Collision detection method for space objects, and device, electronic equipment and program product thereof

Abstract

The application discloses a space target collision detection method and device, electronic equipment and program product, and relates to the technical field of spacecraft orbit safety, wherein the collision detection method comprises the following steps: acquiring orbit ephemeris data of space targets, mapping the orbit ephemeris data into an orbit parameter tensor, determining a relative distance tensor of each pair of space targets based on the orbit parameter tensor, interpolating the relative distance tensor of each pair of space targets based on a preset Chebyshev basis function matrix to obtain an expected relative distance tensor of each pair of space targets, performing iterative search on each expected relative distance tensor to determine a target extreme time point between each pair of space targets, and determining a collision state between the pair of space targets based on the target extreme time point. The application solves the technical problem of low accuracy of collision detection of space targets in the related art.

Description

Technical Field

[0001] This invention relates to the field of spacecraft orbital safety technology, and more specifically, to a collision detection method, apparatus, electronic equipment, and program product for space targets. Background Technology

[0002] With the exponential growth in the number of spacecraft in low Earth orbit, space target collision early warning has become a core technology for ensuring the safety of on-orbit assets. Currently, more than 55,000 space targets have been cataloged, with the highest density in low Earth orbit, where approximately 70% of spacecraft operate at altitudes between 200 and 2000 km. The accelerated deployment of mega-constellations, represented by Starlink and OneNet, with single constellations now exceeding tens of thousands of objects, has led to frequent rendezvous events in space.

[0003] In related technologies, collision detection relies on serial computation using a CPU (Central Processing Unit). It employs SGP4 / SDP4 (Simplified General Perturbations Model 4 / Simplified Drag Model 4) orbital models to predict ephemeris values ​​per target and time node, and performs relative distance analysis on pairs of targets, resulting in low computational efficiency. Taking 20,000 targets as an example, it requires processing... For single-target pairs, the warning cycle can last from tens of minutes to several hours, which cannot meet the real-time requirements of minute-level response in scenarios such as space stations. Furthermore, the extreme value search relies on high-precision time interpolation techniques. The current binary search iteration leads to a decrease in the parallel efficiency of GPU (Graphics Processing Unit) threads. While cubic spline interpolation can achieve continuous interpolation, it increases computational complexity and cannot process thousands of targets in batches. In addition, taking an application scenario with a warning duration of 3 days and an ephemeris data interval of one minute as an example, directly calculating the collision distance of 20000×20000 target pairs requires storing 20000×20000×4320=1.728×1012 data points (single-precision floating-point format occupying 6.4TB of GPU memory), far exceeding the current GPU hardware capabilities. The current block-based calculation scheme uses a static data segmentation strategy, which does not consider the temporal correlation caused by orbital perturbations, requiring frequent host-device data transfers. The measured throughput is less than 200GB / s, leading to increased computational latency.

[0004] Therefore, the collision detection in the related art cannot balance the contradiction among efficiency, accuracy and scale: the CPU scheme is limited by the failure of Moore's law, and the computing power growth stagnates; the current GPU scheme is difficult to break through the real-time early warning bottleneck of ten thousand targets due to the lack of parallelism of the algorithm and the defects of the video memory management. With the application of heavy transport vehicles, the number of satellites deployed by a single launch will reach more than 400, and the current collision detection technology cannot meet the management needs of one million targets in the next ten years. Therefore, a parallelized early warning method supporting large-scale target real-time processing, controllable video memory occupation and up-to-standard accuracy is urgently needed to ensure the safety of spacecraft in orbit.

[0005] At present, no effective solution has been proposed for the above problems. SUMMARY

[0006] Embodiments of the present application provide a space target collision detection method and device, electronic equipment and program product, to solve the technical problem of low accuracy of collision detection of space targets in related technologies.

[0007] According to an aspect of an embodiment of the present application, a space target collision detection method is provided, comprising: obtaining orbit ephemeris data of space targets, wherein the orbit ephemeris data at least includes position information and velocity information of each space target at each preset time interval point within a preset time period, and the time interval step between adjacent two preset time interval points is an initial step; mapping the orbit ephemeris data into an orbit parameter tensor, and determining a relative distance tensor of each space target pair based on the orbit parameter tensor, wherein the space target pair is composed of any two space targets, and the relative distance tensor records the relative distance of the two space targets indicated by the space target pair at each preset time interval point; interpolating the relative distance tensor of each space target pair based on a preset Chebyshev basis function matrix, to obtain an expected relative distance tensor of each space target pair, wherein the expected relative distance tensor records the relative distance of the two space targets indicated by the space target pair at each expected time interval point, and the time interval step between adjacent two expected time interval points is an expected step; performing iterative search on each expected relative distance tensor to determine a target extreme value time point between each space target pair, and judging the collision risk between the space target pair based on the relative distance at the target extreme value time point.

[0008] Further, the step of obtaining the orbit ephemeris data of the space object comprises: collecting two-line element data of the space object, wherein the two-line element data is preset fixed format data used for describing an orbit state of the space object; and performing analysis on the two-line element data to obtain orbit six-element data, wherein the orbit six-element data at least comprises a semi-major axis, an eccentricity, an inclination, an ascending node right ascension, a perigee amplitude, and a mean anomaly of the space object; and generating orbit ephemeris data of the space object in a preset time period based on the orbit six-element data.

[0009] Further, the step of determining the relative distance tensor of each group of space object pairs based on the orbit parameter tensor comprises: determining a block granularity parameter based on a target quantity of all space objects and a memory space, wherein the memory space is a memory space of a GPU used for storing the orbit parameter tensor and calculating the relative distance tensor; decomposing all space object pairs into a plurality of blocks based on the block granularity parameter, wherein a quantity of space object pairs included in each block is determined based on the target quantity and the block granularity parameter; and for each block, calculating the relative distance tensor of each group of space object pairs in the block based on the orbit parameter tensor.

[0010] Further, before interpolating the relative distance tensor of each group of space object pairs based on the preset Chebyshev basis function matrix, the method further comprises: performing standardization processing on a timestamp indicated by each preset time interval point in the relative distance tensor to obtain a standardized relative distance tensor; and generating the preset Chebyshev basis function matrix based on a parallel processing quantity of the GPU for processing the space object pairs.

[0011] Further, the step of interpolating the relative distance tensor of each group of space object pairs based on the preset Chebyshev basis function matrix to obtain an expected relative distance tensor of each group of space object pairs comprises: determining a coefficient matrix of the space object pairs by least square fitting based on the relative distance tensor of each group of space object pairs and the preset Chebyshev basis function matrix; determining a relative distance of the space object pairs at each expected time interval point based on an expected step length and the coefficient matrix; and interpolating the relative distance at an expected step length interval between each two adjacent preset time interval points in the relative distance tensor to obtain the expected relative distance tensor.

[0012] Further, the step of interpolating the relative distance at an expected step length interval between each two adjacent preset time interval points in the relative distance tensor comprises: generating an expected time interval point at each interval of an expected step length; and inserting the relative distance corresponding to the expected time interval point at the expected time interval point.

