Multi-level planning and control method for cooperative formation of unmanned watercraft swarm

Abstract

A multi-level planning and control method for the collaborative formation of a swarm of unmanned boats is proposed. Based on the motion characteristics of a single unmanned boat in each degree of freedom, a single boat kinematic model is established. Based on the obstacle avoidance requirements, the A* algorithm is used to dynamically replan the path of the single boat to obtain the path generated by the global path planning. A method for forming a swarm of unmanned boats is established. Using the path generated by the global path planning as a reference line, the planned path points of each unmanned boat in the swarm are generated by an improved Lattice planning algorithm. A single boat obstacle avoidance model is constructed between swarms. When the distance between boats or between boats and obstacles is too close, the obstacle avoidance model is used to ensure that each unmanned boat in the swarm can maintain a safe distance. The present invention plans and controls the motion of a swarm of unmanned boats from three levels: global path planning, local cluster control, and a single boat obstacle avoidance algorithm. Compensation measures are added to the trajectory evaluation. The obstacle avoidance requirements of unmanned boats between swarms are addressed through an unmanned boat swarm obstacle avoidance algorithm based on the improved artificial potential field method.

Description

Technical Field

[0001] The present invention relates to a technology in the field of unmanned boat track planning, specifically a multi-level planning and control method for the coordinated formation of an unmanned boat cluster. Background Art

[0002] Autonomous maritime operations based on unmanned vehicles (USVs) have attracted widespread attention worldwide in recent years. USVs offer advantages such as small size, high speed, and flexible deployment. Since they require no direct human control, they significantly improve the safety of maritime operations and have excellent application potential in scenarios such as maritime rescue and marine surveys. Compared to a single USV, a swarm of USVs can overcome the shortcomings of a single vessel's lack of reliability and limited operational capabilities, offering enhanced mission capabilities. The main challenge with swarms is the complexity of coordination between USVs, which manifests itself in two main aspects: first, the unmanned vehicles within a swarm must meet collision avoidance requirements, and second, the USVs within a swarm must maintain a relatively stable relative distance. Therefore, USV formation technology plays a key role in improving the overall effectiveness of USV swarms. Summary of the Invention

[0003] In view of the shortcomings of the existing technology that cannot handle the real-time update of upper-level path planning and does not provide a solution for the collision avoidance requirements between boat groups, the present invention proposes a multi-level planning and control method for the collaborative formation of unmanned boat clusters. The movement of the unmanned boat cluster is planned and controlled from three levels: global path planning, local cluster control, and single-boat obstacle avoidance algorithm. The problem of large amount of linear interpolation calculation for lattice planning is solved through reduced-order planning, and compensation measures are added in trajectory evaluation. The obstacle avoidance requirements of unmanned boats between clusters are solved through an unmanned boat cluster obstacle avoidance algorithm based on the improved artificial potential field method.

[0004] The present invention is achieved through the following technical solutions:

[0005] The present invention relates to a multi-level planning and control method for the collaborative formation of a cluster of unmanned boats. According to the motion characteristics of a single unmanned boat in each degree of freedom, a single boat kinematic model is established. Based on the obstacle avoidance requirement, the path of the single boat is dynamically replanned using an A* algorithm to obtain a path generated by global path planning. An unmanned boat cluster model is established. The path generated by the global path planning is used as a reference line, and the planned path points of each unmanned boat in the cluster are generated using an improved Lattice planning algorithm. An obstacle avoidance model for single boats between clusters is constructed. When the distance between boats or between boats and obstacles is too close, the obstacle avoidance model is used to ensure that each unmanned boat in the cluster can maintain a safe distance.

[0006] The present invention relates to a multi-level planning and control system for implementing the above-mentioned method, comprising: an unmanned boat single boat simulation module, a cluster planning algorithm module, and a system configuration module, wherein: the unmanned boat system configuration module sets environmental information and outputs it to the cluster planning algorithm module; the cluster planning algorithm module calculates a global optimal path and a local trajectory smoothing path based on the environmental information to obtain global and local trajectory reference paths; the single boat simulation module calculates the real-time path of each unmanned boat in the unmanned boat cluster based on the trajectory reference path and according to the unmanned boat dynamic constraints and collision avoidance constraints.

