An unmanned ship collision avoidance method based on multiple collision avoidance objects, a terminal device, and a medium

Abstract

The application relates to an unmanned ship collision avoidance method based on multiple collision avoidance targets, a terminal device and a medium, and the method comprises the following steps: collecting motion information of all moving targets in a sea area near the unmanned ship; clustering all moving targets to obtain multiple target clusters; calculating collision avoidance areas of the target clusters; when the unmanned ship travels to a collision avoidance waypoint, if the re-navigation route and the collision avoidance area do not have more than one intersection point, the unmanned ship normally re-navigates; otherwise, the unmanned ship keeps the previous heading to continue traveling or deflects to the right to travel according to the deflected route according to the size of the heading angle until the re-navigation route and the collision avoidance area do not have more than one intersection point, and then the unmanned ship re-navigates. The application can safely complete the collision avoidance action for multiple dynamic targets, reduces the complexity of the collision avoidance decision, and improves the collision avoidance efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of unmanned surface vehicle collision avoidance technology, and particularly relates to an unmanned surface vehicle collision avoidance method based on multiple collision avoidance targets, a terminal device and a medium. BACKGROUND

[0002] An unmanned surface vehicle (USV) is a kind of unmanned water vessel capable of performing specific tasks and having autonomous or semi-autonomous capabilities. Due to its good concealment, strong maneuverability, and small formation size, it is widely used in military, scientific research, artificial intelligence and other fields. When navigating in a complex marine environment, the unmanned surface vehicle will be threatened by collisions with sailing ships (including merchant ships, fishing boats, etc.). Therefore, it is of great practical significance to study the autonomous collision avoidance technology of USV.

[0003] The autonomous collision avoidance technology of USV refers to the ability of the USV to quickly and effectively adjust its heading and speed according to the surrounding environmental information, and to autonomously avoid collisions with single or multiple static and dynamic targets encountered while navigating on the water surface. At present, domestic and foreign experts and scholars mainly use local path planning and intelligent algorithms to solve the collision avoidance planning problem of USV. Common algorithms include speed obstacle method, particle swarm optimization, dynamic window method, artificial potential field method, reinforcement learning algorithm, deep learning algorithm and various intelligent algorithms and their improved algorithms. The existing methods can achieve good results in solving the avoidance problem of single ship, two ships and simple three ships, but the avoidance problem of multiple dynamic targets (such as unmanned surface vehicles, fishing boats and other large-scale clusters) is still a current research difficulty. At present, in order to solve the complex problem of USV collision avoidance with multiple dynamic targets, researchers mostly use the above-mentioned methods to conduct related research based on the concept of ship priority and adopt the strategy of avoiding collision with key ships. When there is a dense traffic flow on the sea surface, the collision avoidance efficiency of these intelligent collision avoidance algorithms will be slow, and the USV may be trapped in a complex traffic flow dilemma, failing to obtain the optimal solution for collision avoidance, resulting in an increased risk of collision. SUMMARY

[0004] In order to solve the above problems, the present application provides an unmanned surface vehicle collision avoidance method based on multiple collision avoidance targets, a terminal device and a medium.

[0005] The specific scheme is as follows:

[0006] An unmanned surface vehicle collision avoidance method based on multiple collision avoidance targets, comprising the following steps:

[0007] S1: collecting motion information of all moving targets in the sea area near the unmanned surface vehicle;

[0008] S2: clustering all moving targets based on the collected motion information to obtain multiple target clusters;

[0009] S3: Calculate the collision avoidance area of each target cluster based on the current coordinates of each target included in the target cluster;

[0010] S4: When the unmanned vehicle travels to the collision avoidance waypoint relative to the collision avoidance area of a target cluster, determine whether the homeward route of the unmanned vehicle at this time intersects with the collision avoidance area more than once. If yes, go to S5; otherwise, control the unmanned vehicle to return according to the homeward route;

[0011] S5: Determine whether the heading angle between the unmanned vehicle and the target cluster satisfies If yes, go to S6; otherwise, control the unmanned vehicle to continue traveling in the previous direction until the homeward route of the unmanned vehicle does not intersect with the current collision avoidance area of the target cluster more than once, and then control the unmanned vehicle to return according to the homeward route; R represents the radius of the collision avoidance area of the target cluster, represents the vertical component of the relative distance between the target cluster and the unmanned vehicle.

[0012] S6: Control the unmanned vehicle to turn right and make it travel according to the turned route until the homeward route of the unmanned vehicle does not intersect with the current collision avoidance area of the target cluster more than once, and then control the unmanned vehicle to return according to the homeward route.

[0013] Further, the clustering of all moving targets in step S2 is performed by using an improved DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm. The improved DBSCAN algorithm adds a speed difference constraint and a heading difference constraint on the basis of the traditional DBSCAN algorithm, i.e., the heading difference between any two targets belonging to the same target cluster is less than a heading difference threshold, and the speed difference between any two targets is less than a speed difference threshold.

