A multi-platform cooperative target trajectory tracking method and device

Abstract

The application belongs to the technical field of kinematics modeling, and more particularly relates to a multi-platform cooperative target trajectory tracking method and device, which comprises the following steps: acquiring a trajectory set of an enemy target T moment, constructing a motion trajectory range of the enemy target, controlling a UAV cluster to implement a global area coverage strategy on the motion trajectory range to realize tracking of the enemy target, the UAV cluster cooperatively completes positioning of the enemy target through tracking of the enemy target, a position coordinate of the enemy target at an i moment is solved through a cooperative positioning algorithm, positioning accuracy is obtained according to a positioning accuracy solving algorithm, an original Informer model is improved to obtain an improved Informer model, the improved Informer model is trained to obtain a trained Informer model, and a long sequence of a flight trajectory of the enemy target is predicted in real time by using the trained Informer model to realize accurate positioning and continuous tracking of the enemy target.

Description

Technical Field

[0001] This invention belongs to the field of kinematic modeling technology, and more specifically, relates to a multi-platform collaborative target trajectory tracking method and apparatus. Background Technology

[0002] With the rapid development of UAV technology, UAVs have broad application prospects in various fields such as military and civilian. In the process of modern battlefield combat command, commanders usually need to accurately locate and continuously track enemy targets in order to carry out precise strikes by analyzing the enemy's movement trajectory data. At present, the combat performance of a single unmanned equipment is no longer sufficient to meet the needs of the modern battlefield. Single UAVs usually face problems such as low positioning accuracy, weak tracking mobility, and low level of intelligence. These problems limit the accuracy and tracking distance of UAVs when performing tracking tasks. Therefore, research on multi-UAV collaborative precise positioning and continuous tracking prediction has important theoretical significance and practical value.

[0003] Multi-UAV collaborative precise positioning and continuous tracking refers to improving the mission execution capability of UAVs in complex environments by utilizing the cooperation between multiple UAVs. It improves the positioning accuracy of both the UAVs themselves and the target by using relative position measurements between UAVs and distance or angle measurements of the target through collaborative processing algorithms, and maintains continuous tracking of the target even when the target may maneuver. This method can be applied to various fields such as battlefield reconnaissance, target strike, and field rescue.

[0004] Existing technologies include using Long Short-Term Memory (LSTM) networks to predict pedestrian trajectories, but this algorithm is a sequential processing algorithm with drawbacks such as weak parallel processing capability and low efficiency. Another approach uses Transformer models to improve the accuracy of vehicle trajectory prediction while avoiding the parallelization problem of LTM models. However, Transformer models also suffer from high time complexity, high memory consumption, and slow inference speed. Additionally, Informer networks can be used to predict drone flight trajectories, but this method only considers the drone's latitude, longitude, and altitude, which has limitations. Furthermore, this method has high computational costs and is not suitable for real-time drone trajectory prediction. Summary of the Invention

[0005] The technical problem that this invention aims to solve is: how to achieve accurate positioning and continuous tracking of enemy targets while reducing the limitations of enemy target prediction and tracking and lowering computational costs.

[0006] The present invention achieves the above objectives through the following technical solution: Firstly, a multi-platform collaborative target trajectory tracking method is provided, including: Obtain the trajectory set of the enemy target at time T, and construct the range of the enemy target's movement trajectory; The drone swarm is controlled to implement a global area coverage strategy over the range of the movement trajectory in order to track the enemy target; The drone swarm tracks the enemy target and coordinates to locate it. It calculates the position coordinates of the enemy target at time i using a collaborative positioning algorithm and obtains the positioning accuracy based on a positioning accuracy calculation algorithm, thereby obtaining the movement trajectory data of the enemy target. The original Informer model is improved to obtain an improved Informer model. The motion trajectory data is preprocessed and then trained to obtain a trained Informer model. The trained Informer model is used to predict the long sequence of flight trajectories of the enemy target in real time, so as to achieve accurate positioning and continuous tracking of the enemy target.

