CN116859928A - Cluster target tracking method based on navigator self-adaption - Google Patents
Cluster target tracking method based on navigator self-adaption Download PDFInfo
- Publication number
- CN116859928A CN116859928A CN202310836505.0A CN202310836505A CN116859928A CN 116859928 A CN116859928 A CN 116859928A CN 202310836505 A CN202310836505 A CN 202310836505A CN 116859928 A CN116859928 A CN 116859928A
- Authority
- CN
- China
- Prior art keywords
- target
- tracking
- leader
- cluster
- pilot
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 230000015572 biosynthetic process Effects 0.000 claims abstract description 48
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 39
- 230000003993 interaction Effects 0.000 claims abstract description 23
- 239000011159 matrix material Substances 0.000 claims abstract description 5
- 238000013213 extrapolation Methods 0.000 claims abstract description 4
- 101100057959 Mus musculus Atxn1l gene Proteins 0.000 claims description 46
- 238000005056 compaction Methods 0.000 claims description 9
- 238000013459 approach Methods 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 230000003068 static effect Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000013461 design Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 6
- 230000008901 benefit Effects 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 230000006872 improvement Effects 0.000 description 3
- 239000003795 chemical substances by application Substances 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000008602 contraction Effects 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 229910003460 diamond Inorganic materials 0.000 description 1
- 239000010432 diamond Substances 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000000178 monomer Substances 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000000087 stabilizing effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Landscapes
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
The application discloses a cluster target tracking method based on pilot self-adaption, which belongs to the technical field of unmanned cluster autonomous control and comprises the following steps: in the tracking approaching stage, the clusters adopt a pilot-following mode, real vessels in the clusters are determined to be pilots through a KM algorithm according to a cluster planning route, and the rest vessels in the clusters follow the pilots in a specified formation; entering a tightening matrix stage when the opening angle of the cluster relative to the target boat, namely the tracking angle is larger than a set threshold value; in the compact-array stage, the clusters take the pose of the target boat obtained by extrapolation of a Starnet-based global information interaction track prediction algorithm as a virtual pilot at the next moment, all real boats in the clusters are followers, and the target boats fall down in sequence according to potential point distribution. The application can reduce the redundancy of the planned path, thereby reducing the time consumption of stable tracking of the cluster and solving the problem of target tracking of multi-unmanned-ship cluster formation.
Description
Technical Field
The application belongs to the technical field of unmanned cluster autonomous control, and particularly relates to a cluster target tracking method based on pilot self-adaption.
Background
The problem of target tracking of multi-unmanned ship cluster formation is one of the hot problems in the current control field, and is widely applied to various aspects of unmanned cluster combat. For the cluster dynamic target tracking problem, common solving methods can be divided into two types, one type is a prediction-planning-execution method (PPE method), a tracking position is predicted by acquiring motion information of a target, and the control quantity of the unmanned aerial vehicle is calculated and tracked according to the relative position; the second type is a guidance method, wherein the control quantity of the tracking robot is obtained through the speed and position information of the target and the tracking unmanned ship, and finally, the stable tracking is successfully maintained.
For cluster target tracking, the current common practice is to directly control and track the unmanned ship towards the estimated position of the target for direct tracking. However, this method is often ineffective in the case of a target with escape behavior and a large estimated noise. Therefore, the conventional trapping algorithm only predicts the short-term state of the target, i.e. predicts the position of the target after a certain time interval, and then continuously plans and adjusts the path according to the real-time position of the target. However, this method of path planning by segments is not globally optimal, often contains some redundancy, and the process of forming stable tracking is elongated. Therefore, an algorithm strategy for fully utilizing the existing motion information of the target is sought in the early stage of forming tracking, so that the path of the unmanned ship cluster is accurately and smoothly tracked.
Disclosure of Invention
Aiming at the defects or improvement demands of the prior art, the application provides a cluster target tracking method based on a self-adaptive navigator strategy, which can reduce redundancy of a planned path, thereby reducing time consumption of stable tracking of cluster formation and solving the problem of target tracking of cluster formation of multiple unmanned boats.
