Spacecraft autonomous modeling method for unascertained near-earth asteroid target

Abstract

The invention provides a spacecraft autonomous modeling method for an unascertained near-earth asteroid target, and belongs to the technical field of spaceflight technology and application thereof, and the method comprises the steps: carrying out typical target extraction, BRDF (Bidirectional Reflection Distribution Function) measurement and characteristic measurement, and application scene measurement, and generating a digital target set; carrying out data processing on the spacecraft, wherein the data processing comprises on-board processing and on-board verification; carrying out algorithm ground verification, including algorithm correctness verification, algorithm clipping and acceleration, and algorithm edge deployment; the satellite-ground combined on-orbit application is specifically characterized in that when a spacecraft is close to a target, target modeling is carried out, data is acquired based on a sensor for on-device processing, target modeling is completed, then data is newly acquired by the sensor for on-device verification and model updating, and model reconstruction and ground processing are introduced by limiting the number of verification times. According to the method, the modeling accuracy can be improved, and the efficient autonomous modeling capability of the spacecraft is comprehensively formed.

Description

Technical Field

[0001] This invention belongs to the field of aerospace technology and its application technology, specifically relating to an autonomous modeling method for spacecraft targeting unknown near-Earth asteroids. Background Technology

[0002] Near-Earth asteroids are an important part of the solar system, significantly impacting the evolution of Earth's environment and human survival. The detection and management of asteroids are crucial for mitigating potential asteroid impact risks, ensuring global safety, studying the formation and evolution of the solar system, and providing the technological foundation for asteroid development and utilization. Currently, approximately 30,000 near-Earth asteroids have completed orbital cataloging globally, representing only 1% of their true estimated number. Cataloged asteroids are primarily identified through ground-based observations, acquiring orbital data and inferring their volume and mass using absolute magnitude. The shape, mass, porosity, rotation characteristics, and surface properties of most asteroids remain unknown, hindering the implementation of specific space engineering missions such as asteroid target identification and detection, asteroid landing site selection, and asteroid geomorphological mapping.

[0003] Currently, deep space exploration missions primarily involve a spacecraft first arriving and orbiting the target in a fixed orbit to obtain close-range space-based observation images. Ground-based systems then use these images to invert and model the target's characteristics. Based on the high-altitude target characteristic inversion modeling results, target matching or small-scale adaptive optimization is performed during the impact or landing mission. However, this mission model has high requirements for the specificity of prior information, a high degree of dependence on ground operations, and high demands for ground support. Especially for missions targeting asteroids, the vast distance between the spacecraft and Earth and the significant information transmission delays severely restrict the timeliness of space missions. For unknown asteroid targets, limited data is available before the mission, and most ground-based data processing methods are geared towards ground applications or targets, without addressing the issues of unknown target characteristics or limited characteristic data, resulting in significant differences from the target application scenario. Furthermore, ground-based data processing methods involve large amounts of data from multiple sources and complex processing methods, and without considering the limitations of spacecraft's real-time processing capabilities, they are difficult to directly apply to spacecraft.

[0004] In other words, existing technologies all involve multiple data sources, large data volumes, and ground-based targets, meaning the basic characteristics of the targets are known, and there are no uncertainties regarding these characteristics. Therefore, the need for exploring and experimenting with target characteristics is low. Furthermore, these methods primarily rely on ground-based processing, without needing to consider the limitations of real-time processing capabilities on spacecraft or the specific limitations of asteroid exploration missions. Summary of the Invention

[0005] To address the aforementioned technical problems, improve the timeliness of modeling the characteristics of unknown near-Earth asteroid targets, enhance the autonomous mission capabilities of spacecraft, and reduce the requirements for deep space missions, this invention provides a spacecraft autonomous modeling method for unknown near-Earth asteroid targets.

[0006] Asteroid size estimation is performed using limited prior information such as orbital information and absolute magnitude data. Based on this prior knowledge, information on asteroid shape, material, surface morphology, and structural characteristics is obtained, and typical asteroid targets are extracted. Combining constraints from commonly used optical sensors (e.g., multispectral, hyperspectral, and infrared sensor spectral bands and accuracy) with typical mission scenarios (e.g., distance and illumination constraints during flyby, pass-through, and approach missions), BRDF (Bidirectional Reflectance Distribution Function) measurements are conducted on production test specimens. Characteristic measurements are then performed on scaled-down measurement specimens. A digital target set is generated based on the measurement results and digital simulation results. This addresses the problems of uncertain asteroid target characteristics and limited targeted data.

