A method and system for generating autonomous tasks on a satellite for ocean mobile targets

Abstract

The present invention provides a method and system for generating autonomous tasks on a satellite for an ocean mobile target, by constructing an ocean mobile target motion model; and predicting the trajectory of the ocean mobile target according to the ocean mobile target motion model to obtain a set of tasks to be generated; and assigning task elements to the set of tasks to be generated to form an autonomous task on a satellite. The present invention obtains the possible position of the target task through a target motion prediction model with double constraints under Gaussian distribution, so that tracking and observation of ocean mobile targets is possible, and the uncertainty of target motion is measured. Therefore, necessary input data can be provided for the planning of subsequent multi-satellite tasks, thereby realizing rapid task planning and task observation and tracking.

Description

Technical Field

[0001] The present invention relates to the field of mobile target task management, and in particular to an on-board autonomous task generation method and system for ocean mobile targets. Background Art

[0002] Complete onboard autonomous mission management should be oriented towards the full-cycle and full-life management of the target, including target state maintenance, motion prediction, tracking task generation, and tracking task attribute configuration. For typical dynamic targets such as mobile targets, although there are many prediction models for mobile targets, and the methods cover classic trajectory fitting to machine learning trajectory prediction, there is a lack of mobile target prediction models under adversarial game characteristics, and in the absence of historical data, it is impossible to use machine learning and other methods to predict mobile targets. The many uncertainties of the target need to be transformed and associated with the satellite's tracking task to facilitate the scheduling of the tracking task, but current research is limited to simple motion prediction and task selection mechanisms. Summary of the invention

[0003] The technical problem to be solved by the present invention is how to track ocean mobile targets on a satellite and form the input of an on-board autonomous task. A method and system for generating on-board autonomous tasks for ocean mobile targets are proposed.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] A method for generating autonomous tasks on a satellite for an ocean mobile target comprises the following steps:

[0006] Step 1: Construct a motion model of ocean moving targets;

[0007] Step 2: predicting the trajectory of the ocean mobile target according to the ocean mobile target motion model to obtain a task set to be generated;

[0008] Step 3: Assign task elements to the task set to be generated to form an on-board autonomous task.

[0009] Furthermore, the ocean moving target motion model is:

[0010] Target The corresponding position of the moment under PGFS is , it is necessary to recursively infer the next position of the target based on the current trajectory point state of the target The goal is The plane where the large arc of the motion trajectory in a short time is located is , according to the direction of motion and the right-hand system, the normal vector of the great circle is The Earth's North Pole is ,flat Normal vector determined by right-hand system for

[0011] (4)

[0012] Depend on Rotation angle Can get ,Right now

[0013] (5)

[0014] in, is a quaternion, Its conjugate quaternion satisfies the following conditions

[0015] (6)

[0016] Target elapsed time from Exercise to , and its movement distance is , the corresponding arc of the Earth's center angle is ,therefore, The coordinates of can be obtained by the following formula

[0017] (7)

[0018] in, is a quaternion, Its conjugate quaternion satisfies the following conditions

[0019] (8)

[0020] Furthermore, in step 2, the trajectory of the ocean mobile target is predicted according to the ocean mobile target motion model to obtain the specific method of the task set to be generated is:

[0021] Step 2.1: Grid the area where the ocean moving target is located and construct a grid index;

[0022] Step 2.2: Recognize the optical image based on the satellite-borne image recognition algorithm to obtain the position and heading of the ocean mobile target, and obtain the speed of the ocean mobile target based on the SAR payload;

[0023] Step 2.3: predicting the fan-shaped area where the ocean moving target is located based on the most recently obtained position, heading and speed of the ocean moving target;

[0024] Step 2.4: All grids within the fan-shaped area are taken as the task set to be generated.

[0025] Furthermore, the sector area is:

[0026] According to the recent k In the identification of the ocean moving target, the position of the ocean moving target is obtained P k and azimuth γ k Information, then the ocean moving target is in the time interval κ The position after is P k As the center of the circle, v max κ is the radius and the heading γ k The symmetrical fan angle is δ max κ The fan-shaped area, v max Indicates that in the recent k The maximum speed of the ocean moving target during the measurement process, δ max Indicates that in the recent k The maximum angular velocity of the ocean moving target during the measurement.

[0027] Furthermore, the task elements in step 3 include determining the load mode of the task, the number of times the task is generated, and the priority of the task.

[0028] Furthermore, the load mode of the task is determined by:

[0029] 1) If image recognition is performed in the imaging results of each observation period, and the information of the ocean mobile target can be found and identified, a high-resolution payload mode is used, and the task is the smallest grid of each division in the fan-shaped area;

[0030] 2) If image recognition is performed in the imaging results of the previous observation cycle and no ocean mobile target is found and identified, the load mode is the wide-width mode, and the task is a new grid divided by the width of the wide-width mode determined by the load mode. The level of the load mode and the number of times the task is generated are determined by the following formula;

[0031]

[0032] in, The working mode of the load is l The corresponding load width at level 1; The working mode of the load is l The corresponding load target recognition rate at level 1 is is the minimum target recognition rate; is the radius of the fan area obtained by using double constraints; k Represents the number of observations for a task.