In order to achieve the above object, the present application provides a method for tracking a cluster target based on pilot adaptation, comprising:
in the tracking approaching stage, the clusters adopt a pilot-following mode, real vessels in the clusters are determined to be pilots through a KM algorithm according to a cluster planning route, and the rest vessels in the clusters follow the pilots in a specified formation;
entering a tightening matrix stage when the opening angle of the cluster relative to the target boat, namely the tracking angle is larger than a set threshold value;
in the compact-array stage, the clusters take the pose of the target boat obtained by extrapolation of a Starnet-based global information interaction track prediction algorithm as a virtual pilot at the next moment, all real boats in the clusters are followers, and the target boats fall down in sequence according to potential point distribution.
In some alternative embodiments, the stable tracking mark is that each boat in the cluster is distributed in a semicircular area taking the target boat as a center, and the stable tracking mark has a trapping interception condition.
In some optional embodiments, in the tracking approach stage, the cluster adopts a pilot-following mode, and a real boat in the cluster is determined to be a pilot through KM algorithm according to a cluster planning route, and the remaining boats in the cluster follow the pilot in a specified formation, including:
the navigator is expressed as a Leader, the Target boat is expressed as a Target, the line of sight LOS is a ray formed by pointing the Target by the Leader, and the angle formed by the LOS and the x-axis is the line of sight angle sigma;
the distance vector from the Leader to the Target is defined as r, and the Leader speed is V Leader Target speed of Target boat is V Target ,V Leader The included angle formed by the positive direction of LOS is eta, V Target Form an included angle eta with the LOS positive direction T θ is the heading angle of the pilot;
assuming constant pilot speed, by V Leader 、V Target Eta, eta and eta T And solving an ideal course angle theta of the navigator in real time, so that all the other vessels in the cluster follow the navigator in a specified formation.
In some alternative embodiments, the V Leader 、V Target Eta, eta and eta T Solving the ideal course angle theta of the pilot in real time comprises the following steps:
from the following componentsIs-> Determining a relative motion equation of the navigator-target boat;
by adjusting the speed direction of the pilotSubstituting the relative motion equation of the navigator-target boat to obtain: v (V) Leader ·sinη=V Target ·cosη T The velocity components of the two in the LOS normal direction are always equal, and tracking approaching on the two-dimensional plane is stably converted into the tracking problem in the LOS tangential direction;
and thenI.e. < ->
In some optional embodiments, the cluster takes the next moment pose of the target boat extrapolated by the global information interaction track prediction algorithm based on Starnet as a virtual navigator, and the cluster comprises:
constructing two types of neural networks, namely a Host Network is a track prediction Network based on LSTM, a Hub Network is a global time sequence interaction calculation Network based on LSTM, and a static map module of the Hub Network obtains a fixed-length feature vector s by receiving position information of all targets at the same moment, a full-connection Network and a maximum pooling operation t ;
Dynamic map module uses LSTM network to pair feature vector s t Performing time sequence coding to finally obtain a global interaction vector r t The Host Network firstly uses the current position of the target boatDynamic map r t Middle inquiry about +.>The position of the user is marked by the dot multiplication operation>And interaction->Inputting together the location of moments under LSTM network predictions
In some alternative embodiments, the position (x) of the next moment of the target is derived from a Starnet-based global information interaction trajectory prediction algorithm c ,y c ) Then the virtual pilot plans the route point coordinates (X Leader ,Y Leader ) The method comprises the following steps:t_interval represents a unit time step.