[0013] Further, the step of performing an iterative search on each expected relative distance tensor to determine the target extreme time point between each group of pairs of space targets comprises: traversing the expected relative distance tensor with an initial step length to determine a first candidate interval, wherein the first candidate interval has a first time point and a second time point, the relative distance at the first time point is less than the preset relative distance, and the relative distance at the second time point is greater than the preset relative distance; for each first candidate interval, performing a correction on the first candidate interval with a first step length to obtain a second candidate interval, wherein the first step length is less than the initial step length and greater than an expected step length, and the length of the second candidate interval is less than the length of the first candidate interval; for each second candidate interval, performing a correction on the second candidate interval with a second step length to obtain an initial extreme time point, wherein the second step length is less than the first step length and greater than the expected step length; based on the initial extreme time point, determining a third candidate interval, and for each third candidate interval, performing a correction on the third candidate interval with the expected step length to obtain a preset extreme time point; shielding a neighborhood of the preset extreme time point, and re-performing an iterative search on the expected relative distance tensor to obtain a new preset extreme time point, until the number of iterations reaches a preset number of times, to obtain a preset number of preset extreme time points, wherein the preset number is equal to the preset number of times; and determining the preset extreme time point corresponding to the minimum relative distance as the target extreme time point.

[0014] According to another aspect of the embodiment of the present application, a collision detection device for space targets is also provided, comprising: an acquisition unit configured to acquire orbit ephemeris data of the space targets, wherein the orbit ephemeris data at least includes position information and velocity information of each space target at each preset time interval point within a preset time period, and the time interval step length between adjacent two preset time interval points is an initial step length; a determination unit configured to map the orbit ephemeris data into an orbit parameter tensor, and determine a relative distance tensor of each group of pairs of space targets based on the orbit parameter tensor, wherein the pair of space targets is composed of any two space targets, and the relative distance tensor records the relative distance between the two space targets indicated by the pair of space targets at each preset time interval point; an interpolation unit configured to perform interpolation on the relative distance tensor of each group of pairs of space targets based on a preset Chebyshev basis function matrix to obtain an expected relative distance tensor of each group of pairs of space targets, wherein the expected relative distance tensor records the relative distance between the two space targets indicated by the pair of space targets at each expected time interval point, and the time interval step length between adjacent two expected time interval points is an expected step length; and a search unit configured to perform an iterative search on each expected relative distance tensor to determine a target extreme time point between each group of pairs of space targets, and determine a collision state between the pair of space targets based on the target extreme time point.

[0015] Further, the acquisition unit comprises: a first collection module, configured to collect two-line element data of the space target, wherein the two-line element data is preset fixed format data used to describe an orbit state of the space target; a first analysis module, configured to analyze the two-line element data to obtain orbit six-element data, wherein the orbit six-element data at least comprises a semi-major axis, an eccentricity, an inclination, an ascending node right ascension, a perigee amplitude, and a mean anomaly of the space target; and a first generation module, configured to generate orbit ephemeris data of the space target in a preset time period based on the orbit six-element data.

[0016] Further, the determination unit comprises: a first determination module, configured to determine a block granularity parameter based on a target quantity of all space targets and a memory space, wherein the memory space is a memory space of a GPU used to store an orbit parameter tensor and calculate a relative distance tensor; a first decomposition module, configured to decompose all space target pairs into a plurality of blocks based on the block granularity parameter, wherein a quantity of space target pairs included in each block is determined based on the target quantity and the block granularity parameter; and a first calculation module, configured to calculate, for each block, a relative distance tensor of each group of space target pairs in the block based on the orbit parameter tensor.

[0017] Further, the collision detection apparatus further comprises: a first processing module, configured to perform standardization processing on a timestamp indicated by each preset time interval point in the relative distance tensor to obtain a standardized relative distance tensor before interpolating the relative distance tensor of each group of space target pairs based on a preset Chebyshev basis function matrix; and a second generation module, configured to generate the preset Chebyshev basis function matrix based on a parallel processing quantity of the GPU for processing the space target pairs.

[0018] Further, the interpolation unit comprises: a second determination module, configured to determine a coefficient matrix of the space target pair based on observation position data of each space target in the space target pair and the preset Chebyshev basis function matrix by using least square fitting; a third determination module, configured to determine a relative distance of the space target pair at each expected time interval point based on an expected step length and the coefficient matrix; and a first interpolation module, configured to interpolate the relative distance at an expected step length interval between each two adjacent preset time interval points in the relative distance tensor to obtain an expected relative distance tensor.

[0019] Further, the first interpolation module comprises: a first generation submodule, configured to generate an expected time interval point every expected step length interval; and a first insertion submodule, configured to insert the relative distance corresponding to the expected time interval point at the expected time interval point.

[0020] Further, the searching unit comprises: a fourth determination module, configured to traverse the expected relative distance tensor by an initial step, and determine a first candidate interval, wherein the first candidate interval is determined by a first time point and a second time point, the relative distance at the first time point is less than the preset relative distance, and the relative distance at the second time point is greater than the preset relative distance; a first correction module, configured to correct each first candidate interval by a first step to obtain a second candidate interval, wherein the first step is less than the initial step and greater than the expected step, and the length of the second candidate interval is less than the length of the first candidate interval; a second correction module, configured to correct each second candidate interval by a second step to obtain an initial extreme time point, wherein the second step is less than the first step and greater than the expected step; a third correction module, configured to determine a third candidate interval based on the initial extreme time point, and correct each third candidate interval by the expected step to obtain a preset extreme time point; and a first searching module, configured to shield a neighborhood of the preset extreme time point, and re-iteratively search a new preset extreme time point in the expected relative distance tensor until the number of iterations reaches a preset number, to obtain a preset number of preset extreme time points, wherein the preset number is equal to the preset number; and determine the preset extreme time point corresponding to the minimum relative distance as the target extreme time point.

[0021] According to another aspect of the embodiments of the present application, a computer program product is also provided, comprising a non-volatile computer readable storage medium, the non-volatile computer readable storage medium storing a computer program, the computer program being executed by a processor to implement the space target collision detection method of any one of the above.

[0022] According to another aspect of the embodiments of the present application, an electronic device is also provided, comprising one or more processors and a memory, the memory being configured to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the space target collision detection method of any one of the above.

[0023] In the present application, the orbit ephemeris data of the space target is acquired, the orbit ephemeris data is mapped into an orbit parameter tensor, the relative distance tensor of each pair of space targets is determined based on the orbit parameter tensor, the relative distance tensor of each pair of space targets is interpolated based on a preset Chebyshev basis function matrix to obtain an expected relative distance tensor of each pair of space targets, the target extreme time point between each pair of space targets is determined by iteratively searching each expected relative distance tensor, and the collision state between the pair of space targets is determined based on the target extreme time point, thereby solving the technical problem of low accuracy of collision detection of space targets in the related art.