[0033] Furthermore, the task priority needs to comprehensively consider the probability of a target appearing in a certain grid, the change in target priority caused by the mobility of the target, and the probability of multiple targets appearing in the same grid:

[0034] 1) The probability of a target appearing in a certain grid, based on the target’s maximum movement speed v max and maximum azimuth deviation velocity δ max Predict the target's position in the most recent target recognition P k and azimuth γ k information, it can be determined in the interval time κ After P k As the center of the circle, v max κ is the radius, and the fan angle symmetric about the heading is δ max κ In the fan-shaped area, assuming that the probability of the target appearing in the grids in the fan-shaped area is equal, the target appears in the grid g i,j The probability is:

[0035]

[0036] For speed, is the azimuth, is the variance of the velocity that follows a Gaussian distribution, is the mean of the azimuth angles that obey the Gaussian distribution, is the variance of the azimuth that follows the Gaussian distribution; is the mean of the distances that follow the Gaussian distribution, is the variance of the distance that follows the Gaussian distribution, Representation Grid g i,j The center point is the closest to the target position after target recognition when the target is observed. P k Distance, time After that, the target is at a distance of The distance is , The target is identified in the grid after the latest observation. g i,jThe azimuth of the center point, is the mean of the distances that follow the Gaussian distribution, is the mean value of the velocity that obeys the Gaussian distribution. Considering that the probability of the target being distributed in the grid in the fan-shaped area is 1, the normalization is used to obtain the probability of the target being distributed in the grid. g i,j The probability is:

[0037]

[0038] v max Indicates the maximum moving speed of the target. δ max Indicates the maximum azimuth deviation speed of the target γ k Indicates the azimuth of the target at the last observation; κ represents Interval time;

[0039] 2) Dynamic priority is introduced due to the change of target priority caused by the mobility of the target;

[0040]

[0041] is the maximum upper limit of the target dynamic priority, Indicates the basic priority predefined by the user. ot k For the k The target information time obtained by the target observation is: Desired tracking interval defined by the user;

[0042] Then at time t the target is in the grid g i,j The priority is expressed as

[0043] ;

[0044] 3) Consider the case where multiple targets appear in the same grid:

[0045] Indicates that at time t, the mth target appears in the i-th row and j-th column grid of the target observation rectangular area g i,j The probability within The value of the target appears in the grid after considering the dynamic priority in 2) g i,j The probability of is calculated;

[0046] but Indicates the overlay priority of the grids formed when multiple targets appear in the same grid during multi-target tracking.i, j The first i Row, No. j A grid of columns, m Indicates the first i Row, No. j The first column in the grid m goals, t Indicates time, | tgt | is the tracking target set;

[0047] Indicates the normalized priority of grid stacking when considering multiple targets appearing in the same grid;

[0048] 4) The target appears on the grid g i,j After coordination, it is changed to:

[0049] Furthermore, the task generation cycle of the target is:

[0050]

[0051] in, Indicates the user's desired tracking time interval for the target;

[0052] Indicates the timeliness of target motion prediction.

[0053] The present invention also provides an on-board autonomous mission generation system for ocean mobile targets, comprising the following modules:

[0054] Model building module: used to build the motion model of ocean moving targets;

[0055] Task set generation module: used to predict the trajectory of the ocean mobile target according to the ocean mobile target motion model constructed by the model construction module to obtain the task set to be generated;

[0056] Task element assignment module: used to assign task elements to the task set to be generated by the task set generation module to form an on-board autonomous task.

[0057] By adopting the above technical solution, the present invention has the following beneficial effects:

[0058] The present invention provides an on-board autonomous task generation method and system for ocean mobile targets. For non-time-sensitive mobile targets, a mobile target motion model is constructed based on the Gaussian distribution characteristics, and a target motion prediction model with dual constraints under Gaussian distribution is proposed to obtain the possible location of the target task, that is, the fan-shaped area, to form a satellite observation task; and assuming that the probability of the target appearing in each grid in the fan-shaped area is equal, the probability of the target appearing in the fan-shaped area grid can be obtained, and then based on whether the target appears after target recognition in each satellite observation image, the load mode of the task and the number of task generation times are selected, and the mobility of the task is taken into account, and the dynamic priority is introduced, and the situation where multiple targets appear in a grid at the same time is considered, and the priority is coordinated, thereby determining the various elements of the task. It makes it possible to track and observe ocean mobile targets, and realizes the measurement of target motion uncertainty. Therefore, it can provide necessary input data for the planning of subsequent multi-satellite tasks, thereby realizing rapid task planning and task observation tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flow chart of the system of the present invention;