In some alternative embodiments, the position leader (k+1) of the leader at the next moment defines the target (k) as the current position of the target boat, the leader (k+1) projects as a leader image on the y-axis where the target (k) is located, and two equidistant auxiliary lines L are made by the leader image each 1 ,L 2 The distance is d, the line formed by leader (k+1) and leader image is the formation radius R, the circle formed by taking the leader image as the center and R as the radius is L 1 L 2 Minor arc p between 1 p 2 Namely, the target position set that the tracking boat group should reach at the next moment is set with the tracking angle epsilon=const, and the leader image is p
In some alternative embodiments, during the approach phase of tracking, the angle p 1 pp 2 <The tracking angle epsilon, d is a set value, R is relatively large,the two are far apart, and the abscissa of the planned position point at the next moment of each unmanned ship is as follows:
in some alternative embodiments, during the matrix compaction phase, +_p- 1 pp 2 >The tracking angle epsilon, d is a set value, R is relatively small, the distance between the tracking angle epsilon and the R is sufficiently short, and the horizontal coordinate point and the vertical coordinate point of the planned position of each unmanned ship at the next moment are as follows: wherein, the pilot is virtual pilot at this moment, and all real boats are the follower:
after entering the array type compaction stage, the cluster formation is converted into a semicircle, unmanned boats positioned at two sides of the formation are finally distributed at the left side and the right side, interception and capturing conditions are formed, the radius of the semicircle formation is set to be R_final, when the chord length corresponding to two ends of the arc in the array type compaction stage is smaller than 2R_final, interception tail conditions are considered to be provided, and at the moment, the corresponding formation radius is as follows:
in general, the above technical solutions conceived by the present application, compared with the prior art, enable the following beneficial effects to be obtained:
the cluster target tracking design can effectively control the boat cluster to track, approach and stabilize the tracked target, and then the cluster target tracking design is further contracted to create conditions for capturing and intercepting. The design is optimized in terms of planning path length, stabilizing tracking time consumption, compared to conventional designs.
Drawings
FIG. 1 is a schematic diagram of a hybrid control-based cluster formation control algorithm according to an embodiment of the present application;
FIG. 2 is a schematic diagram of defining potential points according to an embodiment of the present application;
FIG. 3 is a schematic diagram of a single point tracking problem according to an embodiment of the present application;
FIG. 4 is a schematic diagram of a tracking model according to an embodiment of the present application;
FIG. 5 is a schematic diagram of a first stage tracking approach stage according to an embodiment of the present application;
FIG. 6 is a schematic diagram of a second stage array compaction stage according to an embodiment of the present application;
FIG. 7 is a flowchart of a cluster target tracking algorithm according to an embodiment of the present application;
fig. 8 is a test scenario diagram of a cluster target tracking algorithm according to an embodiment of the present application.
Detailed Description
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application. In addition, the technical features of the embodiments of the present application described below may be combined with each other as long as they do not collide with each other.
The existing algorithm used by the application is as follows:
(1) Global information interaction cluster formation control algorithm based on hybrid control
The virtual navigator is set as a virtual moving rigid body with a position practically related to a certain real boat, and a virtual structure is established by taking the virtual moving rigid body as a basic coordinate system, and each boat in the formation advances towards a reserved landing site (determined by a built-in KM algorithm) in the virtual structure.
Determining pose information of a virtual navigator, as shown in fig. 1, calculating a point with the shortest distance between the mass center of the cluster and a desired navigation line, and marking the point as a reference point 1; for each boat in the cluster, determining a boat closest to the range of the reference point by a KM algorithm aiming at the reference point 1, and marking the boat as the boat 1. The set point 2 is the projection point of the boat 1 on the expected route, namely the landing after the transverse error is eliminated. The virtual pilot takes the position of the point 2 as a real-time position, and the tangential direction of the point 2 on the route is the course of the virtual pilot.
The virtual structure design and the virtual structure method refer to that a pilot is not designated in unmanned ship formation to which the virtual structure method is applied, and an unmanned ship cluster is regarded as a unified virtual rigid body. The unmanned vessels are formed into a virtual geometric center, and all unmanned vessels in the formation are formed into sails and execute tasks according to a certain control strategy and formation by referring to the virtual geometric center. The virtual structure method has the advantages that dependence on a pilot in the pilot following algorithm is relieved, meanwhile, the problem of error conduction is solved, and accuracy of formation control is improved.