[0007] Technical Effects

[0008] The present invention takes into account both global optimality and local constraints, and uses global path planning as a reference line for local path planning, which can reflect the real trajectory characteristics of the unmanned boat cluster. The local trajectory planning based on the improved Lattice can improve the real-time performance of the trajectory planning of the unmanned boat cluster, and the trajectory replanning time is no more than 0.5s. At the same time, taking into account the maneuverability and energy consumption constraints of the unmanned boat, the generated reference trajectory avoids excessive changes in the amplitude of the unmanned boat's action output. Single-boat obstacle avoidance takes into account the collaborative relationship between the unmanned boats in the cluster. The proposed method makes the control signal output to the power system smoother, which can better meet the real-time obstacle avoidance needs between unmanned boat clusters. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 Flowchart of the present invention;

[0010] Figure 2 Schematic diagram of the dynamic model of a single unmanned boat in the embodiment;

[0011] Figure 3 A* algorithm path dynamic replanning schematic diagram in the embodiment;

[0012] Figure 4 Schematic diagram of the leader-follower formation mode of the unmanned boat cluster in the embodiment;

[0013] Figure 5 This is a schematic diagram of the input and output of the Lattice planner in the embodiment;

[0014] Figure 6 Schematic diagram of the relationship between different coordinate systems of Lattice planning in the embodiment;

[0015] Figure 7 Schematic diagram of obstacle avoidance using the improved artificial potential field method for unmanned boat swarms in the embodiment;

[0016] Figures 8-10 Schematic diagram of the embodiment effect. DETAILED DESCRIPTION

[0017] like Figure 1As shown, this embodiment relates to a hierarchical planning and control method for a coordinated formation of an unmanned boat swarm, including:

[0018] Step 1) Create Figure 2 The unmanned boat single boat model shown in the figure: For unmanned boats, the motion of the three degrees of freedom of heave, roll and pitch is usually ignored. The motion of the unmanned boat is generally described in two coordinate systems: the earth coordinate system X is selected E -Y E As a reference coordinate system, the coordinate system X fixed on the unmanned boat B -Y B As the coordinate system of the unmanned boat, in the figure: ψ i is the bow angle, v i =[u i , v i , r i ] is the linear velocity and angular velocity vector of the i-th unmanned boat, u i and υ i X B -Y B The longitudinal and transverse velocities in the coordinate system, r i is the yaw angular velocity, β i is the sideslip angle.

[0019] Considering a swarm system consisting of N unmanned boats, the kinematic and dynamic equations of the i-th unmanned boat can be described by a three-degree-of-freedom nonlinear model: η i =R(ψ i )v i , η i =[x i ,y i , ψ i ] is the position and bow angle of the i-th unmanned boat in the reference coordinate system, τ i is the control input, τ dis is the disturbance input caused by wind, waves and ocean environment, g i (v i ) including unmodeled fluid dynamics and modeling errors, the rotation matrix of the UAV coordinate system relative to the reference coordinate system Inertial mass matrix Centripetal force and Coriolis force matrix Damping Matrix m ij ,c ij ,d ij are all constant coefficients. The sideslip angle β generated by the drift speed of the unmanned boat during navigation is i =atan2(υ i ,u i ).

[0020] The control input satisfies the constraint: -τ ic ≤τ i ≤τ ic , where: τ ic is τ i upper and lower limits.

[0021] Step 2) Dynamically replan the path of a single boat using the A* algorithm to obtain the path generated by the global path planning: Figure 3 As shown in the figure, according to the coordinates of the starting point, the target point, the obstacle coordinates and the grid size occupied, the A* algorithm is used to select the successor point by minimizing the sum of the actual cost and the heuristic cost, and a collision-free path point from the starting point to the target point is obtained; and when the unmanned boat detects surrounding obstacles, the obstacle position and the size of the occupied map grid are added to the environmental information according to the perception and recognition results. At this time, the A* algorithm is called again to calculate a new collision-free planning path until no new obstacles are detected.