[0060] Figure 2 It is a motion model for non-time-sensitive moving targets;

[0061] Figure 3 Schematic diagram of grid division for target area;

[0062] Figure 4 Motion prediction for time-insensitive targets under different conditions. DETAILED DESCRIPTION

[0063] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] In satellite mission planning, the satellite capability and payload constraints are taken into consideration to realize the autonomous generation of on-board missions, and the priority of generated tasks is coordinated based on the multi-attribute analysis of the tasks. The on-board autonomous mission management module is constructed by predicting the motion of mobile targets, generating on-board autonomous tasks, and coordinating priorities, so as to prepare input data for subsequent multi-satellite mission planning and scheduling. For mobile targets at sea, how to automatically generate on-board autonomous tasks is of great significance for rapid tracking.

[0065] Figures 1 to 4 A specific embodiment of a method for generating autonomous tasks on a satellite for an ocean mobile target according to the present invention is shown. Figure 1As shown, the following steps are included:

[0066] Step 1: Construct a motion model of ocean moving targets;

[0067] In this embodiment, to facilitate subsequent modeling, the mobile target at sea records track points at regular intervals while sailing, and the track between adjacent track points follows the large arc trajectory and the target speed is along the tangent direction of the large arc. When the mobile target at sea is sailing in the sea, the movement between adjacent track points is uniform and straight. The entire track of the target can be regarded as multiple sections of uniform, nearly uniform or perturbed curved motion. In order to quantitatively model and analyze the adversarial characteristics of the target, its adversarial nature is mainly reflected in the changes in heading and speed, and it is assumed that such changes obey a Gaussian distribution.

[0068] The ocean moving target motion model constructed in this embodiment uses a six-tuple to describe the target t The state at the moment can also be regarded as

[0069] The target trajectory is t A track point in time TP , ,in, t Indicates the time value. α Indicates the latitude of the ocean mobile target. β Indicates the longitude of the location. h Indicates the height of the location. v express t The speed at the moment, γ express t The azimuth of the heading at the moment.

[0070] Target i The running trajectory can be represented by a series of equal time intervals κ The trajectory point description is

[0071] Traj i = {TP j |κ ; j = 0 , 1 , ..., np i } , np i For the goal i According to the above assumptions, the target speed v Azimuth to target γ Changes δ It obeys Gaussian distribution, that is Compared with the complicated mechanical model, the present invention uses quaternion transformation to establish the motion model of the target. Quaternion can be used for spatial coordinate rotation transformation, and the subsequent motion trajectory of the target can be quickly obtained. Quaternion transformation comes from the document "Xiao Yelun. Modeling of Aerospace Vehicle Motion: Theoretical Basis of Flight Dynamics [M]. Modeling of Aerospace Vehicle Motion: Theoretical Basis of Flight Dynamics, 2003".

[0072] like Figure 2 As shown, the target is The corresponding position at the time under the proposed ground fixed system PGFS is , it is necessary to recursively infer the next position of the target based on the current trajectory point state of the target The goal is The plane where the large arc of the motion trajectory in a short time is located is , according to the direction of motion and the right-hand system, the normal vector of the great circle is The Earth's North Pole is ,flat Normal vector determined by right-hand system for

[0073] (4)

[0074] Depend on Rotation angle Can get ,Right now

[0075] (5)

[0076] in, For the four elements, Its conjugate quaternion satisfies the following conditions

[0077] (6)

[0078] Target elapsed time from Exercise to , and its movement distance is , the corresponding arc of the Earth's center angle is ,therefore, The coordinates of can be obtained by the following formula

[0079] (7)

[0080] in, For the four elements, Its conjugate quaternion satisfies the following conditions

[0081] (8)

[0082] According to the above motion model, the target position and direction at the next moment can be deduced from the target position and direction at the previous moment, and then the motion trajectory of the ocean moving target can be deduced. Compared with the target motion trajectory constructed by the traditional mechanical model, the derivation method of this motion trajectory uses the two-dimensional information of speed and direction under Gaussian distribution to generate, which is more concise and efficient.

[0083] Step 2: Predict the trajectory of the ocean mobile target according to the ocean mobile target motion model to obtain a task set to be generated.

[0084] The trajectory prediction method for the ocean moving target in this embodiment is:

[0085] Step 2.1: Grid the area where the ocean moving target is located and construct a grid index;

[0086] Users usually pay more attention to the situation in a certain key area. For non-time-sensitive mobile target tracking, the focus is also on the trajectory of the target in the area of ​​interest to facilitate the acquisition of its intention. Different satellites carry different payloads. The payload carried by the same satellite has multiple working modes. SAR satellite imaging has multiple modes corresponding to different widths. Therefore, when performing grid decomposition in the area where the mobile target is located, the following two principles need to be considered: 1) Grid division granularity principle. The positioning accuracy of the target, the task solution time, and the effective width of the sensor are all considered in the grid division granularity. High positioning accuracy requires a small grid division granularity, but the corresponding solution time is long. If the grid granularity is too large, although the required solution time is short, the accuracy is not enough, and a single strip cannot achieve coverage of the grid. 2) Grid division direction principle. The grid division direction needs to ensure the consistency of grid maintenance for different satellites and the convenience of calculation.