[0024] In the present application, the device end parallel storage is realized by tensorizing the orbit data, the high-efficiency parallel interpolation is realized by using the least square fitting method of Chebyshev basis function matrix, the data processing delay is effectively reduced by combining the block dynamic tensor calculation strategy and the multi-resolution extreme value search algorithm, the real-time performance and accuracy of the early warning are ensured, and the purpose of real-time processing of large-scale space target collision early warning is achieved, so that the technical effects of improving the calculation efficiency and accuracy are realized. BRIEF DESCRIPTION OF DRAWINGS

[0025] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:

[0026] Figure 1 is a flow chart of an optional space target collision detection method according to an embodiment of the present application;

[0027] Figure 2 is a schematic diagram of an optional space target collision detection system architecture according to an embodiment of the present application;

[0028] Figure 3 is a schematic diagram of an optional data preprocessing module flow according to an embodiment of the present application;

[0029] Figure 4 is a schematic diagram of an optional block dynamic matrix strategy according to an embodiment of the present application;

[0030] Figure 5 is a schematic diagram of an optional GPU-Tensor calculation chain according to an embodiment of the present application;

[0031] Figure 6 is a schematic diagram of an optional multi-resolution extreme value search flow according to an embodiment of the present application;

[0032] Figure 7 is a schematic diagram of an optional space target collision detection device according to an embodiment of the present application;

[0033] Figure 8 is a hardware structure block diagram of an electronic device (or mobile device) for a space target collision detection method according to an embodiment of the present application. DETAILED DESCRIPTION

[0034] In the following, the technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application in order to make the technical personnel in the technical field better understand the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by the ordinary technical personnel in the technical field without creative labor should belong to the scope of protection of the present application.

[0035] It should be noted that the terms "first", "second" and the like in the present application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0036] It should be noted that the information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) collected and related to the present application are all authorized by the user or authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of related data comply with relevant laws, regulations and standards in relevant regions, necessary security measures are taken, public order and good customs are not violated, and appropriate operation portals are provided for users to choose authorization or refusal. For example, interfaces are provided between the system and related users or agencies. Before obtaining the relevant information, the interface needs to send a request to the aforementioned user or agency, and after receiving the consent information feedback from the aforementioned user or agency, the relevant information is obtained.

[0037] The application provides a large-scale space target collision warning parallel computing method, which is suitable for real-time collision warning, dense threat target screening and space debris risk assessment of a spacecraft. The method comprises the following steps: constructing an orbit parameter tensorization model based on a GPU-Tensor (a computing architecture combining the parallel computing capability of a GPU and a tensor computing framework), mapping satellite orbit data into a three-dimensional tensor structure to realize parallel storage on a device side; designing a batch Chebyshev polynomial interpolation operator, and realizing high-precision parallel interpolation with millisecond-level time resolution through least square fitting of an orthogonal basis function matrix; proposing a block dynamic tensor computing strategy, decomposing large-scale targets into tensor subblocks suitable for memory capacity, and realizing efficient computation in combination with pipelined data dynamic loading; and constructing a multi-resolution extreme value search algorithm, and improving extreme value positioning accuracy to the order of hundreds of meters and eliminating the risk of missed detection through coarse-grained screening and three-iteration interpolation correction. In this way, the problems of low CPU serial computing efficiency, high GPU memory occupation and insufficient interpolation accuracy can be solved, and efficient and reliable technical support is provided for real-time safety warning of a spacecraft such as a space station and a satellite constellation in a dense threat environment.

[0038] The application will be described in detail below in conjunction with various embodiments.

[0039] Embodiment one

[0040] According to the embodiment of the application, an embodiment of a space target collision detection method is provided. It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0041] Figure 1 is a flowchart of an optional space target collision detection method according to the embodiment of the application, as shown in Figure 1 The method comprises the following steps:

[0042] In step S101, orbit ephemeris data of space targets is acquired, wherein the orbit ephemeris data at least comprises position information and velocity information of each space target corresponding to each preset time interval point in a preset time period, and a time interval step between adjacent two preset time interval points is an initial step.

[0043] In the embodiment of the application, the orbit ephemeris data of the space targets can be acquired first, and the orbit ephemeris data comprises position information and velocity information of each space target corresponding to each preset time interval point (such as a time interval point with a time step of 1 minute) in a preset time period (such as the next 3 days).

[0044] Here, the space object can be a spacecraft, a spacecraft fragment, a rocket fragment, etc.

[0045] In the embodiments of the present application, TLE (Two-Line Elements) data of all relevant space objects can be obtained from a space monitoring network or a database. TLE is a standard format containing six parameters (semi-major axis, eccentricity, inclination, longitude of ascending node, argument of perigee, and mean anomaly) describing the orbital state of a satellite, as well as the number and other identification information of the satellite. The orbital position and velocity information of the space object predicted by the SGP4 / SDP4 model using the TLE data within a preset time period (such as the next 3 days). The time step (initial step) corresponding to the preset time interval point is set for generating the orbital ephemeris, for example, every minute is a time point, which is used for subsequent collision warning calculation.

[0046] In step S102, the orbital ephemeris data is mapped into an orbital parameter tensor, and based on the orbital parameter tensor, the relative distance tensor of each pair of space objects is determined, wherein the pair of space objects is composed of any two space objects, and the relative distance tensor records the relative distance of the two space objects indicated by the pair of space objects at each preset time interval point.

[0047] In the embodiments of the present application, the orbital ephemeris data is converted into a three-dimensional tensor structure, and the orbital parameter tensor can be represented as X ∈ R^(N×M×3), wherein X represents the orbital parameter tensor, R is a tensor representation, the first dimension represents the number of space objects (N), the second dimension represents the number of time nodes (M), and the third dimension is used to store three-dimensional position coordinates (x, y, z). Then, the orbital parameter tensor is expanded and element-wise subtraction operation is performed by using the tensor operation capability of the GPU to generate a relative position tensor, and the relative distance tensor D is calculated from the relative position tensor. The relative distance tensor D can be represented as D ∈ R^(N×M×1), which is used to record the relative distance of each pair of targets (pair of space objects) at each time node.

[0048] Here, the number of time nodes (M) is the number of all preset time interval points within the preset time period. The orbital parameter tensor is obtained by tensorizing the position and velocity information of each target at each time point, so as to utilize the GPU for parallel calculation. The relative distance tensor D records the distance of any two space objects at each time point, which is used for subsequent collision risk analysis.

[0049] In some optional embodiments, the tensor is migrated to the GPU memory through a programming interface provided by a tensor calculation function library such as PyTorch (a machine learning library) and TensorFlow (a software library).

[0050] In step S103, the relative distance tensor of each group of space object pairs is interpolated based on the preset Chebyshev basis function matrix to obtain an expected relative distance tensor of each group of space object pairs, where the expected relative distance tensor records the relative distance of the two space objects indicated by the space object pairs at each expected time interval point, and the time interval step between adjacent two expected time interval points is an expected step.

[0051] In the embodiment of the present application, a Chebyshev polynomial interpolation operator can be designed, which is based on the Chebyshev basis function matrix (the preset Chebyshev basis function matrix) and performs parallel interpolation on the GPU by the least square fitting method to generate the expected relative distance tensor with millisecond-level time resolution, which records the relative distance of the target pairs at a more intensive time interval, i.e., the relative distance of the two space objects at each expected time interval point, where the time interval step between adjacent two expected time interval points is an expected step, such as 1 ms.

[0052] Here, the Chebyshev basis function matrix is a set of a series of predefined Chebyshev polynomial functions, which is used for high-precision interpolation calculation in the candidate extreme value time interval. The expected relative distance tensor is obtained after interpolation and contains the relative distance information of the target pairs at each expected time interval point (expected step).