[0007] Target identification and matching based on a digital target set are performed through feature comparison to generate a preliminary model. Real-time data acquired by sensors on the spacecraft is then used to refine the model, reducing the requirements for onboard data processing. For example, based on an ellipsoidal asteroid model, dimensions can be adjusted, local depressions added, and surface conditions enriched by incorporating real-time data. Ground target matching and model refinement algorithms are validated in conjunction with application scenarios to ensure algorithm correctness. Fully utilizing onboard computing power, an on-orbit verification mechanism is introduced to validate the generated model against newly acquired sensor data, improving modeling accuracy and comprehensively forming a highly efficient and autonomous modeling capability for the spacecraft.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] An autonomous modeling method for spacecraft targeting uncertain near-Earth asteroids includes the following steps:

[0010] Step 1: Generate a digital target set through typical target extraction, BRDF measurement and characteristic measurement, and application scenario measurement; BRDF represents the bidirectional reflectance distribution function.

[0011] Step 2: Perform on-board data processing, including on-board processing and verification.

[0012] Step 3: Conduct ground-based validation of the algorithm, including algorithm correctness verification, algorithm pruning and speed-up, and algorithm edge deployment;

[0013] Step 4: Conduct joint space-ground on-orbit application. When the spacecraft is close to the target, target modeling is performed. First, the data acquired by the sensors is processed on-board to complete the target modeling. Then, the newly acquired data from the sensors is used for on-board verification and model update. By limiting the number of verifications, model reconstruction and ground processing are introduced.

[0014] Furthermore, in step 1, the typical target extraction is based on existing knowledge of asteroid targets, extracting typical characteristics of the targets, including orbital characteristics, physical characteristics, material and structural characteristics, and obtaining typical targets through different combinations of characteristics.

[0015] Furthermore, in the typical target extraction, orbital characteristics are used to obtain the spacecraft mission orbit and mission scenario in combination with mission requirements, and are obtained from a public database.

[0016] Furthermore, in the extraction of typical targets, physical characteristics include size, shape, and surface morphology, wherein size is estimated based on absolute magnitude, typical shapes include contact binary star, flattened round, near-spherical, slender, gyroscope, dumbbell, or tooth-like, and typical surface morphology includes flat, rocky, crater-like, hill, or ridge.

[0017] Furthermore, in the typical target material, the material includes ferromagnesian rock, carbides, nickel iron ore, basalt, granite, or sulfides.

[0018] Furthermore, in the extraction of typical targets, structural characteristics include monoliths and piles of gravel.

[0019] Furthermore, in step 1, the BRDF measurement and characteristic measurement are based on the extracted typical targets. After analysis, incompatible characteristic combinations are eliminated, and the test scheme is refined by introducing load spectral constraints and illumination condition constraints in the mission scenario. The application scenario measurement introduces the relative motion relationship between the target and the spacecraft during the mission process to enrich the data verification model.

[0020] Furthermore, in step 2, the on-board processing performs multi-source data normalization processing, target shape and rotation characteristic calculation, information fusion, target shape and rotation characteristic calculation, intelligent recognition based on digital target set, and target modeling; the on-board verification is based on newly acquired data from sensors on the spacecraft and the established target model. The on-board verification shares the multi-source data normalization processing, information fusion, and target shape and rotation characteristic calculation modules with the aforementioned on-board processing. The on-board verification performs characteristic verification and model verification, and updates the model after passing the verification.

[0021] Furthermore, in step 3, the algorithm ground verification uses the digital target set as the sensor data source to acquire detection data through sensors, and compares the algorithm running results with the selected scaled-down measurement device to ensure correctness; in the test, the algorithm runs on the ground computer and after the algorithm is deployed at the edge, it runs on the spacecraft hardware product.

[0022] Furthermore, in step 4, the specific operation after limiting the number of verifications is as follows: set the maximum number of verifications to N; when the verification fails, the number of verifications is incremented by 1; when it passes, the number of verifications is reset to zero; if the number of verifications exceeds N, output the original sensor data, and the spacecraft performs a close-range flyby or switches to a backup orbit for ground processing; if the number of verifications does not exceed N, rebuild the target model.

