accurate and smooth trajectory reconstruction, eliminates abrupt trajectory changes, and improves the stability of unmanned surface vessel path tracking.

[0104] The invention will be further explained below with reference to specific application examples.

[0105] In this application example, a swarm of three unmanned surface vessels (USVs) was constructed for obstacle avoidance experiments. One USV served as the lead vessel, and the other two acted as follower vessels. For ease of distinction, as follows... Figure 5 As shown, the lead boat is marked in blue, follower boat 1 and follower boat 2 are represented in orange and yellow respectively, black dots represent obstacles, and "×" marks the location of the target point. In particular, several obstacles are set up near the target point. The unmanned boat will still be affected by these obstacles when it reaches the destination, thus verifying the effectiveness of the improved artificial potential field method in solving the problem of target unreachability.

[0106] Figure 5 A comparison diagram of the traditional artificial potential field method and the improved artificial potential field method; Figure 5 In (a), as the unmanned surface vessel (USV) moves towards the target point and gradually approaches it, the repulsive force generated by obstacles near the target point has a significant impact, preventing the two following USVs from successfully reaching the target point. In (b), however, due to adjustments made to address the repulsive force near the target point, all three USVs not only successfully reached the target point but also avoided any collisions during their journey. This comparative result demonstrates the effectiveness of the improved artificial potential field method in solving the problem of target unreachability during USV swarm obstacle avoidance.

[0107] To verify the effectiveness of the fusion method in improving the oscillation problem, a cooperative navigation task for unmanned surface vessels (USVs) was set up in the experimental scenario. The formation adopted a triangular configuration, with the lead vessel as the vertex and the two follower vessels maintaining a relatively fixed offset distance on either side behind it. The lead vessel's starting point was (0,0) and the target point was (7.5,10). The experimental environment was set up with a complex obstacle zone consisting of 16 obstacles to simulate real navigation challenges.

[0108] When the unmanned surface vessels (USVs) convoy navigated into an area densely packed with obstacles, the artificial potential field method exhibited significant oscillations. Particularly near the narrow passage formed by obstacles at coordinates (4,7), (4.8,5), and (5.2,6), the USVs were subjected to repulsive forces from multiple obstacles, gravitational forces from the target point, and formation-maintaining forces and repulsive forces between USVs, causing them to oscillate repeatedly and become unable to move steadily forward in this area. Figure 6 As can be seen in (a), the trajectory of follower boat 2 is obviously sawtooth-shaped and cannot move forward stably under the combined influence of environmental forces and inter-boat interactions. In (b), by introducing the dynamic window method for fusion optimization, the system can dynamically evaluate and select the optimal speed range under the premise of considering the kinematic constraints of the unmanned surface vessel, thereby effectively guiding the unmanned surface vessel to smoothly pass through the oscillation region and solving the local oscillation problem under the traditional method.

[0109] To verify the effectiveness of the fusion method in improving the local minimum problem, an obstacle (6,6) was placed on the line connecting the starting point (0,0) and the target point (10,10) of the pilot boat in the experimental scenario. When the pilot boat sailed near the obstacle, the attraction and repulsion forces of the artificial potential field reached equilibrium, and the pilot boat got stuck in a local minimum point, unable to continue planning its route and could only stop moving forward.

[0110] exist Figure 7 In (a), when the unmanned surface vessel (USV) reaches the local minimum point (5.66, 5.66), the gravitational and repulsive forces reach equilibrium, causing the USV to stop moving and fail to reach the target point. In (b), however, the fusion algorithm of this invention successfully solves this problem; the USV does not get trapped in the local minimum dilemma and ultimately reaches the target smoothly. The experimental results fully demonstrate the effectiveness of the fusion method for solving local minimum problems, significantly improving the reliability and effectiveness of USV navigation and possessing higher practical value.

[0111] To test the path planning capability of the Bessel optimization-based fusion algorithm, simulations were performed, and the path length and smoothness were statistically analyzed as indicators for comparison with the improved artificial potential field method to evaluate the effectiveness and reliability of the path planning method. In this experiment, the smoothness index is the sum of the changes in the unmanned surface vessel's heading angle.