The formation control method used in the application combines a virtual structure method and a pilot following algorithm, and can adaptively and rapidly form formation, namely, a plurality of intelligent agents start to form a regular shape, such as a horizontal line shape, a vertical line shape, a diamond shape, a triangle shape, a square shape and the like, according to formation control at a completely irregular starting point; the formation can be automatically judged, the formation holding refers to the task that the formation holding is completed by the individual in the system according to the interaction of the adopted method (L-L control method) when the fixed formation formed according to the premise continuously advances towards the target point in the movement process of the multi-agent system; when the formation integrally encounters an obstacle, reasonable avoiding movement is required from the aspect of safety, if a team with too narrow allowable passing path cannot pass safely on the basis of keeping the original formation, the formation of the formation is required to be changed to pass through the obstacle area safely in sequence and then restored to the original formed fixed formation, and finally the target point is reached.
(2) Potential point-based formation capture strategy
Potential points are based on the discretization of an artificial potential field into a series of points distributed around the tracked target to facilitate the dynamic containment task to resolve the tracked boat's traveling target.
As shown in FIG. 2, the set of potential points of the tracking target isThe radius of potential field is R, the arc length between adjacent potential points is a fixed value, +.>For the sequence->The reference "potential point" in (a) has a direction which is consistent with the moving direction of the tracking target, and comprises:
let xi l =arctan((y l -y e ),(x l -x e ) To be xi) l Sequencing from small to large to generate a sequence { ζ } l '}. Without losing generality, let Construction of the sequence { ζ } l ″},{ξ l "is { ζ } l ' sequence shift, ζ A Is { xi } l First element of } ". The sequence { ζ } l "andthe elements in the table are matched in sequence, and xi is calculated l The corresponding "potential point" of "is +.>The two generating and matching processes are task decision processes.
The application provides a cluster target tracking method based on pilot self-adaption, which comprises the following steps:
(1) Cluster tracking system model design
As shown in fig. 4, taking a tracking system composed of 5 unmanned vessels and 1 Target vessel as an example, consider a case where the tracking vessel Follower starts tracking from the side of the Target vessel Target motion direction. The tracking process is divided into two stages in the design of the tracking model, namely a tracking approaching stage and a tightening matrix stage.
1) In the tracking approaching stage, the clusters adopt a pilot-following mode, a real boat (a specific boat is determined through a KM algorithm according to a cluster planning route) in the boat group is taken as a pilot Leader, and the rest boats follow in a specified formation, so that the clusters can approach a tracking target quickly, stably, coordinately and synchronously.
2) In the compact-array stage, the clusters take the pose of the target boat obtained by extrapolation of a Starnet-based global information interaction track prediction algorithm as a virtual pilot at the next moment, all real boats in the clusters are followers, and the target boats fall down in sequence according to 'potential point' distribution.
3) The decision condition for the transition from the first stage to the second stage (i.e. pilot change) is the opening angle of the cluster relative to the target boat, i.e. angle P in FIG. 5 1 PP 2 And if the value is larger than the set threshold value, the tracking angle is recorded.
4) The stable tracking mark is that each boat is distributed in a semicircular area taking a Target of a Target boat as a circle center, and the stable tracking mark has a trapping and intercepting condition. The two-dimensional information of all boats at the current moment in the system is designed to be observable in real time, the observation results are shared globally, and the clusters make decisions according to the position information. The system time interval (unit time step) is set to t_interval, and the current time period is denoted as k. The initial value of k is 1, and when the target starts to move, the value of k is added with 1 every system period; the increase is stopped after the stable tracking is formed, and the k value is the time consuming tracking. Fig. 6 shows a tracking process from the kth cycle to the kth+n cycle.