[0087] This embodiment uses the single scene width corresponding to the minimum payload field of view in the constellation to decompose the region, and uses an equidistant decomposition method along the longitude and latitude directions to ensure that each square grid is of the same size, which facilitates the consistency maintenance and calculation of the grid. Figure 3 ,The grid division for non-time-sensitive moving targets is mainly divided into three processes: regional boundary feature acquisition, regional grid division based on the minimum single scene, and fast grid index construction.

[0088] Step 2.1.1: Get region boundary features.

[0089] The region boundary feature acquisition is to obtain the region's feature vector, and then construct the outer rectangle of the target region to define the regular boundary for grid division. The region's feature vector is a four-dimensional vector composed of the maximum and minimum longitude and latitude coordinates of the region's vertices.

[0090]

[0091] By v feature Can obtain Figure 3 (a) The circumscribed rectangle shown ABCD coordinate.

[0092]

[0093] When the target crosses the 180° longitude, a 360° compensation operation can be performed on the negative longitude value.

[0094] In this embodiment, the regional grid division is based on the minimum single scene, such as Figure 3 As shown in (b), the regional width of the smallest single scene is used to realize the equidistant division of the regional grid. Firstly, it is to reduce the computational complexity, and secondly, it is also for the generation of subsequent tasks. When a large-scale search is not required, the single scene grid decomposition is actually equivalent to a single imaging task. A single scene refers to the geometric area on the earth where the field of view of the payload is projected. The grid division consists of a double-layer loop, which corresponds to the equidistant division of latitude and longitude respectively. The latitude increments of equal distances are equal, but the longitude increments of equal distances are not equal, so the longitude increment needs to be calculated based on the latitude. The equidistant division can ensure that the extension of the decomposed grid in longitude and latitude is consistent and will not cause distortion.

[0095] Step 2.1.2: Build the grid index.

[0096] First, the grid information vector is extracted. Each unit of the grid information vector stores the information of the center node and four vertices of a grid. The introduction of this vector is to improve the utilization of computer storage. Since the grid division is the division of the circumscribed rectangle of any polygonal area, many grids may not be included in the area, so the vector only stores the grid information in the area. Then, the grid index matrix is ​​established. The grid index matrix is ​​a two-dimensional matrix. The constructed two-dimensional list can be used to quickly retrieve the grid, determine the grid where the target is located, and obtain the grid that meets the target motion constraints. Figure 3 (c) shows the relationship between the grid index matrix and the grid information vector. Each element of the grid index matrix records the position of the corresponding grid in the grid information vector. When the value is -1, it means that the corresponding grid is not in the area of ​​interest. Therefore, relying on the increment of the grid index matrix in both row and column dimensions, the index of the grids around the target can be quickly realized, which greatly facilitates the prediction of the target potential area.

[0097] Step 2.2: Based on the satellite-borne image recognition algorithm, the optical image is identified to obtain the position and heading of the ocean mobile target, and based on the SAR payload, the speed of the ocean mobile target is obtained. In this embodiment, in order to facilitate and quickly predict the movement of the target, the present invention currently intends to adopt a simple method to realize the prediction of the movement of the mobile target: double-constrained motion prediction based on Gaussian distribution. The satellite-borne image recognition algorithm can identify the position and heading of the target based on the optical image, and for the SAR payload, the speed of the target can be measured. Therefore, through the optical image and SAR payload, the distribution of the target can be well predicted, and the prediction is based on two types of constraints: maximum movement speed v max and the maximum azimuth deviation velocity δ max For example, for general ships, the target movement speed has an upper limit, and its maximum speed is generally 30 knots (1 knot = 1 nautical mile / hour = 1.852 km / hour = 0.514 m / second, and the maximum speed is 30 knots = 55.56 km / hour = 15.433 m / second). For large ships, it is not easy to make a large turn in a short time. Therefore, these two constraints are considered to predict the target movement.

[0098] Step 2.3: Predict the fan-shaped area where the ocean moving target is located based on the position, heading and speed of the ocean moving target observed in the most recent observation period. Figure 4 (a) shows that in the recent k In the process of identifying the target, its position is determined P k and azimuth γ k information, it can be determined in the interval time κ After P k As the center of the circle, v max κ is the radius, and the fan angle symmetric about the heading is δ max κ In the fan-shaped area, the distribution of targets in the fan-shaped area is uniform (the distribution of targets in each grid is uniform).

