Unmanned ship cooperative target surrounding method

Abstract

The invention discloses an unmanned ship cooperative target surrounding method, and belongs to the technical field of unmanned ship cooperative control. According to the method, a YOLOv8 target detection method is adopted to train a data set of a collaborative surrounding target, and a real-time target detection model is obtained; a particle filter is utilized to realize collaborative target path determination and conversion, target state distribution is approximated through weighted particles, particle weights are updated in real time according to recognized target states, weighted averaging is carried out, a target trajectory is finally formed in combination with time recursion, and continuity and robustness of the trajectory are ensured. According to the method, the target needing to be surrounded can be identified, the trajectory of the target is obtained through a particle filter algorithm, and the trajectory is converted into a determined path.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned ship cooperative control, and in particular to an unmanned ship cooperative target encirclement method. Background Art

[0002] As an automated and intelligent offshore operation platform, unmanned surface vehicles (USVs) have developed rapidly in recent years and are widely used in many fields such as marine monitoring, search and rescue, environmental protection, waterway management, and military. The marine environment is complex and changeable. Many tasks are not only costly but also have great safety risks under traditional manned operations. Unmanned ships, with their high efficiency, flexibility, and safety, can perform various complex tasks in harsh marine environments and have become one of the important tools in marine development and management. In recent years, with the rapid development of artificial intelligence, communication technology, sensor technology, and automatic control technology, the autonomy and intelligence level of unmanned ships have been significantly improved. By integrating a variety of advanced technologies, modern unmanned ships have more accurate navigation, environmental perception, target recognition, and autonomous decision-making capabilities. For example, based on the combination of satellite positioning systems and inertial navigation systems, unmanned ships can achieve high-precision positioning in a global range; advanced sensor systems can help unmanned ships perceive obstacles, currents, and meteorological conditions in the marine environment; the application of artificial intelligence algorithms allows unmanned ships to make autonomous decisions in complex scenarios and optimize task execution efficiency. With the increase in task complexity, the operating capabilities of a single unmanned ship become insufficient in some scenarios. The technology of multi-unmanned ship collaborative operation provides an effective solution to this problem. The so-called multi-unmanned ship collaborative operation refers to the process in which multiple unmanned ships share information through a communication network and collectively complete complex tasks through distributed control and collaborative planning according to task requirements. This collaborative method can give full play to the advantages of swarm intelligence and show significant performance improvements when performing large-scale monitoring, encirclement, tracking, search and rescue and other tasks. At present, a variety of unmanned ship collaborative encirclement methods have been proposed, but these methods still have certain limitations when facing complex scenes such as movable dynamic targets. Therefore, there is an urgent need for an unmanned ship collaborative encirclement method that can be used for dynamic target trajectories to prevent the inability to encircle dynamic targets well in complex environments and to complete established tasks.

[0003] The invention patent "A method for collaborative encirclement of multiple unmanned ships" with application publication number CN 118655884A uses a CNN-BiLSTM model based on GWO parameter optimization to predict the trajectory of the escaping ship, and considers the distance cost and turning cost from the pursuit ship to the defense point at the same time, and solves the dynamic target tracking problem through a hierarchical solution method. This method improves the ability of unmanned ships to encircle dynamic targets to a certain extent, but in a complex marine environment, the trajectory prediction of the ship is easily affected by the environment, etc., and there may be certain deviations, resulting in inaccurate tracking and encirclement paths for subsequent unmanned ships.

[0004] The invention patent "An unmanned ship path tracking control method" with application publication number CN 114610023A discretizes a given path to be tracked into a set of sites, and uses a new guidance algorithm to determine the reference point set to be tracked in the next time window according to the model prediction window size and the current longitudinal speed of the unmanned ship, construct a target optimization function, calculate the optimal rudder angle sequence, and finally calculate the rudder angle sequence ship motion system to achieve target path tracking. This method improves the unmanned ship's ability to track the target trajectory to a certain extent, but it can only track a given path and cannot track ships with unknown paths. Therefore, this method still has certain limitations.

[0005] Therefore, how to better identify the trajectory of the target ship that needs to be surrounded and achieve coordinated encirclement of the target is a key issue that needs to be solved urgently at this stage. Summary of the invention

[0006] The purpose of the present invention is to provide an unmanned ship cooperative target encirclement method to solve the related problems existing in the prior art. The present invention can identify the target to be encircled, obtain the trajectory of the target through a particle filter algorithm, and convert it into a determined path.