(2) Tracking target track prediction method
The conventional track algorithm has the problem of calculating pairwise interactions, and even Message Passing based on the Attention mechanism of Attention is difficult to consider global interactions. Based on improvement of the defect, the track prediction algorithm adopts an unmanned ship track interaction prediction algorithm based on a neural network StarNet, and global interaction among unmanned clusters of ships is fully considered. The core thought is as follows: the locations of all targets at each instant in time may form a static "map" that becomes a dynamic map with timing information over time. The dynamic diagram records the motion information of the targets in each area, wherein the motion information is obtained by mutually fusing all the targets together instead of being formed by mutually interacting two by two independently. In the prediction stage of each target, the obstacle movement information of the area at the historical moment can be queried in the map only according to the position of the target. By constructing the neural network in a mode of sharing the global interaction map and individual inquiry, the global interaction and the compression of the calculation cost can be achieved.
The algorithm mainly builds two types of neural networks. Host Network is an LSTM-based trajectory prediction Network, and Hub Network is an LSTM-based global timing interaction computing Network. The static map module of Hub Network is a fixed-length feature vector s obtained by receiving the position information of all targets at the same time, fully-connected Network and maximum pooling operation t The method comprises the steps of carrying out a first treatment on the surface of the Second, the dynamic map module uses LSTM network to make the above-mentioned feature vector s t Performing time sequence coding to finally obtain a global interaction vector r t . The Host Network first predicts the current position of the target according to the target (i.e. when predicting the target's next time position)Dynamic map r t Middle inquiry about +.>The position of the user is marked by the dot multiplication operation>And interaction->Inputting the location +.of the moment under LSTM network prediction together>
(3) Pilot adaptive strategy design
The first stage: tracking model for navigator in tracking approaching stage
The navigator of the tracking cluster is a single tracking problem with respect to the target. First, a description is given of a monomer tracking model. Referring to fig. 3, the Leader is a navigator, the Target is a Target, and the Leader points to a line of sight (LOS) which is a ray formed by the Target. The angle formed by the LOS and the x-axis is the line of sight angle sigma. The distance vector from the Leader to the Target is defined as r, and the Leader speed is V Leader Target speed of the Target boat is defined as V Target ,V Leader The included angle formed by the positive direction of LOS is eta, V Target Form an included angle eta with the LOS positive direction T θ is the heading angle of the pilot. The model assumes that the speed of the pilot is constant, and the core purpose of the tracking model is to be able to solve the ideal course angle theta of the pilot in real time.
The relative equation of motion of the "navigator-target boat" is:
by adjusting the speed direction of the pilotSubstituting the above formula to obtain:
V Leader ·sinη=V Target ·cosη T (3)
that is, the velocity components of the two in the LOS normal direction are always equal, and the tracking approach on the two-dimensional plane is stably converted into the tracking problem in the LOS tangential direction, and the following problem is obtained by the following formula (3):
namely:
and a second stage: compact array stage virtual pilot location planning
The tracking model shows that the value range of eta is [ -90, 90]For the simplified model, only when V Leader >V Target The stable tracking can be ensured; obtaining the position (x) of the next moment of the target by using a Starnet-based global information interaction track prediction algorithm c ,y c ) Then the virtual pilot plans the route point coordinates (X Leader ,Y Leader ) The method comprises the following steps:
(4) Cluster formation transformation design (guidance) based on potential points
The position leader (k+1) of the navigator leader at the next moment; defining a target (k) as a current position of the target; the leader (k+1) projects as a leader image on the y-axis where the target (k) is located, and two equidistant auxiliary lines L are respectively made by the leader image 1 ,L 2 The distances are d. The link between leader (k+1) and leader image is the formation radius R. A circle formed by taking the LeaderImage as a circle center and R as a radius is L 1 L 2 Minor arc p between 1 p 2 The target position set which is needed to be reached by the following boat group at the next moment is obtained. Let tracking angle epsilon=const, leader image be p, then:
the following two cases are discussed:
1) The first stage: at the tracking approaching stage, the angle p 1 pp 2 <Tracking angle epsilonD is a set value, R is relatively large, the distance between the two is relatively long, and the abscissa of the planned position point at the next moment of each unmanned ship is as follows:
wherein, the following is a follower x ,follower y Respectively represent the horizontal and vertical coordinates of the tracking boat, and the LeaderImage x ,LeaderImage y The abscissa and ordinate of the loaderimage are respectively indicated.