[0053] In step S104, each expected relative distance tensor is iteratively searched to determine the target extreme value time point between each group of space object pairs, and based on the target extreme value time point, the collision state between the space object pairs is determined.

[0054] In the embodiment of the present application, a multi-resolution extreme value search algorithm is adopted, which first traverses the expected relative distance tensor with a relatively coarse time granularity (such as 1 minute) to mark possible extreme value candidate intervals. Subsequently, these candidate intervals are iteratively refined multiple times (such as 10 seconds, 1 second, 1 ms) until the accurate time of each extreme value point is determined, and whether the distance at this time point is lower than the set warning threshold is evaluated to determine whether there is a collision risk.

[0055] Here, the target extreme value time point is determined by the multi-resolution search algorithm, i.e., the time point corresponding to the minimum relative distance found on the expected relative distance tensor. The collision state refers to whether the space object pairs are in a potential collision danger based on the relative distance of the target extreme value time point.

[0056] Figure 2 is a schematic diagram of an optional space object collision detection system architecture according to an embodiment of the present application, as shown in Figure 2As shown, it comprises: an input module for inputting track data; a data preprocessing module for preprocessing the input data; a GPU-Tensor calculation chain for realizing end-to-end GPU parallelization from data input to threat output, breaking through the CPU-GPU data transmission bottleneck; a block dynamic tensor calculation strategy: combining video memory multiplexing and pipeline loading, compressing video memory demand, supporting large-scale target full-volume calculation; a multi-resolution extreme value search module, which reduces the false negative rate through a three-iteration convergence mechanism; and an output module for threat target output.

[0057] In summary, the device end parallel storage can be realized by tensorizing the track data, the efficient parallel interpolation can be realized by using the least square fitting method of Chebyshev basis function matrix, the data processing delay can be effectively reduced by combining the block dynamic tensor calculation strategy and the multi-resolution extreme value search algorithm, the real-time and accuracy of the early warning are ensured, the purpose of real-time processing of large-scale space target collision warning is achieved, the technical effects of improving the calculation efficiency and accuracy are achieved, and the technical problem of low accuracy of space target collision detection in related technologies is solved.

[0058] In order to improve the accuracy of obtaining the track ephemeris data of the space target, in the space target collision detection method provided by the first embodiment of the present application, two-line element data of the space target is collected, wherein the two-line element data is preset fixed format data for describing the track state of the space target; the two-line element data is analyzed to obtain track six-element data, wherein the track six-element data at least includes: semi-major axis, eccentricity, inclination, longitude of ascending node, argument of perigee, and mean anomaly of the space target; and based on the track six-element data, track ephemeris data of the space target in a preset time period is generated.

[0059] In the embodiment of the present application, the two-line element data of the space target, i.e. TLE data, can be collected first. The TLE data is a standard preset fixed format data composed of two lines of text, which is used to describe the track state of the space target (such as a spacecraft, a space debris, etc.). Each line of data contains specific orbital parameters and identifiers. The first line usually contains the number of the target, the launch date, the longitude of the ascending node of the orbit plane, etc., while the second line provides key parameters describing the shape and position of the orbit, such as semi-major axis, eccentricity, inclination, etc.

[0060] In the embodiment of the present application, in order to convert the TLE data into directly processable orbit ephemeris data, the TLE data can be parsed to extract orbit six-element data. The orbit six-element data is six core parameters describing the orbit state, including: semi-major axis (the semi-major axis length of the orbit ellipse, usually in kilometers, reflecting the size of the orbit); eccentricity (the eccentricity of the orbit ellipse, dimensionless, the value is between 0 and 1, 0 represents a circular orbit, and 1 represents a parabolic orbit); inclination (the angle between the orbit plane and the equatorial plane, in degrees, describing the inclination of the orbit relative to the Earth's equator); longitude of ascending node (the Earth's equatorial longitude from the north point to the ascending node of the orbit, in degrees, used to determine the position of the orbit); argument of periapsis (the orbit plane angle from the ascending node to the periapsis, in degrees, reflecting the change of the satellite position in the orbit plane); mean anomaly (the angular distance from the ascending node to the current position of the satellite on the orbit, in degrees, used to determine the instantaneous position of the satellite).

[0061] In the embodiment of the present application, the orbit six-element data can be processed using the SGP4 / SDP4 model to generate the orbit ephemeris data of the space target in a preset time period (such as the next 3 days). The orbit ephemeris data contains the position information and velocity information of each space target at a preset time interval (such as 1 minute). By inputting the orbit six-element data into the SGP4 / SDP4 model, the orbit position and velocity of the space target changing with time can be calculated, and a complete orbit ephemeris data set is generated. This data set is the premise of subsequent collision warning analysis, and can provide accurate position description of the on-orbit target in time sequence.

[0062] Figure 3 is a schematic diagram of an optional data preprocessing module flow according to the embodiment of the present application, as shown in Figure 3 The TLE data can be parsed to obtain the orbit ephemeris data, and then the orbit ephemeris data is mapped to a three-dimensional tensor structure, the orbit parameters are tensorized, and then the tensorized data is stored in the GPU memory.

[0063] In the embodiment, the conversion of the orbit state of the space target from the TLE data to the orbit ephemeris data is realized. Through accurate parsing and prediction of the orbit of the space target, the accuracy and reliability of subsequent calculation are ensured.

[0064] To improve the accuracy of determining the relative distance tensor, in the space target collision detection method provided in Embodiment One of the present application, a block granularity parameter is determined based on the number of targets of all space targets and the memory space of a GPU, wherein the memory space is the memory space of the GPU, the GPU is used to store the orbit parameter tensor and calculate the relative distance tensor; based on the block granularity parameter, all pairs of space targets are decomposed into a plurality of blocks, wherein the number of pairs of space targets included in each block is determined based on the number of targets and the block granularity parameter; for each block, the relative distance tensor of each group of pairs of space targets in the block is calculated based on the orbit parameter tensor.

[0065] In the embodiment of the present application, a reasonable block granularity parameter K can be first determined according to the number of targets N of all space targets and the memory space limit of the GPU, so as to decompose the N×N target pairs into K×K sub-blocks, and the size of a single block is (N / K)×(N / K). The setting of this parameter needs to ensure that the amount of calculation data of each block does not exceed the processing capability of the GPU memory. For example, assuming that the GPU has a 48 GB memory, in order to ensure the calculation efficiency and data throughput, based on preliminary testing and resource planning, the block granularity parameter K can be set so that the number of pairs of space targets contained in each block reaches an optimal balance point, and the amount of data of a single block does not exceed 2 GB, so that the calculation capability of the GPU can be fully utilized, and memory overflow can be avoided.

[0066] In the embodiment of the present application, according to the determined block granularity parameter K, all pairs of space targets are decomposed into K×K blocks, and (N / K)×(N / K) pairs of space targets are contained in each block. This decomposition process is dynamic and depends on the number of targets and the actual available memory of the GPU. After decomposition, the number of pairs of space targets in each block is moderate, which neither occupies too much memory nor ensures sufficient parallel computing load.