[0007] In order to achieve the above object, the present invention provides the following scheme:

[0008] An unmanned ship cooperative target encirclement method is used to enable M unmanned ships to cooperatively encircle a target, M≥3, and the unmanned ships maintain communication, including the following steps:

[0009] Step 1: The first unmanned ship collects water surface images and uses the YOLOv8 target detection algorithm to identify and detect targets on the water surface that need to be collaboratively surrounded;

[0010] Step 2: After the first unmanned ship detects the target that needs to be surrounded collaboratively, it obtains the target information in real time, including the position, speed and angular velocity of the target, and then transmits the target information to other unmanned ships;

[0011] Step 3: After each unmanned ship obtains the target information, it predicts the target trajectory based on the particle filter. The specific method is as follows:

[0012] S31, continuously obtain target information at N moments and initialize N particles, where the tth particle is expressed as:

[0013]

[0014] in, represents the state of the target in the i-th dimension at time t, i = 1, 2, 3, 4, i = 1 corresponds to the x component of the target position, i = 2 corresponds to the y component of the target position, and i = 3 corresponds to the initial velocity of the target.

[0015] degrees, i = 4 corresponds to the initial angular velocity of the target;

[0016] S32, initialize the weight of each particle to

[0017] S33, for particle x t To make a forecast update:

[0018]

[0019] Among them, the superscript (1) indicates after the update, the superscript (0) indicates before the update, v t Represented as Gaussian distributed noise;

[0020] S34, update the weight of each particle:

[0021]

[0022] Among them, w t Represents x t The weight of , the superscript (1) indicates after the update, and the superscript (0) indicates before the update, represents the prediction likelihood, that is, the particle from Update to probability;

[0023] S35, normalize the weights of all particles:

[0024]

[0025] in, is the normalized x t The weight of

[0026] S36, weighted average the target position information in all particles to obtain a predicted position:

[0027]

[0028] S37, repeating steps S33 to S36 to obtain a series of predicted positions q0, q1, q2, ..., averaging every 2k+1 consecutive predicted positions to achieve smoothing of the predicted positions, where k represents the size of the smoothing window, and the smoothed predicted positions constitute the predicted target trajectory;

[0029] Step 4: Each unmanned ship tracks the linear velocity of the target through the PID algorithm, and calculates the angular velocity required for encirclement based on the predicted target trajectory:

[0030]

[0031] Where Ξ represents the angular velocity required for encirclement, q t Indicates the current location coordinates of the target. Indicates the current speed of the target, q i represents the position coordinates of the i-th unmanned ship, Ψ i represents the heading angle of the i-th unmanned ship, R i represents the motion radius of the ith unmanned ship, v i represents the moving speed of the i-th unmanned ship;

[0032] Step 5: Each unmanned ship moves according to its own linear velocity and angular velocity to achieve coordinated encirclement of the target.

[0033] The following beneficial effects can be obtained by adopting the present invention:

[0034] 1. The present invention determines the path of the target and performs kinematic control and dynamic control on the unmanned ship that performs the target encirclement task, so that it can better keep up with the angular velocity of the target to be encircled.

[0035] 2. The present invention adopts a particle filter algorithm, which can effectively track dynamic targets in complex environments. The unmanned ship can identify the target and accurately estimate its trajectory through this algorithm, and can still maintain a high tracking accuracy even if the target is affected by environmental noise or sudden motion changes.

[0036] 3. The present invention can respond to the dynamic changes of the target in real time during the encirclement process, calculate the angular velocity of the unmanned boat in real time, and ensure that multiple unmanned boats can synchronously perform the encirclement task. No matter how fast the target moves or changes direction suddenly, the new motion instructions can be quickly calculated, so that the unmanned boat can adjust its posture in the shortest time and maintain effective encirclement of the target.

[0037] 4. The present invention realizes information sharing among multiple unmanned ships through the network topology structure. Different from the traditional centralized control method, the topology method has the characteristics of distribution, and each unmanned ship makes autonomous decisions and collaborates based on local shared information. This method can reduce information transmission delay and improve the response speed of the system, especially in large-scale unmanned ship collaboration scenarios, which improves the scalability and robustness of the system.

[0038] 5. The present invention can also realize the coordinated encirclement of multiple dynamic targets at the same time. This multi-target support can simulate complex mission requirements, such as multi-level tasks such as defense and monitoring. Compared with the traditional single-target encirclement system, it has more comprehensive functions. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a reference coordinate system of an unmanned ship cooperative target encirclement method in an embodiment of the present invention.