[0007] Preferably, the step of acquiring the trajectory set of the enemy target at time T and constructing the range of the enemy target's movement trajectory specifically involves: Establish a coordinate system, taking the initial position of the enemy target as the origin, with the X-axis representing the flight direction; The movement of the enemy target includes two stages. The first stage is that the enemy target moves in an arc along the X-axis from the initial position until the direction of the enemy target's movement reaches a preset angle with the X-axis. The second stage is that the enemy target continues to move in a straight line along the direction of the preset angle. The range of the enemy target's movement trajectory is obtained by changing the preset direction angle.

[0008] Preferably, the control of the drone swarm implements a global area coverage strategy over the movement trajectory range to achieve tracking of the enemy target, including: The detection range of the drone is a sampling circle, and the drone cluster is controlled to conduct detection within the range of the movement trajectory. The range of the motion trajectory is rasterized to obtain the intersection points between the grids; Each time, select an intersection point between the grids as the center of the sampling circle, and repeat this process until all intersection points between the grids are covered by the sampling circle.

[0009] Preferably, the drone swarm tracks the enemy target and collaboratively locates it. A collaborative positioning algorithm calculates the enemy target's position coordinates at time i, and a positioning accuracy calculation algorithm is used to obtain the positioning accuracy, thus acquiring the enemy target's trajectory data. Specifically: The overlapping portion of the sampling circles is taken as the overlapping region. Within this overlapping region, the enemy target is located using collaborative positioning among UAVs to obtain positioning accuracy. Specifically, this includes: When only one UAV locates the enemy target, the positioning accuracy is the single radar positioning accuracy. When multiple UAVs locate the enemy target, the position coordinates of the enemy target at time i are obtained through the cooperative positioning algorithm, and the positioning accuracy between each pair of UAVs is calculated through the positioning accuracy calculation algorithm. When the positioning accuracy is optimal, the position coordinates of the enemy target at time i are the most accurate.

[0010] Preferably, the cooperative localization algorithm is as follows: The cooperative localization algorithm is used between at least two drones, including a first drone and a second drone; Obtain the position coordinates of the first UAV, the position coordinates of the second UAV, and the distance between the two UAVs. The distance between the first UAV and the enemy target and the distance between the second drone and the enemy target. The position coordinates of the enemy target are calculated based on the obtained data; When at least two UAVs are locating and tracking the same enemy target, the range error and the angle between the enemy target and the two UAVs are obtained respectively, and the positioning accuracy at time i is calculated based on the range error and the angle between the two UAVs.

[0011] Preferably, the positioning accuracy calculation algorithm includes: Based on the single-radar positioning accuracy and the range error, the dividing angle between the single-radar positioning accuracy and the dual-radar cooperative positioning accuracy is obtained. ; When the angle between the enemy target and the two drones is in the range [0, ... ] or [π- When π], the positioning accuracy of the single radar is taken as the positioning accuracy of the enemy target at time i; When the angle between the enemy target and the two drones is within the range of ( ,π- When the distance error is calculated and the angle between the two UAVs is calculated, the positioning accuracy of the enemy target at time i is obtained.

[0012] Preferably, the original Informer model is improved to obtain an improved Informer model. After preprocessing the motion trajectory data, the improved Informer model is trained to obtain a trained Informer model. The preprocessing includes: The position coordinates of the enemy target at time i and the optimal positioning accuracy are substituted into the formula of the Kalman filter, and the result is used as the estimated coordinates at time i. Collect the metadata of the enemy target at time i, the metadata including the estimated coordinates of the enemy target; The metadata dataset is normalized, and then the normalized metadata dataset is split into a training set, a validation set, and a test set.

[0013] Preferably, the improved Informer model includes: The Informer model network architecture includes an Encoder module and a Decoder module, with the output of the Encoder module serving as the input of the Decoder module; The Encoder module is used to encode the input sequence, extract features from the time series data, and use a self-attention mechanism to capture the relationships in the input sequence. The Decoder module is used to generate future prediction sequences, and a multi-head attention mechanism module is used to process the fusion of historical sequence data and location embeddings.

[0014] Preferably, the step of using the trained Informer model to predict the long sequence of the enemy target's flight trajectory in real time, so as to achieve accurate positioning and continuous tracking of the enemy target, includes: The enemy target is located and tracked by the trained Informer model network architecture. When a new meta-dataset is obtained, it is also normalized to obtain a new training set, validation set and test set. The Informer model network architecture was retrained, validated, and tested using new training, validation, and test sets. Incremental learning is used during training, updating only some parameters of the Informer model network architecture to improve training efficiency.

