A GNSS optical angle measurement autonomous orbit determination observation scheduling and consistency adaptive evaluation method for high-orbit spacecraft
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INNOVATION ACAD FOR MICROSATELLITES OF CAS
- Filing Date
- 2026-02-27
- Publication Date
- 2026-06-23
Smart Images
Figure CN122258933A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of autonomous navigation technology for spacecraft, and more specifically, to a method for autonomous orbit determination observation scheduling and consistency adaptive evaluation of Global Navigation Satellite System (GNSS) optical angle measurement for high-orbit spacecraft. Background Technology
[0002] As space missions become increasingly complex, the need for long-term autonomous operation is becoming more urgent for high-orbit spacecraft located in Medium-High Earth Orbit (MEO), Inclined Geosynchronous Orbit (IGSO), and Geostationary Orbit (GEO). In scenarios with weak ground support, such as insufficient telemetry and control coverage, sparse ground arcs, or interrupted communication links, the ability of spacecraft to continuously and accurately acquire its own orbital status is the foundation and prerequisite for it to complete tasks such as on-orbit servicing, rendezvous and docking, and Earth observation.
[0003] Currently, orbit determination and navigation of high-orbit spacecraft mainly rely on the following two types of technical solutions: The first type involves directly using a spaceborne GNSS receiver for navigation. This approach receives signals transmitted by GNSS satellites (such as GPS and BeiDou) and uses pseudorange or carrier phase observations to achieve positioning. However, this approach has inherent limitations in high-orbit environments. Because high-orbit spacecraft are typically located above or at similar altitudes to GNSS constellations, the main lobe beams of GNSS satellite signals are primarily pointed towards Earth, meaning the spacecraft receives mostly sidelobe signals with severely attenuated signal strength. Furthermore, the limited number of observable GNSS satellites and their poor spatial geometry lead to a significant decrease in navigation accuracy, sometimes even preventing positioning altogether, making it difficult to meet the requirements for long-term, continuous, and highly reliable autonomous navigation.
[0004] The second type is the traditional ground-based telemetry, tracking, and command (TT&C) orbit determination mode. This method relies on a ground-based TT&C network to track and measure the spacecraft, and then calculates the orbit through a ground data processing center before transmitting it to the spacecraft. Although this method has high accuracy, its real-time performance and continuity are limited by the geographical distribution of tracking stations, the arrangement of TT&C arcs, and the latency of the space-to-ground link. For spacecraft that seek a high degree of autonomy, this ground-dominated mode cannot provide continuous and timely orbit information updates. Once the TT&C link is interrupted or resources are scarce, the spacecraft will face the risk of losing navigation information.
[0005] Against this backdrop, utilizing spaceborne optical payloads to perform angle measurement observations of space targets with known orbits (such as GNSS satellites) provides a novel technological approach for autonomous orbit determination of high-orbit spacecraft. By using GNSS satellites with known orbits as space reference points, the spacecraft can acquire azimuth and elevation angle information relative to these reference points, and combine this with its own orbital dynamics model to achieve autonomous orbit determination through filtering estimation. This method does not rely on the direct reception of weak GNSS signals and can theoretically solve the challenges of high-orbit GNSS navigation.
[0006] However, there are currently no mature optical angle measurement autonomous orbit determination methods and systems for high-orbit spacecraft on the market, leaving this technological field largely unexplored. To realize this vision, a series of key technical challenges need to be overcome: First, the information content of angle measurements decreases with the square of the target distance; how to efficiently select observation targets from numerous GNSS satellites to maximize information acquisition efficiency is a pressing issue. Second, the maneuverability of onboard turntables is limited, and both camera imaging and data processing are time-consuming; how to plan the optimal observation sequence under hard resource constraints lacks a systematic scheduling method. Finally, during long-arc high-orbit operation, dynamic model errors (such as solar radiation pressure perturbations and third-body gravitational disturbances) gradually accumulate. If filter parameters remain fixed, it can easily lead to inconsistent or even divergent estimation results; currently, there is no targeted adaptive safeguard mechanism. Summary of the Invention
[0007] The purpose of this invention is to address the aforementioned shortcomings in the existing technology by providing a GNSS optical angle measurement autonomous orbit determination observation scheduling and consistency adaptive evaluation method for high-orbit spacecraft. This method aims to solve the technical problem of achieving long-term, continuous, highly reliable, and high-precision autonomous orbit determination for high-orbit spacecraft under conditions where GNSS signals are difficult to utilize directly and ground telemetry and control support is limited.