[0014] Further, when clustering all moving targets by using the improved DBSCAN algorithm, a neighborhood radius constraint is included, i.e., the neighborhood radius ε between any two targets belonging to the same target cluster should be less than or equal to 4 times the safety distance of the target.

[0015] Further, when the unmanned vehicle is controlled to turn right in step S6, the calculation formula of the turning angle is:

[0016]

[0017] where θ represents the turning angle, v U represents the speed of the unmanned vehicle, and v C represents the speed of the target cluster.

[0018] An unmanned surface vehicle collision avoidance terminal device based on multi-collision avoidance targets, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method of the above-mentioned embodiments of the application when executing the computer program.

[0019] A computer readable storage medium stores a computer program, wherein the computer program is executable by a processor to implement the steps of the method of the above-mentioned embodiments of the application.

[0020] The application can safely complete collision avoidance actions for multiple dynamic targets, reduce collision avoidance decision complexity, and improve collision avoidance efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 A flowchart of the method of the embodiment of the application is shown.

[0022] Figure 2 A schematic diagram of the clustering result in this embodiment is shown.

[0023] Figure 3 A schematic diagram of the collision avoidance area in this embodiment is shown.

[0024] Figure 4 A schematic diagram of the collision avoidance of a single target by the traditional geometric collision avoidance algorithm in this embodiment is shown.

[0025] Figure 5 A schematic diagram of the collision avoidance of a target cluster by the traditional geometric collision avoidance algorithm in this embodiment is shown.

[0026] Figure 6 A schematic diagram when the heading angle does not meet the condition in this embodiment is shown.

[0027] Figure 7 A schematic diagram when the heading angle meets the condition in this embodiment is shown.

[0028] Figure 8 A schematic diagram of the target distribution in this embodiment is shown.

[0029] Figure 9 A schematic diagram of the target clustering result in this embodiment is shown.

[0030] Figure 10 A schematic diagram of the USV collision avoidance process based on the traditional collision avoidance method in the encounter scenario in this embodiment is shown.

[0031] Figure 11 A schematic diagram of the USV collision avoidance process based on the method of this embodiment in the encounter scenario is shown.

[0032] Figure 12Fig. 4 shows a schematic diagram of the USV collision avoidance process based on the traditional collision avoidance method in the cross encounter scenario in this embodiment.

[0033] Figure 13 Fig. 5 shows a schematic diagram of the USV collision avoidance process based on the method of this embodiment in the cross encounter scenario in this embodiment.

[0034] Figure 14 Fig. 6 shows a schematic diagram of the USV collision avoidance process based on the traditional collision avoidance method in the overtaking scenario in this embodiment.

[0035] Figure 15 Fig. 7 shows a schematic diagram of the USV collision avoidance process based on the method of this embodiment in the overtaking scenario in this embodiment.

[0036] Figure 16 Fig. 8 shows a comparison diagram of the USV collision avoidance paths of the two collision avoidance methods in the three encounter scenarios in this embodiment.

[0037] Figure 17 Fig. 9 shows a comparison diagram of the collision avoidance efficiency in this embodiment.

[0038] Figure 18 Fig. 10 shows a comparison diagram of the total sum of the closest approach distances in this embodiment.

[0039] Figure 19 Fig. 11 shows a diagram of the encounter distance changes of the USV and the target in this embodiment. DETAILED DESCRIPTION

[0040] To further illustrate the embodiments, the present application provides accompanying drawings. These drawings are part of the disclosure of the present application, mainly used to illustrate the embodiments, and can be used to explain the operating principles of the embodiments in conjunction with the related descriptions in the specification. Those of ordinary skill in the art should be able to understand other possible implementations and advantages of the present application by referring to these contents.

[0041] The present application will be further described in conjunction with the accompanying drawings and specific embodiments.

[0042] Embodiment one:

[0043] The embodiment of the present application provides a USV collision avoidance method based on multiple collision avoidance targets, which comprises the following steps: Figure 1 As shown in the figure, the method comprises the following steps:

[0044] S1: Collect the target motion information of each target in the sea area around the USV.

[0045] The target can be a merchant ship, a fishing boat, etc. sailing on the sea surface.

[0046] The collected target motion information is information needed in the subsequent steps, such as speed, heading, coordinates, etc.

[0047] S2: clustering all moving targets based on the collected motion information to obtain a plurality of target clusters.