[0015] Secondly, a multi-platform collaborative target trajectory tracking device is proposed, including: a construction module, a control module, an acquisition module, a training module, and a prediction module; The construction module is used to obtain the trajectory set of the enemy target at time T and construct the range of the enemy target's movement trajectory; The control module is used to control the UAV swarm to implement a global area coverage strategy over the range of the movement trajectory in order to track the enemy target; The acquisition module is used to coordinate the positioning of the enemy target by the UAV cluster through tracking the enemy target, calculate the position coordinates of the enemy target at time i through a cooperative positioning algorithm, and obtain the positioning accuracy according to the positioning accuracy calculation algorithm, so as to obtain the motion trajectory data of the enemy target; The training module is used to improve the original Informer model to obtain an improved Informer model. After preprocessing the motion trajectory data, the improved Informer model is trained to obtain the trained Informer model. The prediction module is used to predict the long sequence of the enemy target's flight trajectory in real time using the trained Informer model, so as to achieve accurate positioning and continuous tracking of the enemy target.

[0016] The beneficial effects of this invention are as follows: This invention calculates the possible trajectories of an enemy target after time T, obtains the maximum set of trajectories of the enemy target at time T, and constitutes the range of the enemy target's trajectory. Then, a global area coverage strategy is implemented for this trajectory range to ensure that the enemy target is within the detection range of the UAV swarm. The UAV swarm then obtains the enemy target's position coordinates and calculates the positioning accuracy through a cooperative positioning algorithm. The improved and trained Informer model is used to realize real-time prediction of long sequences of enemy target trajectories, reducing the limitations of UAVs in predicting and tracking enemy targets, making the calculation simpler, and ultimately achieving accurate positioning and continuous tracking of enemy targets. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. The drawings are used to provide a further understanding of this disclosure and constitute a part of the specification. They are used together with the following detailed description to explain this disclosure, but do not constitute a limitation of this disclosure. In the drawings: Figure 1 This is a schematic flowchart illustrating a multi-platform collaborative target trajectory tracking method provided in an embodiment of the present invention. Figure 2 This is a flowchart illustrating a multi-platform collaborative target trajectory tracking method provided in an embodiment of the present invention; Figure 3 This invention provides a multi-platform collaborative target trajectory tracking method for enemy targets along the direction angle. A schematic diagram of the motion trajectory; Figure 4 This is a schematic diagram showing the range of enemy target movement trajectories in a multi-platform collaborative target trajectory tracking method provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of dual-machine collaborative positioning for a multi-platform collaborative target trajectory tracking method provided in an embodiment of the present invention; Figure 6 This is a flowchart of a multi-platform collaborative target trajectory tracking method based on an improved Informer model for real-time long sequence trajectory prediction, provided by an embodiment of the present invention. Figure 7 This is a schematic diagram of the Informer model network architecture of a multi-platform collaborative target trajectory tracking method provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of the long-time prediction results of a multi-platform collaborative target trajectory tracking method provided in an embodiment of the present invention; Figure 9 This is a schematic diagram of the structure of a multi-platform collaborative target trajectory tracking device provided in an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other. The invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0019] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0020] Example 1 In the field of existing real-time tracking and positioning technology, some existing tracking, positioning, and prediction technologies may rely on complex mathematical models and algorithms, which increases computational complexity and resource consumption, potentially posing a challenge for real-time applications or resource-constrained systems.

[0021] The trajectory of enemy targets is often nonlinear and non-Gaussian, while some traditional single prediction models (such as Kalman filters) are based on linear and Gaussian assumptions, which may lead to prediction bias or inaccuracy. Furthermore, they have certain limitations in calculating and predicting the motion state of enemy targets.

[0022] To solve the above problems, in this embodiment, as follows: Figure 1 and Figure 2 As shown, a multi-platform collaborative target trajectory tracking method is proposed, including: Step 101: Obtain the set of trajectories of the enemy target at time T, and construct the range of the enemy target's movement trajectory.

[0023] A coordinate system is established, with the initial position of the enemy target as the origin and the X-axis representing the flight direction. The movement of the enemy target includes two stages. The first stage is that the enemy target moves in an arc along the X-axis from the initial position until the direction of the enemy target's movement reaches a preset angle with the X-axis. The second stage is that the enemy target continues to move in a straight line along the direction of the preset angle. The range of the enemy target's movement trajectory is obtained by changing the preset angle.