[0022] Step 3) Figure 4 As shown in the figure, a Leader-Follower model of an unmanned boat cluster is established: taking two unmanned boats as an example, are the position and heading of the pilot boat and the following boat respectively, l is the relative distance between the pilot boat and the following boat, l x and l y is the longitudinal and transverse distance between the two boats, and are the headings of the pilot boat and the following boat respectively. In actual navigation, when the position of the pilot boat is given and the relative distance and angle between the two boats are determined, then l x and l y The value of is fixed, that is, the position of the following boat is unique. Therefore, the formation spacing (l x ,l y ) to realize the formation control of multiple unmanned boats, including: Taking the derivative, we get the relative motion equation between the leader and the follower: Then we get the kinematic target Dynamic Target Where: u l , v l is the speed of the pilot boat in the longitudinal and transverse directions, u f , v f is the speed of the following boat in the longitudinal and transverse directions, r l and r f are the angular velocities of the lead boat and the following boat, are the x-axis and y-axis components of the distance between the two boats, e x and e yis the distance tracking error in the x and y directions, so that during the march, the relative positions of the boats in the formation remain unchanged as much as possible, and the motion error with the leader boat is as small as possible.

[0023] The Leader-Follower model of the unmanned boat cluster further includes a safety distance constraint, specifically: v -l obs ||≥d safe , where: d safe is the predetermined safety distance, l v is the x,y coordinate of the unmanned boat v, l obs The x, y coordinates of the obstacle boundary are used to ensure that the motion paths of the boats coincide with each other during the longitudinal formation, while maintaining a certain safety distance from the surrounding obstacles.

[0024] Step 4) Figure 5 As shown in the figure, the path generated by the global path planning in step 2 is used as the reference line (Referenceline), and the improved Lattice planner is used to complete the local trajectory planning of the unmanned boat. That is, after the unmanned boat coordinate system is converted to the Frenet coordinate system, the d-axis and s-axis of Frenet are planned in the Frenet coordinate system respectively to form the planned trajectory in the Frenet coordinate system. Then, the trajectory in the Frenet coordinate system is synthesized into the world coordinate system and restored to the trajectory in the world coordinate system. Specifically, it includes:

[0025] 4.1) If Figure 6 As shown in the figure, the current position information of the unmanned boat is converted to the Frenet coordinate system to obtain the initial state of the unmanned boat in the Frenet coordinate system; the forward distance is calculated according to the current speed to obtain the forward point and the target state of the unmanned boat at the forward point in the Frenet coordinate system; the coordinate point P of the unmanned boat is projected onto the reference line to obtain a projection point R on the reference line. The path length from the starting point of the reference line to the projection point is the longitudinal offset s of the unmanned boat in the Frenet coordinate system, and the distance l(s) from the projection point R to the position of the unmanned boat is the lateral offset of the unmanned boat in the Frenet coordinate system.

[0026] In this embodiment, the lateral offset l(s) is designed to be a function of the longitudinal offset s. The purpose of converting the coordinate points to the Frenet coordinate system is to facilitate the generation of planning curves and trajectory sampling in the lateral and longitudinal directions of the road line, thereby obtaining a smooth sampling trajectory covering the entire waterway.

[0027] 4.2) Using the trajectory running time t, target speed v, and lateral displacement d from the reference line as planning parameters, the sampling state is obtained. The Lattice planner performs lateral sampling and longitudinal sampling to form different lateral offsets, as well as trajectory time period sampling to sample the trajectory running cycle time. Specifically, the lateral offset l(s) and longitudinal offset s(t) are generated by polynomial fitting and solving, and the planning function is generated based on the sampling state. The longitudinal fifth-order polynomial function is:

[0028] Constraint function: Where: s(t), v(t) and a(t) are the longitudinal offset, velocity and acceleration of the UAV at time t, respectively; c1~c6 are the coefficients of linear interpolation, respectively; s(t0), v(t0), a(t0) are the position, velocity and acceleration of the UAV at the initial moment of trajectory planning, respectively; s(t1), v(t1), a(t1) are the longitudinal offset, velocity and acceleration of the UAV at the end moment of trajectory planning, respectively.

[0029] This embodiment adopts a downsampling method, that is, a cubic function is used to fit the trajectory, and only the velocity and trajectory are fitted to ensure the smoothness of the generated trajectory. For acceleration, a penalty term is added to the objective function of the trajectory evaluation to constrain the acceleration of the unmanned boat during the execution of the action, and try to ensure that its changes will not be too fast. The expression of the downsampling curve is: s(t) = c1t 3 +c2t 2 +c3t+c4,v(t)=3c1t 2 +2c2t 1 +c3; Constraints: Among them: a max It is the maximum acceleration of the unmanned boat and is related to the performance of the unmanned boat itself.