[0048] DBSCAN algorithm is a density-based clustering algorithm. Among them, the neighborhood refers to the area with a distance less than ε from the selected object, and the other member set in the neighborhood is expressed as formula (1):

[0049] N ε (q)={p belong to D|dist(q,p)≤ε} (1)

[0050] In the formula, q is the selected object; D is the neighborhood of the selected object; ε is the neighborhood radius; p is the object other than the selected object. In this algorithm, q is called the core object, and if it satisfies formula (2):

[0051] |N ε (q)|≥Minpts (2)

[0052] In the formula, |N ε (q)| is the number of members in the set; Minpts refers to the minimum number of neighborhood points required when defining the core object. That is, in the ε neighborhood of the core object q, the number of non-core objects p is greater than or equal to the minimum number of neighborhood points, then all objects p in the ε neighborhood are directly density reachable from q, and finally these objects are clustered into a cluster. The target clustering schematic diagram based on the traditional DBSCAN algorithm is shown in Figure 2 .

[0053] When the distance between multiple targets is large enough for the USV to safely avoid collision between targets, it is obviously not necessary to avoid it as a whole cluster. Therefore, when clustering targets, the value of neighborhood radius ε should be determined. The selection of neighborhood radius ε parameter for clustering is closely related to the safe distance (Safe Distance, SD) of the target. Generally, the safe distance of a ship above 10,000 tons should be not less than 1 nm in good visibility; the safe distance SD of the USV can be calculated by formula (3):

[0054] SD=R U +R T +d s (3)

[0055] In the formula, R U is half the length of the USV, R T is half the length of the target; d s is the safety margin, and the value is calculated as follows:

[0056] d s =v R ·t+L x (4)

[0057] where t is time step; L x is the distance of the one with poor maneuvering performance to change course 90°.

[0058] When the USV can avoid collision from any two targets with a distance more than 2 times of SD, it can be considered that the USV has enough safe distance to avoid collision. Therefore, when clustering dynamic targets, the neighborhood radius ε should be less than or equal to 4 times of SD.

[0059] When the difference of the course of each member in the target cluster is too large, the USV and the members in the target cluster may have multiple encounter situations, and it is obviously not suitable to avoid collision as a whole. If the difference of the speed of the members in the target cluster is too large, the distance between the members in the target cluster will gradually expand over time, and individual cluster members may be separated from the target cluster in the process of forming a collision situation. Obviously, the cluster collision avoidance model cannot be used to avoid collision at this time. Therefore, the neighborhood radius ε and the motion situation of the members of the cluster are restricted as follows when clustering dynamic targets in the embodiment.

[0060] ε≤4SD (5)

[0061]

[0062] where and are the course of the ith target and the course of the jth target, respectively; and are the speed of the ith target and the speed of the jth target, respectively; E c is the course difference threshold, E v is the speed difference threshold; equation (5) indicates that the neighborhood radius ε should be less than or equal to 4 times of the safety distance SD; equation (6) indicates that the difference of the course of any two members in the target cluster should be less than or equal to E c ; equation (7) indicates that the difference of the speed of any two members in the target cluster should be less than or equal to E v .

[0063] where E c and E v are not fixed values and can have different values in different encounter situations. In maritime practice, the avoidance opportunity can be determined by considering the distance of closest point of approach (DCPA) and the distance (D) when taking collision avoidance action. When the relationship satisfies equation (8), avoidance action should be taken.

[0064] DCPA / D<0.5 (8)

[0065] When avoiding a 10,000-ton ship cluster, the DCPA required to be maintained is large. When the relative distance D between the two is large under the condition of formula (8), the collision avoidance process is relatively long. In order to prevent the target cluster from being too discrete, E c and E v The value of can be smaller. In the existing research literature, E is used to avoid collision with 10,000-ton ships. c Take 1°, E v Take it as 0.5kn; In this embodiment, for the USV collision avoidance problem, the DCPA required to be maintained is relatively small, so the entire collision avoidance process is faster, and the discrete change amplitude of the object cluster is relatively small, so it can be appropriately larger. After a lot of experimental analysis, this embodiment will E c Take 5°, E v Take 2kn.

[0066] Based on the above constraints of dynamic object clustering, this embodiment improves the traditional DBSCAN algorithm and introduces the heading difference threshold E c and speed difference threshold E v After clustering, the heading difference and speed difference between members of each cluster are detected. If the heading difference and speed difference between members of the same cluster are both less than the set threshold, they are density-reachable and will be clustered into one cluster. The set of all objects in the same cluster can be expressed by formula (9):

[0067]

[0068] Where, represents the heading of the i-th object, represents the heading of the j-th object, represents the speed of the i-th object, Represents the speed of the jth object. Based on the improved DBSCAN algorithm, the cluster of objects that are clustered into the same cluster can be used to avoid collisions with the cluster collision avoidance model.

[0069] S3: Calculating the collision avoidance area of ​​each object cluster based on the current coordinates of each object cluster.

[0070] The collision avoidance area of ​​the object cluster is used to determine the avoidance range of the object cluster. In this embodiment, the calculation of the collision avoidance area is implemented using an existing algorithm.