[0024] The first stage is when the enemy target arrives at the direction angle along a circular arc at its maximum flight speed Vmax. The turn is completed, with a turning radius of r and a first-stage time of t1; the second stage is the enemy target's angle along the direction. Traveling in a straight line at the maximum speed Vmax, the second stage lasts for t2.

[0025] The enemy target's current position is taken as the origin (0,0), its flight direction is taken as the x-axis, its tilt angle is θ, its gravity is g, and its turning radius r is directly proportional to the square of its velocity and inversely proportional to the tangent tanθ of its tilt angle θ. The formula is as follows: like Figure 3 As shown, the flight direction angle of the enemy target The angle at which the enemy target flies with a turning radius of r is equal to that of the target, that is: Arrival direction and angle Phase 1 time t1: The formula shows that the time t1 in the first stage is related to the direction angle. The maximum flight speed Vmax is directly proportional to the enemy target's tilt angle θ.

[0026] Enemy target arrives at flight direction angle Coordinates after turning: From the above formula, it can be deduced that the enemy target, after turning, from... The expression for traveling in a straight line at maximum speed Vmax is: time :T=t1+t2, where x and y represent the x-coordinate and y-coordinate of the enemy target's position after turning, respectively.

[0027] In a preferred embodiment, it is assumed that the maximum flight speed Vmax of the enemy target is 200 m / s, the enemy target tilt angle θ is π / 3, and the gravity is g = 9.8 m / s². 2 Given that time T is 200s, the angle of the enemy target can be calculated by combining the above equations. The maximum range of motion M1 is determined by the value of M1 gradually increasing from the preset direction angle (0~π), and the maximum range of motion M2 is the mirror image of M1 along the x-axis. From this, the flight direction angle of the enemy target can be obtained. The range M of the motion trajectory as it gradually increases from (-π to π) is as follows: Figure 4 As shown.

[0028] Step 102: Control the drone swarm to implement a global area coverage strategy over the range of the movement trajectory in order to track the enemy target.

[0029] The drone swarm contains at least one drone, and the detection range of each drone on the motion trajectory range is the area covered by a sampling circle.

[0030] By implementing a global area coverage strategy using a swarm of drones equipped with radar, saturation coverage of enemy targets can be achieved. In the next time T, the enemy targets will be kept within the detection range of the drone swarm, and the tracking of the enemy targets will be completed.

[0031] The detection range of the UAV is a sampling circle. The UAV cluster is controlled to detect within the range of the motion trajectory. The range of the motion trajectory is rasterized to obtain the intersection points between the grids. Each time, an intersection point between the grids is selected as the center of the sampling circle. This process is repeated until all intersection points between the grids are covered by the sampling circle.

[0032] The above detection method can be simplified to: using N sampling circles with radius R to cover an approximately elliptical region. This scenario is an NP problem, which can generally be solved using a greedy algorithm.

[0033] The greedy algorithm can be designed as follows: (1) First, the irregular shape, i.e. the range M of the motion trajectory, is rasterized. The intersection of the raster is the point that the sampling circle needs to cover. (2) Each time, select the intersection of a grid as the center of the sampling circle to cover the most uncovered points; (3) Then repeat the above process until all points are covered; (4) The final result will have a large number of overlapping areas between sampling circles. These overlapping areas can be located collaboratively by multiple unmanned platforms to improve accuracy.

[0034] The proposed cooperative localization and cooperative localization algorithm will be described in the following steps.

[0035] Step 103: The UAV cluster tracks the enemy target and coordinates to locate the enemy target. It calculates the position coordinates of the enemy target at time i using a collaborative positioning algorithm and obtains the positioning accuracy based on a positioning accuracy calculation algorithm to obtain the movement trajectory data of the enemy target.

[0036] Specifically, the motion trajectory data includes multiple data points, such as the enemy target's speed, flight direction, position coordinates at time i, and positioning accuracy. Based on these data, the position and motion of the enemy target can be predicted and tracked. The specific prediction and tracking methods will be described below.