[0008] To achieve this objective, the technical solution adopted by the present invention is as follows: A method for autonomous orbit determination observation scheduling and consistency adaptive evaluation using GNSS optical angle measurement for high-orbit spacecraft is characterized by the following steps: Step S1: System initialization establishes the spacecraft's state vector in the geocentric inertial coordinate system, which includes at least position and velocity components; establishes an orbital dynamics model for state extrapolation, which includes at least the Earth's central gravitational term; sets key operating parameters such as the filtering period, single-epoch observation time budget, and the maximum angular velocity and maximum angular acceleration of the onboard turntable.
[0009] Step S2: GNSS Target Orbit Acquisition and Visibility Determination At the beginning of each filtering cycle, the orbital information of navigation satellites from multiple GNSS constellations is read from the onboard ephemeris database and unified into the inertial coordinate system; based on the spacecraft's current predicted position and the orbital positions of each candidate navigation satellite, the line-of-sight direction is calculated and visibility is determined, targets that do not meet constraints such as solar avoidance angle, Earth edge avoidance angle, and camera field of view are eliminated, forming a set of visible candidate targets for the current epoch.
[0010] Step S3: Resource Constraint Modeling For each navigation satellite target in the set of visible candidate targets generated in step S2, the total time cost required to complete one observation is calculated. The total time cost consists of two parts: fixed overhead and turntable maneuvering time. The fixed overhead includes at least turntable stabilization time, camera exposure time, image data readout time, and image processing and centroid extraction time. The turntable maneuvering time is calculated using a trapezoidal angular velocity profile model, based on the angle between the current turntable pointing direction and the target's line-of-sight direction, combined with the turntable's maximum angular velocity and maximum angular acceleration limits.
[0011] Step S4: Observation Scheduling and Sequence Generation Based on the resource constraint model established in step S3, an iterative loop mechanism of "candidate evaluation - greedy addition - state update" is adopted to generate the optimal multi-target observation sequence within the single epoch observation time budget. In each iteration, all candidate targets that meet the remaining time budget are scored and ranked according to a preset scoring function. The target with the highest score is selected and added to the observation sequence, and the turntable pointing state and remaining time budget are updated accordingly. The iteration terminates when the remaining budget is insufficient or all candidate targets have been selected. The input parameters of the scoring function include at least the observation time cost, the target apparent magnitude, the geometric dispersion with the selected target set, and the information gain calculated based on covariance virtual update.
[0012] Step S5: Imaging and Angle Measurement According to the observation sequence generated in step S4, the turntable is driven to point at the target, the camera is triggered to image, the image data is read out, image processing and centroid extraction are performed, and finally the azimuth and elevation angles of the target navigation satellite relative to the spacecraft are calculated. At the same time, a measurement noise model related to the observation conditions is established. The measurement noise is composed of at least the imaging centroid extraction noise, the platform attitude determination noise and the camera calibration noise.
[0013] Step S6: Extended Kalman Filter Prediction and Update An extended Kalman filter is used for orbital state estimation. In the prediction phase, the state vector and covariance matrix are numerically extrapolated based on the orbital dynamics model. In the update phase, each angle measurement obtained in step S5 is processed sequentially, the measurement prediction residuals and innovation covariance are calculated, and the state vector and covariance matrix are updated. Before updating the measurement, the normalized squared innovation statistic is calculated; if it exceeds a preset chi-square test threshold, the update of that measurement is rejected.
[0014] Step S7: Consistency Assessment and Closed-Loop Tuning A closed-loop mechanism for consistency statistical verification and adaptive tuning of process noise for autonomous orbit determination based on optical angle measurement is established. Normalized innovation square sequences are collected within a sliding time window. The mean of the statistics within the window is compared with the confidence interval of the theoretical chi-square distribution. When the statistical mean deviates from the confidence interval, the consistency ratio is calculated, and the intensity scaling factor of the process noise is adjusted accordingly. After amplitude limiting, the result is fed back to the filtering and prediction stage in step S6.
[0015] Step S8: Output and Loop Execution Output the spacecraft's position vector, velocity vector, and covariance matrix at the current epoch as the solution for autonomous orbit determination, and proceed to the next filtering cycle, repeating steps S2 to S7.
[0016] Furthermore, as a preferred embodiment, the orbital dynamics model in step S1 further includes one or more of the following: J2 perturbation, solar radiation pressure perturbation, and gravitational disturbance of the Sun and Moon as a third body; and the solar radiation pressure reflection coefficient is synchronously estimated as an extended state variable during the filtering process.
[0017] Furthermore, as a preferred option, the scoring function in step S4, when using information gain for scoring, needs to perform a virtual update on the covariance after each target is added, in order to reflect the diminishing marginal returns of the observation.
[0018] Furthermore, as a preferred embodiment, the consistency evaluation in step S7, under the condition of having a reference track or simulation verification, also collects the normalized estimation error square sequence, which, together with the normalized innovation square sequence, is used to determine the statistical consistency of the filter.
[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: 1) By using GNSS navigation satellites as optical observation targets, the weak GNSS signals in high orbit can be received directly without relying on spacecraft, bypassing the inherent difficulties of receiving GNSS signals in high orbit. This provides a new and stable source of observation updates for high orbit spacecraft, filling a technological gap in this field.
[0020] 2) By explicitly incorporating turntable maneuverability, fixed observation overhead, and time budget into resource constraint modeling, and constructing a scoring function that comprehensively considers time cost, target quality, and geometric configuration, this invention can generate the optimal multi-target observation sequence within limited onboard resources through a greedy iterative algorithm, significantly improving the information acquisition efficiency per unit time and accelerating the orbit determination and convergence speed.
[0021] 3) A statistical testing framework based on normalized squared innovation was established, which was used to drive the closed-loop adaptive tuning of process noise. This mechanism enables the filter to automatically sense and adapt to dynamic model errors and environmental changes, maintaining statistical consistency throughout long-arc operation, effectively avoiding filter divergence, and significantly improving the long-term reliability and accuracy assessability of the system.
[0022] 4) It features a high degree of modularity, with clear interfaces for each step, facilitating engineering implementation and subsequent upgrades and maintenance. Furthermore, its compatibility with multi-constellation GNSS targets allows it to fully utilize the ever-expanding space-based navigation resources, giving it broad application prospects. Attached Figure Description
[0023] Figure 1 This is a flowchart illustrating the overall process of the present invention: an autonomous orbit determination observation scheduling and consistency adaptive evaluation method for Global Navigation Satellite System (GNSS) optical angle measurement for high-orbit spacecraft.
[0024] Figure 2 This is a schematic diagram of the internal process and resource constraints of the observation scheduling module in this invention. Figure 3 This is a schematic diagram of the consistency assessment and process noise closed-loop tuning process in this invention. Detailed Implementation
[0025] The technical solutions of the present invention will now be described in detail with reference to the accompanying drawings. To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0026] This invention provides a method for autonomous orbit determination observation scheduling and consistency adaptive evaluation of GNSS optical angle measurement for high-orbit spacecraft. It utilizes onboard optical payloads to take GNSS navigation satellites with known orbits as space reference points, and obtains their angle measurement information. Combined with orbital dynamics models and adaptive filtering algorithms, it achieves highly reliable autonomous orbit determination of spacecraft under GNSS signal blind zones and weak ground support conditions.
[0027] like Figure 1 As shown, the overall process of this invention includes the following steps: Step S1: System Initialization First, a state vector for the spacecraft in a geocentric inertial coordinate system (such as the J2000.0 coordinate system) is established. This vector contains six state variables, including position and velocity components: position components r_x, r_y, and r_z along the three coordinate axes, and velocity components v_x, v_y, and v_z. To improve the accuracy of orbit determination over long arc segments, parameters related to the dynamic model, such as the solar radiation pressure reflection coefficient, can be added as extended-dimensional state variables to the filter for estimation.
[0028] Secondly, an orbital dynamics model is established for state extrapolation. This model should at least include the Earth's central gravitational term and can be extended to include J2 perturbations, solar radiation pressure perturbations, and third-body gravitational disturbances from the Sun and Moon, depending on accuracy requirements. To enhance the adaptability of long-arc orbit determination, the solar radiation pressure reflection coefficient can be simultaneously estimated as an extended state variable during the filtering process.
[0029] Finally, key operating parameters need to be set, including the filter period, single-epoch observation time budget, and the maximum angular velocity and maximum angular acceleration of the onboard turntable.
[0030] Step S2: GNSS target orbit acquisition and visibility determination At the start of each filtering cycle, the system reads the orbital information of the GNSS constellation's navigation satellites from the onboard ephemeris database. This information can be derived from broadcast ephemeris, predicted ephemeris, or precise ephemeris. For multi-constellation applications, the orbital information of different constellations needs to be uniformly converted to an inertial coordinate system.
[0031] Subsequently, based on the spacecraft's current predicted position and the orbital positions of each candidate navigation satellite, the line-of-sight direction is calculated. Based on the line-of-sight direction, combined with the spacecraft's attitude, camera mounting matrix, and field of view, visibility is assessed. Targets that do not meet the constraints of solar avoidance angle, Earth edge avoidance angle, camera field of view, or are obstructed are eliminated, forming the set of visible candidate targets for the current epoch. Specifically, the solar avoidance angle constraint requires that the angle between the line-of-sight direction and the sun's direction be greater than a certain threshold (e.g., 30°) to prevent direct sunlight from damaging the camera or causing image saturation. The Earth edge avoidance angle constraint requires that the line-of-sight direction avoid the Earth's edge entering the field of view to prevent interference from Earth's background radiation or Earth obstruction. The camera field of view constraint requires that the target be within the camera's field of view. The Earth's shadow / illumination constraint requires that the target be illuminated by the sun to provide sufficient brightness. Through the above screening, invisible targets are eliminated, forming the set of visible candidate targets for the current epoch.
[0032] Step S3: Resource Constraint Modeling: Explicitly incorporate the onboard hardware resource constraints into the observation planning mathematical model for optical angle measurement autonomous orbit determination.
[0033] For each navigation satellite target in the visible candidate target set, the total time cost required to complete one observation is calculated. This cost consists of two parts: fixed overhead and turntable maneuvering time. Fixed overhead includes turntable stabilization time, camera exposure time, image data readout time, and image processing and centroid extraction time. Turntable maneuvering time is calculated using a trapezoidal angular velocity profile, based on the angle between the current turntable pointing direction and the target's line of sight, combined with the turntable's maximum angular velocity and maximum angular acceleration limits. Through this modeling, each candidate target is assigned a clear time cost, providing a quantitative basis for subsequent scheduling optimization.
[0034] Step S4: Observation scheduling and sequence generation: A method for optimizing the generation of multi-objective observation sequences under single-epoch time budget constraints is proposed.
[0035] The scheduling process employs an iterative loop mechanism of "candidate evaluation - greedy addition - state update", such as... Figure 2 As shown.
[0036] Before the iteration begins, initialize the remaining time budget, set the selected target set to empty, and set the current turntable pointer to either the initial pointer or the pointer at the end of the previous epoch. Perform the following operations during each iteration: 1. Candidate Evaluation: All candidate targets that meet the remaining time budget are scored and ranked according to the scoring function. The scoring function is designed to take into account multiple factors: the observation time cost reflects the degree of resource consumption, the target apparent magnitude characterizes the observation quality and the probability of missed detection, the geometric dispersion with the selected target set reflects the quality of the observation configuration, and the information gain calculated based on covariance virtual update quantifies the marginal contribution of a single observation to the orbit determination accuracy.
[0037] 2. Greedy addition: Select the target with the highest score to add to the observation sequence, then update the turntable pointing status and deduct the remaining time budget.
[0038] 3. Status Update: Update the remaining time budget and update the current turntable direction to point towards the target in preparation for the next maneuver calculation.
[0039] The iteration terminates when the remaining time budget is insufficient to complete the observation of any candidate targets or when all candidate targets have been selected, and the final observation sequence is output. For the strategy using information gain scoring, after each target is added, the current filtered covariance matrix needs to be replaced with a virtually updated covariance matrix for the scoring calculation in the next iteration, thus correctly reflecting the diminishing marginal returns.
[0040] Step S5: Imaging and Angle Measurement Following the observation sequence generated in step S4, the system sequentially observes each target navigation satellite. Each observation includes driving the turntable to point towards the target, triggering camera exposure and imaging, reading out image data, and performing image processing and centroid extraction. By combining the centroid coordinates of the satellite points with camera calibration parameters and platform attitude information, the azimuth and elevation angles of the target navigation satellite relative to the spacecraft can be calculated. To ensure the reliability of the filtered estimation, a measurement noise model related to the observation conditions needs to be established. The optical angle measurement noise is a synthesis of imaging centroid extraction noise, platform attitude determination noise, and camera calibration noise, while the centroid extraction noise typically increases with the target's apparent magnitude. Furthermore, the equivalent noise contribution after the navigation satellite orbital information error is projected into angular space must also be considered.