[0071] In this embodiment, each object is first puffed to construct an object avoidance area model, and then the dense multiple object avoidance areas are covered with circular envelopes based on the circular envelope algorithm to determine the collision avoidance area of ​​the object cluster, such as Figure 3 As shown in the figure, the dotted circle represents the avoidance range of the individual object, and the solid circle represents the cluster avoidance area after being covered by the circular envelope, which is used as the collision avoidance area of ​​the object cluster.

[0072] When the USV avoids the cluster of targets, the position relationship should satisfy equation (10):

[0073] (x U -x C ) 2 +(y U -y C ) 2 ≥R C 2 (10)

[0074] where x U and y U are the horizontal and vertical coordinates of the USV position respectively; x C and y C are the horizontal and vertical coordinates of the cluster safety region center respectively; R C is the cluster safety region radius, which can be calculated by equation (11).

[0075]

[0076] where x O and y O are the horizontal and vertical coordinates of the target respectively; SD is the safety distance.

[0077] S4: When the USV travels to the collision avoidance waypoint relative to the collision avoidance region of a certain target cluster, it is determined whether the re-navigation route of the USV and the collision avoidance region have more than one intersection point. If yes, go to S5; otherwise, control the USV to re-navigate according to the re-navigation route.

[0078] S5: Determine whether the heading angle between the USV and the target cluster satisfies If yes, go to S6; otherwise, control the USV to continue to travel in the previous direction until the re-navigation route of the USV and the current collision avoidance region of the target cluster have no more than one intersection point, and then control the USV to re-navigate according to the re-navigation route; R represents the radius of the collision avoidance region of the target cluster, represents the vertical component of the relative distance between the target cluster and the USV, and in this embodiment, the vertical component of the relative distance between the center point of the target cluster and the USV is used.

[0079] S6: Control the USV to deflect to the right and make it travel according to the deflected route until the re-navigation route of the USV and the current collision avoidance region of the target cluster have no more than one intersection point, and then control the USV to re-navigate according to the re-navigation route.

[0080] In this embodiment, the calculation formula of the deflection angle is:

[0081]

[0082] Where θ represents the deflection angle, v U Indicates the speed of the unmanned boat, v C Indicates the speed of the object cluster.

[0083] The traditional geometric collision avoidance algorithm determines the safe course of the unmanned boat to avoid the oncoming ship through geometric drawing based on the current position, heading, speed and other motion information of the unmanned boat and the oncoming ship. When using the traditional geometric collision avoidance algorithm to avoid a single target ship, the unmanned boat can calculate the collision avoidance waypoints of the unmanned boat and the target ship in the future (usually the right tangent point of the target ship's collision avoidance area) based on its own and the target ship's position, heading, speed and other motion information. When the unmanned boat reaches the collision avoidance waypoint, it completes the safe avoidance of the oncoming ship and then resumes sailing to continue to the destination. Figure 4 As shown, the resumed flight route is the line connecting the initial position and the destination when resuming the flight.

[0084] However, when using the geometric collision avoidance algorithm to avoid the cluster of objects, the unmanned boat uses the center of the cluster as the reference point for geometric construction to obtain relevant motion parameters, and uses it as a reference for judging collision risks and avoiding collisions. Since there are multiple target ships in the cluster, if there is no collision risk between the unmanned boat and the center of the cluster, it will resume its original route and continue to sail towards the destination. Figure 5 As shown in the figure, the unmanned boat may pass through the middle of the object cluster and collide with some members of the object cluster. Therefore, when the unmanned boat uses the traditional geometric collision avoidance algorithm to avoid collision with the object cluster, there are certain defects and it cannot meet the requirements of safe collision avoidance of the object cluster. It should be improved.

[0085] This embodiment introduces a method for detecting the risk of re-entry based on the navigation line of sight of the unmanned boat, that is, whether there is a risk of re-entry collision between the unmanned boat and the object cluster according to whether there is more than one intersection between the ray of the unmanned boat's heading and the collision avoidance area of ​​the object cluster. When the unmanned boat travels to the collision avoidance waypoint and is ready to re-enter, the intersection of the unmanned boat's re-entry route and the collision avoidance area is judged. If there is only one intersection, the unmanned boat is tangent to the object cluster, and the unmanned boat will not cross the collision avoidance area of ​​the object cluster, and the unmanned boat can safely re-enter; if there are two intersections, the unmanned boat will cross the collision avoidance area of ​​the object cluster. At this time, the unmanned boat may have a risk of re-entry collision, and the unmanned boat should not re-enter immediately, but should re-enter based on the heading angle between the unmanned boat and the object cluster. The angle between the heading of the unmanned ship and the heading of the object cluster is further determined. When it is detected that the unmanned ship and the object cluster have a collision risk, if the unmanned ship keeps the original heading and continues to travel, and there is no intersection between the unmanned ship and the object cluster, the unmanned ship will keep the original heading and continue to travel; if the unmanned ship keeps the original heading and continues to travel, and will inevitably enter the object cluster, the unmanned ship will turn right to avoid entering the object cluster.