[0030] 4.3) Construct polynomial planning functions l(s) and s(t) for the lateral and longitudinal offsets. After obtaining the planning functions for the lateral and longitudinal displacements, perform time interpolation to obtain the trajectory points in the reference line Frenet coordinate system. Finally, convert the trajectory points from the Frenet coordinate system to the Cartesian coordinate system to obtain the physical world sampling trajectory.

[0031] Since both the horizontal and vertical directions are obtained through high-order polynomial interpolation, the trajectory in the Cartesian coordinate system is also smooth.

[0032] 4.4) The trajectory evaluation function is used to score the collision detection, curvature constraints, and optimal trajectory of the sampled trajectory to obtain an optimal collision-free smooth trajectory that meets the constraints. This is the trajectory that the Lattice outputs to the controller for the unmanned vehicle to follow.

[0033] The trajectory evaluation function is: Where: s(t), a(t), j(t) are the position, acceleration and jerk of the unmanned boat at time t, s opt , a0, j0 are the reference position, acceleration and jerk of the unmanned boat, respectively, which are obtained by linear interpolation. is a constant coefficient. k lat Cost lat is the sum of the deviations between the actual speed / acceleration / jerk of the UAV and the ideal speed / acceleration / jerk. R is the obstacle threat radius, x(t), y(t) are the x and y coordinates of the UAV at time t, x obs ,y obs is the obstacle coordinate, k obs Cost obs is the distance penalty between the UAV and the obstacle. This term is positive when the UAV enters the obstacle threat radius R. After generating the desired output trajectory, the UAV's underlying controller invokes PID control to ultimately generate the UAV's control signal, including jerk (propeller output power) and angular velocity (rudder angle).

[0034] Step 5) An improved artificial potential field method is used to design an obstacle avoidance algorithm for the unmanned boat. A single-boat obstacle avoidance model is established to avoid the problem that the unmanned boat may deviate from the reference path and get too close to other boats during the actual simulation process. Specifically, the starting point, end point, obstacles, and robots are simulated into artificial potential fields. The movement of the robot is abstracted into moving particles in an electronic potential field with different signs or two magnetic poles in a magnetic field. At the same time, the collaborative obstacle avoidance actions of multiple unmanned boats are considered to complete the obstacle avoidance action at the minimum cost.

[0035] For example, if both the unmanned boat and the obstacle particles are positively charged, the obstacle will produce a repulsive potential field on the unmanned boat; the target will produce a gravitational potential field on the unmanned boat, then the combined potential field generated by the obstacle and the target will cause the unmanned boat to move toward the target and spontaneously avoid the obstacle.

[0036] like Figure 7 As shown, the repulsive potential field is Where: λ1>0 is a constant, d(X ij )=||X i -X tar || is the distance between the unmanned boat i and the target point X tar distance. is the impact distance of the ith obstacle, d ij (X ij ) is the vertical distance from the unmanned boat to the obstacle. The potential field force is proportional to the negative gradient of the potential field function, and the potential field attraction and potential field repulsion are: Among them: Δd is to consider the power output of the unmanned boat in the actual engineering application process as linear as possible, set a certain operation dead zone, that is, in the dead zone, the power output and rudder angle of the unmanned boat are not adjusted. Considering the driving force of the unmanned boat in the forward direction The combined power of unmanned boats For a formation cluster consisting of more than two unmanned boats, the resultant force expression is similar, that is, considering the superposition of the potential field force vectors between multiple unmanned boats:

[0037] After specific actual experiments, this method was simulated on a simulation platform, and 5 unmanned boats were set to form an unmanned boat formation. The initial position was located in a circle with a radius of 150 meters and (200, 200) as the center in the Cartesian coordinate system. The target point that the unmanned boats headed for was (3200, 5000), and the cruising speed of the unmanned boat cluster was set to 6 knots. The unmanned boat cluster was set to adopt a diamond formation in open waters. In terms of environmental parameters, the scene of the simulated boat group passing through a narrow waterway was set with an obstacle threat radius of 100 meters. The 5 unmanned boats needed to pass through the narrow waterway before heading to the target point. Running the simulation program, the test results obtained are as follows Figures 8-10 As shown. Figure 8 It can be seen that the unmanned boat swarm can maintain a basically unchanged formation spacing in open waters. After entering a narrow waterway, the unmanned boat swarm cannot pass in parallel due to the need to meet the constraints of obstacle avoidance and collision avoidance between boats. Therefore, it adopts a straight line to pass through the narrow waterway. Figure 9 It can be seen that the distance between the boats is kept within a certain range by improving the artificial potential field method, and the speed of the boats is basically consistent. Figure 10 It can be seen that in open waters, the unmanned boat group returns to the diamond formation. According to the reference trajectory given by the proposed trajectory planning method, the unmanned boats smoothly switch the formation according to the reference trajectory.