[0067] Then, the relative distance tensor of each group of pairs of space targets in each block can be loaded and calculated in a pipelined manner, that is, the data loading, calculation and result return processes are overlapped: Stream 1: load the i-th block data to the memory; Stream 2: perform the i-th block distance calculation; Stream 3: return the i-1-th block result to the host memory. Specifically, for each decomposed block, the orbit parameter tensor is expanded and an element-by-element subtraction operation is performed by using the tensor operation capability of the GPU to generate a relative position tensor, and then the relative distance tensor D is further calculated. When these calculations are performed on the GPU, the block data is stored and reused through the pre-allocated memory pool, which avoids frequent memory application and release operations, and improves the calculation speed and memory utilization.

[0068] Figure 4 is a schematic diagram of an optional block dynamic matrix strategy according to the embodiment of the present application, as Figure 4As shown, the matrix can be divided into 200x200 blocks, and then each block with memory <=2GB is processed by the dynamic batch processing engine to optimize the calculation / transmission overlap of the relative distance of each group of target pairs in each block.

[0069] Figure 5 is a schematic diagram of an optional GPU-Tensor calculation chain according to an embodiment of the application, as Figure 5 As shown, tensor broadcasting calculation can be performed first to generate a relative distance tensor, and then Chebyshev interpolation is parallelized.

[0070] In this embodiment, by determining the block granularity parameter and performing the block dynamic tensor calculation strategy, the problems of memory management and parallel calculation in large-scale space target collision warning are effectively solved. This strategy not only reduces the occupancy of GPU memory, but also improves the calculation efficiency through pipelining data loading and calculation, laying a foundation for realizing a minute-level warning period.

[0071] In order to accurately generate the Chebyshev basis function matrix, in the space target collision detection method provided in Embodiment One of the application, before interpolating the relative distance tensor of each group of space target pairs based on the preset Chebyshev basis function matrix, the method further includes: performing standardization processing on the time stamp indicated by each preset time interval point in the relative distance tensor to obtain a standardized relative distance tensor; and generating the preset Chebyshev basis function matrix based on the number of parallel processing of the GPU for processing the space target pairs.

[0072] In this embodiment of the application, after the relative distance data of each group of space target pairs is calculated, the time stamp corresponding to each preset time interval point in the relative distance tensor D is standardized, that is, it is mapped to the interval [-1, 1]. This process can be realized by linear transformation. Considering the start time and end time of the warning period, the standardized value of the time stamp is calculated. The standardized relative distance tensor D retains the time sequence information of the original data, and is converted into a format suitable for interpolation calculation.

[0073] In this embodiment of the application, the preset Chebyshev basis function matrix B is generated based on the number of parallel processing of the GPU for processing the space target pairs (i.e. BatchSize), and the preset Chebyshev basis function matrix B can be an 11-order Chebyshev basis function matrix, which is represented as B∈R^(Batch×11×11). This matrix contains a set of orthogonal Chebyshev polynomials, which are used for subsequent least squares fitting interpolation calculation.

[0074] In this embodiment, the standardization process ensures the effectiveness and stability of the interpolation algorithm, and the generation of the Chebyshev basis function matrix is the key to implementing batch interpolation calculation. In this way, not only does it solve the efficiency problem caused by the traditional bisection iteration, but it also overcomes the limitations of cubic spline interpolation in large-scale target pair processing, providing a high-precision and efficient calculation solution for real-time collision warning. In addition, through parallel processing and batch interpolation, the computational complexity is also reduced, so that even when processing tens of thousands of target pairs, the calculation can still maintain real-time and accuracy, providing strong technical support for the safe operation of spacecraft in a dense threat environment.

[0075] To improve the accuracy of determining the expected relative distance tensor, in the space target collision detection method provided in Embodiment One of the present application, based on the observation position data of each space target in the space target pair and the preset Chebyshev basis function matrix, the least squares fitting is used to determine the coefficient matrix of the space target pair; based on the expected step length and the coefficient matrix, the relative distance of the space target pair at each expected time interval point is determined; and between each two adjacent preset time interval points in the relative distance tensor, interpolation is performed with the expected step length as the interval to obtain the expected relative distance tensor.

[0076] In the embodiment of the present application, the observation position data of each space target in the space target pair can be used as the basis, combined with the preset Chebyshev basis function matrix, to determine the corresponding coefficient matrix C of each group of space target pairs by using the GPU parallel computing capability to perform the least squares fitting algorithm. The least squares fitting is a statistical method used to find a set of parameters that minimize the sum of squared differences between predicted and observed values. In this scenario, the coefficient matrix C reflects the relationship between the Chebyshev basis function and the actual observation position data, which is used for subsequent high-precision interpolation calculation.

[0077] Exemplarily, a linear equation system B·C=X can be constructed, where C is the coefficient matrix and X is the observation position data. The coefficient matrix C is solved by QR decomposition (i.e. a matrix decomposition method).

[0078] Here, the observation position data refers to the accurate position information of the space target obtained through the orbital ephemeris data at the preset time interval points. The preset Chebyshev basis function matrix is a set of Chebyshev polynomial functions standardized by time, which is used for interpolation calculation. The coefficient matrix C is a set of parameters obtained by fitting the observation position data and the basis function matrix, which is used for interpolation calculation to predict the relative position of the target at any time point.

[0079] After obtaining the coefficient matrix C, the expected step length (usually much smaller than the step length of the preset time interval, for example, 1 ms) is used as the time sampling interval, and the relative distances of the space target pairs at each expected time interval point are calculated according to the coefficient matrix. This process is executed in parallel on the GPU, accelerating the interpolation calculation time.

[0080] Here, the expected step length refers to the time step length used in the interpolation calculation to generate the expected relative distance tensor. Compared with the step length of the preset time interval point, the expected step length is more refined and can capture the small details of the relative distance changes.

[0081] Then, the relative distance values at a series of new time points are obtained by using the expected step length to interpolate the adjacent time periods between each preset time interval point. These relative distance values are reorganized into a matrix according to the time sequence, thereby forming the expected relative distance tensor. This matrix contains the relative distance data between all space target pairs at a more refined time scale, providing detailed data support for subsequent extreme value search and collision state evaluation.

[0082] In this embodiment, the coefficient matrix is determined by least squares fitting using the Chebyshev basis function matrix, and the target relative distance is calculated using the coefficient matrix, and finally the expected relative distance tensor is generated, effectively solving the problems of insufficient interpolation accuracy and low computational efficiency in large-scale space target real-time collision warning. Through the combination of GPU parallel computing capability and Chebyshev polynomial interpolation method, not only the speed of interpolation calculation is improved, but also high-precision parallel interpolation with millisecond-level time resolution is realized.

[0083] In order to accurately interpolate at the expected step length intervals, in the space target collision detection method provided in Embodiment One of the present application, an expected time interval point is generated every expected step length interval, and the relative distance corresponding to the expected time interval point is inserted at the expected time interval point.

[0084] In the present embodiment, a more refined expected step length (for example, 1 ms) is introduced on the basis of the preset time interval points (such as 1-minute intervals) to generate the expected time interval points. That is, in each 1-minute preset time interval, 60,000 expected time interval points are additionally generated, greatly increasing the time resolution. The generation of each expected time interval point is based on an expected step length increment, ensuring the continuity of the time sequence. After obtaining the expected time interval points, interpolation calculation is performed on each expected time interval point to obtain the corresponding relative distance values. These relative distance values are then inserted into the corresponding time positions in the expected relative distance tensor, filling the gaps between the preset time interval points.