[0086] As shown in Figure 6 the heading angle does not meet the above conditions, at this time the unmanned ship travels to the collision avoidance waypoint and keeps the original heading to continue to travel until the unmanned ship and the object cluster have more than one intersection, and then the unmanned ship resumes the voyage. As shown in Figure 7 the heading angle meets the above conditions, after the unmanned ship travels to the collision avoidance waypoint, if it continues to keep the original heading, it will inevitably enter the object cluster, at this time the unmanned ship will turn right according to the calculated right turning angle to avoid entering the object cluster, thereby avoiding collision danger. The improved geometric collision avoidance algorithm effectively solves the problem that the unmanned ship will collide with individual members of the object cluster after reaching the collision avoidance waypoint, and can be applied to the cluster collision avoidance problem occupying a large water space.

[0087] Simulation analysis

[0088] 1. Dynamic object clustering simulation based on improved DBSCAN

[0089] To verify the effectiveness of the improved DBSCAN algorithm, this embodiment selects the distribution of ships near latitude 32°26'N and longitude 123°23'E at 1514 on August 12, 2024, as shown in Figure 8 (picture and data from Shipnet), and the specific motion data of the ships is shown in Table 1. Based on the improved DBSCAN algorithm, a clustering simulation experiment is performed on multiple dynamic objects in the red circle in the figure.

[0090] Table 1

[0091]

[0092] Based on the existing related research and the above analysis, this embodiment takes the heading difference threshold E c as 5°, and the speed difference threshold E vTake 2kn. Since any two moving targets close to each other in terms of moving state and position can be clustered into a target cluster, the avoidance complexity can be reduced when avoiding the target cluster, so take 2 for Minpts, and the clustering results of dynamic targets based on the improved DBSCAN algorithm are as shown in Figure 9

[0093] As shown in the first graph in Figure 9 , when the neighborhood radius ε = 400 m, since the distances between target 1, target 2 and target 3 are less than 400 m, and the heading difference and speed difference between them do not exceed the limit value, they are clustered into one class, and the three dynamic targets can be regarded as a target cluster for overall and unified collision avoidance. As shown in the second graph in the figure, when the neighborhood radius ε = 500 m, since target 6, target 7 and target 8 also satisfy the distance relationship and the heading difference and speed difference between them are less than the set threshold, they are clustered into one class and regarded as a target cluster. As shown in the third graph in Figure 9 , when the neighborhood radius ε = 800 m, although the distances between target 1, 2, 3, 4 and 5 are less than 800 m, since the speed difference between target 4 and other targets in the class cluster exceeds the speed difference threshold E v , the clustering constraint condition is not met, and the improved DBSCAN algorithm cannot cluster them into one class, so the overall avoidance method of the cluster cannot be used in collision avoidance. As shown in the fourth graph in Figure 9 , when the neighborhood radius ε = 1000 m, the relative distances of all targets except target 9 are less than 1000 m, but the speed difference between these targets exceeds E v , so they cannot be clustered into a target cluster.

[0094] It can be found from the simulation results that the improved DBSCAN algorithm proposed in the embodiment can effectively cluster multiple targets close in moving state and position into a target cluster, and the unmanned ship can use the overall avoidance strategy to avoid the target cluster to reduce the difficulty of collision avoidance decision.

[0095] 2. Target cluster collision avoidance simulation

[0096] For multiple dynamic targets that can be clustered into a target cluster, the embodiment uses the MATLAB platform and combines the three encounter situations of meeting, crossing and overtaking to compare and simulate the traditional collision avoidance method of key ships and the method proposed in the embodiment. The basic parameters of the unmanned ship prototype used in the simulation are shown in Table 2, the safety distance SD of the target is taken as 100 m in the simulation experiment, the neighborhood radius ε is taken as 400 m, Minpts Minpts is taken as 2, the heading difference threshold E c is taken as 5°, and the speed difference threshold E v is taken as 2 kn.

[0097] Table 2​

[0098]

[0099] (1) Simulation verification of the encountered situation

[0100] A situation in which a USV and multiple objects meet head-on is designed. Simulation experiments are conducted on the USV using the traditional collision avoidance method for key vessels and the method of this embodiment to avoid collisions with densely packed objects. The simulation data are shown in Table 3.

[0101] Table 3

[0102]

[0103] (a) USV collision avoidance simulation results based on the traditional collision avoidance key ship method in the encounter situation are shown in the figure below. Figure 10 shown. Figure 10 During the navigation process, the USV will encounter multiple objects. It will use the geometric collision avoidance algorithm to avoid collisions with object 1, object 2, and object 3 in turn and then head towards the destination. Figure 10 As shown in the first figure, the USV and object 1 collided head-on. The USV turned right to avoid the collision. However, during the collision avoidance process, the USV was too close to object 5 and had invaded the safety area of ​​object 5. Figure 10 As shown in the second figure, when the USV resumed its flight, it encountered the object 2 again and turned right again to avoid collision with it. Figure 10 As shown in the third figure, after the USV avoided object 2, it again encountered object 3 during its return journey. The USV turned right again to avoid it and then continued on its way to the destination. Throughout the entire navigation process, the USV completed three collision avoidance maneuvers. However, the safe distance margin between the USV and object 1 was too small, with the closest approach distance being only 77 meters. The closest approach distances between the USV and objects 2 and 3 were also less than the safe approach distance of 100 meters, indicating that the USV was unable to safely avoid multiple objects during the collision avoidance process.

[0104] (b) The USV collision avoidance simulation results obtained based on the method of this embodiment under the encounter situation are as follows Figure 11 shown. Figure 11 The USV clustered multiple objects into one object cluster based on the improved DBSCAN algorithm and avoided it as a whole. Figure 11 At the moment shown in the first figure, the USV detects a collision risk with the object cluster and turns right according to COLREGs to avoid the collision. Figure 11 At the moment shown in the second figure, the USV detects that resuming the route at this point would pose a collision risk. Therefore, the USV steers again to avoid the collision, resuming the route toward the destination after the collision is no longer a risk. Using the collision avoidance model proposed in this embodiment, the closest approach distance between the USV and each object remains above 100 meters, making the USV's collision avoidance actions safer.

[0105] (2) Simulation verification of intersection encounter situation

[0106] A situation in which the USV and multiple objects intersect is designed. Simulation experiments are conducted on the USV using the traditional collision avoidance method and the method of this embodiment to avoid collisions with dense multiple objects. The simulation data are shown in Table 4.

[0107] Table 4

[0108]

[0109] (a) USV collision avoidance simulation results based on the traditional collision avoidance key ship method in the case of crossing encounters are shown in the figure. Figure 12 shown. Figure 12 In the video, the USV successively encountered objects 6, 2, 3, and 1 at right angles. According to the collision avoidance rules, the USV made four turns to avoid them. During actual navigation at sea, continuous avoidance may affect the difficulty of collision avoidance prediction for other objects and increase navigation risks.

[0110] (b) The USV collision avoidance simulation results based on the method of this embodiment in the case of intersection encounter are as follows: Figure 13 shown. Figure 13 In the simulation, the USV completed the overall avoidance of the cluster of objects with only one turn, avoiding the problem that the USV may cross and avoid between multiple dynamic objects and cannot avoid them at a sufficiently safe distance, reducing the complexity of collision avoidance decision-making and improving collision avoidance efficiency.

[0111] (3) Simulation verification of overtaking situation

[0112] The USV is designed to overtake multiple dynamic targets during navigation. The simulation data are shown in Table 5.

[0113] Table 5

[0114]

[0115] (a) USV collision avoidance simulation results based on the traditional collision avoidance key ship method in the overtaking situation are shown in the figure. Figure 14 shown. Figure 14 In the video, the USV overtook Object 5. As the overtaking vehicle, it turned left to give way to Object 5, weaving through the multi-object system. The USV then overtook Object 3 and again turned left to avoid it. Finally, the USV overtook Object 1 again, again turning left to avoid it. The USV completed three collision avoidance maneuvers before continuing toward the finish line.

[0116] (b) The USV collision avoidance simulation results based on the collision avoidance model of this embodiment in the overtaking situation are as follows: Figure 15 shown. Figure 15In this embodiment, the USV detects that the overtaking situation is formed with the target cluster, and the USV selects the right turn to avoid the collision because the right turn collision avoidance path is shorter when the USV overtakes the target cluster. The USV only performs one collision avoidance operation to complete the avoidance of the entire target cluster, which reduces the complexity of the collision avoidance decision, improves the collision avoidance efficiency, greatly shortens the collision avoidance path of the USV, and increases the safety distance between the USV and the target.

[0117] By arranging the distance change diagrams of the USV and each dynamic target in the three encounter scenarios, it is found that, as shown in Figure 16 The left three diagrams are the USV collision avoidance path diagrams based on the traditional collision avoidance key ship method, and the right three diagrams are the USV collision avoidance path diagrams based on the collision avoidance model of this embodiment. It is found by comparison that, when the USV approaches the dense multi-targets in the collision avoidance navigation situation, the USV may pass through the multi-targets multiple times and turn to avoid at a close distance to avoid collision by using the traditional collision avoidance key ship method, and when the USV avoids collision based on the method of this embodiment, the number of collision avoidance decisions of the USV is significantly reduced, and the collision avoidance path length is reduced in the three different encounter scenarios. As shown in Figure 17 It is found that the USV collision avoidance path and the collision avoidance time based on the method of this embodiment (CCAMDG in the diagram) are reduced compared with the traditional collision avoidance method in the three scenarios, and it can be seen that the collision avoidance efficiency of the USV is higher when the USV avoids the dense multi-targets in the collision avoidance navigation situation by using the method of this embodiment.