[0023] Beneficial effects:

[0024] Compared to general multi-source data fusion, this invention focuses on addressing the problem of insufficient prior information when asteroid targets are unknown. It expands upon ground data based on prior knowledge to establish a ground-based digital target set, adjusting testing and experimental schemes according to specific application scenarios. Considering the limited amount of data available on the spacecraft and the time-varying positions of the spacecraft and asteroid targets, it fully utilizes the limited computing power and available data resources on the spacecraft, proposing a model building method based on a basic model plus corrections. Through extensive ground-based experiments and tests, a digital target set is established. Based on this digital target set, typical target identification and matching combined with real-time data correction can effectively reduce the computing power requirements for on-orbit modeling, enabling edge deployment on the spacecraft and thus forming autonomous mission capabilities.

[0025] This invention enables spacecraft to autonomously model unknown asteroid targets, improves the spacecraft's adaptability to different asteroid target missions, reduces dependence on ground systems, improves the real-time performance of mission response, facilitates the spacecraft to independently complete the detection and identification of unknown targets, achieves autonomous guidance of targets, and autonomously executes deep space missions such as asteroid impact and close-range exploration. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of a spacecraft autonomous modeling method for unknown asteroid targets, as described in an embodiment of the present invention.

[0027] Figure 2 This is a schematic diagram of a data processing method on a spacecraft according to an embodiment of the present invention;

[0028] Figure 3 This is a schematic diagram of the algorithm ground verification method in an embodiment of the present invention;

[0029] Figure 4 This is a schematic diagram of the satellite-ground joint on-orbit application method in an embodiment of the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0031] like Figure 1 As shown, an embodiment of the present invention provides a spacecraft autonomous modeling method for unknown asteroid targets, which includes establishing a digital target set, on-spacecraft data processing, algorithm ground verification, and joint on-orbit application between spacecraft and ground. Specifically, it includes the following steps:

[0032] Step 1: Establish a digital target set. Establishing a digital target set mainly involves extracting typical targets, performing BRDF and characteristic measurements, measuring application scenarios, and generating the digital target set.

[0033] 1) Extraction of typical targets

[0034] Typical target extraction is based on existing knowledge of near-Earth asteroid targets. Typical asteroid target characteristics are extracted, and by combining these characteristics, typical targets are obtained. Numerical simulation models are then built, and measurement components are produced for BRDF (Body-Range Functional Detection) measurements and characteristic measurements. The main characteristics of near-Earth asteroid targets include orbital characteristics, physical characteristics, and material and structural characteristics.

[0035] Near-Earth Asteroid Orbital Characteristics: Based on ground-based observation data, the orbital characteristics and absolute magnitudes of known near-Earth asteroids have been discovered. For a given asteroid target, orbital information can be obtained from various publicly available asteroid databases, yielding orbital characteristics such as semi-major axis, eccentricity, orbital dynamics, and stability. Based on the target asteroid's orbit and mission requirements, spacecraft mission orbits and mission scenarios can be derived through engineering design.

[0036] Physical characteristics of near-Earth asteroids: The main physical characteristics of near-Earth asteroids include size, shape, and surface morphology. The relationship between an asteroid's absolute magnitude and size can be described by the following empirical formula:

[0037]

[0038] in, For absolute magnitude, The diameter of the asteroid. This represents the albedo of the asteroid. It represents the integral of the ratio of reflected energy to incident energy across all frequencies, and it is related to the properties of the asteroid's surface material (composition, particle size, surface structure, etc.).

[0039] Regarding the shape of near-Earth asteroids, due to collisions, mergers, rotation and structural instability, and orbital evolution, most near-Earth asteroids have irregular shapes, exhibiting elongated, forked, or contact binary forms, with a few approaching spherical shapes. According to publicly available radar observation data, nearly 15% of near-Earth asteroids with diameters exceeding 180 meters are contact binary, such as 1999JD6, 2014HQ124, and 2006DP14. Elongated shapes include 1996HW1 and 2023SD220. Near-spherical shapes include 036 Ganymed and 433 Eros. Contact binary, oblate, near-spherical, elongated, gyratory, dumbbell, and tooth-like shapes are considered typical shapes.

[0040] Near-Earth asteroids exhibit diverse surface morphologies, including features such as craters, grooves, and ridges. For example, the surface of asteroid 433 Eros consists of blocks, impact craters, and walls, while asteroid 253 Mathilde is covered with impact craters. Taking all factors into account, the typical surface morphology of an asteroid is characterized by flatness, rockiness, numerous impact craters, hills, and ridges.