2) And a second stage: the array contraction stage is shown in FIG. 6, where +.p 1 pp 2 >The tracking angle epsilon, d is a set value, R is relatively small, the distance between the tracking angle epsilon and the R is sufficiently short, and the horizontal coordinate point and the vertical coordinate point of the planned position of each unmanned ship at the next moment are as follows: wherein, the pilot is virtual pilot at this moment, and all real boats are the follower:
wherein, the following is a follower x ,follower y Respectively represent the horizontal and vertical coordinates of the tracking boat and the Leader x ,Leader y Respectively representing the abscissa and the ordinate of the virtual pilot;
after entering the array type compaction stage, the cluster formation is converted into a semicircle, unmanned boats positioned at two sides of the formation are finally distributed at the left side and the right side, interception and capturing conditions are formed, the radius of the semicircle formation is set to be R_final, when the chord length corresponding to two ends of the arc in the array type compaction stage is smaller than 2R_final, interception tail conditions are considered to be provided, and at the moment, the corresponding formation radius is as follows:
the unmanned ship cluster controller is built as shown in fig. 7, a homemade algorithm test training system is connected after codes are written, a 10-sea-area task area is designed under a simulation environment, a warning area is shown as a large square area in the following figure, a ship cluster is responsible for regional warning patrol, two unmanned ships are respectively arranged on each side of the area and are responsible for patrol warning, and when an unknown identity ship enters the area, two unmanned ships are dispatched for tracking. And testing whether the cluster target tracking algorithm can effectively control the unmanned ship to stably track the target.
After 1000 times of testing, the result shows that the cluster target tracking algorithm can effectively realize the cluster target tracking function and stably track time consumption as shown in fig. 8. When the target moves linearly, the method has little advantage compared with the existing algorithm (the existing algorithm is 74.8s, and the algorithm of the application is 71.6 s); when the target is maneuvered, the method has certain advantages compared with the existing algorithm (the existing algorithm is 102.8s, and the algorithm is 93.6 s).
It should be noted that each step/component described in the present application may be split into more steps/components, or two or more steps/components or part of operations of the steps/components may be combined into new steps/components, according to the implementation needs, to achieve the object of the present application.
It will be readily appreciated by those skilled in the art that the foregoing description is merely a preferred embodiment of the application and is not intended to limit the application, but any modifications, equivalents, improvements or alternatives falling within the spirit and principles of the application are intended to be included within the scope of the application.
Claims (9)
1. The cluster target tracking method based on the pilot self-adaption is characterized by comprising the following steps of:
in the tracking approaching stage, the clusters adopt a pilot-following mode, real vessels in the clusters are determined to be pilots through a KM algorithm according to a cluster planning route, and the rest vessels in the clusters follow the pilots in a specified formation;
entering a tightening matrix stage when the opening angle of the cluster relative to the target boat, namely the tracking angle is larger than a set threshold value;
in the compact-array stage, the clusters take the pose of the target boat obtained by extrapolation of a Starnet-based global information interaction track prediction algorithm as a virtual pilot at the next moment, all real boats in the clusters are followers, and the target boats fall down in sequence according to potential point distribution.
2. The method of claim 1, wherein the stable tracking is marked by distributing all boats in the cluster in a semicircular area with a target boat as a center, and the stable tracking is provided with a trapping interception condition.