[0118] In order to quantify the collision avoidance safety of the two collision avoidance methods, the encounter distance changes between the USV and each target in the three encounter scenarios are arranged, as shown in Figure 19 The left three diagrams are the distance change diagrams of the USV and each target in the three encounter scenarios based on the traditional collision avoidance key ship method, and the right three diagrams are the distance change diagrams of the USV and each target in the three encounter scenarios based on the method of this embodiment. It is found by comparison that, when the USV avoids the dense multi-targets based on the traditional collision avoidance key ship method, the closest encounter distance between the USV and individual targets is less than 100m in the three encounter scenarios, which indicates that the USV cannot avoid the targets safely enough, and when the USV avoids collision based on the method of this embodiment, the closest encounter distance between the USV and each target is greater than 100m, and the collision avoidance action of the USV is safer.

[0119] By calculating and arranging the sum of the closest encounter distances between the USV and each target, it is found by comparison that, as shown in Figure 18 The sum of the closest encounter distances between the USV and each target based on the method of this embodiment is greater than the sum of the closest encounter distances based on the traditional collision avoidance key ship method in the three encounter scenarios, which indicates that the collision avoidance safety of the USV is improved by using the collision avoidance model of this embodiment.

[0120] Based on the simulation results of the three different encounter scenarios, the following conclusions can be drawn:

[0121] (1) In the three different encounter scenarios, based on the traditional collision avoidance method of focusing on the key ship, the USV may pass through multiple targets and need to make multiple collision avoidance decisions. However, when using the method of the embodiment, the USV avoids the close dense targets as a cluster and avoids passing through multiple ships at close range, as shown in Figs. Figure 11 、 Figure 13 and Figure 15 , and the number of collision avoidance decisions of the USV is reduced, thereby improving the navigation efficiency.

[0122] (2) The path length and time consumed for collision avoidance of the USV based on the method of the embodiment are reduced compared to the traditional collision avoidance method of focusing on the key ship, as shown in Fig. Figure 17 . In the meeting, crossing and overtaking scenarios, the USV collision avoidance path is reduced by 13.2%, 30.3% and 23.0%, respectively, and the collision avoidance time is reduced by 13.8%, 29.8% and 22.9%, respectively. Therefore, when avoiding close dense targets based on the method of the embodiment, the collision avoidance efficiency of the USV is improved.

[0123] (3) In the three encounter scenarios, based on the traditional collision avoidance method of focusing on the key ship, the closest encounter distance between the USV and the targets is less than 100m, which is lower than the minimum safe encounter distance. However, when using the collision avoidance model in the embodiment, the closest encounter distance between the USV and each target is greater than 100m, and the sum of the closest encounter distances with all targets is greater, as shown in Fig. Figure 18 . In the meeting, crossing and overtaking scenarios, the sum of the closest encounter distances between the USV and the targets is increased by 142.5%, 231.0% and 181.5%, respectively. Therefore, the collision avoidance model in the embodiment has better safety in solving the collision avoidance problem of close dense targets.

[0124] In summary, the method of the embodiment can effectively reduce the complexity of collision avoidance decision-making, and the USV has higher collision avoidance efficiency and better safety in solving the collision avoidance problem of close dense targets.

[0125] The embodiment of the present application introduces the heading difference threshold and the speed difference threshold into the classical DBSCAN algorithm, clusters the dense dynamic targets in the surrounding of the unmanned surface vehicle which meet the clustering condition into the target cluster which needs to be avoided, and establishes the collision avoidance area of the target cluster by using the circular envelope algorithm; introduces the collision detection method based on the line of sight into the geometric collision avoidance algorithm, and realizes the safe collision avoidance of the unmanned surface vehicle to the target cluster. The experiment shows that the embodiment can safely complete the collision avoidance action to the multiple dynamic targets, reduces the complexity of the collision avoidance decision, improves the collision avoidance efficiency, and can be applied to the avoidance decision in the dense traffic flow scene such as the sea avoidance formation cluster and the fishing boat cluster.

[0126] Embodiment two:

[0127] The present application also provides an unmanned surface vehicle collision avoidance terminal device based on multiple collision avoidance targets, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps in the above method embodiments of the embodiment one of the present application when executing the computer program.