[0099] Step 2.4: All grids in the fan-shaped area are used as the task set to be generated. v max and the maximum azimuth deviation velocity δ maxThe fan-shaped area obtained under the constraint is used as the position where the target may reach at the next moment, which reduces the uncertainty of the target to a certain extent. However, for planning, the tasks generated based on this will still be numerous. Therefore, this study introduces Gaussian distribution measurement to further reduce the uncertainty of the target. The movement of the target is affected by many external factors such as ocean currents and internal factors such as confrontation strategies. Its movement is difficult to predict, but the purpose of the target will eventually guide it to move in the target direction. The present invention uses Gaussian distribution measurement to measure this uncertainty, and the distribution parameters can be appropriately relaxed, and the distribution parameters can be updated as the tracking data accumulates.

[0100] Step 3: Assign task elements to the task set to be generated to form an on-board autonomous task.

[0101] On-board autonomous task generation is one of the important modules of on-board autonomous task management. It is responsible for converting uncertain targets into tasks with attributes. It not only prepares the necessary data basis for subsequent task planning, but also is the key to guiding the optimization direction of the system. On-board autonomous task generation needs to solve two important sub-problems: task generation time decision (when to generate tasks) and task generation element decision (how to generate tasks). In this application, the definition of a task is: an imaging task generated according to user observation needs or observation targets, as a separate observation task, does not depend on satellite resources and can be executed on multiple resources. It is mainly described by a single longitude and latitude coordinate or a sequence of longitude and latitude coordinates of the target. At the same time, the task also includes necessary attributes such as task priority, imaging duration, and imaging resolution. In this application, on-board autonomous task generation is mainly aimed at task generation at the task level, and the generation of subsequent meta-tasks is the input for the single-satellite autonomous scheduling problem. The concept of meta-task is that the meta-task is an observation task that can be executed by a satellite in a single transit, which depends on the task description of satellite resources and is associated with satellite resources. It is usually generated after the satellite performs task preprocessing on the observation task set, corresponding to an observation strip of the satellite, and is described by the four vertex coordinates of the strip. Unlike tasks, meta-tasks contain visible time window attributes, which correspond one-to-one with satellite resources. Therefore, the present invention only generates on-board tasks.

[0102] The task elements in this embodiment include determining the task generation time interval, the task load mode and the task priority.

[0103] In this embodiment, the target task generation cycle is:

[0104]

[0105] in, Indicates the user's desired tracking time interval for the target;

[0106] Indicates the timeliness of target motion prediction.

[0107] Regarding the decision on the task generation time, that is, the task generation cycle, consider when to generate new tasks to facilitate subsequent planning. In fact, during the onboard image processing, the acquired target information will be continuously fed back to the satellite for task planning, but frequent triggering of planning is not necessary. First, the target movement speed is not fast enough, and there is no need to respond to its movement changes in near real time; second, frequent triggering of planning will increase the system's computing and communication costs and increase costs. The decision on the task generation time is mainly considered from the following two aspects:

[0108] one is the user's desired tracking interval for the target . Under the condition of limited space-based satellite resources, it is difficult to track the target for a long time without interruption, and it is difficult to take into account the situation of multi-target tracking. Users generally have the need to track the target, and the tracking time interval has a certain utility for the user. When the tracking time interval is too small, useful target information may not be obtained. When the tracking time is too long, the target may be lost and its strategic intention cannot be judged. Therefore, the generation of tasks is considered by referring to the user's expected tracking time interval for the target.

[0109] Second, the timeliness of target motion prediction . Since the execution of the tracking task is the result of subsequent scheduling, when generating tasks, if the interval period is too long, it will cause the target to leave the original distribution grid, and the effectiveness of the target's motion prediction will be reduced. Taking the maximum speed of 30 knots (about 55.56 kilometers per hour) as an example, if the granularity of the grid division is 10km, then the target may leave the prediction grid in about one-fifth of an hour, which means the motion prediction timeliness is about eleven or twelve minutes, and then the target's motion prediction efficiency will be reduced.

[0110] Based on the above two considerations, the present invention aims at generating a time interval for a task The smaller of the user's expected tracking time interval for the target and the timeliness of the target motion prediction is used as the task generation time period.

