Underwater robot cluster formation motion and experiment method architecture

Abstract

The invention discloses an underwater robot cluster formation motion and experiment method, and belongs to the field of underwater robot cluster formation motion. The method comprises the following steps: establishing an individual virtual force control model of an underwater robot cluster; establishing a local sensing information model of the individual underwater robot, and acquiring three-dimensional relative distance and azimuth information of the individual underwater robot and the neighbor underwater robot; based on the virtual force control model and the local sensing information model, an expected underwater robot cluster formation is obtained through adjustment; performing an underwater robot cluster formation motion experiment in an open water area; and performing underwater robot cluster formation completion degree calculation based on experimental acquisition data to verify and achieve an underwater robot cluster formation effect. According to the invention, the problem that accurate distance and azimuth information are difficult to obtain to complete formation tasks when cluster motion for realizing communication by adopting a global information interaction technology is applied to an underwater environment is solved.

Description

Technical Field

[0001] The invention belongs to the field of underwater robot cluster formation movement, and in particular relates to an underwater robot cluster formation movement and an experimental method. Background Art

[0002] In nature, many organisms develop or evolve in the form of populations, achieving coordinated movement through local interactions. For example, schools of fish often swim in cylindrical groups to avoid the threat of predators. Many fish rely on visual observation of nearby neighbors to achieve collective coordinated movement. This behavior does not require any explicit communication mechanism, but rather spontaneously generates collective responses through local sensory information. This cluster movement mechanism based on local information provides important biological inspiration for the control of multi-agent systems.

[0003] Existing multi-agent swarm control methods mainly include behavior-based methods, leader-follower methods, virtual structure methods, and artificial potential field methods. These methods have been relatively maturely applied in multi-agent swarm motion. For example, the behavior-based method realizes the overall motion of the swarm by defining the interaction rules between individuals; the leader-follower method relies on the guidance of a specific individual (leader), and other individuals move according to the instructions of the leader; the virtual structure method guides the movement of individuals by constructing a virtual overall structure; and the artificial potential field method enables individuals to achieve swarm motion under the drive of the potential field by designing the potential field function.

[0004] However, most existing studies rely on global information interaction between robots when implementing multi-agent swarm control. For example, when an aerial drone swarm forms a specific swarm operation form, it usually relies on a central controller and a global positioning system (GPS) to obtain and exchange the location information of neighbors in real time. This global information reliance works well in an open environment with good communication conditions, but in practical applications, especially in underwater environments, due to limited communication, multi-agent formation control methods that rely on global information and positioning will be difficult to apply. Signal propagation is limited in underwater environments, and communication delays and packet loss problems are common, which makes control methods based on global information face significant challenges. Summary of the invention

[0005] Technical issues to be solved:

[0006] In order to avoid the shortcomings of the prior art, the present invention provides an underwater robot cluster formation movement and experimental method, drawing on the fish school to complete efficient cluster movement through local visual observation of its neighbors. An underwater robot cluster formation model is established based on a virtual force model and local visual perception information, and the orderly movement of the cluster is achieved by simulating the virtual force between individuals; then, based on the yolox model with an increased attention mechanism and the binocular ranging principle, the relative distance and orientation information of the underwater robot neighbors is acquired, providing local perception information for the realization of cluster formation. Finally, a deployment planning method for underwater robot cluster movement experiments is established, and the cluster formation experiments are evaluated with formation completion indicators. The present invention solves the problem that when cluster movement using global information interaction technology to achieve communication is applied to underwater environments, it is difficult to obtain accurate distance and orientation information to complete formation tasks.

[0007] The technical solution of the present invention is: an underwater robot cluster formation movement and experimental method, the specific steps are as follows:

[0008] Establish an individual virtual force control model for the underwater robot cluster, set up a formation motion model for N underwater robots, and update the position of the individual underwater robot by obtaining the direction and magnitude of the combined force of the N underwater robots;

[0009] A local perception information model of an individual underwater robot is established. Based on the target detection model, the ORB feature extraction and matching method is used to obtain the three-dimensional relative distance and orientation information between the individual underwater robot and its neighboring underwater robots.

[0010] Based on the virtual force control model and the local perception information model, the desired underwater robot cluster formation is obtained through adjustment;

[0011] Conduct underwater robot swarm formation movement experiments in open waters;

[0012] Based on the experimental collected data, the underwater robot cluster formation completion calculation is carried out to verify the underwater robot cluster formation effect.

[0013] A further technical solution of the present invention is: the individual virtual force control model includes individual underwater robot repulsion and individual underwater robot attraction; when the distance between adjacent robots is less than the expected relative distance, the robots exert repulsion on each other, and the repulsion increases as the distance decreases to avoid robot collision; when the distance between the robots is greater than the expected distance, the robots exert attraction on each other, attract each other, and tend to the set expected position, and finally maintain a fixed distance to stably form a cluster.

[0014] A further technical solution of the present invention is: a method for obtaining the combined force of the N underwater robots is that an individual underwater robot obtains the first-order derivative of the distance to its neighbor robots, and then sums the derivatives of all neighbor distances to obtain all combined forces with all neighbors.

[0015] A further technical solution of the present invention is: the position update rule of the individual underwater robot is:

[0016]

[0017] Where N is the total number of underwater robots, d is the relative distance between adjacent underwater robots, and d t is the expected relative distance, q ij is the relative position of adjacent underwater robots; a is the variable defining repulsion, and b is the variable defining attraction.

[0018] A further technical solution of the present invention is: the local perception information model includes a yolox target detection model and a binocular distance and orientation estimation model;

[0019] The binocular camera in the binocular distance and orientation estimation model is used to obtain the image data of various motion modes of the underwater robot as a sample data set;

[0020] Input the sample data set into the Yolox target detection model and output the position coordinates of the neighbor underwater robot on the image;

[0021] The binocular distance and orientation estimation model adopts the ORB feature extraction and matching method to obtain the disparity of the binocular left and right views in the target detection frame on the image. According to the binocular ranging principle, the three-dimensional relative distance and orientation information of the neighboring underwater robot are calculated from the obtained disparity.

[0022] A further technical solution of the present invention is: a coordinated attention mechanism is set in the backbone feature extraction network of the yolox target detection model, and the dependencies between data channels and in spatial dimensions are captured simultaneously through the coordinated attention mechanism, thereby increasing the global perception ability of the model and maintaining the real-time performance of the model calculation.

[0023] A further technical solution of the present invention is: the experimental method for performing underwater robot cluster formation movement in open waters is:

[0024] Calibrate the binocular vision system of each underwater robot and check whether each underwater robot is in normal condition;

[0025] Arrange the underwater robots in the designated waters in order, initialize the underwater robot cluster formation, plan the starting position and expected formation of the underwater robot cluster, and mark the pilot robot and follower robot;

[0026] Start the underwater robot control system, conduct necessary preliminary debugging, and use the perception system to identify and track the target; perform underwater robot movement cluster control according to the predetermined plan, set the cluster formation, and plan a reasonable formation layout;

[0027] After collecting data on various motion conditions of the test target, the cluster formation experiment was completed, and all experimental equipment was recovered and its status checked.

[0028] A further technical solution of the present invention is: the underwater robot cluster formation completion calculation method is:

[0029] Assume that the theoretical position coordinates of the underwater robot on the plane are P i (x i ,y i ), the actual coordinate position photographed by the swarm experiment drone is P i '(x i ',y i '), then the distance between the theoretical position and the actual position is

[0030] Calculate the individual completion degree g of the underwater robot i ,

[0031]

[0032] Where R is the distance deviation threshold between the given theoretical position and the actual position of the underwater robot;

[0033] According to the individual completion degree g of the underwater robot i Calculate the formation completion degree G,

[0034] .

[0036] A cluster formation of underwater robots, including N underwater robots constituting the formation, namely virtual force control models and local perception information models carried thereon, wherein the relative positions between adjacent underwater robots are controlled by the virtual force control model, thereby completing the motion control of the cluster formation; the relative distance and orientation information of adjacent underwater robots are obtained by the local perception information model; and the invention also includes at least one processor and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the underwater robot cluster formation motion and experimental method.

[0037] A computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the underwater robot cluster formation movement and experimental method when executed.

[0038] Beneficial Effects

[0039] The beneficial effects of the present invention are as follows: the present invention adopts a cluster formation motion model based on a virtual force model and local perception information, verifies the feasibility of the method in forming a desired formation through simulation experiments, and verifies the cluster formation algorithm on an underwater robot. The formation completion evaluation index is established for the results after the experiment to conduct experimental evaluation, and the effectiveness of the underwater robot in achieving the desired formation is further verified. At the same time, the present invention establishes a deployment planning method for large-scale cluster experiments in open waters, taking into account the requirements of experimental scenarios and experimental condition restrictions, and provides a practical implementation path for realizing underwater robot cluster experiments.

[0040] Through experimental verification, the present invention has verified the effectiveness of establishing a cluster formation algorithm through modeling, simulation verification and open water experiments of underwater robot cluster motion, and completed the experimental task of forming a formation of 15 underwater manta ray-like robots on a lake. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 : Simulation results of relative motion position of underwater robot cluster self-organization;

[0042] Figure 2 : Simulation diagram of relative position change of underwater robot cluster self-organized movement;

[0043] Figure 3 : Simulation result diagram of underwater robot cluster motion orientation;

[0044] Figure 4 : Result diagram of underwater robot cluster motion orientation error;

[0045] Figure 5 : Framework diagram of local perception method for underwater robots;

[0046] Figure 6 : Schematic diagram of the underwater robot's perception of its neighbors at different distances;

[0047] Figure 7 : Flowchart of deployment of open water swarm formation movement experiment;

[0048] Figure 8 : Schematic diagram of cluster formation completion calculation;

[0049] Fig. 9 : Calculation of the completion of the formation of 3 and 6 underwater robots in clusters;

[0050] Fig.10 : Completion calculation of 15 underwater robots. DETAILED DESCRIPTION

[0051] The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0052] Based on existing research, when implementing multi-agent cluster control, most of them rely on global information interaction between robots. For example, when an aerial drone cluster forms a specific cluster operation form, it usually relies on a central controller and a global positioning system (GPS) to obtain and exchange the location information of neighbors in real time. This global information dependence is effective in an open environment with good communication conditions, but in practical applications, especially in underwater environments, due to limited communication, multi-agent formation control methods that rely on global information and positioning will be difficult to apply. The present invention provides an underwater robot cluster formation movement and experimental method, and the specific steps are as follows:

[0053] Step 1: Establish an individual virtual force control model for the underwater robot cluster, set up a formation motion model for N underwater robots, and update the position of the individual underwater robot by obtaining the direction and magnitude of the combined force of the N underwater robots;

[0054] Step 2: Establish a local perception information model of an individual underwater robot, and use the ORB feature extraction and matching method based on the target detection model to obtain the three-dimensional relative distance and orientation information between the individual underwater robot and its neighboring underwater robots;

[0055] Step 3: Based on the virtual force control model and the local perception information model, the desired underwater robot cluster formation is obtained through adjustment;

[0056] Step 4: Conduct underwater robot cluster formation movement experiments in open waters;

[0057] Step 5: Based on the experimental data collected, the underwater robot cluster formation completion degree is calculated to verify that the underwater robot cluster formation effect is achieved.

[0058] A cluster formation of underwater robots, including N underwater robots constituting the formation, namely virtual force control models and local perception information models carried thereon, wherein the relative positions between adjacent underwater robots are controlled by the virtual force control model, thereby completing the motion control of the cluster formation; the relative distance and orientation information of adjacent underwater robots are obtained by the local perception information model; and the invention also includes at least one processor and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the underwater robot cluster formation motion and experimental method.

[0059] A computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the underwater robot cluster formation movement and experimental method when executed.

[0060] The present invention draws on the behavioral mechanism of fish cluster movement based on visual observation of neighbors, and uses the local visual perception information of the underwater robot to study and realize the underwater robot cluster formation movement. Specifically, the present invention first establishes an underwater robot cluster formation model based on the virtual force model and local visual perception information, and realizes the orderly movement of the cluster by simulating the virtual force between individuals.

[0061] The above technical solution is further described below with reference to the accompanying drawings and examples:

[0062] In one embodiment, the underwater robot uses an underwater manta ray-like robot. When the manta ray-like robot swims at a low speed, it has high body stability, less shaking, strong maneuverability and swimming efficiency, and can achieve complex movements such as multi-degree-of-freedom movement, low-speed position holding, and direction change. These characteristics make the manta ray-like robot an ideal platform for studying underwater robot cluster motion. The specific steps of an underwater robot cluster formation motion and experimental method in this embodiment are as follows:

[0063] Step 1: Modeling and simulation of self-organized motion of underwater robot swarm

[0064] Inspired by the clustering behavior of certain groups in nature (fish schools, bird flocks, etc.), this paper controls the movement of robots by simulating the repulsion and attraction of adjacent robots based on the virtual force control model and the local perception information model.

[0065] Step 1.1: Construct a cluster formation control model based on virtual force

[0066] In this embodiment, the virtual force control model is divided into two parts, repulsion and attraction. targetWhen the distance d between the robots is greater than the expected distance, the robots will exert a gravitational force on each other, causing them to attract each other and tend to the set expected position, eventually maintaining a fixed distance and forming a stable cluster.

[0067] The sum of the forces F of all N neighbors ij is the Lennard-Jones potential V ij The average value of the resultant force is multiplied by the relative position q between robot i and its neighbor robot j. ij Determines the next movement p of robot i i The motion update rule of underwater robot i is expressed as,

[0068]

[0069] Where N is the total number of underwater robots, d is the relative distance between the underwater robot and its neighboring robots, and d t is the expected relative distance, q ij is the relative position of the adjacent robots. Variable a defines the magnitude of the repulsive force, and variable b defines the magnitude of the attractive force. Through this formula, robot i will update its position according to the direction and magnitude of the resultant force, and gradually adjust to a stable desired formation.

[0070] In one embodiment, in order to achieve a stable desired formation, the study is completed with a triangular formation as the desired formation. This paper controls the movement according to the relative position between the underwater robot and its neighbors to ensure that the relative positions between the robots conform to the layout of the desired formation. By setting the desired position of each robot relative to its neighbors, a stable triangular formation is formed between the robot and the adjacent robots. If there are three robots i, j, and k, the desired distance between robot i and j is d target , the distance between robots i and k is also d target , and ∠ijk=60°, forming an equilateral triangle.

[0071] Step 1.2: Construct a local perception information model of underwater robots based on binocular vision

[0072] Reference Figure 5 As shown, in this embodiment, a Yolox target detection model with an added attention mechanism and a binocular distance and orientation estimation method are used to convert the image information of the neighbor underwater robot under binocular perception into three-dimensional relative distance and orientation information with the neighbor, providing local perception information for cluster formation control.

[0073] First, a binocular camera is used to obtain image data of multiple motion modes (gliding, swooping and diving) of the underwater bionic robot. Each mode of the underwater bionic robot provides a different motion trajectory and perception angle. By collecting image data of multiple motion modes, the perception system can provide information to accurately capture the characteristics of neighboring robots. At the same time, the collected image data is annotated with the category "manta" to create a data set for training the detection model.

[0074] Next, in order to improve the detection model's perception accuracy of neighbor targets, the present invention adds a Coordinate Attention mechanism to the main feature extraction network of yolox. This attention mechanism can simultaneously capture the dependencies between data channels and spatial dimensions, increase the model's global perception ability, and this attention mechanism will not reduce the model's computational efficiency, and can maintain the real-time performance of the model calculation. The improvement of this model aims to enhance the perception system's ability to perceive neighbor targets in a fuzzy and complex underwater environment.

[0075] Furthermore, after obtaining the position coordinates of the neighbor robot on the image through the target detection method, the disparity of the binocular left and right views is obtained in the target detection frame. In order to improve the calculation efficiency of the perception system, the present invention adopts the ORB feature extraction and matching method. The biggest advantage of this method is that the calculation speed is fast. After obtaining the left and right disparity, according to the binocular ranging principle, the three-dimensional relative distance and orientation information with the neighbor underwater robot are calculated by the obtained disparity.

[0076] Finally, the method was deployed on an underwater bionic robot prototype for experimental verification, verifying the method's real-time computing efficiency, target detection accuracy of neighboring robots, distance measurement accuracy, and azimuth measurement accuracy to meet the requirements of cluster formation control. Figure 6 The underwater robot's perception of its neighbors at different distances is shown.

[0077] In one embodiment, using local perception technology, robot i adjusts its movement direction to maintain the desired triangle formation based on its relative position to neighbors j and k. If the relative positions of adjacent robots deviate, the robot calculates the offset and adjusts its position through virtual forces.

[0078] Based on the virtual force control model and the local perception information model, this model establishes an underwater robot cluster formation motion model, so that each robot can dynamically adjust its own motion direction according to the surrounding environment, so as to stably form the desired formation cluster. During the adjustment process, the robot will ensure that the distance and angle with its neighbors meet the set expected values, and finally achieve a steady-state formation.

[0079] In one embodiment, a simulation experiment is conducted on the established underwater robot cluster formation motion model. In the experiment, the number of underwater robots is set to N=6, and the local visual perception range of each underwater robot is set to 4 meters. For neighbor robots that are beyond the visual range, the system removes them from the neighbor set and does not consider the local decisions of these robots. Through local visual perception technology, the robot can obtain the relative position, distance and orientation information of the neighbor robots within its line of sight. This information is used to calculate the virtual force model to control the movement of the robot cluster.

[0080] The simulation parameters of the experiment are:

[0081] Number of robots: N=6.

[0082] Visual perception range: The perceptible range of each underwater robot is 4 meters.

[0083] Simulation time: 120 seconds.

[0084] Initial position: The initial positions of all underwater robots are set randomly.

[0085] Neighbor detection: For neighbors within the visual range, the robot calculates their relative position, relative distance, and relative orientation based on the visual perception algorithm.

[0086] In the experiment, the motion state of all underwater robots is adjusted in real time based on the position and motion information of neighboring robots. The desired formation is set to an isosceles triangle formation. According to the calculated virtual force, the robot will perform corresponding motion control and finally form the target cluster shape. The result of the self-organized motion position of the underwater robot cluster is shown in the figure below. Figure 1 shown.

[0087] The simulation results show that this method can effectively realize the movement of underwater robot clusters, and the robots successfully move towards the set cluster shape. Figure 1 Relative position and Figure 2 Relative orientation simulation results show that when the robots move in a cluster, their headings will be dynamically adjusted according to the information of their neighbors within their line of sight. Figure 2 It can be seen that the robot makes a navigation adjustment at 10 seconds. The relative position and orientation between robots will change over time, but as the simulation time increases, the robot's heading tends to be stable, and the cluster gradually tends to a stable desired formation shape.

[0088] Reference Figure 3 As shown in the figure, the heading change diagram of 6 underwater robots during cluster movement. According to the simulation results, there is a certain fluctuation in the heading change of the robots during cluster movement, but it eventually tends to a stable orientation. Figure 4The figure shows the heading error diagram of the 6 underwater robot cluster during movement. The standard deviation of the heading error is 1.8°. The heading error is mainly affected by the neighbor detection accuracy, the calculation error of the relative position between robots, and the accuracy of neighbor recognition within the visual range. Although there is a certain error, the error is acceptable within a certain range because the virtual force model can effectively control the relative position between robots. After a long period of movement, the movement of the robot cluster gradually stabilizes, the heading error decreases, and finally forms the desired cluster shape.

[0089] The simulation experiment verifies the feasibility and effectiveness of the underwater robot cluster motion method based on local perception. Through the combination of local perception algorithm and virtual force model, the underwater robot can autonomously adjust the motion direction according to the relative position and orientation information of neighboring robots, thereby stably forming a cluster. Although there is a certain orientation error, the simulation results show that this method can effectively realize cluster motion and form the expected cluster morphology, providing a theoretical basis for the underwater robot cluster motion in practical applications.

[0090] Step 2: Deployment planning for the open water underwater robot swarm formation motion experiment

[0091] (1) Experimental purpose: To study the swarm behavior and underwater motion performance of underwater robots, test the swarm motion ability of underwater robots in lake environments, and improve the control algorithm and performance evaluation of underwater robots.

[0092] (2) Experimental location: Select an outdoor lake that ensures water quality and lighting conditions, ensures local perception and can be photographed using drones, and has a moderate lake area and no potential sources of danger.

[0093] (3) Experimental preparation:

[0094] Personnel arrangement: 1 project leader, who will coordinate the overall work allocation of personnel and the preparation of various equipment for the bionic fish experiment, and conduct risk assessment before the experiment to ensure the safety of the experimenters and bionic fish. 5 bionic fish operators, who will be responsible for the debugging and control operation of the bionic fish on the shore. 1 drone operator, who will observe and record the process of each bionic fish cluster experiment through the drone. 1 data recorder, who will record the experimental data and the corresponding experimental results after each bionic fish cluster experiment. 5 safety officers, who will ensure the safety of on-site personnel and equipment, handle the corresponding emergencies, and promptly recover the equipment and rescue personnel. 5 equipment managers, who will be responsible for the safety inspection of the bionic fish before each experiment on the shore, and will be responsible for putting the bionic fish into the lake, as well as the recovery and inspection of the bionic fish after the experiment.

[0095] Equipment and tools: 15 underwater bionic fish, with necessary control and perception communication equipment. Control equipment such as radio, antenna, remote control and corresponding host computer program for each bionic fish. Data recording equipment (drone, camera, notebook, etc.). Rescue equipment (such as life jackets, life buoys, kayaks, long sticks, long ropes, etc.).

[0096] On-site arrangement: delineate the lake swimming test area and mark the lake swimming test boundary. Set up the equipment placement point, ensure the stability of the placement point to ensure the safety of the bionic fish equipment and prevent water from entering the electronic equipment. Arrange relevant personnel to guard the warning area to prevent unauthorized personnel from approaching and causing risks.

[0097] (4) Implementation steps:

[0098] First, before the cluster lake test, arrange the corresponding operators to check the status of all bionic fish and control equipment, calibrate the binocular vision system, and ensure that it can accurately identify the target in the underwater environment. Perform an airtightness check on each bionic fish and check the battery power of each device, and ensure that the device and the bionic fish can be effectively controlled. Debug the actuator of the bionic fish in the air to ensure that it works normally underwater. After the electric adjustment in the air, put the bionic fish into the water for basic swimming inspection to confirm that the binocular vision system is started normally.

[0099] Next, arrange the bionic fish in order in the designated waters, initialize the formation of the bionic fish, plan the starting position and expected formation of the bionic fish, and describe the role of each bionic fish in the cluster, such as the leading bionic fish and the following bionic fish.

[0100] Furthermore, the bionic fish control system was started, necessary preliminary debugging was performed, and the target was identified and tracked using the perception system. The bionic fish movement cluster control was performed according to the predetermined plan, the cluster formation was set, and reasonable formation layouts such as vertical columns and triangles were planned. In the experiment, various movement conditions of the target were tested, such as linear movement and accelerated movement.

[0101] Finally, after completing the cluster formation experiment, all experimental equipment was recovered and their status was checked. The aerial photography drone was recovered to ensure that the experimental data was completely preserved.

[0102] (5) Data analysis

[0103] After the experiment is completed, the formation completion degree is used as the evaluation index to calculate the formation completion degree of different numbers of underwater bionic robot clusters to verify that the expected formation effect is achieved. Figure 7 The figure shows the open water test process of the open water underwater robot cluster established in this embodiment.

[0104] Step 3: Calculation of underwater robot formation completion

[0105] Reference Figure 8The figure shows a schematic diagram of the calculation of the completion degree of the cluster formation. In this embodiment, it is assumed that the theoretical position coordinates of the underwater robot on the plane are P i (x i ,y i ), the actual coordinate position photographed by the swarm experiment drone is P i '(x i ',y i '), then the distance between the theoretical position and the actual position is Then calculate the individual completion degree g of the underwater robot i ,

[0106]

[0107] Finally, according to the individual completion degree g of the underwater robot i Calculate the formation completion degree G,

[0108]

[0109] Reference Fig. 9 The results of the cluster formation experiment of 3 and 6 underwater robots on the lake are shown. According to the above calculation steps, the cluster formation completion of 3 underwater robots is 91.7%. The cluster formation completion of 6 underwater robots is 85%.

[0110] Reference Fig.10 The figure shows the results of the cluster formation experiment of 15 underwater robots on the lake. According to the above calculation, the completion rate of the formation of 15 underwater robots in a group of "one" shapes is 63.6%, and the completion rate of the formation of 15 underwater robots in two groups of "one" shapes is 81.3%. It can be seen from the formation completion rate that as the number of underwater robots increases, the degree of underwater robots completing the predetermined formation tasks decreases, and the effect of achieving the expected formation goals decreases.

[0111] The underwater robot cluster formation movement and experimental method described in the present invention was used to realize the cluster movement experiment of 15 underwater manta ray-like robots on a lake for the first time, which verified the feasibility of the cluster formation method adopted in this paper and provided certain technical support for large-scale cluster applications.

[0112] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention without departing from the principles and intent of the present invention.

Claims

1. An underwater robot swarm formation movement and experimental method, characterized in that The specific steps are as follows: Establish an individual virtual force control model for the underwater robot cluster, set up a formation motion model for N underwater robots, and update the position of the individual underwater robot by obtaining the direction and magnitude of the combined force of the N underwater robots; A local perception information model of an individual underwater robot is established. Based on the target detection model, the ORB feature extraction and matching method is used to obtain the three-dimensional relative distance and orientation information between the individual underwater robot and its neighboring underwater robots. Based on the virtual force control model and the local perception information model, the desired underwater robot cluster formation is obtained through adjustment; Conduct underwater robot swarm formation movement experiments in open waters; Based on the experimental collected data, the underwater robot cluster formation completion calculation is carried out to verify the underwater robot cluster formation effect.