[0041] Step S6: Extended Kalman Filter Prediction and Update An extended Kalman filter (EKF) is used for orbital state estimation. In the prediction phase, based on the dynamic model established in step S1, the state vector and covariance matrix are numerically integrated and extrapolated to obtain the prior state estimate. In the update phase, the angular measurements obtained in step S5 are processed sequentially. The measurement prediction residuals and innovation covariance are calculated, and the posterior covariance is updated using the Joseph form to enhance numerical stability. Before updating the measurements, the normalized squared innovation statistic is calculated; if it exceeds a preset chi-square threshold, the update of that measurement is rejected, thereby suppressing the divergence risk caused by missed detections, mismatches, or anomalous observations.
[0042] Step S7: Consistency Assessment and Closed-Loop Tuning: Establish a closed-loop mechanism for consistency statistical verification and process noise adaptive tuning for autonomous orbit determination based on optical angle measurement, ensuring that the filter maintains statistical consistency during long-arc operation, such as... Figure 3 As shown.
[0043] The system collects a normalized innovation square sequence within a sliding time window and, under conditions of reference trajectory or simulation verification, simultaneously collects a normalized estimation error square sequence. The mean of the statistics within the window is compared with the confidence interval of the theoretical chi-square distribution to determine if the filter is in a consistent state. When the statistical mean deviates from the confidence interval, a consistency ratio is calculated, and the intensity scaling factor of the process noise is adjusted accordingly. The adjusted scaling factor, after amplitude limiting, is fed back to the filter's prediction stage.
[0044] Through this closed-loop tuning mechanism, when there is a deviation between the dynamic model and reality, the process noise will automatically increase to absorb the model error; when the model is well matched, the process noise will converge appropriately to improve the estimation accuracy. That is, the filter can automatically adapt to model errors and environmental changes, always maintaining statistical consistency, so that the output covariance matrix P can truly reflect the confidence level of the estimation error.
[0045] Step S8: Output and Loop Execution After completing the above steps, the system outputs the spacecraft's position vector, velocity vector, and covariance matrix for the current epoch as the solution for autonomous orbit determination. Then, it enters the next filtering cycle and repeats steps S2 to S7, thereby achieving continuous autonomous orbit determination over long arc segments.
[0046] Example 1: Application scenarios seeking rapid convergence under resource allocation Consider a communication satellite located near geostationary orbit, equipped with an optical camera and a multi-axis precision turntable. The filtering period is equal to the single-epoch time budget, meaning the observation task is completed within each period. The turntable has a certain maximum angular velocity and maximum angular acceleration capability. Fixed observation overhead includes turntable stabilization time, exposure time, readout time, and processing time. With the above configuration, sequential observations of multiple navigation satellites can be completed within a single filtering period.
[0047] For this scenario, a scoring strategy prioritizing bright stars or combining brightness with maneuver time penalties is recommended. When there are sufficient visible navigation satellites, targets with smaller apparent magnitudes (i.e., higher brightness) and closer distances to the spacecraft should be prioritized. These targets not only have high imaging signal-to-noise ratios and low false negative probabilities, but also, due to their proximity, angle measurements are more sensitive to position, providing a greater amount of effective information per observation. Under this operating condition, the orbit determination system can converge from a large initial error to steady-state accuracy within hours. Typical steady-state orbit determination accuracy is a root mean square error in the range of tens to hundreds of meters, and a root mean square error in the range of centimeters per second. The consistency closed-loop tuning mechanism automatically adjusts process noise during convergence to ensure that the covariance envelope matches the actual error level.
[0048] Example 2: Application scenarios emphasizing consistency and robustness under resource constraints Considering that spacecraft may be degraded due to limited attitude maneuvering or turntable malfunction, resulting in a reduced time budget per epoch, only a small number of navigation satellites can be observed per filtering cycle. Meanwhile, the dynamic model may contain uncalibrated system biases, and the solar radiation pressure coefficient may deviate from the actual value.