[0085] In this embodiment, by means of Chebyshev polynomial interpolation technology, parallel computing is carried out on the GPU, and the desired relative distance tensor can be generated at a higher time resolution, that is, a desired time interval point is generated every interval of a desired step length, and the corresponding relative distance value is accurately calculated and inserted at these time points. This not only ensures the high accuracy of the collision warning analysis, but also improves the real-time response capability of the warning, because more intensive time series data can timely reveal the dynamic changes between pairs of space targets, and early warning of possible close approach or collision events can be made, thereby providing sufficient time for the operator of the spacecraft to take necessary avoidance measures.

[0086] In order to improve the accuracy of determining the target extreme value time point, in the space target collision detection method provided in Embodiment One of the present application, the desired relative distance tensor is traversed with an initial step length to determine a first candidate interval, wherein the first candidate interval has a first time point and a second time point determined, the relative distance at the first time point is less than a preset relative distance, and the relative distance at the second time point is greater than the preset relative distance; for each first candidate interval, the first candidate interval is modified with a first step length to obtain a second candidate interval, wherein the first step length is less than the initial step length and greater than the desired step length, and the length of the second candidate interval is less than the length of the first candidate interval; for each second candidate interval, the second candidate interval is modified with a second step length to obtain an initial extreme value time point, wherein the second step length is less than the first step length and greater than the desired step length; based on the initial extreme value time point, a third candidate interval is determined, and for each third candidate interval, the third candidate interval is modified with the desired step length to obtain a preset extreme value time point; the neighborhood of the preset extreme value time point is shielded, a new preset extreme value time point is searched again in the desired relative distance tensor, until the iteration number reaches a preset number, a preset number of preset extreme value time points are obtained, wherein the preset number is equal to the preset number; the preset extreme value time point corresponding to the minimum relative distance is determined as the target extreme value time point.

[0087] In the embodiment of the present application, coarse-grained screening can be performed first, and a larger initial step length (such as 1 minute) is used to traverse the desired relative distance tensor D to find the first time point t1 and the second time point t2, wherein the relative distance of t1 is less than the preset relative distance threshold D_threshold (i.e., the preset relative distance, a safety distance threshold preset in advance), and the relative distance of t2 is greater than the threshold. This pair of time points defines the first candidate interval, indicating that there may be a minimum value in this interval, i.e., a potential collision risk.

[0088] Then, multiple iterations of refinement are performed, specifically, for each first candidate interval, the step size is adjusted to a smaller first step size (e.g. 10 seconds), and the expected relative distance tensor is re-traversed within the interval to find a more accurate candidate location of the extremum. This process further narrows down the first candidate interval to a second candidate interval, improving the accuracy of the extremum search. A smaller second step size (e.g. 1 second) is then used to search again within the second candidate interval until the turning point where the relative distance change trend changes from decreasing to increasing, i.e. the initial extremum time point t_initial, is found. This time point is the starting point for further pinpointing the extremum. After that, a shorter third candidate interval is determined around the initial extremum time point t_initial, for example, set to a range of ±0.5 seconds. Within this interval, interpolation is performed with a desired step size (e.g. 1 ms) to accurately find the minimum point of the relative distance, i.e. the preset extremum time point t_preset.

[0089] In embodiments of the present application, to eliminate the risk of repeatedly detecting the same extremum point, the relative distance values within a neighborhood (e.g. ±5 minutes) of the preset extremum time point t_preset are masked, i.e. increased to be beyond D_threshold (e.g. increased by 1000 km offset), ensuring that this point will not be marked as an extremum candidate again. Then, the search algorithm is re-executed within the unmasked interval until the search number reaches a preset number (e.g. 3 times), thereby obtaining a preset number of preset extremum time points.

[0090] After multiple searches and refinements are completed, all extremum points are sorted by relative distance, and the preset extremum time point with the smallest relative distance is selected as the target extremum time point t_target for the final collision warning output.

[0091] Figure 6 is a schematic diagram of an optional multi-resolution extremum search process according to embodiments of the present application, as shown in Figure 6 A coarse-grained screening can be performed first, such as traversing the distance matrix D with a step size of 1 minute to mark candidate intervals. Then a first iteration is performed, in which interpolation is performed within the candidate intervals with a resolution of 10 seconds to narrow down the extremum range, and a second iteration is performed: interpolation is performed within the narrowed extremum range with a resolution of 1 second to locate the initial extremum point, and a third iteration is performed after that: interpolation is performed within the interval determined by the initial extremum point with a resolution of 1 ms to determine the final extremum time.

[0092] In the embodiment, through the way of gradually narrowing the time range and improving the time resolution in stages, efficient and accurate extreme value search in real-time collision warning of large-scale space targets is realized. This process not only improves the accuracy of extreme value positioning to the level of hundreds of meters, but also effectively avoids the risk of repeated calculation and missed detection through the shielding mechanism, providing strong protection for the safe operation of spacecraft. With the help of GPU parallel computing capability, tens of thousands or even hundreds of thousands of target pair data can be processed in a very short time, realizing real-time warning response in minutes and meeting the urgent safety needs of space stations, satellite constellations and other spacecraft.

[0093] In the embodiment of the application, through the GPU-Tensor heterogeneous computing architecture, Chebyshev interpolation parallelization, block dynamic tensor calculation strategy and multi-resolution extreme value search algorithm, the problems of low computing efficiency, high memory occupation and insufficient interpolation accuracy are solved, ensuring the real-time and accuracy of the warning, achieving the purpose of real-time processing of large-scale space target collision warning, and realizing the technical effects of improving the computing efficiency and accuracy.

[0094] The following will be described in detail in combination with another embodiment.

[0095] Embodiment two

[0096] The collision detection device for space targets provided in the embodiment comprises a plurality of implementation units, each of which corresponds to each implementation step in the above embodiment one.

[0097] Figure 7 is a schematic diagram of an optional collision detection device for space targets according to the embodiment of the application, as Figure 7 shown, the collision detection device can include an acquisition unit 70, a determination unit 71, an interpolation unit 72, and a search unit 73.

[0098] The acquisition unit 70 is configured to acquire orbit ephemeris data of space targets, wherein the orbit ephemeris data at least includes position information and velocity information of each space target at each preset time interval point within a preset time period, and a time interval step between adjacent two preset time interval points is an initial step.

[0099] The determination unit 71 is configured to map the orbit ephemeris data into an orbit parameter tensor, and determine a relative distance tensor of each group of space target pairs based on the orbit parameter tensor, wherein the space target pair is composed of any two space targets, and the relative distance tensor records relative distances of the two space targets indicated by the space target pair at each preset time interval point.

[0100] An interpolation unit 72 is configured to interpolate, based on a preset Chebyshev basis function matrix, the relative distance tensor of each group of space object pairs to obtain an expected relative distance tensor of each group of space object pairs, wherein the expected relative distance tensor records the relative distance of two space objects indicated by the space object pair at each expected time interval point, and the time interval step between two adjacent expected time interval points is an expected step.

[0101] A searching unit 73 is configured to iteratively search each expected relative distance tensor to determine a target extreme value time point between each group of space object pairs, and determine a collision state between the space object pairs based on the target extreme value time point.