2. The underwater robot swarm formation movement and experimental method according to claim 1, characterized in that: The individual virtual force control model includes individual underwater robot repulsion and individual underwater robot attraction; when the distance between adjacent robots is less than the expected relative distance, the robots exert repulsion on each other, and the repulsion increases as the distance decreases to avoid robot collision; when the distance between robots is greater than the expected distance, the robots exert attraction on each other, attract each other, tend to the set expected position, and finally maintain a fixed distance to stably form a cluster.

3. The underwater robot swarm formation movement and experimental method according to claim 2, characterized in that: The method for obtaining the combined force of the N underwater robots is that an individual underwater robot obtains the first-order derivative of the distance to its neighbor robots, and then sums the derivatives of all neighbor distances to obtain all combined forces with all neighbors.

4. The underwater robot cluster formation movement and experimental method according to claim 3, characterized in that: The position update rule of the individual underwater robot is: Where N is the total number of underwater robots, d is the relative distance between adjacent underwater robots, and d t is the expected relative distance, q ij is the relative position of adjacent underwater robots; a is the variable defining repulsion, and b is the variable defining attraction.

5. The underwater robot swarm formation movement and experimental method according to claim 1, characterized in that: The local perception information model includes a yolox target detection model and a binocular distance and orientation estimation model; The binocular camera in the binocular distance and orientation estimation model is used to obtain the image data of various motion modes of the underwater robot as a sample data set; Input the sample data set into the Yolox target detection model and output the position coordinates of the neighbor underwater robot on the image; The binocular distance and orientation estimation model adopts the ORB feature extraction and matching method to obtain the disparity of the binocular left and right views in the target detection frame on the image. According to the binocular ranging principle, the three-dimensional relative distance and orientation information of the neighboring underwater robot are calculated from the obtained disparity.

6. The underwater robot swarm formation movement and experimental method according to claim 5, characterized in that: A coordinated attention mechanism is set in the backbone feature extraction network of the Yolox target detection model. The coordinated attention mechanism simultaneously captures the dependencies between data channels and in spatial dimensions, increases the global perception capability of the model, and maintains the real-time performance of the model calculation.

7. The underwater robot swarm formation movement and experimental method according to claim 1, characterized in that: The experimental method for conducting underwater robot cluster formation movement in open water is as follows: Calibrate the binocular vision system of each underwater robot and check whether each underwater robot is in normal condition; Arrange the underwater robots in the designated waters in order, initialize the underwater robot cluster formation, plan the starting position and expected formation of the underwater robot cluster, and mark the pilot robot and follower robot; Start the underwater robot control system, conduct necessary preliminary debugging, and use the perception system to identify and track the target; perform underwater robot movement cluster control according to the predetermined plan, set the cluster formation, and plan a reasonable formation layout; After collecting data on various motion conditions of the test target, the cluster formation experiment was completed, and all experimental equipment was recovered and its status checked.

8. The underwater robot swarm formation movement and experimental method according to claim 7, characterized in that: The underwater robot cluster formation completion calculation method is: Assume that the theoretical position coordinates of the underwater robot on the plane are P i (x i ,y i ), the actual coordinate position photographed by the swarm experiment drone is P i '(x i ',y i '), then the distance between the theoretical position and the actual position is Calculate the individual completion degree g of the underwater robot i , Where R is the distance deviation threshold between the given theoretical position and the actual position of the underwater robot; According to the individual completion degree g of the underwater robot i Calculate the formation completion degree G, 。 9. An underwater robot cluster formation, characterized by: It includes N underwater robots constituting a formation, namely, virtual force control models and local perception information models carried thereon, and the relative positions between adjacent underwater robots are controlled by the virtual force control model, thereby completing the motion control of the cluster formation; the relative distance and orientation information of adjacent underwater robots are acquired by the local perception information model; it also includes at least one processor, and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor, so that the at least one processor can execute the underwater robot cluster formation motion and experimental method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the underwater robot cluster formation movement and experimental method described in any one of claims 1-8 when executed.

Images