[0037] In a preferred embodiment, the overlapping portion of the sampling circles is taken as the overlapping region. The enemy target is located in the overlapping region by cooperative positioning among UAVs to obtain the positioning accuracy. Specifically, when only one UAV locates the enemy target, the positioning accuracy is the single radar positioning accuracy. When multiple UAVs locate the enemy target, the position coordinates of the enemy target at time i are obtained by the cooperative positioning algorithm, and the positioning accuracy between each pair of UAVs is calculated by the positioning accuracy calculation algorithm. When the positioning accuracy is optimal, the position coordinates of the enemy target at time i are the most accurate.

[0038] As can be seen from the maximum trajectory area coverage in step 102, at least one UAV covers the enemy target's movement trajectory, enabling the positioning and tracking of the enemy target. If only one UAV completes the positioning of the enemy target, the positioning accuracy is the same as that of a single radar. If two or more UAVs can simultaneously locate and track enemy targets, a cooperative positioning algorithm can be used to calculate the optimal positioning accuracy between each pair of UAVs.

[0039] Specifically, the cooperative localization algorithm is as follows: The cooperative localization algorithm is used between at least two drones, including a first drone and a second drone, to obtain the position coordinates of the first drone, the position coordinates of the second drone, and the distance between the two drones, respectively. The distance between the first UAV and the enemy target and the distance between the second drone and the enemy target. The position coordinates of the enemy target are calculated based on the obtained data. When at least two UAVs are tracking and locating the same enemy target, the range error and the angle between the enemy target and the two UAVs are obtained respectively. The positioning accuracy at time i is calculated based on the range error and the angle between the two UAVs.

[0040] like Figure 5 As shown, when the position of the first UAV A in the coordinate system is The position of the second drone B in the coordinate system is: The distance at which the first UAV A detected the enemy target was... The second UAV B detected the enemy target at a distance of [distance missing]. The distance between the two drones is In this case, the cooperative localization algorithm can be expressed as: Through distance error Range error Angle between the enemy target and the two drones The positioning accuracy at the current moment can be calculated. In the above formula and These represent the x-coordinate and y-coordinate of the enemy target's coordinates, respectively.

[0041] The positioning accuracy calculation algorithm is as follows: Based on the single radar positioning accuracy and the range error, the boundary angle between the single radar positioning accuracy and the dual radar cooperative positioning accuracy is obtained. When the angle between the enemy target and the two drones is in the range [0, ... ] or [π- When π], the positioning accuracy of the single radar is taken as the positioning accuracy of the enemy target at time i. When the angle between the enemy target and the two UAVs is in the range of (π], ,π- When i), the positioning accuracy of the enemy target at time i is calculated based on the distance error and the angle between the two UAVs.

[0042] Among them, the dividing angle between single-radar positioning accuracy and dual-radar cooperative positioning accuracy. When the included angle Values At that time, because the positioning algorithm result had too large an error and was lower than the positioning accuracy of a single radar, the positioning coordinates of the single radar with higher positioning accuracy were taken as the joint positioning accuracy. When the included angle Values At that time, the positioning results are calculated through algorithms, and the dividing angle between the positioning accuracy of a single radar and the cooperative positioning accuracy of two radars is determined. for: Positioning accuracy It can be represented as: Based on the above cooperative localization algorithm, we can derive the... Location data at the specified time and the positioning accuracy at that moment .

[0043] Step 104: Improve the original Informer model to obtain the improved Informer model. After preprocessing the motion trajectory data, train the improved Informer model to obtain the trained Informer model.

[0044] The preprocessing includes: substituting the position coordinates of the enemy target at time i and the optimal positioning accuracy into the formula of the Kalman filter, and using the result as the estimated coordinates at time i; collecting the metadata of the enemy target at time i, the metadata including the estimated coordinates of the enemy target; normalizing the metadata so that the values ​​of the metadata after normalization are within a preset range; and splitting the normalized metadata into a training set, a validation set, and a test set.

[0045] In a preferred embodiment, refer to Figure 6 Substituting the enemy target's position coordinates and optimal positioning accuracy at time i into the Kalman filter is a method that uses the linear system state equation to estimate the system state optimally through the system input and output observation data. Since the observation data includes the influence of noise and interference in the system, the optimal estimation can also be regarded as a filtering process. The estimation of the enemy target's position using the Kalman filter will not be explained in detail in this embodiment.