[0112] The simulation used three unmanned surface vessels (USVs). The starting coordinates of the lead USV were (0,0) and the ending coordinates were (11,14). The starting coordinates of follower USV 1 were (-1,-1) and the ending coordinates were (10,13). The coordinates of follower USV 2 were (1,-1) and the ending coordinates were (12,13). The coordinates of the obstacles were (3.5, 4.0), (3.0, 2.0), (5.5, 6.0), (4.0, 8.0), (7.0, 5.0), (10.0, 10.0), (11.0, 15.1), and (6.0, 9.0). The initial velocity and direction were both set to 0.

[0113] The simulation parameters are shown in Table 1.

[0114] Table 1

[0115] The evaluation indicators for the two methods are shown in Table 2.

[0116] Table 2

[0117] Table 2 shows the statistical table of evaluation indicators for the two methods. Compared with the improved artificial potential field method, the fusion path planning algorithm based on Bezier optimization improves the path smoothness of the pilot boat, follower boat 1, and follower boat 2 by 90.78%, 84.85%, and 94.32%, respectively. Figure 8 In (a), the unmanned surface vessel (USV) is near an obstacle, resulting in a significant inflection point and a drastic change in the heading angle of the path. In (b), the path planned by the fusion algorithm based on Bezier optimization does not have obvious inflection points and is generally smoother. In particular, follower vessel 2 exhibits the greatest smoothness in the improved artificial potential field method, with the least smooth path and the most significant improvement effect, achieving a smoothness improvement of up to 94.32%. This indicates that the fusion algorithm can effectively overcome the shortcomings of the improved artificial potential field method in path planning and greatly optimizes the navigation path. The significant improvement in path smoothness means a more stable and safer navigation process, reducing unnecessary turns and turbulence, which helps the USV complete its mission more efficiently, highlighting the effectiveness and practicality of the fusion path planning algorithm based on Bezier optimization.

[0118] Of course, the above description is only a preferred embodiment of the present invention. The present invention is not limited to the above-described embodiments. It should be noted that any equivalent substitutions or obvious modifications made by those skilled in the art under the guidance of this specification fall within the scope of this specification and should be protected by the present invention.

Claims

1. A method for obstacle avoidance in unmanned surface vessel swarms that integrates artificial potential field method and dynamic window method, characterized in that... Includes the following steps: S1. Divide the unmanned surface vessel (USV) cluster into a lead vessel and several follower vessels. Both the lead vessel and the follower vessels are equipped with lidar and GNSS modules to acquire information and construct a unified geographic coordinate system of USV-obstacle-target point. S2. The improved artificial potential field method is used to calculate the resultant force on the pilot boat, and the motion trend direction of the unmanned surface vessel swarm is determined by the resultant force, and the reference speed of the pilot boat is determined. S3. Determine the optimal speed of the pilot boat using the improved dynamic window method; S4. The reference speed and the optimal speed of the pilot boat are fused by adaptive weights to obtain the fused speed of the pilot boat, and the position of the pilot boat is updated with the fused speed of the pilot boat. S5. Repeat S2-S4 to obtain the fusion speed of each following vessel and complete the obstacle avoidance of the unmanned vessel swarm.

2. The obstacle avoidance method for unmanned surface vessels (USVs) swarms that integrates the artificial potential field method and the dynamic window method according to claim 1, characterized in that, S1 includes the following steps: S11. Perform a 360° omnidirectional scan using lidar to collect point cloud data within the area surrounding the lead boat and follower boats; S12. Preprocess the point cloud data to obtain preprocessed point cloud data; S13. Based on the preprocessed point cloud data, construct a local environmental grid map shared by the unmanned surface vessel cluster, and extract feature information of static and dynamic obstacles. S14. First, establish a global coordinate system, and then use the GNSS module to obtain the current position, speed, heading angle and yaw rate of the lead boat and follower boat respectively, and construct a unified geographic coordinate system of unmanned boat, obstacle and target point.