[0111] The mode selection of the mission payload is a decision on the multi-mode of the payload under the target uncertainty state. This decision is mainly for the multi-imaging mode of the SAR payload. The target resolution brought by different imaging modes is inconsistent, and the recognition probability of the target is also different. The wider the tracking width, the wider the coverage, but the lower the resolution, the lower the probability of target recognition. In the usual tracking mode, the high-resolution imaging mode is generally used. When the target uncertainty increases, the wide-width imaging mode can be used. However, increasing the width will lead to a decrease in the target recognition rate. Therefore, the target recognition rate can only be improved by increasing the number of imaging times. Assume that the working mode of the payload is divided intol The corresponding width and target recognition rate are width l and pd l (Width varies with level l Increase, recognition rate increases with level l Reduced, the corresponding value in normal high score mode is width 0 and pd 0), it is necessary to ensure the minimum target recognition resolution pd α ( pd 0 > pd α ) conditions to achieve the generation of the target task. Therefore, in this embodiment, the method for determining the load mode of the task is:

[0112] 1) If image recognition is performed in the imaging results of each observation cycle, and the information of the ocean mobile target can be found and identified, a high-resolution payload mode is used, and the task is the smallest grid of each division in the fan-shaped area. Each observation cycle is a task generation cycle, and each observation will generate a new task;

[0113] 2) If image recognition is performed in the imaging results of the previous observation period and no ocean mobile target is found and identified, the load mode is the wide-width mode, and the task is a new grid divided by the width of the wide-width mode determined by the load mode. The level of the load mode and the number of times the task is generated are determined by the following formula;

[0114]

[0115] in, width l The working mode of the load is l The corresponding load width at level 1;

[0116] The working mode of the load is l The corresponding load target recognition rate at level 1 is is the minimum target recognition rate; is the radius of the fan area obtained by using double constraints; k Represents the number of observations for a task.

[0117] In this embodiment, the task priority needs to comprehensively consider the probability of the target appearing in a certain grid, the change of the target priority caused by the mobility of the target, and the probability of multiple targets appearing in the same grid:

[0118] 1) The probability of a target appearing in a certain grid, based on the target’s maximum movement speed v maxand maximum azimuth deviation velocity δ max Predict the target in the most recent target recognition ， Determine its location P k and azimuth γ k information, it can be determined in the interval time κ After P k As the center of the circle, v max κ is the radius, and the fan angle symmetric about the heading is δ max κ In the fan-shaped area, assuming that the probability of the target appearing in the grids in the fan-shaped area is equal, the target appears in the grid g i,j The probability is:

[0119]

[0120] For speed, is the azimuth, is the variance of the velocity that follows a Gaussian distribution, is the mean of the azimuth angles that obey the Gaussian distribution, is the variance of the azimuth that follows the Gaussian distribution; is the mean of the distances that follow the Gaussian distribution, is the variance of the distance that follows the Gaussian distribution, r i,j Representation Grid g i,j The center point is the closest to the target position after target recognition when the target is observed. P k Distance, time After that, the target is at a distance of The distance is , r i,j The target is identified in the grid after the latest observation. g i,j The azimuth of the center point, is the mean of the distances that follow the Gaussian distribution, is the mean value of the velocity that obeys the Gaussian distribution. Considering that the probability of the target being distributed in the grid in the fan-shaped area is 1, the normalization is used to obtain the probability of the target being distributed in the grid. g i,j The probability is:

[0121]

[0122] v max Indicates the maximum moving speed of the target. δmax Indicates the maximum azimuth deviation speed of the target γ k Indicates the azimuth of the target at the last observation; κ represents Interval time.

[0123] In this embodiment, the target appears in grid g i,j The probability formula is calculated through the following analytical reasoning.

[0124] According to the recent k The target position obtained by target motion recognition , azimuth γ k Information and speed , assuming that after the prior information, the speed and azimuth of the target movement obey the Gaussian distribution, that is , , the prior information refers to the prior information, that is, the target position, heading and speed obtained from the most recent target motion recognition. Then, the probability distribution functions of its speed and azimuth can be obtained as formula (11) and formula (12):

[0125] (11)

[0126] (12)

[0127] Elapsed time After that, the target is at a distance of The distance is , then the distance Also follows the Gaussian distribution , whose probability distribution satisfies the following formula (13)

[0128] (13)

[0129] in, , .

[0130] Then the joint probability density distribution function of the distance and azimuth of the target in the next observation period is

[0131] (14)

[0132] For a certain grid g i,j For example, the probability distribution of the target in the next observation period on this grid is

[0133] (15)

[0134] In this embodiment, the calculation of the azimuth angle, taking Figure 2 as an example, the known position P and P′ ,but P arrive

[0135] P′ Azimuth γ for

[0136] (16)

[0137] In order to improve the computational efficiency of prediction, it is assumed that the target probability density distribution in the same grid is equal, and the grid center point G i, j As a reference, the grid g i,j For example, the probability distribution of the target in the next observation period in this grid is approximately as follows:

[0138] (17)

[0139] Normalization is used to ensure the conservation of distribution probability, that is, the sum of the target distribution probability is equal to 1. Under the double constraint condition, the target is obtained by normalization on the grid g i,j The final probability distribution is

[0140] (18)

[0141] According to the current state of the target, the probability distribution of the next observation period is obtained, and its probability distribution graph is constructed. Through the double-constraint prediction model based on Gaussian distribution, the high-probability potential distribution position of the target can be obtained, such as Figure 4 As shown in (b), the dual-constraint prediction model can further reduce the uncertainty of the target and improve the utilization efficiency of resources.