[0046] The preset range is specifically between 0 and 1. When normalizing the metadata dataset, the minimum and maximum values ​​in the metadata set are first determined, and then the normalized value is calculated. For each specific value of each metadata field, the normalized value is calculated using the following formula: Normalized value = (Original value - Minimum value) / (Maximum value - Minimum value).

[0047] The original value is the specific value of the metadata field, and the minimum and maximum values ​​are the minimum and maximum values ​​of the field, respectively.

[0048] Finally, update the metadata value by replacing the original metadata value with the calculated normalized value. In this way, the value range of all metadata fields will be normalized to (0~1).

[0049] The specific normalization operations involved will not be explained in detail in this embodiment.

[0050] Once the training, validation, and test sets are obtained, the Informer model network architecture is constructed: like Figure 7 As shown, the Informer model network architecture includes an Encoder module and a Decoder module. The output of the Encoder module serves as the input of the Decoder module. The Encoder module is used to encode the input sequence, extract features from the time series data, and use a self-attention mechanism to capture the relationships in the input sequence. The Decoder module is used to generate future prediction sequences and employs a multi-head attention mechanism module to process the fusion of historical sequence data and location embeddings.

[0051] The Informer model is a neural network model used for time series forecasting, and its network architecture includes an Encoder module and a Decoder module.

[0052] The Encoder module encodes the input sequence and extracts features from the time series data. It uses a self-attention mechanism to capture the relationships in the input sequence. The self-attention mechanism can calculate the representation of each position based on the importance of different positions in the sequence, thereby better capturing the long-range dependencies in the sequence. The encoded features output by the Encoder module are used as the input to the Decoder module.

[0053] The Decoder module is used to generate future predicted sequences. It uses a multi-head attention mechanism to process the fusion of historical sequence data and location embeddings. Multi-head attention can improve the model's representational ability by considering information from multiple different locations at the same time. In the Informer model, multi-head attention is applied to the decoder module to fuse historical sequence data and location embeddings to generate future predicted sequences.

[0054] By combining the Encoder and Decoder modules, the Informer model network architecture can extract key features from the input sequence and use these features to generate accurate future prediction sequences. This gives the Informer model network architecture good performance and generalization ability in time series prediction tasks.

[0055] Step 105: Use the trained Informer model to predict the long sequence of the enemy target's flight trajectory in real time, so as to achieve accurate positioning and continuous tracking of the enemy target.

[0056] After constructing the Informer model network architecture, it needs to be trained using the training set. After training, the performance of the trained Informer model network architecture is evaluated using the validation set to determine its fit and generalization ability. The trained Informer model network architecture is then tested and evaluated using the test set, and its metrics on the test set are calculated to assess its predictive ability after training.

[0057] The specific process of training, testing, and evaluating the Informer model network architecture is not described in detail in this embodiment.

[0058] After training, testing, and evaluating the Informer model network architecture, it can be directly used to predict the long-term flight trajectory of enemy targets. The long flight trajectory sequence contains various parameter data of the enemy target during its flight, such as position coordinates, flight speed, and heading. The Informer model network architecture constructed in this embodiment can predict the long flight trajectory sequence of enemy targets in real time. The specific calculation process of the Informer model network architecture will not be explained in detail in this embodiment.

[0059] In a preferred embodiment, the enemy target is located and tracked using the trained Informer model network architecture. When new metadata is obtained, it is also normalized to obtain new training, validation, and test sets. The Informer model network architecture is then retrained, validated, and tested using these new sets. During training, an incremental learning method is employed, updating only some parameters of the Informer model network architecture to improve training efficiency. After the model is updated, the updated model can be used to predict the latest data, obtaining updated long-term time-series prediction results for the enemy target. Figure 8 As shown above, the specific operation has been explained and will not be repeated in this embodiment.

[0060] This embodiment calculates the possible trajectories of an enemy target after time T, obtaining the maximum set of trajectories of the enemy target at time T, which constitutes the range of the enemy target's trajectory. Then, a global area coverage strategy is implemented for this trajectory range to ensure that the enemy target is within the detection range of the UAV swarm. The UAV swarm then obtains the enemy target's position coordinates and calculates the positioning accuracy through a cooperative positioning algorithm. The improved and trained Informer model is used to achieve real-time prediction of long sequences of enemy target trajectories, reducing the limitations of UAVs in predicting and tracking enemy targets, making the calculation simpler, and ultimately achieving accurate positioning and continuous tracking of the enemy target.