[0040] Figure 2 It is a communication topology diagram of multiple unmanned ships in an embodiment of the present invention.

[0041] Figure 3 and Figure 4 They are respectively diagrams of simulation results of single unmanned ship target tracking and multiple unmanned ships collaborative target encirclement in embodiments of the present invention.

[0042] Figures 5 to 14 This is a diagram showing the simulation results of the speed / angular velocity tracking of multiple unmanned ships in cooperative target encirclement in an embodiment of the present invention. Figure 5 and Figure 6 This is the simulation result diagram of the speed and angular velocity tracking of the first ship. Figure 7 and Figure 8 This is the speed and angular velocity tracking simulation result diagram of the second ship. Fig. 9 and Fig.10 This is the speed and angular velocity tracking simulation result diagram of the third ship. Fig.11 and Fig.12 This is the simulation result diagram of the speed and angular velocity tracking of the fourth ship. Fig.13 and Fig.14 This is the speed and angular velocity tracking simulation result diagram of the fifth ship. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] An unmanned ship cooperative target encirclement method is used to enable M unmanned ships to cooperatively encircle a target, M≥3, and the unmanned ships maintain communication, including the following steps:

[0045] Step 1: The first unmanned ship collects water surface images and uses the YOLOv8 target detection algorithm to identify and detect targets on the water surface that need to be collaboratively surrounded;

[0046] Step 2: After the first unmanned ship detects the target that needs to be surrounded collaboratively, it obtains the target information in real time, including the position, speed and angular velocity of the target, and then transmits the target information to other unmanned ships;

[0047] Step 3: After each unmanned ship obtains the target information, it predicts the target trajectory based on the particle filter. The specific method is as follows:

[0048] S31, continuously obtain target information at N moments and initialize N particles, where the tth particle is expressed as:

[0049]

[0050] in, represents the state of the target in the i-th dimension at time t, i = 1, 2, 3, 4, i = 1 corresponds to the x component of the target position, i = 2 corresponds to the y component of the target position, and i = 3 corresponds to the initial velocity of the target.

[0051] degrees, i = 4 corresponds to the initial angular velocity of the target;

[0052] S32, initialize the weight of each particle to

[0053] S33, for particle x t To make a forecast update:

[0054]

[0055] Among them, the superscript (1) indicates after the update, the superscript (0) indicates before the update, v t Represented as Gaussian distributed noise;

[0056] S34, update the weight of each particle:

[0057]

[0058] Among them, w t Represents x t The weight of , the superscript (1) indicates after the update, and the superscript (0) indicates before the update, represents the prediction likelihood, that is, the particle from Update to The probability of

[0059] S35, normalize the weights of all particles:

[0060]

[0061] in, is the normalized x t The weight of

[0062] S36, weighted average the target position information in all particles to obtain a predicted position:

[0063]

[0064] S37, repeating steps S33 to S36 to obtain a series of predicted positions q0, q1, q2, ..., averaging every 2k+1 consecutive predicted positions to achieve smoothing of the predicted positions, where k represents the size of the smoothing window, and the smoothed predicted positions constitute the predicted target trajectory;

[0065] Step 4: Each unmanned ship tracks the linear velocity of the target through the PID algorithm, and calculates the angular velocity required for encirclement based on the predicted target trajectory:

[0066]

[0067] Where Ξ represents the angular velocity required for encirclement, q t Indicates the current location coordinates of the target. Indicates the current speed of the target, q i represents the position coordinates of the i-th unmanned ship, Ψ i represents the heading angle of the i-th unmanned ship, R i represents the motion radius of the ith unmanned ship, v i represents the moving speed of the i-th unmanned ship;

[0068] Step 5: Each unmanned ship moves according to its own linear velocity and angular velocity to achieve coordinated encirclement of the target.

[0069] Among them, the unmanned ship is equipped with four parts: target recognition and detection part, path determination and tracking part, unmanned ship kinematic control and unmanned ship dynamic control.

[0070] The target recognition and detection part constructs a target detection method based on the improved YOLOv8 (You Only Look Once) and the dataset of the collaboratively surrounded target, trains the target detection network, and obtains a model for real-time detection of collaborative targets.

[0071] Specifically, a collaborative encirclement target image dataset is constructed, which includes the target's unique appearance, size, and key parts. Labelme is used to annotate the collected image dataset, and the training set and validation set are divided into training set and validation set according to a certain ratio. The training is performed in the YOLOv8 network structure to obtain a model for real-time detection of collaborative encirclement targets, and the model is deployed on an unmanned ship that performs the encirclement mission.