[0049] For this scenario, a comprehensive scoring strategy combining information gain and geometric dispersion while also considering time costs is recommended. Due to limited observation resources, the marginal contribution of each observation is particularly critical, and targets that maximize the improvement of orbital covariance should be prioritized. When selecting targets, sufficient geometric diversity in observation directions across different periods must be ensured to maintain system observability. In this usage state, the orbit determination convergence time will be appropriately longer than in Example A, but the consistency closed-loop tuning mechanism plays a crucial role. When the dynamic model deviation causes the actual error to grow faster than the covariance prediction, the consistency ratio will be too high, and the system will automatically increase process noise to absorb model errors; as observation updates are continuously injected, the filter state is gradually corrected. This adaptive adjustment mechanism effectively avoids the risk of filter divergence under conditions of resource scarcity and model mismatch, ensuring the long-term robust operation of the system.
[0050] Example 3: Application scenarios requiring strategy switching under conditions of superimposed error sources Considering the multiple adverse factors encountered during mission execution: the navigation satellite ephemeris degrades from precise ephemeris to only broadcast ephemeris available, significantly increasing ephemeris error; space environment disturbances lead to a decrease in image quality for some observation cycles, increasing the target miss rate; the optical system experiences a certain degree of defocusing due to changes in the thermal environment, increasing centroid extraction noise.
[0051] For this scenario, it is recommended to increase the weight of the target brightness and observation distance related terms in the scoring function, or introduce a minimum observation quality constraint, such as setting the target apparent magnitude threshold and target distance threshold to not exceed the corresponding upper limit. When observation conditions deteriorate further, a more conservative scheduling strategy can be switched to, abandoning the pursuit of maximizing information gain and instead ensuring that at least one reliable high-quality observation can be obtained in each cycle. If necessary, a strategy adaptive switching mechanism can be adopted, using a bright star priority strategy to achieve rapid convergence in the early stage of orbit determination, switching to an information gain priority strategy to maintain optimal accuracy after entering a steady state, and switching back to the bright star priority strategy to ensure robustness when a significant decrease in observation quality is detected. In this usage state, the innovative squared threshold test mechanism can effectively eliminate poor-quality measurements caused by missed detections or abnormal centroid extraction, avoiding contamination of the filtering state. The consistency closed-loop tuning mechanism automatically adjusts the process noise according to the statistical characteristics of the remaining effective measurements, appropriately relaxing the covariance envelope when observation information decreases, and maintaining the statistical reliability of the estimation results. Through the synergistic effect of multiple mechanisms, this scheme can still maintain acceptable orbit determination accuracy and reliable error assessment capability under the adverse conditions of superimposed error sources.
[0052] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for autonomous orbit determination observation scheduling and consistency adaptive evaluation using GNSS optical angle measurement for high-orbit spacecraft, characterized in that, Includes the following steps: Step S1: System Initialization: Establish the spacecraft's state vector and orbital dynamics model in the geocentric inertial coordinate system, and set the filtering period, single-epoch observation time budget, and the maximum angular velocity and maximum angular acceleration of the onboard turntable; Step S2: GNSS target orbit acquisition and visibility determination: At the beginning of each filtering cycle, the orbital information of navigation satellites of the GNSS constellation is acquired, visibility is determined based on the current predicted position of the spacecraft, targets that do not meet the observation constraints are eliminated, and a set of visible candidate targets for the current epoch is formed. Step S3: Resource constraint modeling considering the dynamic performance of the turntable: For each target in the set of visible candidate targets, based on the angle between the current turntable pointing and the target's line of sight, and combined with the maximum angular velocity and maximum angular acceleration, a trapezoidal angular velocity profile model is used to calculate the turntable maneuver time, thereby determining the total time cost required for one observation. The total time cost consists of the turntable time and fixed observation overhead. Step S4: Optimal sequence generation under time budget constraints: Using the single-epoch observation time budget as the upper limit and the total time cost as the resource consumption unit, the comprehensive score of the target is evaluated iteratively to score and select the target in the set of visible candidate targets, thereby generating a multi-target observation sequence that maximizes information acquisition efficiency within a limited time. Step S5: Target imaging and angle measurement extraction based on the observation sequence: According to the multi-target observation sequence, the turntable is driven to point at the target and the camera is triggered to image. The target centroid coordinates are extracted through image processing, and the azimuth and pitch angles of the target relative to the spacecraft are calculated by combining the platform attitude information. At the same time, a measurement noise model adapted to the observation conditions is established. Step S6: Extended Kalman Filter Prediction and Update: An extended Kalman filter is used to predict the state based on the orbital dynamics model, and the state is updated using the angle measurements. Before updating, the measurement consistency is checked by the normalized squared innovation statistic. Step S7: Consistency Assessment and Closed-Loop Tuning Within the sliding time window, a normalized innovation square sequence is collected, its statistical characteristics are compared with the theoretical confidence interval, and the process noise intensity is adaptively adjusted based on the comparison results and fed back to the filtering and prediction stage in step S6. Step S8: Output and loop execution. Output the spacecraft position, velocity and covariance matrix of the current epoch, and enter the next filtering cycle, repeating steps S2 to S7.