[0041] Near-Earth asteroid material: The main characteristics of near-Earth asteroids are their metallicity and the types of metals they contain. The main components of a near-Earth asteroid affect spectral imaging, and the albedo parameter influences the size assessment. Based on spectral type, the most common types are C-type (rich in carbon, albedo around 0.05), S-type (mainly composed of silicate materials and nickel-iron, albedo 0.1~0.25), and M-type (high metallicity, albedo 0.1~0.18). There are also 11 other less common types. In the initial estimation, an average value of 0.15 is used, with a range of [0.03 0.3] to estimate the size of the target asteroid. Considering the main components of the asteroid, typical materials can be treated as ferromagnesian rocks, carbides, nickel-iron ore, basalt, granite, and sulfides.

[0042] Structural characteristics of near-Earth asteroids: The main consideration is whether the near-Earth asteroid itself has a monolithic or rubble pile structure, with porosity as the characterization parameter. Most near-Earth asteroids are not dense monoliths, but rather have a loose, porous structure, or even exhibit a rubble pile structure, i.e., formed by the gravitational accretion of numerous small rocks. The porosity distribution of asteroids is generally 0%–80%, reflecting information about the asteroid's internal structure. It is generally believed that if the porosity is below 30%, it is closer to a monolithic structure. If the porosity exceeds 30%, it is closer to a rubble pile structure, such as asteroid 25143 Itokawa. Different structural characteristics will be modeled separately when extracting typical targets. Porosity affects the mass and density of asteroids. Considering all factors, both monolithic and rubble pile structures can be considered.

[0043] Typical targets for BRDF measurement are obtained by combining different surface morphologies and materials. Typical targets are obtained by combining different surface morphologies, materials, shapes, and structural features.

[0044] 2) BRDF Measurement and Characteristic Measurement

[0045] Both BRDF measurements and characteristic measurements aim to obtain optical inversion modeling data and verify the correctness of the simulation model. The BRDF function describes the reflection of light on opaque objects. BRDF measurements use typical targets with different combinations of surface morphology and materials as references for production measurement, such as flat surfaces with mafic rocks, rocky surfaces with basalt, and impact craters with carbides. Five surface morphologies and six typical materials were extracted, resulting in 30 typical BRDF measurement targets: 5 typical surface morphologies × 6 typical materials. The measurement spectrum is selected based on the optical payload on the spacecraft, such as the 400-900nm visible light camera's commonly used spectral band plus the mid-infrared spectral band. The measurement system resolution is selected based on the spectral resolution of the optical payload.

[0046] In target characteristic measurement, typical surface morphologies include five types: flat, rocky, impact crater-ridden, hills, and ridges; typical shapes include seven types: contact bispherical, elongated, oblate, gyroscopic, dumbbell-shaped, near-spherical, and tooth-like; and materials include six types: ferromagnesian rock, carbides, nickel-iron ore, basalt, granite, and sulfides. By combining the asteroid's surface morphology and material with its shape and structural characteristics, typical targets for target characteristic measurement are formed.

[0047] A typical target can be formed by combining multiple impact craters (surface morphology), carbides (material), monoliths (structural features), and dumbbells (shape). The maximum number of typical target characteristic measurements is 30 combinations of surface morphology and material × 6 typical shapes × 2 typical structural features = 360. Considering that only multi-rock surfaces can be formed under the rubble pile structural features, and the shapes are mostly near-spherical or contact double-spherical, this is reduced to 12. Combining the BRDF measurement results of different surface morphologies and materials, the target characteristic measurement focuses on measuring shapes with significant BRDF differences, materials with large BRDF differences, and prominent surface characteristics, using scaled-down samples from production. Considering measurement convenience, 45 scaled-down measurement samples were produced, with the volume of each sample constrained to 30 cm³. 3 Within this range. In target characteristic measurements, mission lighting conditions should be fully considered. For example, if the lighting conditions within the flight window are stable at around 10°, test redundancy can be considered. , , As the testing lighting conditions, the actual test conditions consisted of 45 scaled-down measurement pieces × 3 lighting conditions = 135 conditions.

[0048] Digital modeling of all typical targets is performed based on simulation and character modeling, BRDF measurement, and characteristic measurement results.