[0102] The collision detection device can realize parallel storage of the device end through the tensorization of the orbit data, perform efficient parallel interpolation by using the least square fitting method of the Chebyshev basis function matrix, and effectively reduce the data processing delay by combining the block dynamic tensor calculation strategy and the multi-resolution extreme value search algorithm, thereby ensuring the real-time performance and accuracy of the early warning and achieving the purpose of real-time processing of large-scale space object collision early warning, thereby achieving the technical effects of improving the calculation efficiency and accuracy, and further solving the technical problem of low accuracy of collision detection of space objects in the related art.

[0103] Optionally, the acquisition unit comprises: a first acquisition module configured to acquire two-line element data of the space object, wherein the two-line element data is preset fixed format data used to describe the orbit state of the space object; a first analysis module configured to analyze the two-line element data to obtain orbit six-element data, wherein the orbit six-element data at least comprises a semi-major axis, an eccentricity, an inclination, an ascending node right ascension, a perigee amplitude angle and a mean anomaly of the space object; and a first generation module configured to generate orbit ephemeris data of the space object in a preset time period based on the orbit six-element data.

[0104] Optionally, the determination unit comprises: a first determination module configured to determine a block granularity parameter based on the target quantity of all space objects and a memory space, wherein the memory space is a memory space of a GPU, and the GPU is used to store the orbit parameter tensor and calculate the relative distance tensor; a first decomposition module configured to decompose all space object pairs into a plurality of blocks based on the block granularity parameter, wherein the number of space object pairs included in each block is determined based on the target quantity and the block granularity parameter; and a first calculation module configured to calculate, for each block, the relative distance tensor of each group of space object pairs in the block based on the orbit parameter tensor.

[0105] Optionally, the collision detection device further comprises: a first processing module, configured to perform standardization processing on a time stamp indicated by each preset time interval point in the relative distance tensor to obtain a standardized relative distance tensor, before interpolating the relative distance tensor of each group of space target pairs based on the preset Chebyshev basis function matrix; and a second generation module, configured to generate the preset Chebyshev basis function matrix based on a parallel processing number of the GPU for processing the space target pairs.

[0106] Optionally, the interpolation unit comprises: a second determination module, configured to determine a coefficient matrix of the space target pair based on observation position data of each space target in the space target pair and the preset Chebyshev basis function matrix; a third determination module, configured to determine a relative distance of the space target pair at each expected time interval point based on the expected step length and the coefficient matrix; and a first interpolation module, configured to interpolate the relative distance at the expected step length between every two adjacent preset time interval points in the relative distance tensor to obtain an expected relative distance tensor.

[0107] Optionally, the first interpolation module comprises: a first generation submodule, configured to generate an expected time interval point every expected step length; and a first insertion submodule, configured to insert the relative distance corresponding to the expected time interval point at the expected time interval point.

[0108] Optionally, the searching unit comprises: a fourth determination module, configured to traverse the expected relative distance tensor at an initial step length to determine a first candidate interval, wherein the first candidate interval is determined by a first time point and a second time point, the relative distance at the first time point is less than the preset relative distance, and the relative distance at the second time point is greater than the preset relative distance; a first correction module, configured to correct each first candidate interval at a first step length to obtain a second candidate interval, wherein the first step length is less than the initial step length and greater than the expected step length, and the length of the second candidate interval is less than the length of the first candidate interval; a second correction module, configured to correct each second candidate interval at a second step length to obtain an initial extreme time point, wherein the second step length is less than the first step length and greater than the expected step length; a third correction module, configured to determine a third candidate interval based on the initial extreme time point, and correct each third candidate interval at the expected step length to obtain a preset extreme time point; a first searching module, configured to shield a neighborhood of the preset extreme time point, re-iteratively search a new preset extreme time point in the expected relative distance tensor until a preset number of iterations is reached to obtain a preset number of preset extreme time points, wherein the preset number is equal to the preset number of iterations; and a determination module, configured to determine a target extreme time point corresponding to a minimum relative distance.

[0109] The collision detection device can further include a processor and a memory, and the acquisition unit 70, the determination unit 71, the interpolation unit 72, the searching unit 73, and the like are stored in the memory as program units, and the corresponding functions are realized by the processor executing the program units stored in the memory.

[0110] The processor includes a core, and the core calls the corresponding program units in the memory. The core can be one or more, and the relative distance tensor of each group of space target pairs is iteratively searched by adjusting the core parameters, the target extreme value time point between each group of space target pairs is determined, and the collision state between the space target pairs is determined based on the target extreme value time point.

[0111] The memory can include a non-permanent memory in a computer readable medium, a random access memory (RAM), and / or a non-volatile memory such as a read-only memory (ROM) or a flash memory (flash RAM), and the memory includes at least one memory chip.

[0112] The application further provides a computer program product adapted to execute the program of the following method steps when executed on a data processing device: acquiring orbit ephemeris data of space targets, mapping the orbit ephemeris data into an orbit parameter tensor, determining a relative distance tensor of each group of space target pairs based on the orbit parameter tensor, interpolating the relative distance tensor of each group of space target pairs based on a preset Chebyshev basis function matrix to obtain an expected relative distance tensor of each group of space target pairs, iteratively searching each expected relative distance tensor to determine a target extreme value time point between each group of space target pairs, and determining a collision state between the space target pairs based on the target extreme value time point.

[0113] According to another aspect of the embodiments of the application, a computer program product is also provided, including a non-volatile computer readable storage medium, the non-volatile computer readable storage medium storing a computer program, and the computer program is executed by a processor to implement the collision detection method of space targets according to any one of the above.

[0114] According to another aspect of the embodiments of the application, an electronic device is also provided, including one or more processors and a memory, and the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the collision detection method of space targets.

[0115] Figure 8 is a hardware structure block diagram of an electronic device (or mobile device) for a collision detection method of space targets according to an embodiment of the application. As shown in Figure 8 the electronic device can include one or more processors (for example, Figure 8The processor 802a, the processor 802b,..., the processor 802n, etc. in the figure can include, but are not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA, a memory 804 for storing data. In addition, it can also include a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which can be included as one of the ports of the I / O interface), a network interface, a keyboard, a power supply and / or a camera. Those skilled in the art can understand that Figure 8 The structure shown in the figure is only schematic and does not limit the structure of the above-mentioned electronic device. For example, the electronic device can also include more or fewer components than those shown in the figure, or have a different configuration from that shown in the figure. Figure 8 The structure shown in the figure is only schematic and does not limit the structure of the above-mentioned electronic device. For example, the electronic device can also include more or fewer components than those shown in the figure, or have a different configuration from that shown in the figure. Figure 8 The structure shown in the figure is only schematic and does not limit the structure of the above-mentioned electronic device. For example, the electronic device can also include more or fewer components than those shown in the figure, or have a different configuration from that shown in the figure.

[0116] The above-mentioned embodiment numbers of the present application are only for description, not representing the advantages and disadvantages of the embodiments.