[0142] 2) Dynamic priority is introduced due to the change of target priority caused by the mobility of the target;

[0143] (19)

[0144] is the maximum upper limit of the target dynamic priority, Indicates the basic priority predefined by the user. ot k For the k The target information time obtained by the target observation is: is the user-defined desired tracking time interval; then t The target is on the grid g i,j The priority is expressed as

[0145] ; (20)

[0146] 3) Consider the case where multiple targets appear in the same grid:

[0147] Indicates at time t The mth target appears in the rectangular area of ​​the target observation. i Row, No. j Column Grid i,j The probability within The value of the target appears in the grid g after considering the dynamic priority in 2). i,j The probability of is calculated;

[0148] but Indicates the overlay priority of the grids formed when multiple targets appear in the same grid during multi-target tracking. i , j The first i Row, No. j A grid of columns, m Indicates the first i Row, No. j The first column in the grid m goals, t Indicates time, | tgt | is the tracking target set;

[0149] Indicates the normalized priority of grid stacking when considering multiple targets appearing in the same grid;

[0150] 4) The target appears on the grid The overall priority is changed to: .

[0151] The present invention also provides an on-board autonomous mission generation system for ocean mobile targets, comprising the following modules:

[0152] Model building module: used to build the motion model of ocean moving targets;

[0153] Task set generation module: used to predict the trajectory of the ocean mobile target according to the ocean mobile target motion model constructed by the model construction module to obtain the task set to be generated;

[0154] Task element assignment module: used to assign task elements to the task set to be generated by the task set generation module to form an on-board autonomous task.

[0155] For non-time-sensitive mobile targets, the present invention constructs a mobile target motion model based on the Gaussian distribution characteristics, and proposes a target motion prediction model with dual constraints under Gaussian distribution to obtain the possible location of the target task, i.e., the fan-shaped area, to form a satellite observation task; and assuming that the probability of the target appearing in each grid in the fan-shaped area is equal, the probability of the target appearing in the fan-shaped area grid can be obtained, and then based on whether the target appears after target recognition in each satellite observation image, the load mode of the task and the number of task generation times are selected, and at the same time, considering the mobility of the task, dynamic priority is introduced, and considering the situation where multiple targets appear in a grid at the same time, priority coordination is performed, thereby determining the various elements of the task. This makes it possible to track and observe mobile targets in the ocean, and realizes the measurement of target motion uncertainty. Therefore, it can provide necessary input data for the planning of subsequent multi-satellite tasks, thereby realizing rapid task planning and task observation tracking.

[0156] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating autonomous onboard missions for ocean mobile targets, characterized in that: The following steps are involved: Step 1: Construct a motion model of ocean moving targets; Step 2: predicting the trajectory of the ocean mobile target according to the ocean mobile target motion model to obtain a task set to be generated; Step 3: Assign task elements to the task set to be generated to form an on-board autonomous task; The ocean moving target motion model is: Target The corresponding position of the moment under PGFS is , it is necessary to recursively infer the next position of the target based on the current trajectory point state of the target , the target is The plane where the large arc of the motion trajectory in a short time is located is , according to the direction of motion and the right-hand system, the normal vector of the great circle is , the Earth's North Pole is ,flat Normal vector determined by right-hand system for Depend on Rotation angle Can get ,Right now in, is a quaternion, Its conjugate quaternion satisfies the following conditions Target elapsed time from Exercise to , and its movement distance is , the corresponding arc of the Earth's center angle is ,therefore, The coordinates of can be obtained by the following formula in, is a quaternion, Its conjugate quaternion satisfies the following conditions 。 2. The method according to claim 1, characterized in that In step 2, the trajectory of the ocean mobile target is predicted according to the ocean mobile target motion model to obtain the specific method of the task set to be generated: Step 2.1: Grid the area where the ocean moving target is located and construct a grid index; Step 2.2: Recognize the optical image based on the satellite-borne image recognition algorithm to obtain the position and heading of the ocean mobile target, and obtain the speed of the ocean mobile target based on the SAR payload; Step 2.3: predicting the fan-shaped area where the ocean moving target is located based on the most recently obtained position, heading and speed of the ocean moving target; Step 2.4: All grids within the fan-shaped area are taken as the task set to be generated.

3. The method according to claim 2, characterized in that The sector areas are: According to the recent k In the identification of the ocean moving target, the position of the ocean moving target is obtained P k and azimuth γ k Information, then the ocean moving target is in the time interval κ The position after is P k As the center of the circle, v max κ is the radius and the heading γ k The symmetrical fan angle is δ max κ The fan-shaped area, v max Indicates that in the recent k The maximum speed of the ocean moving target during the measurement process, δ max Indicates that in the recent k The maximum angular velocity of the ocean moving target during the measurement.