[0049] 3) Application scenario measurement

[0050] Application scenario measurements aim to enrich data and validate models by combining engineering mission scenarios. Input information from typical application scenarios is used, such as changes in the relative position of the spacecraft and asteroid within the spacecraft's flight window during long-distance approach scenarios, the rotation of the target asteroid, and changes in illumination during the approach. Using a scaled-down measurement component as the target for characteristic measurement, sunlight is simulated by moving a light source. The relative motion between the spacecraft and the target can be simulated by moving the scaled-down measurement component or an optical sensor (i.e., an optical camera), thus conducting measurements in typical application scenarios. For example, an asteroid with an equivalent diameter of 50 km is characterized by multiple impact craters (surface morphology), carbides (material), monoliths (structural features), and a dumbbell shape, with a length of symmetry around the dumbbell axis. Rotation speed. Application scenario: asteroid impact defense mission; illumination angle: The relative speed between the spacecraft and the asteroid is 6 km / s. Typical scenario measurements are performed during the spacecraft's approach to the target near-Earth asteroid to verify the accuracy of the digital model, introduce the influence of typical scenarios, and improve the simulation-based modeling.

[0051] 4) Generate a digital target set

[0052] Based on the extracted typical targets, digital simulation modeling is performed for each typical target. BRDF measurement and characteristic measurement data as well as application scenario measurement data are introduced to verify and improve the digital model and generate a digital target set.

[0053] Step 2: Onboard data processing, which includes two parts: onboard processing and onboard verification.

[0054] 1) On-device processing

[0055] like Figure 2 As shown, the onboard data processing mainly involves multi-source data normalization, target shape and rotation characteristic calculation, information fusion, intelligent identification based on a digital target set, and target modeling. The spacecraft can use optical sensors such as infrared cameras, multispectral cameras, and hyperspectral cameras to image asteroid targets and use laser rangefinders for ranging. This invention uses infrared cameras, multispectral cameras, and laser rangefinders as data sources. If microwave detection or other detection methods are used, different data processing methods should be selected to achieve multi-source data normalization and information fusion.

[0056] Multi-source data normalization primarily completes image preprocessing, unifying image format, bit depth, data modality, and denoising methods to facilitate the extraction of effective target characteristics. Data acquired by different sensors on a spacecraft at the same time constitutes the same set of data; therefore, data fusion within the same set is performed first, followed by fusion of data from different sets.

[0057] For infrared and multispectral cameras, exposure time can be controlled to group simultaneously exposed data into the same group and other data into different groups. For two sets of inputs from different times, there will be differences in temporal and spatial dimensions. To perform spatiotemporal fusion, the relative position and illumination conditions are deduced using information timestamps. Data normalization can be achieved using models such as the Fit-FC model based on spatial filtering and hybrid pixel decomposition, and the spatial and temporal adaptive reflectivity fusion model (STARFM). For data within the same group, the installation position constraints of different sensors (i.e., the spacecraft installation position transformation matrix, which can be obtained through spacecraft precision measurement and on-orbit calibration) are introduced for intra-group fusion. Information fusion mainly involves feature matching. Features are extracted from the input image using feature stacking fusion algorithms, and various feature matching or similarity measurement algorithms are used for feature matching, such as Euclidean distance, gradient-based algorithms, and cross-correlation-based methods, to obtain target feature data for intelligent recognition and target modeling based on digital target sets.

[0058] Calculating the target shape and rotation characteristics requires incorporating multiple sets of sensor data to separately calculate the calibrated target shape and rotation characteristics. Infrared cameras, in particular, offer significant advantages for target shape observation. The target shape is calculated using multiple sets of observation data individually, and the rotation characteristics are calculated by combining shape changes with the relative motion between the spacecraft and the target. The calculation results can be used for multi-source data normalization and on-board verification.

[0059] The fused data is then used to introduce an intelligent recognition algorithm based on a digital target set for intelligent identification. This algorithm can be based on YOLOv5, convolutional neural networks, or similar methods. The algorithm selects the most matching typical target model from the digital target set and completes target modeling based on this typical target model and the target feature data obtained during information fusion.

[0060] 2) On-device verification

[0061] Onboard calibration is based on newly acquired data from sensors on the spacecraft and the established target model. Onboard calibration shares the same multi-source data normalization processing module, information fusion module, and target shape and rotation characteristic calculation module as the aforementioned onboard processing. Onboard calibration performs characteristic verification and model calibration.