[0072] The path determination and tracking part constructs a collaborative target path determination and transformation algorithm based on a particle filter. It uses a series of weighted particles to approximate the real state distribution of the collaborative encirclement target. The weights of all particles are updated in real time according to the state of the collaborative encirclement target identified. The weighted average of all the particles with updated weights is performed. Through time recursion, the target trajectory is finally formed in combination with the collaborative encirclement target model. Finally, the trajectory is smoothed to obtain the mathematical expression of the collaborative encirclement target trajectory. The weighted particles are characterized by being a discretized representation of the target state space, where the "weight" is used to represent the closeness of the particle to the true target state, and the "weight" is used to adjust the influence of the particle. A large "weight" means that the particle contributes more to the state estimation of the collaborative encirclement target, while particles with small "weights" are gradually eliminated, and finally a cluster of high-weight particles is retained, thereby effectively estimating the trajectory of the collaborative encirclement target; the "weighted averaging of particles" is characterized in that the weighting process not only considers the state distribution of the particles, but also dynamically adjusts the weight of the particles based on the Bayesian update rule, which is not only applicable to the state estimation of the current time step, but also performs multi-step prediction through time recursion to ensure the continuity and robustness of the trajectory.

[0073] Specifically, the target information of N moments is obtained continuously, and N particles are initialized, where the tth particle is expressed as:

[0074]

[0075] in, represents the state of the target in the i-th dimension at time t, i = 1, 2, 3, 4, i = 1 corresponds to the x component of the target position, i = 2 corresponds to the y component of the target position, and i = 3 corresponds to the initial velocity of the target.

[0076] degrees, i = 4 corresponds to the initial angular velocity of the target;

[0077] S32, initialize the weight of each particle to

[0078] S33, for particle x t To make a forecast update:

[0079]

[0080] Among them, the superscript (1) indicates after the update, the superscript (0) indicates before the update, v t Represented as Gaussian distributed noise;

[0081] S34, update the weight of each particle:

[0082]

[0083] Among them, w t Represents x t The weight of , the superscript (1) indicates after the update, and the superscript (0) indicates before the update, represents the prediction likelihood, that is, the particle from Update to The probability of

[0084] S35, normalize the weights of all particles:

[0085]

[0086] in, is the normalized x t The weight of

[0087] S36, weighted average the target position information in all particles to obtain a predicted position:

[0088]

[0089] S37, repeat steps S33 to S36 to obtain a series of predicted positions q0, q1, q2, ..., average every 2k+1 consecutive predicted positions to smooth the predicted positions, where k is the size of the smoothing window. The smoothed predicted positions constitute the predicted target trajectory.

[0090] The motion control system of the unmanned ship includes kinematic control and dynamic control. Kinematic control is to design the required posture and speed; dynamic control obtains the thrust and torque required by the ship according to the posture and speed information given by the kinematic guidance law. The kinematic control part of the unmanned ship designs the posture and speed required by the unmanned ship through kinematic control of the unmanned ship. The dynamic control part of the unmanned ship is designed based on the PID controller. The posture and speed information of the unmanned ship given by the kinematic control are used to obtain the corresponding thrust and torque required by the unmanned ship, so that the unmanned ship can better keep up with the speed and angular velocity of the cooperative encirclement target and achieve better cooperative encirclement.

[0091] Specifically, the kinematic model of the i-th unmanned ship can be expressed as follows:

[0092]

[0093] The kinematic model of the i-th unmanned ship can be expressed as follows:

[0094]

[0095] Among them, x and y represent the axis coordinates, Ψ represents the bow roll angle of the unmanned ship, r represents the bow roll angular velocity of the unmanned ship, τ u represents the longitudinal thrust of the unmanned ship, τ r Expressed as the moment in the bow rolling direction of the unmanned ship, m u ,m v ,m r Represents the mass and inertia of the unmanned ship.