[0038] Compared with the existing technology, the present invention adopts the idea of ​​hierarchical planning and control, combining upper-level path planning, middle-level cluster planning, trajectory planning and bottom-level single-boat obstacle avoidance control. Under the premise of global path optimization, the local path is planned with the global path as a reference, and the generated local path is more practical. The present invention simplifies and improves the Lattice algorithm, improves the speed of local path planning, and at the same time, considering the problem of insufficient smoothness of acceleration caused by linear interpolation reduction, designs corresponding penalty terms in the evaluation function to ensure that the control action of the unmanned boat does not change drastically. The present invention considers the synergistic effect of unmanned boats between clusters, improves the traditional artificial potential field method, and proposes a method for calculating the potential field force of unmanned boats between clusters. The proposed algorithm can reflect the potential field force relationship between the cluster and the unmanned boat, and guide the obstacle avoidance action of each unmanned boat. The proposed method distributes the obstacle avoidance action to all relevant unmanned boats in the cluster, optimizes the action amplitude of a single unmanned boat, and makes the output of the unmanned boat smoother.

[0039] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principles and purpose of the present invention. The scope of protection of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. All implementation schemes within its scope shall be subject to the constraints of the present invention.

Claims

1. A multi-level planning and control method for the coordinated formation of a swarm of unmanned boats, characterized in that: include: Step 1) Establish a single unmanned boat model: For the unmanned boat, ignore the motion of the three degrees of freedom of heave, roll and pitch, and describe the motion of the unmanned boat in two coordinate systems: select the earth coordinate system X E -Y E As a reference coordinate system, the coordinate system X fixed on the unmanned boat B -Y B As the UAV coordinate system, consider a cluster system consisting of N UAVs. The kinematic and dynamic equations of the i-th UAV are described by a three-degree-of-freedom nonlinear model: , ,in: is the bow angle, are the linear velocity and angular velocity vectors of the i-th unmanned boat, and X B -Y B The longitudinal and transverse velocities in the coordinate system are: is the yaw angular velocity, is the sideslip angle, is the position and bow angle of the i-th unmanned boat in the reference coordinate system, is the control input, is the disturbance input caused by wind, waves and ocean environment, Including unmodeled fluid dynamics and modeling errors, the rotation matrix of the UAV coordinate system relative to the reference coordinate system , inertial mass matrix , centripetal force and Coriolis force matrices , the damping matrix , m ij, c ij , d ij are all constant coefficients. The sideslip angle generated by the drift speed of the unmanned boat during navigation is ; Step 2) Dynamically replan the path of the single boat using the A* algorithm to obtain a path generated by global path planning: Based on the coordinates of the starting point, the coordinates of the target point, the coordinates of the obstacle, and the grid size occupied, the A* algorithm is used to select successor points by minimizing the sum of the actual cost and the heuristic cost to obtain a collision-free path from the starting point to the target point. When the unmanned boat detects surrounding obstacles, the obstacle location and the grid size occupied by the obstacle are added to the environmental information based on the perception and recognition results. At this time, the A* algorithm is re-calculated to calculate a new collision-free planned path until no new obstacles are detected. Step 3) Establish an unmanned boat cluster model: Take two unmanned boats as an example, ( )、( ) are the positions and headings of the pilot boat and the following boat respectively, is the relative distance between the lead boat and the following boat, and is the longitudinal and transverse distance between the two boats, and are the headings of the pilot boat and the following boat respectively. In actual navigation, when the position of the pilot boat is given and the relative distance and angle between the two boats are determined, then and The value of is fixed, that is, the position of the following boat is unique, so by adjusting the formation spacing ( , ) to realize the formation control of multiple unmanned boats, including: , and the derivative of it gives the relative motion equation between the leader and the follower: ; and then get the kinematic target , dynamic target ,in: is the speed of the pilot boat in the longitudinal and transverse directions, To follow the speed of the boat in the longitudinal and transverse directions, and are the angular velocities of the lead boat and the following boat, are the x-axis and y-axis components of the distance between the two boats, and is the distance tracking error in the x and y directions, so that the relative positions of the boats in the formation remain unchanged during the march, and the motion error with the leader boat is small; Step 4) Using the path generated by the global path planning in step 2 as a reference line, the improved Lattice planner is used to complete the local trajectory planning of the unmanned vehicle. That is, after converting the unmanned vehicle coordinate system to the Frenet coordinate system, the d-axis and s-axis of the Frenet coordinate system are planned in the Frenet coordinate system to form a planned trajectory in the Frenet coordinate system. The trajectory in the Frenet coordinate system is then synthesized into the world coordinate system and restored to the trajectory in the world coordinate system. Step 5) An improved artificial potential field method is used to design an obstacle avoidance algorithm for the unmanned boat and establish a single-boat obstacle avoidance model to avoid the problem of the unmanned boat deviating from the reference path and being too close to other boats during the actual simulation process. Specifically, the starting point, end point, obstacles, and robots are simulated into an artificial potential field, and the robot's motion is abstracted into a moving particle in an electronic potential field with different signs or two magnetic poles in a magnetic field. At the same time, the collaborative obstacle avoidance actions of multiple unmanned boats are considered to complete the obstacle avoidance action at the minimum cost.