[0062] In the characteristic verification process, the target shape and spin characteristics are independently calculated based on the newly acquired data from the sensor, resulting in updated data for the target shape and spin characteristics. The calculated results are then compared with the previously acquired characteristics. If the data does not show any contradictions (e.g., the change in the spin axis between the two calculations is within a reasonable range, and the target shape does not undergo any unreasonable abrupt changes), the characteristic verification is considered successful.

[0063] During model validation, multi-source data normalization and information fusion are performed on newly acquired sensor data to obtain updated target feature data. This updated data is compared with the original model. If no contradictions are found, the updated data is introduced to update the model. If contradictions are found, such as a previously identified impact crater area showing a flat topography in the updated data, the model is not updated immediately. Instead, multiple sets of data are introduced for comparison or ground processing is performed to ensure model accuracy.

[0064] Step 3, algorithm ground verification, includes three steps: algorithm correctness verification, algorithm pruning and speed-up, and algorithm edge deployment.

[0065] like Figure 3 As shown, a digital target set is used as the data source for the sensor, and sensor detection data is obtained through sensor sensitivity. In the algorithm correctness verification, the sensor detection data is input into a ground processing algorithm that primarily performs the functions and runs on ground-based computer hardware. Based on the algorithm results, inversion modeling is performed on the ground and compared with scaled-down measurement samples to verify the algorithm's correctness.

[0066] Algorithm pruning and acceleration aim to reduce computational requirements, improve responsiveness, and adapt to spacecraft deployment. Algorithm optimization involves tailoring functionality or reducing accuracy requirements, resulting in a prototype on-board data processing algorithm. A digital target set is used as the data source for sensors. Sensor detection data is obtained through sensor sensitivity and input into the prototype on-board data processing algorithm to obtain the target model. The target model is then compared with data from scaled-down measurement devices to ensure the on-board data processing algorithm meets requirements.

[0067] In edge deployment of the algorithm, the prototype data processing algorithm on the spacecraft is modified to adapt to the spacecraft hardware and runs on the spacecraft hardware products to simulate the algorithm's operation in orbit. After edge deployment, the digital target set still serves as the sensor data source. A target model is established and compared with data from scaled-down measurement devices to ensure the algorithm's correctness and reliability.

[0068] Step 4: Space-Ground Joint On-Orbit Application

[0069] When the spacecraft is far from the target asteroid (e.g., tens to millions of kilometers), the target asteroid is a point target. Considering the detection range and accuracy of different detection methods, optical detection is mainly used at long distances. The estimated position and motion characteristics of the asteroid are obtained by extrapolating its orbit, enabling spacecraft pointing control. The spacecraft primarily extracts the target by matching background star charts within its field of view, visually observing motion characteristics, and analyzing point target brightness changes, achieving target-oriented guidance. When the spacecraft is relatively close to the target, such as during the orbital phase when the distance is within a few hundred kilometers, the target asteroid becomes a surface target, and this method can be used for target modeling.

[0070] like Figure 4 As shown, during time slot T0, after the spacecraft's onboard sensors acquire asteroid target information, the data is processed onboard to complete target modeling, resulting in the target model and its characteristics. During time slot T1, the spacecraft's onboard sensors obtain new data and perform onboard verification based on the target model, characteristics, and the new data. A maximum number of verification executions, N, is set onboard. The verification count is incremented by 1 if the verification fails, and resets to zero if it passes. If the number of verifications is less than or equal to N, the spacecraft returns to time slot T0 to rebuild the target model. If the number of verifications exceeds N, indicating situations such as significant differences in multiple results for the spin characteristics, and multiple onboard verifications fail, the original sensor data is output. The spacecraft then performs a close flyby or switches to a backup orbit for ground processing. If the verification passes, the model is updated and transmitted to the spacecraft's control unit. This updated model will also be used for onboard verification during time slot T2. The spacecraft's work during time slot T2 is similar to that of time slot T1. If the onboard verification fails and the number of verifications does not exceed the verification execution limit, the spacecraft returns to time slot T0 to rebuild the target model.

[0071] The following sections use uncertain asteroid landing missions and asteroid impact defense missions as examples to illustrate the application. First, during the spacecraft development process, a digital target set is established based on the mission requirements, forming an on-board data processing algorithm, which is then validated on the ground. The following sections provide examples of the space-ground joint on-orbit application method.

[0072] Example 1:

[0073] In missions involving landing on unknown asteroids, the spacecraft identifies and guides itself from a distance, approaches the asteroid, and then orbits it before attempting a landing. The spacecraft is equipped with infrared and high-resolution cameras. Onboard data primarily comes from the approach and orbit phases after the ground and asteroid reach a surface level.

[0074] During the T0 period, the spacecraft's high-resolution camera and infrared camera simultaneously expose the target to acquire information, and then perform onboard data processing. The data acquired by the infrared and high-resolution cameras are processed onboard. Depending on the lighting conditions during the exposure period, if the asteroid's lighting conditions are good, panoramic imaging can be performed. The high-resolution camera is primarily used to determine the asteroid's shape, while the infrared camera is used for verification. Multiple exposures are performed by the high-resolution and infrared cameras, and after receiving multiple sets of data, the target's shape and rotation characteristics are calculated. Normalization processing, information fusion, and intelligent recognition based on a digital target set are performed on the data from exposures at different times to complete target modeling, obtaining the target model and its characteristics.

[0075] During the T1 period, on-board calibration verification is performed based on newly acquired data from the high-resolution camera and infrared camera. If the on-board verification passes, the model built during the T0 period is updated and sent to the control unit for landing area selection. If the verification fails, it is determined whether the number of verification executions N has been exceeded. If the number of executions is less than N, the spacecraft control repeats the T0 period operation, reconstructing the target model based on all sensor data and calculating the target characteristics. If the number of executions is greater than N, the raw sensor information is output, and the spacecraft transmits the information downlink to the ground for processing. The spacecraft maintains a close-range flyby state, awaiting ground commands.

[0076] After selecting the landing area, confirmation is performed either by ground control or autonomously by the spacecraft. Ground control or the spacecraft autonomously plans the landing path. During landing, the distance between the spacecraft and the target asteroid continuously decreases. Based on newly acquired sensor information, subsequent periods repeat the work of period T1, continuously updating the model, enriching model details, correcting target spin characteristics, etc., to support mission execution.

[0077] Example 2:

[0078] This paper takes the application of a kinetic impactor spacecraft (i.e., an impactor) in an uncertain asteroid defense mission as an example. Due to the high relative velocity between the impactor and the asteroid target, the onboard guidance camera is mainly used for asteroid detection and identification at ultra-long distances and for asteroid aiming point extraction at closer distances. The spacecraft's sensors consist of two heterogeneous high-resolution cameras. At long distances, faint targets are detected and identified through star map matching, visual motion characteristics, and changes in point target brightness, thus achieving target guidance. At closer distances, when the image is of a surface target, the method of this invention is used for joint space-ground application. During the development of the impactor, the establishment of a digital target set for the target asteroid, the design of onboard data processing, and ground verification of the algorithm have been completed.

[0079] During the T0 period, two high-resolution cameras simultaneously exposed multiple times to acquire data. Based on the bright spots and the illumination conditions during the exposures, the shape and rotation characteristics of the asteroid target were calculated. Since the camera positions were fixed and approaching each other before impact, and the illumination conditions were consistent, the rotation of the asteroid target could be determined from the two sets of characteristic data, thus obtaining its spin characteristics. After multi-source data normalization and information fusion, intelligent recognition based on a digital target set was performed to match typical shapes and complete the target modeling.

[0080] During period T1, the two high-resolution cameras perform on-board verification after acquiring new data. If the verification passes, the model is updated and transmitted to the control unit. If the verification fails, it is determined whether the number of verification executions (N) has been exceeded. If the number of executions is less than N, the impactor control repeats the operation of period T0, reconstructing the target model based on all sensor data and calculating the target characteristics. If the number of executions is greater than N, the raw sensor information is output and transmitted down to the ground via the impactor for processing. The impactor then switches to a backup orbit to acquire more data, and the ground control repeats the mission. If the verification passes, different aiming point extraction algorithms are matched based on the shape, spin characteristics, and pre-impact illumination conditions to ensure that the aiming point is located in the central solid part of the asteroid target for better impact results.

[0081] During the approach to the target asteroid, the T1 phase is repeated to continuously update the model, enrich the model details, and continuously transmit the updated model to the control unit, whereby the impactor performs data downlink until the impactor stops controlling the system.