3. The obstacle avoidance method for unmanned surface vessels (USVs) swarms, integrating the artificial potential field method and the dynamic window method according to claim 2, is characterized in that... S2 includes the following steps: S21. Divide the area of ​​influence between the pilot boat and the obstacle into a danger zone, a controllable zone, and a non-impact zone; Dangerous area: Controllable area: Unaffected area: ; In the formula: The coordinates of the pilot boat's current position in the global coordinate system. The range of influence of the obstacle. Distance to the collision hazard zone of the obstacle. The Euclidean distance between the pilot boat and the obstacle; S22. Divide the area of ​​effect between the pilot vessel and the target point into near range and long range; Near range: Long distance: ; L is the threshold value of the effective range between the pilot boat and the target point. These are the position coordinates of the target point in the global coordinate system. The Euclidean distance between the pilot boat and the target point; S23. The repulsive potential field function acting on the pilot boat is: ； In the formula, Let be the repulsive potential field function acting on the pilot boat. , This is a weighted index representing the rate of change of the intensity of the repulsive potential field with respect to the distance between the pilot vessel and the target point. The coefficient of the repulsive potential field; The repulsive force experienced by the pilot boat is: ； In the formula, The repulsive force experienced by the pilot boat. The gradient operator represents the repulsive potential field function. Find the gradient. The basic obstacle avoidance repulsion force component is related to the distance to the obstacle; The corrected repulsive force component is related to the distance to the target point; S24. When within close range of the target point, i.e. ; When the pilot boat is within a controllable area, that is hour; ； When in a danger zone, i.e. hour; ； in, The direction is from the obstacle towards the pilot boat. The direction is from the pilot boat towards the target point; pass The value of to Control is exercised to control the direction of the resultant force; when the pilot boat is in a dangerous area, control is exercised. ; S25. When at a distance from the target point, i.e. ; When it is within a controllable area, that is hour; ； When in a danger zone, i.e. hour; ； When the pilot boat is in a danger zone, it needs to increase the repulsive force in order to better avoid obstacles. At that time, it should meet the following requirements. .

4. The unmanned surface vessel swarm obstacle avoidance method integrating artificial potential field method and dynamic window method according to claim 3, characterized in that, S2 also includes the following steps: S26. Obstacle avoidance of unmanned surface vessels swarms based on an improved artificial potential field method; The combined force on the pilot boat Represented as: ； In the formula, The attraction of the target point to the pilot boat, The repulsive force experienced by the pilot boat; S27. Calculate the reference speed of the pilot boat; 1； In the formula, For the reference speed of the pilot boat, The maximum speed that the pilot boat can reach; 1 represents the unit direction vector. , The combined force experienced by the pilot boat, It is the modulus of the resultant force.

5. The unmanned surface vessel swarm obstacle avoidance method according to claim 4, characterized in that, S3 includes the following steps: S31, The speed search space is: ； In the formula, Let be the linear velocity component of the unmanned surface vessel in the x-direction in the global coordinate system. Let be the linear velocity component of the unmanned surface vessel in the y-direction in the global coordinate system; This is the maximum speed that the unmanned surface vessel can achieve. S32, The evaluation function is: ； In the formula, Let be the linear velocity component of the unmanned surface vessel in the x-direction in the global coordinate system. Let be the linear velocity component of the unmanned surface vessel in the y-direction in the global coordinate system; It is the cost function for the change in heading angle. Let the target distance cost function be... It is a speed cost function. It is the obstacle collision cost function. , , d and d represent the weights of each part, respectively. This indicates normalization processing; The cost function for the change in heading angle is: ； In the formula, It is the heading angle for predicting speed. It is the heading angle at the current speed; The target distance cost function is: ； In the formula, For the reason The simulated trajectory end position The coordinates of the target point; The speed cost function is: ； In the formula, v is the magnitude of the linear velocity of the currently sampled unmanned surface vessel; The obstacle collision cost function is: ； In the formula, Indicates the distance to the nearest obstacle. Indicates the safe distance from obstacles; S33. Generate all feasible velocity pairs in the velocity search space. For each candidate velocity pair among all feasible velocity pairs, calculate all cost function values ​​and substitute them into the evaluation function to calculate the comprehensive score. Select the velocity pair with the highest comprehensive score as the optimal speed of the unmanned surface vessel. When the unmanned surface vessel is the lead vessel, it obtains the optimal speed of the lead vessel.

6. The unmanned surface vessel swarm obstacle avoidance method according to claim 5, characterized in that, In S4, speed fusion is performed using adaptive weights: ； In the formula: The reference speed for the pilot boat; The optimal speed for the pilot boat; For fusion weights; The fusion speed of the pilot boat; According to the fusion speed of the pilot boat Update the position of the lead boat.