4. The method according to claim 3, characterized in that The task elements in step 3 include determining the load pattern of the task, the number of times the task is generated, and the priority of the task.

5. The method according to claim 4, characterized in that The load mode determination method of the task is: 1) If image recognition is performed in the imaging results of each observation period, and the information of the ocean mobile target can be found and identified, a high-resolution payload mode is used, and the task is the smallest grid of each division in the fan-shaped area; 2) If image recognition is performed in the imaging results of the previous observation cycle and no ocean mobile target is found and identified, the load mode is the wide-width mode, and the task is a new grid divided by the width of the wide-width mode determined by the load mode. The level of the load mode and the number of times the task is generated are determined by the following formula; in, The working mode of the load is l The corresponding load width at level 1; The working mode of the load is l The corresponding load target recognition rate at level 1 is is the minimum target recognition rate; is the radius of the fan area obtained by using double constraints; k Represents the number of observations for a task.

6. The method according to claim 5, characterized in that Task priority needs to take into account the probability of a target appearing in a certain grid, the change in target priority caused by the mobility of the target, and the probability of multiple targets appearing in the same grid: 1) The probability of a target appearing in a certain grid, based on the target’s maximum movement speed v max and maximum azimuth deviation velocity δ max Predict the target's position in the most recent target recognition P k and azimuth γ k information, it can be determined in the interval time κ After P k As the center of the circle, v max κ is the radius, and the fan angle symmetric about the heading is δ max κ In the fan-shaped area, assuming that the probability of the target appearing in the grids in the fan-shaped area is equal, the target appears in the grid g i,j The probability is: For speed, is the azimuth, is the variance of the velocity that follows a Gaussian distribution, is the mean of the azimuth angles that obey the Gaussian distribution, is the variance of the azimuth that follows the Gaussian distribution; is the mean of the distances that follow the Gaussian distribution, is the variance of the distance that follows the Gaussian distribution, Representation Grid g i,j The center point is the closest to the target position after target recognition when the target is observed. P k Distance, time After that, the target is at a distance of The distance is , The target is identified in the grid after the latest observation. g i,j The azimuth of the center point, is the mean of the distances that follow the Gaussian distribution, is the mean value of the velocity that obeys the Gaussian distribution. Considering that the probability of the target being distributed in the grid in the fan-shaped area is 1, the normalization is used to obtain the probability of the target being distributed in the grid. g i,j The probability is: v max Indicates the maximum moving speed of the target. δ max Indicates the maximum azimuth deviation speed of the target γ k Indicates the azimuth of the target at the last observation; κ represents Interval time; 2) Dynamic priority is introduced due to the change of target priority caused by the mobility of the target; is the maximum upper limit of the target dynamic priority, Indicates the basic priority predefined by the user. ot k For the k The target information time obtained by the target observation is: Desired tracking time interval defined by the user; Then at time t the target is in the grid The priority is expressed as ； 3) Consider the case where multiple targets appear in the same grid: Indicates at time t The mth target appears in the rectangular area of ​​the target observation. i Row, No. j Column Grid g i,j The probability within The value of the target appears in the grid after considering the dynamic priority in 2) g i,j The probability of is calculated; but Indicates the overlay priority of the grids formed when multiple targets appear in the same grid during multi-target tracking. i, j The first i Row, No. j A grid of columns, m Indicates the first i Row, No. j The first column in the grid m goals, t Indicates time, | tgt | is the tracking target set; Indicates the normalized priority of grid stacking when considering multiple targets appearing in the same grid; 4) The target appears on the grid g i,j After coordination, it is changed to: .

7. The method according to claim 5, characterized in that The task generation cycle of the target is: in, Indicates the user's desired tracking time interval for the target; Indicates the timeliness of target motion prediction.

8. An onboard autonomous mission generation system for ocean mobile targets, characterized in that: Includes the following modules: Model building module: used to build the motion model of ocean moving targets; Task set generation module: used to predict the trajectory of the ocean mobile target according to the ocean mobile target motion model constructed by the model construction module to obtain the task set to be generated; Task element assignment module: used to assign task elements to the task set to be generated by the task set generation module to form an on-board autonomous task; The ocean moving target motion model is: Target The corresponding position of the moment under PGFS is , it is necessary to recursively infer the next position of the target based on the current trajectory point state of the target , the target is The plane where the large arc of the motion trajectory in a short time is located is , according to the direction of motion and the right-hand system, the normal vector of the great circle is , the Earth's North Pole is ,flat Normal vector determined according to the right-hand system for Depend on Rotation angle Can get ,Right now in, is a quaternion, Its conjugate quaternion satisfies the following conditions Target elapsed time from Exercise to , and its movement distance is , the corresponding arc of the Earth's center angle is ,therefore, The coordinates of can be obtained by the following formula in, is a quaternion, Its conjugate quaternion satisfies the following conditions 。

Images