[0061] Example 2 In Example 1, a multi-platform collaborative target trajectory tracking method was proposed. In this example, a device is proposed, such as... Figure 9 As shown, the device is used to implement a multi-platform collaborative target trajectory tracking method as described in Embodiment 1. The device includes: a construction module, a control module, an acquisition module, a training module, and a prediction module.

[0062] The construction module is used to acquire the trajectory set of the enemy target at time T and construct the trajectory range of the enemy target's movement. The control module is used to control the UAV swarm to implement a global area coverage strategy over the trajectory range to achieve tracking of the enemy target. The acquisition module is used to coordinately locate the enemy target based on the tracking of the enemy target by the UAV swarm, calculate the position coordinates of the enemy target at time i using a collaborative positioning algorithm, and obtain the positioning accuracy based on a positioning accuracy calculation algorithm to obtain the trajectory data of the enemy target. The training module is used to improve the original Informer model to obtain an improved Informer model, preprocess the trajectory data, and train the improved Informer model to obtain a trained Informer model. The prediction module is used to use the trained Informer model to predict the long sequence of the enemy target's flight trajectory in real time to achieve accurate positioning and continuous tracking of the enemy target.

[0063] The specific implementation process of the construction module, control module, acquisition module, training module and prediction module can be referred to the multi-platform collaborative target trajectory tracking method in Embodiment 1, and will not be elaborated in this embodiment.

[0064] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-platform collaborative target trajectory tracking method, characterized in that, include: Obtain the trajectory set of the enemy target at time T, and construct the range of the enemy target's movement trajectory; The drone swarm is controlled to implement a global area coverage strategy over the range of the movement trajectory in order to track the enemy target; The drone swarm tracks the enemy target and coordinates to locate it. It calculates the position coordinates of the enemy target at time i using a collaborative positioning algorithm and obtains the positioning accuracy based on a positioning accuracy calculation algorithm, thereby obtaining the movement trajectory data of the enemy target. The positioning accuracy calculation algorithm includes: Based on the single-radar positioning accuracy and range error, the dividing angle between single-radar positioning accuracy and dual-radar cooperative positioning accuracy is obtained. ; When the angle between the enemy target and the two drones is in the range [0, ... ] or [π- When , π], the positioning accuracy of a single radar is taken as the positioning accuracy of the enemy target at time i; When the angle between the enemy target and the two drones is within the range of ( ,π- When the distance error and the angle between the two UAVs are calculated, the positioning accuracy of the enemy target at time i is obtained. The original Informer model is improved to obtain an improved Informer model. The motion trajectory data is preprocessed and then trained to obtain a trained Informer model. The trained Informer model is used to predict the long sequence of flight trajectories of the enemy target in real time, so as to achieve accurate positioning and continuous tracking of the enemy target.

2. The multi-platform collaborative target trajectory tracking method according to claim 1, characterized in that, The acquisition of the trajectory set of the enemy target at time T, and the construction of the movement trajectory range of the enemy target, specifically involves: Establish a coordinate system, taking the initial position of the enemy target as the origin, with the X-axis representing the flight direction; The movement of the enemy target includes two stages. The first stage is that the enemy target moves in an arc along the X-axis from the initial position until the direction of the enemy target's movement reaches a preset angle with the X-axis. The second stage is that the enemy target continues to move in a straight line along the direction of the preset angle. The range of the enemy target's movement trajectory is obtained by changing the preset direction angle.

3. The multi-platform collaborative target trajectory tracking method according to claim 1, characterized in that, The control of the drone swarm employs a global area coverage strategy over the range of the movement trajectory to track the enemy target, including: The detection range of the drone is a sampling circle, and the drone cluster is controlled to conduct detection within the range of the movement trajectory. The range of the motion trajectory is rasterized to obtain the intersection points between the grids; Each time, select an intersection point between the grids as the center of the sampling circle, and repeat this process until all intersection points between the grids are covered by the sampling circle.