[0082] The descriptions of the various embodiments have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

Claims

1. A spacecraft autonomous modeling method for unknown near-Earth asteroid targets, characterized in that, Includes the following steps: Step 1: Generate a digital target set through typical target extraction, BRDF measurement and characteristic measurement, and application scenario measurement; BRDF represents the bidirectional reflectance distribution function. Step 2: Perform on-board data processing, including on-board processing and on-board verification; Step 3: Conduct ground-based validation of the algorithm, including algorithm correctness verification, algorithm pruning and speed-up, and algorithm edge deployment; Step 4: Conduct joint space-ground on-orbit application. When the spacecraft is close to the target, target modeling is performed. First, the data acquired by the sensors is processed on-board to complete the target modeling. Then, the newly acquired data from the sensors is used for on-board verification and model update. By limiting the number of verifications, model reconstruction and ground processing are introduced.

2. The spacecraft autonomous modeling method for unknown near-Earth asteroid targets according to claim 1, characterized in that, In step 1, the typical target extraction is based on the existing knowledge of asteroid targets. Typical characteristics of the targets are extracted, including orbital characteristics, physical characteristics, material and structural characteristics. Typical targets are obtained by combining different characteristics.

3. The spacecraft autonomous modeling method for unknown near-Earth asteroid targets according to claim 2, characterized in that, In the typical target extraction, orbital characteristics are used to obtain the spacecraft mission orbit and mission scenario in combination with mission requirements, and are obtained from a public database.

4. The spacecraft autonomous modeling method for unknown near-Earth asteroid targets according to claim 2, characterized in that, In the extraction of typical targets, physical characteristics include size, shape, and surface morphology. Size is estimated based on absolute magnitude. Typical shapes include contact binary, flattened round, near-spherical, slender, gyroscope, dumbbell, or tooth-like. Typical surface morphology includes flat, rocky, crater-filled, hill, or ridge.

5. The spacecraft autonomous modeling method for unknown near-Earth asteroid targets according to claim 2, characterized in that, The typical target material includes ferromagnesian rock, carbides, nickel iron ore, basalt, granite, or sulfides.

6. The spacecraft autonomous modeling method for unknown near-Earth asteroid targets according to claim 2, characterized in that, In the extraction of typical targets, structural characteristics include monoliths and piles of gravel.

7. The spacecraft autonomous modeling method for unknown near-Earth asteroid targets according to claim 1, characterized in that, In step 1, the BRDF measurement and characteristic measurement are based on the extracted typical targets. After analysis, incompatible characteristic combinations are eliminated, and load spectral constraints and illumination condition constraints in the task scenario are introduced to refine the test scheme. Application scenario measurement introduces the relative motion relationship between the mission process target and the spacecraft to enrich the data verification model.

8. The spacecraft autonomous modeling method for unknown near-Earth asteroid targets according to claim 1, characterized in that, In step 2, the on-device processing performs multi-source data normalization processing, target shape and rotation characteristics calculation, information fusion, target shape and rotation characteristics calculation, intelligent recognition based on digital target set, and target modeling. Onboard verification is based on newly acquired data from sensors on the spacecraft and the established target model. Onboard verification shares the same multi-source data normalization processing, information fusion, and target shape and rotation characteristic calculation modules as the aforementioned onboard processing. Onboard verification performs characteristic verification and model verification, and updates the model after passing the verification.

9. The spacecraft autonomous modeling method for unknown near-Earth asteroid targets according to claim 1, characterized in that, In step 3, the algorithm ground verification uses the digital target set as the sensor data source, acquires detection data through sensors, and compares the algorithm running results with the data of the selected scaled-down measurement device to ensure correctness. In the algorithm correctness verification, algorithm trimming and speed-up test, the algorithm runs on the ground computer, and after the algorithm is deployed at the edge, the algorithm runs on the spacecraft hardware product.

10. The spacecraft autonomous modeling method for unknown near-Earth asteroid targets according to claim 1, characterized in that, In step 4, the specific operation after limiting the number of verifications is as follows: set the maximum number of verifications to N; when the verification fails, the number of verifications is incremented by 1; when it passes, the number of verifications is reset to zero; if the number of verifications exceeds N, output the original sensor data, and the spacecraft performs a close-range flyby or switches to a backup orbit for ground processing; if the number of verifications does not exceed N, rebuild the target model.

Images