2. The multi-level planning and control method for cooperative formation of a swarm of unmanned boats according to claim 1 is characterized in that: The step 4 specifically includes: 4.1) Convert the current position information of the UAV to the Frenet coordinate system to obtain the initial state of the UAV in the Frenet coordinate system; calculate the forward distance based on the current speed to obtain the forward point and the target state of the UAV at the forward point in the Frenet coordinate system; project the coordinate point P of the UAV onto the reference line to obtain a projection point R on the reference line. The path length from the starting point of the reference line to the projection point is the longitudinal offset s of the UAV in the Frenet coordinate system, and the distance from the projection point R to the position of the UAV is is the lateral offset of the unmanned boat in the Frenet coordinate system; 4.2) Using the trajectory running time t, target speed v, and lateral displacement d to the reference line as planning parameters to obtain the sampling state, the Lattice planner performs lateral sampling, longitudinal sampling, and trajectory time period sampling to generate different lateral offsets. Specifically, the lateral offset generated by polynomial fitting is generated based on the sampling state. and longitudinal offset And planning function, the longitudinal fitting 5th-order polynomial function is: ; Constraint function: ,in: , and are the longitudinal offset, velocity and acceleration of the unmanned boat at time t, are the coefficients of linear interpolation respectively; , , are the position, velocity and acceleration of the unmanned boat at the initial moment of trajectory planning; , , are the longitudinal offset, velocity and acceleration of the unmanned boat at the end of trajectory planning; 4.3) Construct polynomial planning functions s(t) and d(s) for the lateral and longitudinal offsets. After obtaining the planning functions for the lateral and longitudinal displacements, perform time interpolation to obtain the trajectory points in the reference line Frenet coordinate system. Finally, convert the trajectory points from the Frenet coordinate system to the Cartesian coordinate system to obtain the physical world sampling trajectory. 4.4) The trajectory evaluation function is used to score the collision detection, curvature constraints, and optimal trajectory of the sampled trajectory to obtain an optimal collision-free smooth trajectory that meets the constraints. This is the trajectory that the Lattice outputs to the controller for the unmanned vehicle to follow.

3. A multi-level planning control system for implementing the multi-level planning control method for cooperative formation of a swarm of unmanned boats according to claim 1 or 2, characterized in that: include: Unmanned boat single boat simulation module, cluster planning algorithm module and system configuration module, among which: the unmanned boat system configuration module sets the environmental information and outputs it to the cluster planning algorithm module; the cluster planning algorithm module calculates the global optimal path and the local trajectory smooth path according to the environmental information, and obtains the global and local trajectory reference paths; the single boat simulation module calculates the real-time path of each unmanned boat in the unmanned boat cluster based on the trajectory reference path according to the unmanned boat dynamic constraints and collision avoidance constraints.

Images