3. The method according to claim 2, wherein in the tracking approach phase, the cluster adopts a pilot-following mode, and a real boat in the cluster is determined to be a pilot through KM algorithm according to a cluster planning route, and the remaining boats in the cluster follow the pilot in a specified formation, including:
the navigator is expressed as a Leader, the Target boat is expressed as a Target, the line of sight LOS is a ray formed by pointing the Target by the Leader, and the angle formed by the LOS and the x-axis is the line of sight angle sigma;
the distance vector from the Leader to the Target is defined as r, and the Leader speed is V Leader Target speed of Target boat is V Target ,V Leader The included angle formed by the positive direction of LOS is eta, V Target Form an included angle eta with the LOS positive direction T θ is the heading angle of the pilot;
assuming constant pilot speed, by V Leader 、V Target Eta, eta and eta T And solving an ideal course angle theta of the navigator in real time, so that all the other vessels in the cluster follow the navigator in a specified formation.
4. A method according to claim 3, wherein said step of generating a signal is performed by V Leader 、V Target Eta, eta and eta T Solving the ideal course angle theta of the pilot in real time comprises the following steps:
from the following componentsIs-> Determination ofThe relative equation of motion of the pilot-target boat;
by adjusting the speed direction of the pilotSubstituting the relative motion equation of the navigator-target boat to obtain: v (V) Leader ·sinη=V Target ·cosη T The velocity components of the two in the LOS normal direction are always equal, and tracking approaching on the two-dimensional plane is stably converted into the tracking problem in the LOS tangential direction;
and thenI.e. < ->
5. The method of claim 4, wherein the cluster takes a next moment pose of the target boat extrapolated by a Starnet-based global information interaction trajectory prediction algorithm as a virtual pilot, and comprises:
constructing two types of neural networks, namely a Host Network is a track prediction Network based on LSTM, a Hub Network is a global time sequence interaction calculation Network based on LSTM, and a static map module of the Hub Network obtains a fixed-length feature vector s by receiving position information of all targets at the same moment, a full-connection Network and a maximum pooling operation t ;
Dynamic map module uses LSTM network to pair feature vector s t Performing time sequence coding to finally obtain a global interaction vector r t The Host Network firstly uses the current position of the target boatDynamic map r t In the area of the current position of the query itselfThe position of the user is marked by the dot multiplication operation>And interaction->Inputting the location +.of the moment under LSTM network prediction together>
6. The method according to claim 5, wherein the position (x) of the next moment of the object is obtained by a Starnet-based global information interaction trajectory prediction algorithm c ,y c ) Then the virtual pilot plans the route point coordinates (X Leader ,Y Leader ) The method comprises the following steps:t_interval represents a unit time step.
7. The method of claim 6, wherein the position leader (k+1) of the leader at the next moment defines the target (k) as the current position of the target boat, the leader (k+1) projects as a leader image on the y-axis where the target (k) is located, and two equidistant auxiliary lines L are made by the leader image each 1 ,L 2 The distance is d, the line formed by leader (k+1) and leader image is the formation radius R, the circle formed by taking the leader image as the center and R as the radius is L 1 L 2 Minor arc p between 1 p 2 Namely, the target position set that the tracking boat group should reach at the next moment is set with the tracking angle epsilon=const, and the leader image is p
8. The method according to claim 7,it is characterized in that in the tracking approaching stage, the angle p is 1 pp 2 <The tracking angle epsilon, d is a set value, R is relatively large, the distance between the tracking angle epsilon and the R is relatively long, and the abscissa of the planned position point at the next moment of each unmanned ship is as follows:
wherein, the following is a follower x ,follower y Respectively represent the horizontal and vertical coordinates of the tracking boat, and the LeaderImage x ,LeaderImage y The abscissa and ordinate of the loaderimage are respectively indicated.