2. The GNSS optical angle measurement autonomous orbit determination observation scheduling and consistency adaptive evaluation method for high-orbit spacecraft according to claim 1, characterized in that, The iteration in step S4 is a "candidate evaluation - greedy addition - state update" loop, specifically including: Step S41: Initialize the remaining time budget to the single-epoch observation time budget, and set the selected target set to empty; Step S42: Select all unselected targets from the set of visible candidate targets that satisfy the condition that the total time cost is less than or equal to the current remaining time budget; Step S43: Score the selected targets according to the preset scoring function, and select the target with the highest score to add to the observation sequence; Step S44: Update the remaining time budget, subtract the total time cost of the selected target, and update the current turntable direction to point to the target; Step S45: Repeat steps S42 to S44 until the remaining time budget is insufficient to observe any remaining candidate targets or all candidate targets have been selected, and output the final observation sequence.
3. The GNSS optical angle measurement autonomous orbit determination observation scheduling and consistency adaptive evaluation method for high-orbit spacecraft according to claim 2, characterized in that, The scoring function in step S43 has input parameters including at least one or a combination of observation time cost, target apparent magnitude, geometric dispersion with the selected target set, and information gain calculated based on covariance virtual update.
4. The GNSS optical angle measurement autonomous orbit determination observation scheduling and consistency adaptive evaluation method for high-orbit spacecraft according to claim 2, characterized in that, When the scoring function includes an information gain term, after each target is added, the current filter covariance matrix needs to be virtually updated based on the target's measurement matrix. The updated covariance matrix is then used for the scoring calculation in the next iteration to reflect the diminishing marginal returns of the observation.
5. The GNSS optical angle measurement autonomous orbit determination observation scheduling and consistency adaptive evaluation method for high-orbit spacecraft according to claim 2, characterized in that, The consistency assessment and closed-loop tuning in step S7 specifically include: Step S71: Establish a sliding time window of length L. After each filtering cycle, store the normalized squared innovation values of all valid measurements within that cycle into the window to form a statistical sequence. Step S72: Calculate the sample mean of the sequence within the window, and compare the sample mean with the theoretical confidence interval of the chi-square distribution for the corresponding degrees of freedom; Step S73: When the sample mean deviates from the confidence interval, calculate the consistency ratio and adjust the intensity scaling factor of the process noise according to the ratio; Step S74: The adjusted scaling factor is subjected to amplitude limiting and fed back to the prediction stage of the next filtering cycle to update the process noise matrix.
6. The GNSS optical angle measurement autonomous orbit determination observation scheduling and consistency adaptive evaluation method for high-orbit spacecraft according to claim 5, characterized in that, In step S7, under the condition of having a reference track or simulation verification, the normalized estimation error square sequence is also collected within the sliding time window, and its sample mean and theoretical confidence interval are used together as the basis for judging the consistency of the filter.
7. The GNSS optical angle measurement autonomous orbit determination observation scheduling and consistency adaptive evaluation method for high-orbit spacecraft according to claim 1, characterized in that, The orbital dynamics model in step S1 also includes one or more of the following: J2 perturbation, solar radiation pressure perturbation, and third-body gravitational disturbance between the Sun and the Moon; and the solar radiation pressure reflection coefficient is estimated synchronously as an extended state variable during the filtering process.
8. The GNSS optical angle measurement autonomous orbit determination observation scheduling and consistency adaptive evaluation method for high-orbit spacecraft according to claim 1, characterized in that, The measurement consistency check in step S6 is as follows: for each angle measurement to be processed, the normalized innovation square statistic is calculated. When it exceeds the preset chi-square distribution threshold, the measurement is determined to be an outlier and rejected, and is not used for state update.