4. The multi-platform collaborative target trajectory tracking method according to claim 3, characterized in that, The drone swarm tracks the enemy target and collaboratively locates it. A collaborative positioning algorithm calculates the target's coordinates at time i, and a positioning accuracy calculation algorithm is used to obtain the target's trajectory data. Specifically: The overlapping portion of the sampling circles is taken as the overlapping region. Within this overlapping region, the enemy target is located using collaborative positioning among UAVs to obtain positioning accuracy. Specifically, this includes: When only one UAV locates the enemy target, the positioning accuracy is the single radar positioning accuracy. When multiple UAVs locate the enemy target, the position coordinates of the enemy target at time i are obtained through the cooperative positioning algorithm, and the positioning accuracy between each pair of UAVs is calculated through the positioning accuracy calculation algorithm. When the positioning accuracy is optimal, the position coordinates of the enemy target at time i are the most accurate.

5. The multi-platform collaborative target trajectory tracking method according to claim 4, characterized in that, The cooperative localization algorithm is specifically as follows: The cooperative localization algorithm is used between at least two drones, including a first drone and a second drone; Obtain the position coordinates of the first UAV, the position coordinates of the second UAV, and the distance between the two UAVs. The distance between the first UAV and the enemy target and the distance between the second drone and the enemy target. The position coordinates of the enemy target are calculated based on the obtained data; When at least two UAVs are locating and tracking the same enemy target, the range error and the angle between the enemy target and the two UAVs are obtained respectively, and the positioning accuracy at time i is calculated based on the range error and the angle between the two UAVs.

6. The multi-platform collaborative target trajectory tracking method according to claim 4, characterized in that, The original Informer model is improved to obtain an improved Informer model. The motion trajectory data is preprocessed, and then the improved Informer model is trained to obtain a trained Informer model. The preprocessing includes: The position coordinates of the enemy target at time i and the optimal positioning accuracy are substituted into the formula of the Kalman filter, and the result is used as the estimated coordinates at time i. Collect the metadata of the enemy target at time i, the metadata including the estimated coordinates of the enemy target; The metadata dataset is normalized, and then the normalized metadata dataset is split into a training set, a validation set, and a test set.

7. The multi-platform collaborative target trajectory tracking method according to claim 6, characterized in that, The improved Informer model includes: The Informer model network architecture includes an Encoder module and a Decoder module, with the output of the Encoder module serving as the input of the Decoder module; The Encoder module is used to encode the input sequence, extract features from the time series data, and use a self-attention mechanism to capture the relationships in the input sequence. The Decoder module is used to generate future prediction sequences, and a multi-head attention mechanism module is used to process the fusion of historical sequence data and location embeddings.

8. The multi-platform collaborative target trajectory tracking method according to claim 7, characterized in that, The step of using the trained Informer model to predict the long sequence of the enemy target's flight trajectory in real time, so as to achieve accurate positioning and continuous tracking of the enemy target, includes: The enemy target is located and tracked by the trained Informer model network architecture. When a new meta-dataset is obtained, it is also normalized to obtain a new training set, validation set and test set. The Informer model network architecture was retrained, validated, and tested using new training, validation, and test sets. Incremental learning is used during training, updating only some parameters of the Informer model network architecture to improve training efficiency.

9. A multi-platform collaborative target trajectory tracking device, characterized in that, The device is applicable to the multi-platform collaborative target trajectory tracking method as described in any one of claims 1 to 8, comprising: a construction module, a control module, an acquisition module, a training module, and a prediction module; The construction module is used to obtain the trajectory set of the enemy target at time T and construct the range of the enemy target's movement trajectory; The control module is used to control the UAV swarm to implement a global area coverage strategy over the range of the movement trajectory in order to track the enemy target; The acquisition module is used to coordinate the positioning of the enemy target by the UAV cluster through tracking the enemy target, calculate the position coordinates of the enemy target at time i through a cooperative positioning algorithm, and obtain the positioning accuracy according to the positioning accuracy calculation algorithm, so as to obtain the motion trajectory data of the enemy target; The training module is used to improve the original Informer model to obtain an improved Informer model. After preprocessing the motion trajectory data, the improved Informer model is trained to obtain the trained Informer model. The prediction module is used to predict the long sequence of the enemy target's flight trajectory in real time using the trained Informer model, so as to achieve accurate positioning and continuous tracking of the enemy target.

Images