9. The method of claim 7, wherein during the array compaction phase, the angle p 1 pp 2 The tracking angle epsilon, d is a set value, R is relatively small, the distance between the tracking angle epsilon and the tracking angle d is enough, and the horizontal coordinate point and the vertical coordinate point of the planned position of each unmanned ship at the next moment are as follows: wherein, the pilot is virtual pilot at this moment, and all real boats are the follower:
wherein, the following is a follower x ,follower y Respectively represent the horizontal and vertical coordinates of the tracking boat and the Leader x ,Leader y Respectively representing the abscissa and the ordinate of the virtual pilot;
after entering the array type compaction stage, the cluster formation is converted into a semicircle, unmanned boats positioned at two sides of the formation are finally distributed at the left side and the right side, interception and capturing conditions are formed, the radius of the semicircle formation is set to be R_final, when the chord length corresponding to two ends of the arc in the array type compaction stage is smaller than 2R_final, interception tail conditions are considered to be provided, and at the moment, the corresponding formation radius is as follows:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310836505.0A CN116859928A (en) | 2023-07-07 | 2023-07-07 | Cluster target tracking method based on navigator self-adaption |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310836505.0A CN116859928A (en) | 2023-07-07 | 2023-07-07 | Cluster target tracking method based on navigator self-adaption |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116859928A true CN116859928A (en) | 2023-10-10 |
Family
ID=88233628
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310836505.0A Pending CN116859928A (en) | 2023-07-07 | 2023-07-07 | Cluster target tracking method based on navigator self-adaption |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116859928A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115686004A (en) * | 2022-10-27 | 2023-02-03 | 南京长峰航天电子科技有限公司 | Path dynamic planning method based on unmanned ship distributed formation control |
-
2023
- 2023-07-07 CN CN202310836505.0A patent/CN116859928A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115686004A (en) * | 2022-10-27 | 2023-02-03 | 南京长峰航天电子科技有限公司 | Path dynamic planning method based on unmanned ship distributed formation control |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Sang et al. | The hybrid path planning algorithm based on improved A* and artificial potential field for unmanned surface vehicle formations | |
CN111780777B (en) | Unmanned vehicle route planning method based on improved A-star algorithm and deep reinforcement learning | |
Liu et al. | Self-adaptive dynamic obstacle avoidance and path planning for USV under complex maritime environment | |
Kim et al. | Perception-driven navigation: Active visual SLAM for robotic area coverage | |
Wang et al. | A COLREGs-based obstacle avoidance approach for unmanned surface vehicles | |
CN110362089A (en) | A method of the unmanned boat independent navigation based on deeply study and genetic algorithm | |
Yu et al. | Collision-free trajectory generation and tracking for UAVs using Markov decision process in a cluttered environment | |
Blaich et al. | Fast grid based collision avoidance for vessels using A∗ search algorithm | |
CN110069075A (en) | A kind of cluster super maneuver barrier-avoiding method of imitative dove group emergency avoidance mechanism | |
CN116859928A (en) | Cluster target tracking method based on navigator self-adaption | |
Fan et al. | Formation control of multiple unmanned surface vehicles using the adaptive null-space-based behavioral method | |
Li et al. | Toward time-optimal trajectory planning for autonomous ship maneuvering in close-range encounters | |
Chen et al. | Survey on cooperative control for waterborne transport | |
CN112327872A (en) | Double unmanned ship cooperative track tracking method for oil spill containment | |
Liang et al. | Economic MPC-based planning for marine vehicles: Tuning safety and energy efficiency | |
CN114088094A (en) | Intelligent route planning method and system for unmanned ship | |
CN113325852A (en) | Leader follower mode-based control method for formation change of multiple intelligent agents in advancing process | |
CN113093804A (en) | Unmanned ship formation control method and control system based on inversion sliding mode control | |
CN116257067A (en) | Collaborative formation multi-level planning control method for unmanned ship cluster | |
Jose et al. | Navigating the Ocean with DRL: Path following for marine vessels | |
CN114879694A (en) | Unmanned ship automatic collision avoidance method based on probability game theory framework | |
CN116301011A (en) | Multi-agent efficient formation obstacle avoidance method | |
CN114690772A (en) | Meeting time-space prediction method and system for sea area non-tracking navigation ship | |
Yao et al. | UAV/USV Cooperative Trajectory Optimization Based on Reinforcement Learning | |
CN114545971A (en) | Multi-agent distributed flyable path planning method, system, computer equipment and medium under communication constraint |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |