A high-speed target continuous tracking system and method based on inter-satellite cooperative calculation

Abstract

The application discloses a high-speed target continuous tracking system and method based on inter-satellite cooperative calculation, comprising a distributed architecture management module, an inter-satellite communication module, a dual-satellite cooperative tracking module and a distributed calculation coordination module. The system of the application adopts a distributed hierarchical management structure, through a hierarchical management mechanism of local management within a cluster and global coordination between clusters, the global management ability of the centralized layer is introduced while the fast response and autonomy of the distributed system are retained, and the communication overhead is effectively reduced. Moreover, the application designs a multi-type missile trajectory prediction method based on deep learning, and realizes a prediction-driven tracking switching mechanism, through an adaptive threshold to trigger the next round of prediction task, and the visibility window is avoided to be missed. In terms of calculation resource scheduling, the application can reduce the time delay of task scheduling, while ensuring the tracking accuracy, significantly reducing the communication overhead, and ensuring the high timeliness of the tracking target, and realizing the global continuous tracking of the target.

Description

Technical Field

[0001] This application belongs to the field of satellite communication technology, specifically relating to a high-speed continuous target tracking system and method based on inter-satellite collaborative computing. Background Technology

[0002] With the rapid development of low-Earth orbit (LEO) satellite constellations, achieving full-range target tracking through multi-satellite collaboration has become a crucial direction for military reconnaissance and early warning. Targets exhibit characteristics such as sudden launch, high speed, and strong maneuverability, posing significant challenges to existing detection systems. While traditional ground-based, sea-based radars, and high-Earth orbit platforms possess certain detection capabilities, they are limited by terrain obstruction, Earth's curvature, and link distance, making continuous global tracking difficult. Furthermore, transmission delays hinder rapid response to time-sensitive targets. In contrast, LEO satellites, with their low orbit, fast revisit times, and wide coverage, can generate multi-view observations in a short time, providing a new approach to achieving continuous tracking. However, missile flight involves multiple dynamic phases, and some targets possess lift maneuvers or evasive characteristics, significantly increasing the complexity of trajectory prediction and mission scheduling. Moreover, under traditional single-satellite processing modes, limitations in onboard computing resources make it difficult to complete computationally intensive tasks such as real-time target trajectory prediction, multi-target visibility calculation, and dynamic mission scheduling. Meanwhile, in large-scale constellations, relying solely on a fully distributed architecture will cause overhead to increase dramatically with scale due to frequent point-to-point communication, leading to link congestion and latency accumulation, making it difficult to meet the requirements for real-time and reliable continuous tracking. Therefore, how to design an efficient collaborative architecture with low communication overhead and full utilization of the constellation's distributed computing resources while ensuring prediction accuracy and rapid response has become an urgent problem to be solved.

[0003] In related technologies, the National University of Defense Technology proposed a fully distributed mission planning architecture based on a mutually exclusive target pool in its patent application number 202110158011.2. This satellite architecture can achieve multi-satellite relay tracking through dynamic priority mechanisms and request-response cooperation, which improves mission continuity and autonomy to a certain extent. However, this scheme relies on fully distributed communication and negotiation. As the constellation size and the number of targets increase, the communication overhead and cooperation complexity rise sharply, which can easily lead to link congestion and scheduling delays. At the same time, its trajectory prediction part is based on physical models and lacks a high-precision prediction mechanism for the complex dynamics of targets. It is difficult to ensure high accuracy and timeliness of tracking simultaneously under large-scale constellations. Furthermore, this architecture does not consider the scheduling of satellite computing resources, which can lead to uneven distribution of computing load, with some satellites having idle computing resources while others are overloaded, thus affecting the real-time response capability and tracking accuracy of the overall system.

[0004] Harbin Institute of Technology (Shenzhen) proposed a distributed two-layer auction scheduling mechanism in its patent application (application number 202510038095.4) that utilizes spatiotemporal grid coding, on-board resource storage, and a cluster management architecture. This method achieves rapid access to constellation resources by dividing the grid using geographic coordinates and allocates tasks between clusters and satellites through auctions, effectively alleviating the low efficiency of centralized scheduling and improving the resource allocation capabilities of large-scale constellations. However, this method primarily focuses on resource management for general moving targets, lacking consideration for the target's high speed and phased dynamic differences, and failing to effectively solve the coupling problem between trajectory prediction and task scheduling. Its application in time-sensitive, continuous tracking tasks like target tracking is therefore limited.

[0005] In summary, there is an urgent need to provide a method for a continuous target tracking system that achieves low overhead, high accuracy, and high timeliness in tracking. Summary of the Invention

[0006] To address the aforementioned problems in the existing technology, this application provides a high-speed continuous target tracking system based on inter-satellite cooperative computing. The technical problem to be solved by this application is achieved through the following technical solution:

[0007] This application provides a high-speed continuous target tracking system based on inter-satellite collaborative computing, comprising: a distributed architecture management module, an inter-satellite communication module, a dual-satellite collaborative tracking module, and a distributed computing coordination module deployed on satellites;

[0008] The distributed architecture management module is used to manage the allocation of tracking tasks and status monitoring of satellites within each satellite cluster, and to coordinate tracking tasks between satellites in different satellite clusters; each satellite cluster includes multiple satellites, and there is a cluster head satellite among the multiple satellites;

[0009] The inter-satellite communication module is used to forward tracking tasks and tracking data within a satellite cluster, as well as to coordinate tracking tasks between different satellite clusters.

[0010] The dual-satellite collaborative tracking module is used to execute the tracking task to obtain tracking data, and to use the tracking data to calculate the target's spatial position; and to determine the target's historical trajectory based on the target's spatial position.

[0011] The distributed computing coordination module is used to generate computing resource requirements using the target's historical trajectory and tracking task, and select a satellite based on the resource requirements so that the satellite can perform the tracking task to obtain a visible time window of the target's future trajectory.

[0012] This application provides a high-speed continuous target tracking method based on inter-satellite cooperative computing, characterized in that it utilizes a high-speed continuous target tracking system based on inter-satellite cooperative computing, and the high-speed continuous target tracking method based on inter-satellite cooperative computing includes:

[0013] S100, receive the tracking task of the target to be tracked and the target to be tracked by the tracking task;

[0014] S200: The distributed architecture management module selects two satellites within a cluster that are currently visible within the time window from multiple satellite clusters; the inter-satellite communication module forwards the tracking task; the dual-satellite collaborative tracking module controls the two satellites to track the target and obtain tracking data; the target's spatial position is calculated based on the tracking data, and the target's historical trajectory is determined based on the target's spatial position.

[0015] S300: Using the distributed computing coordination module, the system generates computing resource requirements based on the target's historical trajectory and the tracking task, and selects satellites to perform the tracking task based on the resource requirements, thereby obtaining the visible time window of the target's future trajectory for the member satellites within the cluster.

[0016] Beneficial effects:

[0017] 1. This application provides a high-speed continuous target tracking system and method based on inter-satellite collaborative computing. Compared with the existing technology's fully distributed task planning architecture based on mutually exclusive target pools, which relies on frequent point-to-point communication leading to a sharp increase in communication overhead with scale, this application designs a distributed architecture management architecture with inter-satellite collaborative computing capabilities. Through a hierarchical management mechanism of intra-cluster local optimization and inter-cluster global coordination, it retains the rapid response and autonomy of the distributed system while introducing the global management capabilities of the centralized layer, effectively reducing communication overhead. Simultaneously, the inter-satellite collaborative computing framework enables intelligent allocation of computational load, fully utilizing inter-satellite distributed computing resources and avoiding uneven distribution of computational load and overload of some satellites, thereby achieving efficient collaborative tracking with low communication overhead in large-scale constellations.

[0018] 2. This application provides a high-speed continuous target tracking system and method based on inter-satellite collaborative computing. Addressing the issue that existing resource allocation methods for tracking moving targets in the middle layer lack consideration for the differences in target speed and time sensitivity, this application designs a deep learning-based multi-type target trajectory prediction and prediction-driven tracking switching method. By identifying target types through a trajectory classification unit and selecting the corresponding deep learning prediction model, high-precision trajectory prediction for different target types is achieved. Simultaneously, a prediction-driven tracking switching mechanism is proposed, triggering the next round of prediction tasks through a preset time threshold. This effectively avoids visibility window errors and overcomes the limitation of traditional physical model prediction methods in simultaneously ensuring prediction accuracy and response speed in sudden maneuver scenarios, achieving full-domain continuous tracking of the target.

[0019] The present application will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0020] Figure 1 This is a system design block diagram of a high-speed target continuous tracking method based on inter-satellite cooperative computing provided in an embodiment of this application.

[0021] Figure 2 This is a flowchart illustrating a high-speed continuous target tracking method based on inter-satellite collaborative computing, as provided in an embodiment of this application.

[0022] Figure 3 This is a flowchart illustrating the implementation of dual-star cooperative tracking provided in an embodiment of this application.

[0023] Figure 4 This is a line graph showing the communication overhead results of different satellite architectures in a scenario involving 48 satellites and 6 concurrent targets, as provided in the embodiments of this application. Detailed Implementation

[0024] The present application will be described in further detail below with reference to specific embodiments, but the implementation of the present application is not limited thereto.

[0025] refer to Figure 1 As shown, Figure 1 This is a block diagram of a high-speed continuous target tracking method system based on inter-satellite collaborative computing, as proposed in this application. The high-speed continuous target tracking system based on inter-satellite collaborative computing provided in this application includes: a distributed architecture management module, an inter-satellite communication module, a dual-satellite collaborative tracking module, and a distributed computing coordination module deployed on satellites; the distributed architecture management module includes an intra-cluster management unit and an inter-cluster coordination unit; the inter-satellite communication module includes an intra-cluster message routing unit and an inter-cluster message routing unit; the dual-satellite collaborative tracking module includes a tracking execution unit and a tracking satellite switching unit; the distributed computing coordination module includes a deep learning prediction submodule and a resource computing submodule; the deep learning prediction submodule includes a trajectory classification unit and a trajectory prediction unit; and the resource computing submodule includes a computing requirement assessment unit and a visibility computing unit.

[0026] Specifically, the output of the intra-cluster management unit is communicatively connected to the input of the intra-cluster message routing unit and the input of the inter-cluster coordination unit; the output of the inter-cluster coordination unit is communicatively connected to the input of the inter-cluster message routing unit; the output of the intra-cluster message routing unit is communicatively connected to the input of the tracking execution unit and the tracking satellite switching unit, respectively; the input of the inter-cluster message routing unit is communicatively connected to the output of the inter-cluster coordination unit, and the output of the inter-cluster message routing unit is communicatively connected to the input of the intra-cluster management unit of each cluster; the input of the tracking execution unit is communicatively connected to the intra-cluster message routing unit. The output of the unit is connected to the network communication network; the input of the tracking satellite switching unit is connected to the output of the intra-cluster message routing unit; the input of the trajectory classification unit is connected to the output of the tracking execution unit, and the output of the trajectory classification unit is connected to the input of the trajectory prediction unit; the output of the trajectory prediction unit is connected to the input of the computational requirement assessment unit; the output of the computational requirement assessment unit is connected to the input of the visibility computation unit, and the output of the visibility computation unit is connected to the input of the intra-cluster management unit in the distributed architecture management module.

[0027] The distributed architecture management module of this application is used to manage the allocation of tracking tasks and status monitoring of satellites within each satellite cluster, and to coordinate the tracking tasks between satellites in different satellite clusters; each satellite cluster includes multiple satellites, and there is a cluster head satellite among the multiple satellites;

[0028] The inter-satellite communication module is used to forward tracking tasks and tracking data within a satellite cluster, as well as to coordinate tracking tasks between different satellite clusters.

[0029] The dual-satellite collaborative tracking module is used to execute the tracking task to obtain tracking data, and to use the tracking data to calculate the target's spatial position; and to determine the target's historical trajectory based on the target's spatial position.

[0030] The objective of this application is a high-speed space target, which is limited to a speed of not less than 1.3 km / s.

[0031] The distributed computing coordination module is used to generate computing resource requirements using the target's historical trajectory and tracking task, and select a satellite based on the resource requirements so that the satellite can perform the tracking task to obtain a visible time window of the target's future trajectory.

[0032] Continue to refer to Figure 1The cluster management unit is used to: cluster all satellites using the k-means clustering algorithm to obtain multiple satellite clusters, and select a cluster head satellite for each satellite cluster; send intra-cluster task instructions and satellite status information to the satellites within the cluster; initiate a tracking switch task when the visible time window of either of the two satellites performing the tracking task ends; determine whether there is an intra-cluster satellite with the next visible time window corresponding to the tracking switch task; if so, send the tracking switch task and the number of the intra-cluster satellite to the intra-cluster message routing unit; if not, initiate an inter-cluster coordination instruction.

[0033] The criterion for determining the end of the visible time window of either satellite in the two satellites performing the tracking task within the cluster management unit of this application is as follows:

[0034] When the visibility window for any currently tracked satellite has remaining time satisfy ≤ The tracking switching task is triggered at certain times; where the threshold is... Adaptive setting as follows:

[0035]

[0036] In the formula, As the minimum safety threshold, For the current network round-trip latency estimate, Budget for processing time during a forecast and assessment. All are non-negative coefficients, used to balance transmission delay and prediction task processing delay.

[0037] The cluster management unit in this application determines whether there are intra-cluster satellites in the next visible time window corresponding to the tracking handover task, including:

[0038] Calculate the evaluation function for each candidate satellite within the cluster, whereby the evaluation function is expressed as:

[0039]

[0040] in, Indicates candidate satellites within the cluster. Candidate satellites within the cluster The duration of the visible window for the target's future trajectory, For weight parameters, Candidate satellites The abundance of computational resources available for visibility computing tasks. ;

[0041] Using the evaluation function, the intra-cluster satellites corresponding to the next visible time window for the tracking handover mission are determined from the intra-cluster candidate satellites, as expressed as:

[0042]

[0043] in, For clusters Candidate satellites, For end-to-end delay estimation, This is the time-sensitive threshold.

[0044] The cluster management unit of this application has a satellite trajectory data pool within the cluster and outputs the tracking task to the cluster message routing unit through the SS1 interface; when the satellite resources within the cluster are insufficient, the visible time window ends, and a tracking switch task is initiated and output to the inter-cluster coordination unit through the RD2 interface.

[0045] If the intra-cluster management unit of this application determines that there are no intra-cluster satellites in the next visible time window corresponding to the tracking handover task, it initiates an inter-cluster coordination command. The process is as follows:

[0046] a: The REQUEST–RESPONSE–BID (RRB) inter-cluster coordination mechanism is adopted, and the inter-cluster coordination process is triggered when the visible resources within the cluster are insufficient;

[0047] b: The requesting cluster head satellite sends a REQUEST message to neighboring cluster head satellites. The message contains the target identifier (ID), the target's future trajectory, and the visible time window.

[0048] The inter-cluster coordination unit is used to select the pending satellite that meets the next visible time window from other clusters according to the cross-cluster coordination instruction, and send the number of the pending satellite and the tracking switching task to the inter-cluster message routing unit.

[0049] The inter-cluster coordination unit of this application is used to coordinate the cooperative tracking tasks between different satellite clusters and outputs the cooperative signal to the inter-cluster message routing unit through the SS3 interface.

[0050] The process by which the inter-cluster coordination unit of this application selects potential satellites that meet the next visible time window from other clusters according to cross-cluster coordination instructions includes:

[0051] c: Each neighboring cluster head accepts the REQUEST within its own cluster and evaluates it based on the visible time window, mission status, and computing resource usage of member satellites;

[0052] d: If any member satellite meets the coordination conditions, the neighboring cluster head satellite will return an ACCEPT response, along with the candidate satellite identifier and corresponding visibility duration; otherwise, a REJECT response will be returned.

[0053] e: The cluster head performs atomization competition processing according to the order of response arrival and temporarily locks the target during the evaluation period; when all responses are received or the preset timeout period is reached, the evaluation function is applied. The criteria select the best satellite as the relay satellite and the cluster to which the satellite belongs as the winning cluster;

[0054] f: The original cluster head sends an INFORM message to the selected winning cluster head, which then forwards the INFORM to its candidate member satellites. Both parties confirm the collaborative startup through the INFORM, and unselected clusters release their occupied resources.

[0055] The intra-cluster message routing unit is used to establish communication links between satellites within the cluster, and to forward tracking tasks, observation data packets, satellite trajectory pool data, tracking switching tasks, and satellite numbers within the same satellite cluster.

[0056] Continue to refer to Figure 1 The intra-cluster message routing unit is used to process communication message routing between satellites within the cluster, allocate tracking for dual satellites within the cluster, and output task information to the tracking execution unit through the SS2 interface. It also outputs the switching tracking decision to the tracking satellite switching unit through the DD1 interface.

[0057] The inter-cluster message routing unit is used to establish communication links between different satellite clusters and to forward tracking tasks, observation data packets, satellite trajectory pool data, tracking switching tasks, and the numbers of pending satellites between different satellite clusters.

[0058] Continue to refer to Figure 1 The inter-cluster message routing unit is used to handle the routing of communication messages between different satellite clusters, select the cooperative tracking cluster, and feed back the routing results to the intra-cluster management unit through the SS4 interface.

[0059] The tracking execution unit is used to receive tracking tasks, observation data packets, and satellite trajectory pool data, and fuse the observation data packets to execute the tracking task so as to calculate the target's spatial position using the fused observation data packets; and determine the target's historical trajectory based on the target's spatial position.

[0060] Continue to refer to Figure 1 The tracking execution unit outputs the target's historical trajectory to the trajectory classification unit via the SD1 interface.

[0061] The tracking satellite switching unit is used to receive the tracking switching task and the number of the satellite in the cluster or the number of the satellite to be determined, and to control the satellite in the cluster or the satellite to be determined to trigger the switching when the visibility window is about to end.

[0062] The trajectory classification unit is used to identify the stage type of the target's historical trajectory and output the classification result;

[0063] This application loads the trained target type classifier and trajectory prediction model into the trajectory classification unit and trajectory prediction unit, respectively; and loads the computational demand assessment model and the visibility computational physics model into the computational demand assessment unit and the visibility computational unit, respectively.

[0064] The target type classifier uses a one-dimensional temporal convolutional network (TCN). The trajectory prediction model uses a hybrid architecture of long short-term memory network and one-dimensional temporal convolutional network (LSTM-TCN).

[0065] Continue to refer to Figure 1 The trajectory classification unit of this application outputs the classification results to the trajectory prediction unit through the SD2 interface.

[0066] The trajectory prediction unit is used to select the corresponding prediction model according to the classification result, predict the future trajectory of the target according to the target's historical trajectory through the prediction model, and generate a visibility calculation task.

[0067] The prediction process of the deep learning-based trajectory prediction model used in this application is further described below:

[0068] Step S2.1: When satellite trajectory prediction is required, the trajectory prediction unit is initialized, loading pre-trained deep learning model components, including: a data normalizer, a target type classifier, and trajectory prediction models for different target types. The data normalizer standardizes the data based on statistical information from the training data. The target type classifier uses a one-dimensional temporal convolutional network and a multi-head attention mechanism to identify missile types; the trajectory prediction models for different target types predict future trajectory data based on historical target trajectory data.

[0069] Step S2.2: The system extracts a fixed length of historical trajectory data from the current time point backward as the prediction input. The window size for extracting historical trajectory data is 300 time steps, with each time step spaced 1 second apart. The historical trajectory data contains six dimensions: target position information (x, y, z) and velocity information (vx, vy, vz). For cases with insufficient data, zero-padding is used to ensure consistency of input dimensions.

[0070] Step S2.3: Convert the historical trajectory data into tensor format, adjust the dimension order to [batch, features, time], and then input it into the target classifier. The classifier extracts trajectory features through a temporal convolutional network, combines a multi-head attention mechanism to identify key temporal patterns, and outputs the target type classification result.

[0071] Step S2.4: Select the corresponding LSTM-TCN prediction model from the pre-trained models based on the classification results. Each model adopts a hybrid architecture of two-layer LSTM (128 hidden dimensions) + 6-layer TCN (64 channels per layer), which has the ability to perform temporal modeling.

[0072] Step S2.5: Generate future trajectories using a closed-loop autoregressive prediction strategy: First, initialize the prediction buffer, using normalized historical data as the initial state. In each subsequent prediction step, extract the time step of the most recent data window from the buffer as the model input. Predict the 6-dimensional state vector for the next time step using the selected LSTM-TCN model. Add the prediction result to the end of the buffer and update the historical state. Repeat this process until a prediction of the specified length is completed.

[0073] Step S2.6: Stack the normalized results of all prediction steps into a predicted trajectory matrix, and perform denormalization using a data normalizer to restore the original physical dimensions. Output complete future trajectory prediction data, including the three-dimensional position coordinates and three-dimensional velocity vectors for each time step.

[0074] refer to Figure 1 The trajectory prediction unit of this application outputs the future trajectory of the target to the computational requirement assessment unit through the SD3 interface.

[0075] The computational requirement assessment unit is used to generate computational resource requirements based on the target's future trajectory and visibility computational tasks;

[0076] refer to Figure 1 The computational requirement assessment unit of this application assesses the resource requirements of the tracking task, provides computational requirement constraints for the tracking task, and outputs the computational resource requirements to the visibility computation unit through the SD4 interface.

[0077] The visibility calculation unit is used to select a tracking satellite from candidate satellites to perform the tracking task based on the available resources and computing requirements of the satellite, so that the selected satellite can use the trajectory pool data to perform the visibility calculation task and obtain the visible time window of the target's future trajectory for the satellites in the satellite cluster.

[0078] refer to Figure 1 The visibility calculation unit is used to calculate the satellite's visibility window to the target, providing visibility constraints for the tracking mission, and outputting the calculation requirement data and visibility calculation data to the cluster management unit through the RD1 interface.

[0079] The availability calculation unit of this application selects tracking satellites from candidate satellites to perform tracking tasks based on the available resources and computing requirements of the satellites.

[0080] Based on the available satellite resources and computing requirements, a selection score is calculated for each candidate satellite, and the selection score is expressed as follows:

[0081]

[0082] in, Indicates candidate satellites, Communication overhead (data bytes). For weight parameters, For this mission on candidate satellites The end-to-end delay estimate is expressed as:

[0083]

[0084] In the formula, the processing delay is... , This represents the computational requirements for this visibility computation task. For node computing power; according to 1 is approximately ,in For arrival rate, For service rate, , This represents the average computational requirement for the computational task.

[0085] The candidate satellite with the lowest score is selected as the satellite to perform the tracking task. The selection process is as follows:

[0086]

[0087] The candidate satellites include primary observation satellites, secondary observation satellites, and cluster head satellites.

[0088] This application provides a high-speed continuous target tracking method based on inter-satellite cooperative computing, utilizing the aforementioned high-speed continuous target tracking system based on inter-satellite cooperative computing, such as... Figure 2 As shown, the high-speed continuous target tracking method based on inter-satellite cooperative computing includes:

[0089] S100, receive the tracking task of the target to be tracked and the target to be tracked by the tracking task;

[0090] S200: The distributed architecture management module selects two satellites within a cluster that are currently visible within the time window from multiple satellite clusters; the inter-satellite communication module forwards the tracking task; the dual-satellite collaborative tracking module controls the two satellites to track the target and obtain tracking data; the target's spatial position is calculated based on the tracking data, and the target's historical trajectory is determined based on the target's spatial position.

[0091] The specific execution tracking process of this application is as follows: Figure 3 As shown:

[0092] Step a: The cluster management unit analyzes the visibility status of satellites within the cluster based on the target's real-time trajectory, assigns tracking tasks to the corresponding cluster member satellites, and sends this information to the cluster message routing unit via the SS1 interface. The cluster message routing unit then sends the information to the tracking execution unit of the corresponding cluster member satellite via the SS2 interface.

[0093] Step b: The tracking execution unit analyzes the received tracking tasks and allocates a primary observation satellite and a secondary observation satellite based on computing resources when tracking a target for the first time or when switching tracking. The secondary observation satellite sends real-time observation data packets to the primary observation satellite, and the target's spatial position is calculated using a dual-satellite geometric positioning model.

[0094] Step c: The tracking execution unit in the master observation satellite outputs the target spatial position information to the deep learning prediction submodule on the master observation satellite through the SD1 interface.

[0095] S300: Using the distributed computing coordination module, the system generates computing resource requirements based on the target's historical trajectory and the tracking task, and selects satellites to perform the tracking task based on the resource requirements, thereby obtaining the visible time window of the target's future trajectory for the member satellites within the cluster.

[0096] The deep learning prediction module of this application performs trajectory prediction, the computational demand assessment unit evaluates the required computational resources, and the process of selecting the satellite for the visibility calculation task based on the scoring function is as follows:

[0097] Step a: The trajectory classification unit receives target position data, performs data preprocessing, extracts trajectory features, and inputs them into the classifier model. The classifier identifies the missile type (ascent phase / ballistic phase / semi-ballistic phase) and outputs the classification results to the trajectory prediction unit via the SD2 interface.

[0098] Step b: The trajectory prediction unit selects the corresponding prediction model constructed from a long short-term memory network based on the classification results. The prediction model predicts the missile's trajectory over a future period based on the historical trajectory data of the current target, generating prediction result data.

[0099] Step c: Output the prediction results data to the computational requirement assessment unit via the SD3 interface. Based on satellite computing resources and mission computational requirements, select the satellite to execute the visibility computation task, choosing from the primary observation satellite, secondary observation satellites, and cluster head satellites, and selecting the candidate execution node with the lowest current score. The satellite serves as the execution node for the visibility computing task, outputting prediction results data and computing task requests to the satellite's visibility computing unit via the SD4 interface.

[0100] When there are insufficient satellite resources within a cluster, i.e., the visible time window of a satellite within the cluster has ended and there are no satellites within the cluster with the next visible time window corresponding to the tracking handover task, the intra-cluster management unit initiates an inter-cluster coordination command. The inter-cluster coordination unit then selects pending satellites between clusters. The specific process is as follows:

[0101] Step 1: The REQUEST–RESPONSE–BID (RRB) request-response coordination mechanism is adopted to trigger the inter-cluster coordination process when there are insufficient visible resources within the cluster.

[0102] Step 2: The requesting cluster head satellite sends a REQUEST message to the neighboring cluster head satellites. The message contains the target identifier (ID), the target predicted trajectory, and the visibility window information.

[0103] Step 3: Each neighboring cluster head accepts the REQUEST within its own cluster and evaluates it based on the visibility window, mission status, and computing resource usage of the member satellites.

[0104] Step 4: If any member satellite meets the coordination conditions, the neighboring cluster head satellite returns an ACCEPT response, along with the candidate satellite identifier and corresponding visibility duration; if not, a REJECT response is returned.

[0105] Step 5: The cluster head performs atomization competition processing according to the order of response arrival and temporarily locks the target during the evaluation period; when all responses are received or the preset timeout period is reached, the evaluation function is applied. The criteria select the best satellite as the relay satellite and the cluster to which the satellite belongs as the winning cluster;

[0106] Step 6: The original cluster head sends an INFORM message to the selected winning cluster head, which then forwards the INFORM to its candidate member satellites. Both parties confirm the collaborative startup through the INFORM, and unselected clusters release their occupied resources.

[0107] Once a cluster is selected, the satellite tracking process within the cluster begins, as follows:

[0108] Step a: The intra-cluster message routing unit sends a tracking satellite switching task to the tracking switching unit of the intra-cluster member satellite based on the member satellite identifier of the relay tracking.

[0109] Step b: Perform a tracking satellite switch. The satellite whose visible window in the original two-satellite pair is about to end releases its tracking of the target, and the new relay tracking satellite and the satellite that has not yet ended its tracking form a new two-satellite pair.

[0110] This application performs a cyclical tracking and prediction process to achieve continuous tracking and trajectory prediction of the target; dynamically adjusts the tracking strategy and computing resource allocation according to changes in the target's motion state; and terminates the tracking process when the target is out of the visible range of all satellites or when the tracking mission is completed.

[0111] This application discloses a high-speed continuous target tracking system and method based on inter-satellite collaborative computing, comprising: a distributed architecture management module, an inter-satellite communication module, a dual-satellite collaborative tracking module, and a distributed computing coordination module. Addressing the problems of rapidly increasing communication overhead and uneven computational load distribution in traditional fully distributed architectures under large-scale constellations, the system adopts a distributed hierarchical management structure. Through a hierarchical management mechanism of local management within clusters and global coordination between clusters, it retains the rapid response and autonomy of distributed systems while introducing the global management capabilities of a centralized layer, effectively reducing communication overhead. Considering the high speed and phased dynamic differences of missile targets, a multi-type missile trajectory prediction method based on deep learning is designed, and a prediction-driven tracking switching mechanism is implemented. The next round of prediction tasks is triggered by an adaptive threshold, avoiding missed visibility windows. Regarding computational resource scheduling, the optimal execution node is selected based on a scoring function, achieving intelligent allocation of computational load and reducing task scheduling latency. This invention significantly reduces communication overhead while ensuring tracking accuracy and high timeliness of the tracked target, achieving full-domain continuous tracking of missile targets.

[0112] In one optional embodiment of this application, the effects of this application will be further described below in conjunction with simulation experiments.

[0113] 1. Simulation experimental conditions:

[0114] The simulation experiment platform in this application is: Windows 10 operating system and Python 3.8.

[0115] The simulation experiment in this application sets up a satellite network scenario with 48 satellites, using the Walker constellation, and the orbital altitude is... ,inclination In the region The simulation includes six concurrent targets: three lifting missiles, one semi-ballistic missile, and two ballistic missiles. The simulation duration is set to 2000 seconds.

[0116] 2. Simulation content and result analysis:

[0117] In the simulation experiments of this application, the following two target tracking architectures were simulated respectively: the fully distributed satellite missile tracking architecture and the high-speed continuous target tracking system based on inter-satellite cooperative computing designed in this application.

[0118] Depend on Figure 4As can be seen, the communication overhead of the high-speed continuous target tracking system based on inter-satellite collaborative computing designed in this application is lower than that of the existing fully distributed satellite missile tracking architecture. With the increase in the number of concurrent missiles, the communication overhead of the two architectures exhibits significantly different growth patterns. This is because in the fully distributed satellite architecture, each tracking switch requires all satellites to participate in global negotiation, resulting in a large increase in the number of messages with the number of nodes and targets. In contrast, the communication overhead growth of the distributed architecture in this application is more gradual, thanks to its hierarchical collaborative mechanism: intra-cluster tracking transfer only requires intra-cluster satellites to participate in decision-making, while inter-cluster collaboration involves limited information exchange through the cluster head satellite, and optimizes the selection of computation task execution nodes through a scoring function, reducing unnecessary data transmission.

[0119] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of this application and should not be construed as limiting the specific implementation of this application to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of this application, and all such modifications or substitutions should be considered within the scope of protection of this application.

Claims

1. A high-speed continuous target tracking system based on inter-satellite cooperative computing, characterized in that, include: Deployed on satellites are distributed architecture management modules, inter-satellite communication modules, dual-satellite collaborative tracking modules, and distributed computing coordination modules; The distributed architecture management module is used to manage the allocation of tracking tasks and status monitoring of satellites within each satellite cluster, and to coordinate tracking tasks between satellites in different satellite clusters; each satellite cluster includes multiple satellites, and there is a cluster head satellite among the multiple satellites; The inter-satellite communication module is used to forward tracking tasks and tracking data within a satellite cluster, as well as to coordinate tracking tasks between different satellite clusters. The dual-satellite collaborative tracking module is used to execute the tracking task to obtain tracking data, and to use the tracking data to calculate the target's spatial position. The target's historical trajectory is determined based on its spatial location; The distributed computing coordination module is used to generate computing resource requirements using the target's historical trajectory and tracking task, and select a satellite based on the resource requirements so that the satellite can perform the tracking task to obtain a visible time window of the target's future trajectory; wherein, the criterion for determining the end of the visible time window is: When the visibility window for any currently tracked satellite has remaining time satisfy ≤ The tracking switching task is triggered at certain times; where the threshold is... Adaptive setting as follows: In the formula, As the minimum safety threshold, For the current network round-trip latency estimate, Budget for processing time during a forecast and assessment. All are non-negative coefficients, used to balance transmission delay and prediction task processing delay.

2. The high-speed continuous target tracking system based on inter-satellite cooperative computing according to claim 1, characterized in that, The distributed architecture management module includes an intra-cluster management unit and an inter-cluster coordination unit; the inter-satellite communication module includes an intra-cluster message routing unit and an inter-cluster message routing unit; the dual-satellite collaborative tracking module includes a tracking execution unit and a tracking satellite switching unit; the distributed computing coordination module includes a deep learning prediction submodule and a resource computing submodule; the deep learning prediction submodule includes a trajectory classification unit and a trajectory prediction unit, and the resource computing submodule includes a computing requirement assessment unit and a visibility computing unit; Specifically, the output of the intra-cluster management unit is communicatively connected to the input of the intra-cluster message routing unit and the input of the inter-cluster coordination unit; the output of the inter-cluster coordination unit is communicatively connected to the input of the inter-cluster message routing unit; the output of the intra-cluster message routing unit is communicatively connected to the input of the tracking execution unit and the tracking satellite switching unit, respectively; the input of the inter-cluster message routing unit is communicatively connected to the output of the inter-cluster coordination unit, and the output of the inter-cluster message routing unit is communicatively connected to the input of the intra-cluster management unit of each cluster; the input of the tracking execution unit is communicatively connected to the intra-cluster message routing unit. The output of the unit is connected to the network communication network; the input of the tracking satellite switching unit is connected to the output of the intra-cluster message routing unit; the input of the trajectory classification unit is connected to the output of the tracking execution unit, and the output of the trajectory classification unit is connected to the input of the trajectory prediction unit; the output of the trajectory prediction unit is connected to the input of the computational requirement assessment unit; the output of the computational requirement assessment unit is connected to the input of the visibility computation unit, and the output of the visibility computation unit is connected to the input of the intra-cluster management unit in the distributed architecture management module.

3. The high-speed continuous target tracking system based on inter-satellite cooperative computing according to claim 2, characterized in that, The intra-cluster management unit is used to: cluster all satellites using the k-means clustering algorithm to obtain multiple satellite clusters, and select a cluster head satellite for each satellite cluster; send intra-cluster task instructions and satellite status information to the satellites within the cluster; initiate a tracking switch task when the visible time window of either of the two satellites performing the tracking task ends; determine whether there is an intra-cluster satellite with the next visible time window corresponding to the tracking switch task; if so, send the tracking switch task and the number of the intra-cluster satellite to the intra-cluster message routing unit; if not, initiate an inter-cluster coordination instruction. The inter-cluster coordination unit is used to select the pending satellite that meets the next visible time window from other clusters according to the cross-cluster coordination instruction, and send the number of the pending satellite and the tracking switching task to the inter-cluster message routing unit.

4. The high-speed continuous target tracking system based on inter-satellite cooperative computing according to claim 3, characterized in that, The determination of whether there are intra-cluster satellites corresponding to the next visible time window of the tracking switching task includes: Calculate the evaluation function for each candidate satellite within the cluster, whereby the evaluation function is expressed as: in, Indicates candidate satellites within the cluster. Candidate satellites within the cluster The duration of the visible window for the target's future trajectory, For weight parameters, Candidate satellites The abundance of computational resources available for visibility computing tasks. ; Using the evaluation function, the intra-cluster satellites corresponding to the next visible time window for the tracking handover mission are determined from the intra-cluster candidate satellites, as expressed as: in, For clusters Candidate satellites, For end-to-end delay estimation, This is the time-sensitive threshold.

5. The high-speed continuous target tracking system based on inter-satellite cooperative computing according to claim 2, characterized in that, The intra-cluster message routing unit is used to establish communication links between satellites within the cluster, and to forward tracking tasks, observation data packets, satellite trajectory pool data, tracking switching tasks, and satellite numbers within the same satellite cluster. The inter-cluster message routing unit is used to establish communication links between different satellite clusters and to forward tracking tasks, observation data packets, satellite trajectory pool data, tracking switching tasks, and the numbers of pending satellites between different satellite clusters.

6. The high-speed continuous target tracking system based on inter-satellite cooperative computing according to claim 2, characterized in that, The tracking execution unit is used to receive tracking tasks, observation data packets, and satellite trajectory pool data, and to fuse the observation data packets to execute the tracking task so as to calculate the target spatial position using the fused observation data packets; The historical trajectory of the target is determined based on its spatial location; The tracking satellite switching unit is used to receive the tracking switching task and the number of the satellite in the cluster or the number of the satellite to be determined, and to control the satellite in the cluster or the satellite to be determined to trigger the switching when the visibility window is about to end.

7. The high-speed continuous target tracking system based on inter-satellite cooperative computing according to claim 6, characterized in that, The trajectory classification unit is used to identify the stage type of the target's historical trajectory and output the classification result; The trajectory prediction unit is used to select the corresponding prediction model according to the classification result, predict the future trajectory of the target according to the target's historical trajectory through the prediction model, and generate a visibility calculation task. The computational requirement assessment unit is used to generate computational resource requirements based on the target's future trajectory and visibility computational tasks; The visibility calculation unit is used to select a tracking satellite from candidate satellites to perform the tracking task based on the available resources and computing requirements of the satellite, so that the selected satellite can use the trajectory pool data to perform the visibility calculation task and obtain the visible time window of the target's future trajectory for the satellites in the satellite cluster.

8. The high-speed continuous target tracking system based on inter-satellite cooperative computing according to claim 7, characterized in that, Based on the available resources and computing requirements of the satellites, the tracking satellites selected from the candidate satellites to perform the tracking mission include: Based on the available satellite resources and computing requirements, a selection score is calculated for each candidate satellite, and the selection score is expressed as follows: in, Indicates candidate satellites, For communication overhead, For weight parameters, For this mission on candidate satellites The end-to-end delay estimate is expressed as: In the formula, the processing delay is... , This represents the computational requirements for this visibility computation task. For node computing power; ,in For arrival rate, For service rate, , This represents the average computational requirement for the computational task. The candidate satellite with the lowest score is selected as the satellite to perform the tracking task. The selection process is as follows: The candidate satellites include primary observation satellites, secondary observation satellites, and cluster head satellites.

9. A high-speed continuous target tracking method based on inter-satellite cooperative computing, characterized in that, The high-speed continuous target tracking system based on inter-satellite cooperative computing according to any one of claims 1 to 8, wherein the high-speed continuous target tracking method based on inter-satellite cooperative computing includes: S100, receive the tracking task of the target to be tracked and the target to be tracked by the tracking task; S200: The distributed architecture management module selects two satellites within a cluster that are currently visible within the time window from multiple satellite clusters; the inter-satellite communication module forwards the tracking task; the dual-satellite collaborative tracking module controls the two satellites to track the target and obtain tracking data; the target's spatial position is calculated based on the tracking data, and the target's historical trajectory is determined based on the target's spatial position. S300, utilizing the distributed computing coordination module, generates computing resource requirements based on the target's historical trajectory and the tracking task, and selects satellites to execute the tracking task based on these resource requirements, thereby obtaining the visible time window of the target's future trajectory for member satellites within the cluster; wherein, the criterion for determining the end of the visible time window is: When the visibility window for any currently tracked satellite has remaining time satisfy ≤ The tracking switching task is triggered at certain times; where the threshold is... Adaptive setting as follows: In the formula, As the minimum safety threshold, For the current network round-trip latency estimate, Budget for processing time during a forecast and assessment. All are non-negative coefficients, used to balance transmission delay and prediction task processing delay.

Images