[0096] Design an unmanned ship control rate algorithm, whose ultimate goal is to ensure that the unmanned ship can surround the target with a certain movement radius. Through the control rate algorithm, the trajectory of the unmanned ship converges to 0 during the movement, as shown in the following formula:

[0097]

[0098] like Figure 1 As shown in the following formula, q t Indicates the coordinates of the target being surrounded, q i The position coordinates of the i-th unmanned ship, Ψ i represents the heading angle of the i-th unmanned ship, R i represents the motion radius of the ith unmanned ship, v i represents the speed of the i-th unmanned ship, r i represents the angular velocity of the ith unmanned ship. Through mathematical conversion, we can get:

[0099]

[0100]

[0101] Transforming the above formula (8), we can get the following formula:

[0102]

[0103] The final single / multiple unmanned ship control rate algorithm is shown in the following formula:

[0104]

[0105] Among them, q t Represented as the location coordinates of the cooperatively surrounded target, Expressed as the movement speed of the cooperatively surrounded target.

[0106] The kinematic control system of single / multiple unmanned ships is built based on the simulation platform. The kinematic control system consists of three main parts: detection module, actuator module and unmanned ship module. The target detection and particle filter algorithms are run in the detection module to obtain relevant information of the collaborative encirclement target; the control rate algorithm is run in the actuator module, and the results are passed to the unmanned ship module to form a complete kinematic control system of single / multiple unmanned ships.

[0107] Information communication between multiple unmanned ships Figure 2 As shown in the figure, with the topological structure, the unmanned ship performing the encirclement mission can only obtain useful speed, angular velocity and information about the coordinated encirclement target from the adjacent unmanned ships. The simulation results are shown in Figure 3 and Figure 4 As shown, it can be seen that single / multiple unmanned ships can track / surround the target well.

[0108] Based on the kinematic control of the unmanned ship, the unmanned ship dynamic control system was built, using the PID control method. The system is still composed of three parts: the detection module, the actuator module and the unmanned ship module. Each part performs the same function. The final dynamic control result is as follows: Figures 5 to 14 As shown, the unmanned ships performing encirclement tasks can keep up with the speed and angular velocity of the target well and complete the encirclement tasks well.

[0109] In the description of the present invention, it should be understood that the fixed or positional relationship indicated in the relevant description is based on the orientation or positional relationship shown in the accompanying drawings, which is only for the convenience of describing the present patent method, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0110] The embodiments described above are only descriptions of the preferred modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. An unmanned ship cooperative target encirclement method, characterized in that: The method is used to enable M unmanned ships to cooperatively surround a target, where M≥3, and the unmanned ships maintain communication, and includes the following steps: Step 1: The first unmanned ship collects water surface images and uses the YOLOv8 target detection algorithm to identify and detect targets on the water surface that need to be collaboratively surrounded; Step 2: After the first unmanned ship detects the target that needs to be surrounded collaboratively, it obtains the target information in real time, including the position, speed and angular velocity of the target, and then transmits the target information to other unmanned ships; Step 3: After each unmanned ship obtains the target information, it predicts the target trajectory based on the particle filter. The specific method is as follows: S31, continuously obtain target information at N moments and initialize N particles, where the tth particle is expressed as: in, represents the state of the target in the i-th dimension at time t, i=1,2,3,4, i=1 corresponds to the x-component of the target position, i=2 corresponds to the y-component of the target position, i=3 corresponds to the initial velocity of the target, and i=4 corresponds to the initial angular velocity of the target; S32, initialize the weight of each particle to S33, for particle x t To make a forecast update: Among them, the superscript (1) indicates after the update, the superscript (0) indicates before the update, v t Represented as Gaussian distributed noise; S34, update the weight of each particle: Among them, w t Represents x t The weight of , the superscript (1) indicates after the update, and the superscript (0) indicates before the update, represents the prediction likelihood, that is, the particle Update to probability; S35, normalize the weights of all particles: in, is the normalized x t The weight of S36, weighted average the target position information in all particles to obtain a predicted position: S37, repeating steps S33 to S36 to obtain a series of predicted positions q0, q1, q2, ..., averaging every 2k+1 consecutive predicted positions to achieve smoothing of the predicted positions, where k represents the size of the smoothing window, and the smoothed predicted positions constitute the predicted target trajectory; Step 4: Each unmanned ship tracks the linear velocity of the target through the PID algorithm, and calculates the angular velocity required for encirclement based on the predicted target trajectory: Where X represents the angular velocity required for encirclement, q t Indicates the current location coordinates of the target. Indicates the current speed of the target, q i represents the position coordinates of the i-th unmanned ship, Y i represents the heading angle of the i-th unmanned ship, R i represents the motion radius of the i-th unmanned ship, v i represents the moving speed of the i-th unmanned ship; Step 5: Each unmanned ship moves according to its own linear velocity and angular velocity to achieve coordinated encirclement of the target.

Images