[0128] Further, as an executable scheme, the unmanned surface vehicle collision avoidance terminal device based on multiple collision avoidance targets can be a desktop computer, a notebook, a palm computer, a cloud server and other computing devices. The unmanned surface vehicle collision avoidance terminal device based on multiple collision avoidance targets can include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above composition structure of the unmanned surface vehicle collision avoidance terminal device based on multiple collision avoidance targets is only an example of the unmanned surface vehicle collision avoidance terminal device based on multiple collision avoidance targets, and does not constitute a limitation on the unmanned surface vehicle collision avoidance terminal device based on multiple collision avoidance targets, and can include more or fewer components than the above, or combine certain components, or different components, for example, the unmanned surface vehicle collision avoidance terminal device based on multiple collision avoidance targets can also include an input and output device, a network access device, a bus, etc., and the present application does not limit this.

[0129] Further, as an executable solution, the processor can be a central processing unit (CPU), and can also be other general-purpose processors, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor or the like, which is a control center of the multi-collision-object-based unmanned ship collision avoidance terminal device, and connects all parts of the multi-collision-object-based unmanned ship collision avoidance terminal device through various interfaces and lines.

[0130] The memory can be used to store the computer program and / or modules, and the processor realizes various functions of the multi-collision-object-based unmanned ship collision avoidance terminal device by running or executing the computer program and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system and at least one application required by a function; and the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0131] The application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the steps of the method provided in the embodiments of the application.

[0132] The modules / units of the unmanned ship collision avoidance terminal equipment based on multi-collision-avoidance objects, if realized in the form of software function units and sold or used as independent products, can be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware. The computer program can be stored in a computer readable storage medium. When the computer program is executed by a processor, the steps of each method embodiment can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files or some intermediate forms, etc. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM) and software distribution medium, etc.

[0133] Although the present application has been specifically shown and described with respect to the preferred embodiments, it should be understood that various changes in form and details can be made therein without departing from the spirit and scope of the application as defined in the appended claims.

Claims

1. A collision avoidance method for an unmanned surface vehicle based on multi-collision-avoiding objects, characterized in that, The method comprises the following steps: S1: collecting motion information of all moving targets in the sea area near the unmanned ship; S2: clustering all moving targets based on the collected motion information to obtain a plurality of target clusters; S3: calculating the collision avoidance area of each target cluster based on the current coordinates of each target included in the target cluster; S4: when the unmanned ship travels to the collision avoidance waypoint relative to the collision avoidance area of a certain target cluster, determining whether there is more than one intersection point between the homeward voyage route of the unmanned ship at this time and the collision avoidance area, if yes, entering S5; otherwise, controlling the unmanned ship to perform homeward voyage according to the homeward voyage route; S5: judging the heading angle between the unmanned surface vehicle and the cluster of objects whether the condition is satisfied if yes, go to S6; otherwise, control the unmanned surface vehicle to keep the previous heading to continue sailing until the re-navigation route of the unmanned surface vehicle and the current collision avoidance area of the cluster of objects do not have more than one intersection, and then control the unmanned surface vehicle to re-navigate according to the re-navigation route; R represents the radius of the collision avoidance area of the cluster of objects, represents the vertical component of the relative distance between the cluster of objects and the unmanned surface vehicle; S6: controlling the unmanned ship to deflect to the right and make it travel according to the deflected route until there is no more than one intersection point between the homeward voyage route of the unmanned ship and the current collision avoidance area of the target cluster, and then controlling the unmanned ship to perform homeward voyage according to the homeward voyage route.

2. The multi-collision-avoiding-object-based collision avoidance method for an unmanned surface vehicle according to claim 1, characterized in that: In step S2, the improved DBSCAN algorithm is used to cluster all moving targets; the improved DBSCAN algorithm adds speed difference constraints and heading difference constraints on the basis of the traditional DBSCAN algorithm, that is, the heading difference between any two targets belonging to the same target cluster is less than the heading difference threshold, and the speed difference between any two targets is less than the speed difference threshold.

3. The multi-collision-avoiding-object-based collision avoidance method for an unmanned surface vehicle according to claim 1, characterized in that: When clustering all moving targets by using the improved DBSCAN algorithm, a neighborhood radius constraint is included, that is, the neighborhood radius ε of any two targets belonging to the same target cluster should be less than or equal to 4 times the safety distance of the target.

4. The multi-collision-avoiding-object-based collision avoidance method for an unmanned surface vehicle according to claim 1, characterized in that: When the unmanned ship is deflected to the right in step S6, the calculation formula of the deflection angle is: where θ represents the deflection angle, v U represents the speed of the unmanned surface vehicle, v C represents the speed of the target cluster.

5. An unmanned surface vehicle collision avoidance terminal device based on multi-collision-avoiding object, characterized in that: The computer program is executed by the processor to realize the steps of the method according to any one of claims 1-4.

6. A computer readable storage medium storing a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the method according to any one of claims 1-4.

Images