7. The unmanned surface vessel swarm obstacle avoidance method according to claim 6, characterized in that, In S5: The net force acting on a follower vessel in an unmanned surface vessel swarm is the attractive force exerted by the target point on the follower vessel. The attraction generated by the pilot boat The repulsive force generated by the pilot boat The repulsive force generated by the remaining follower boats and the repulsive force exerted by the obstacle on the following vessel. The resultant force of these five forces is as follows: ； In the formula, This represents the net force acting on a particular follower vessel. It is the vector sum of the repulsive forces exerted on the current following vessel by the other k following vessels, where k refers to the number of other following vessels in the unmanned surface vessel cluster besides the current following vessel; Calculate the reference speed of the following vessel: ； In the formula, For the reference speed of the following boat, This is the maximum speed that the follower boat can achieve; It is a unit direction vector. , The net force acting on a particular follower vessel. It is the modulus of the resultant force; Repeat S3. When the unmanned surface vessel is a follower vessel, obtain the optimal speed of the follower vessel. Repeat step S4 to perform velocity fusion, obtain the fused velocity of the following vessel, and update the position of the following vessel based on the fused velocity of the following vessel.

8. The unmanned surface vessel swarm obstacle avoidance method according to claim 3, characterized in that, The weighting index and the repulsive potential field coefficient constitute the repulsive adjustment coefficient, which is determined by the following steps: (1) Set initial parameters, including the initial repulsive potential field coefficient. Initial weight index , ; (2) In a scenario where only the target's gravitational field exists, observe whether the pilot boat can successfully reach the target point: if it cannot reach the target point and oscillates near the target point, it indicates that the repulsive force decays too slowly when approaching the target point, and needs to be increased. ; If there is no repulsive force during the arrival process and the path is entirely dominated by gravity, it indicates that the overall repulsive force is too weak and needs to be increased. ; (3) Observe the obstacle avoidance behavior of the pilot boat in the presence of obstacles: If the pilot boat gets too close to the obstacle and there is a risk of collision, it means the repulsive force is too weak and needs to be increased. Or adjust ; If the pilot boat deviates significantly from the overall path or veers excessively, it indicates that the repulsive force is too strong and needs to be reduced. Or adjust , ; (4) Compare different parameter combinations and select the set of repulsive potential field coefficients with the best overall performance. and weight index , This serves as the final value, thus determining the repulsion adjustment coefficient.

9. The unmanned surface vessel swarm obstacle avoidance method according to claim 8, characterized in that, To prevent collisions between unmanned surface vessel (USV) swarms, a repulsion rule is added. The rule is designed as follows: each USV establishes a radius of... The collision area and radius are The collision hazard zone will have a radius of [missing information]. The annular region serves as a collision buffer zone; As the unmanned surface vessels approach each other, the hulls of the unmanned surface vessel cluster... Located on the hull Outside the collision buffer zone, there is no repulsive force between the two; when the hull... Located in the hull When within the collision buffer zone, the repulsive force increases linearly with decreasing distance; when the hull... Located in the hull When the collision hazard zone is reached, the repulsive force increases rapidly in an exponential relationship. The specific design expression for the repulsive force is as follows: ； In the formula, The repulsive force coefficient, This represents the position vector of hull i in the global coordinate system. This represents the position vector of hull j in the global coordinate system. This is the effective boundary of the inter-boat repulsive force. This is the boundary distance between the collision buffer zone and the danger zone between the vessels. Let be the Euclidean distance between hull i and hull j. It is a piecewise adjustment function; function The expression is: ； In the formula, This is the radius of the safe zone for the hull; anything smaller than this zone is considered a collision. It is the radius of the range of the repulsive force between the boats. Let i be the actual distance between hull i and hull j; assume that the unmanned surface vessel cluster is related to the hull i and hull j. The hulls that communicate with each other have One, then the hull The repulsive force generated by the other hulls in the unmanned surface vessel cluster is The resultant force of the three forces is: ； Using this as a constraint for the motion control of unmanned surface vessels (USVs), the USV is guided to simultaneously avoid all adjacent USVs, thus achieving safe and coordinated navigation of the USV swarm.

10. The obstacle avoidance method for unmanned surface vessels (USVs) swarms, integrating the artificial potential field method and the dynamic window method according to claim 9, is characterized in that... Also includes S6: Bézier curves are used to perform post-processing optimization on the planned path; Given a set of control points The formula for the Bézier curve is shown below: ； In the formula, Let be a point on the Bézier curve. For time, As control points, Let be the combination number, denoted as: ； By downsampling the control points, we obtain: ； in, The sampling interval is... These are the control points after downsampling; The final smoothed trajectory is shown in the following formula: 。

Images