[0117] The embodiments or examples of the present disclosure are not exhaustive, but only a part of the embodiments or examples, and are not specific limitations on the protection scope of the present disclosure. In the case of no contradiction, each step in a certain embodiment or example can be implemented as an independent embodiment, and the steps can be combined arbitrarily, for example, the scheme after removing part of the steps in a certain embodiment or example can also be implemented as an independent embodiment, and the order of the steps in a certain embodiment or example can be exchanged arbitrarily, in addition, the optional ways or optional examples in a certain embodiment or example can be combined arbitrarily; in addition, the embodiments or examples can be combined arbitrarily, for example, part or all steps of different embodiments or examples can be combined arbitrarily, a certain embodiment or example can be combined with the optional ways or optional examples of other embodiments or examples.

[0118] In the above-mentioned embodiments of the present application, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0119] In the several embodiments of the present application, it should be understood that the disclosed technology can be implemented by other means. Among them, the above-mentioned device embodiment is only schematic, for example, the division of the unit can be a logical function division, and actual implementation can have another division way, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point, the coupling or direct coupling or communication connection between the displays or discussions can be through some interfaces, indirect coupling or communication connection between units or modules, which can be electrical or other forms.

[0120] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0121] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0122] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical scheme of the present application or the part of the present application which contributes to the prior art or the whole or part of the technical scheme can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes various media that can store program codes, such as a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0123] The above is only the preferred embodiment of the present application. It should be noted that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A method of collision detection for a space object, characterized by, The method comprises the following steps: acquiring orbit ephemeris data of space targets, wherein the orbit ephemeris data at least comprises position information and velocity information of each space target at each preset time interval point within a preset time period, and a time interval step between adjacent two preset time interval points is an initial step; mapping the orbit ephemeris data into an orbit parameter tensor, and determining a relative distance tensor of each space target pair based on the orbit parameter tensor, wherein the space target pair is composed of any two space targets, and the relative distance tensor records relative distances of the two space targets indicated by the space target pair at each preset time interval point; interpolating the relative distance tensor of each space target pair based on a preset Chebyshev basis function matrix to obtain an expected relative distance tensor of each space target pair, wherein the expected relative distance tensor records relative distances of the two space targets indicated by the space target pair at each expected time interval point, and a time interval step between adjacent two expected time interval points is an expected step; performing iterative search on each expected relative distance tensor to determine a target extreme value time point between each space target pair, and determining a collision state between the space target pair based on the target extreme value time point.

2. The collision detection method of claim 1, wherein, The step of acquiring orbit ephemeris data of space targets comprises: collecting two-line element data of the space targets, wherein the two-line element data is preset fixed format data for describing an orbit state of the space targets; parsing the two-line element data to obtain orbit six-element data, wherein the orbit six-element data at least comprises a semi-major axis, an eccentricity, an inclination, an ascending node right ascension, a perigee amplitude, and a mean anomaly of the space target; generating the orbit ephemeris data of the space target within the preset time period based on the orbit six-element data.

3. The collision detection method of claim 1, wherein, The step of determining a relative distance tensor of each space target pair based on the orbit parameter tensor comprises: determining a block granularity parameter based on a target number of all the space targets and a memory space, wherein the memory space is a memory space of a GPU, the GPU is used to store the orbit parameter tensor and calculate the relative distance tensor; decomposing all the space target pairs into a plurality of blocks based on the block granularity parameter, wherein a number of the space target pairs included in each block is determined based on the target number and the block granularity parameter; for each block, calculating the relative distance tensor of each space target pair in the block based on the orbit parameter tensor.

4. The collision detection method of claim 1, wherein, Before interpolating the relative distance tensor of each space target pair based on a preset Chebyshev basis function matrix, the method further comprises: performing standardization processing on a timestamp indicated by each preset time interval point in the relative distance tensor to obtain a standardized relative distance tensor; generating the preset Chebyshev basis function matrix based on a parallel processing number of the GPU for processing the space target pairs.

5. The collision detection method of claim 1, wherein, The step of interpolating the relative distance tensor of each group of the space target pair based on the preset Chebyshev basis function matrix to obtain an expected relative distance tensor of each group of the space target pair comprises: Based on the observation position data of each space target in the space target pair and the preset Chebyshev basis function matrix, a coefficient matrix of the space target pair is determined by least square fitting; Based on the expected step length and the coefficient matrix, the relative distance of the space target pair at each expected time interval point is determined; The step of interpolating the relative distance tensor of each group of the space target pair based on the preset Chebyshev basis function matrix to obtain an expected relative distance tensor of each group of the space target pair comprises:

6. The collision detection method of claim 5, wherein, The step of interpolating the relative distance tensor of each group of the space target pair based on the preset Chebyshev basis function matrix to obtain an expected relative distance tensor of each group of the space target pair comprises: An expected time interval point is generated every expected step length; The relative distance corresponding to the expected time interval point is inserted at the expected time interval point.

7. The collision detection method of claim 1, wherein, The step of performing iterative search on each expected relative distance tensor to determine the target extreme time point between each group of the space target pair comprises: The expected relative distance tensor is traversed by the initial step length to determine a first candidate interval, wherein the first candidate interval is determined by a first time point and a second time point, the relative distance at the first time point is less than a preset relative distance, and the relative distance at the second time point is greater than the preset relative distance; For each first candidate interval, the first candidate interval is modified by a first step length to obtain a second candidate interval, wherein the first step length is less than the initial step length and greater than the expected step length, and the length of the second candidate interval is less than the length of the first candidate interval; For each second candidate interval, the second candidate interval is modified by a second step length to obtain an initial extreme time point, wherein the second step length is less than the first step length and greater than the expected step length; Based on the initial extreme time point, a third candidate interval is determined, and for each third candidate interval, the third candidate interval is modified by the expected step length to obtain a preset extreme time point; The neighborhood of the preset extreme time point is shielded, and a new preset extreme time point is re-iteratively searched in the expected relative distance tensor until the iteration number reaches a preset number, to obtain a preset number of preset extreme time points, wherein the preset number is equal to the preset number; The preset extreme time point corresponding to the minimum relative distance is determined as the target extreme time point.

8. A collision detection apparatus for a space object, characterized by The method comprises: An acquisition unit is configured to acquire orbit ephemeris data of space targets, wherein the orbit ephemeris data at least comprises position information and velocity information of each space target at each preset time interval point in a preset time period, and the time interval step length between adjacent two preset time interval points is an initial step length; The determining unit is configured to map the orbit ephemeris data into an orbit parameter tensor, and determine a relative distance tensor of each pair of space objects based on the orbit parameter tensor, wherein the pair of space objects is composed of any two of the space objects, and the relative distance tensor records relative distances of the two space objects indicated by the pair of space objects at each preset time interval point. The interpolating unit is configured to interpolate the relative distance tensor of each pair of space objects based on a preset Chebyshev basis function matrix to obtain an expected relative distance tensor of each pair of space objects, wherein the expected relative distance tensor records relative distances of the two space objects indicated by the pair of space objects at each expected time interval point, and a time interval step length between two adjacent expected time interval points is an expected step length. The searching unit is configured to perform iterative search on each expected relative distance tensor to determine a target extreme time point between each pair of space objects, and determine a collision state between the pair of space objects based on the target extreme time point.

9. A computer program product, characterised in that, A non-transitory computer readable storage medium storing a computer program, the computer program being executed by a processor to implement the space object collision detection method of any one of claims 1 to 7.

10. An electronic device, comprising: One or more processors and a memory, the memory being configured to store one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the space object collision detection method of any one of claims 1 to 7.

Images