A feature-targeted rocket recovery guidance system and method

Abstract

This invention discloses a rocket recovery guidance system and method based on feature targeting, belonging to the field of aerospace launch vehicle recovery technology. It addresses the problems of existing rocket recovery systems, such as reliance on satellite navigation, weak anti-interference capabilities, and insufficient autonomous controllability. The system includes fully airborne feature perception, feature targeting matching, guidance control, and closed-loop verification modules, completely excluding satellite navigation and external environmental feature perception modules. The feature perception module is deployed on non-ablation parts of the rocket body and consists of high-precision, high-temperature resistant sensors, collecting only the rocket's own structural and attitude characteristics. The matching module dynamically updates the feature library based on fuel consumption and deformation data from airborne aerodynamic simulation, using an improved SIFT algorithm to achieve high-speed and accurate matching. The guidance control module generates constraint control commands based on an active disturbance rejection PID algorithm, and the closed-loop verification module performs multi-node deviation verification and secondary correction. The method operates entirely without GPS access or external environmental feature input, relying solely on a single calibration using a ground reference before launch. There is no data interaction between the rocket and the ground after launch, covering the entire recovery phase. This invention achieves centimeter-level high-precision recovery without external dependence, with strong anti-interference and anti-denial capabilities, and fully autonomous closed-loop control. It is adaptable to sea and land vertical recovery scenarios for different types of rockets, and has low R&D and modification costs and high robustness.

Description

Technical Field

[0001] This invention relates to the field of aerospace launch vehicle recovery technology, specifically to a rocket recovery guidance system and method based on feature targeting. It is applicable to high-precision vertical recovery and landing closed-loop control of the first and second stages of a launch vehicle in scenarios without GNSS / satellite navigation dependence, and can adapt to the autonomous guidance requirements of the entire stage of rocket reentry into the atmosphere, subsonic gliding, and vertical landing. Background Technology

[0002] The current navigation and guidance system for rocket recovery relies primarily on satellite navigation systems such as GPS / BeiDou, supplemented by inertial navigation to achieve position, velocity, and attitude perception. This technical approach has significant technical bottlenecks and limitations in engineering applications. Satellite signals are susceptible to interference and interruption: During the rocket's reentry into the atmosphere, the plasma sheath formed on the rocket's surface will shield satellite navigation signals. During the landing phase, the rocket is susceptible to the complex electromagnetic environment on the ground and extreme weather conditions, which can lead to deterioration of navigation accuracy or even signal loss, making it impossible to meet the requirements for continuous guidance. High dependence on external infrastructure: Satellite navigation is a space-based infrastructure, which is at risk of being denied access in special environments, causing rocket recovery to lose its core navigation basis and resulting in insufficient independent controllability; Inertial navigation has inherent defects: the position and attitude errors of pure inertial navigation will accumulate over time. Traditional solutions rely on satellite signals for real-time calibration. Without external calibration, the error will quickly exceed the accuracy requirements for vertical rocket recovery. Existing auxiliary navigation solutions have limitations: current vision-based and lidar-based auxiliary navigation solutions all take the external environment (landscape, landmarks, celestial bodies) as the object of perception, which is easily affected by environmental occlusion, complex background and other factors. Moreover, they have not formed a closed-loop control system with the rocket's own characteristics as the core target, and cannot achieve autonomous recovery guidance that is completely independent of external signals.

[0003] There is currently no publicly available high-precision guidance system or method that uses the rocket's own structure / attitude characteristics as the sole targeting reference, requires no satellite navigation or external environmental input, and can be adapted to all stages of rocket recovery. There is an urgent need to solve the problems of autonomous navigation, real-time attitude correction, and high-precision landing control without external signal dependence. Summary of the Invention

[0004] (a) Purpose of the invention To address the technical problems of existing rocket recovery guidance technologies, such as reliance on satellite navigation, weak anti-interference capabilities, insufficient autonomous controllability, and susceptibility of existing auxiliary navigation schemes to external environmental influences, this invention provides a rocket recovery guidance system and method based on feature targeting. Using the rocket's own structural / attitude characteristics as the sole target reference, and combining it with inertial navigation calibrated only before takeoff by a ground reference station, this system achieves high-precision guidance and control throughout the entire rocket recovery process without GPS / satellite navigation dependence, without external environmental feature input, and without any data interaction between the rocket and the ground after takeoff. This improves the autonomy, reliability, anti-interference capabilities, and environmental adaptability of rocket recovery, meeting the requirements for centimeter-level landing positioning and attitude control.

[0005] (II) Technical Solution 1. System Composition A rocket recovery guidance system based on feature targeting completely excludes the access of satellite navigation modules and external environment feature sensing modules. It includes an airborne feature sensing module, an airborne feature targeting matching module, an airborne guidance control module, and an airborne closed-loop verification module, all electrically connected in sequence. Each module is linked with the rocket's attitude control engine, main engine, inertial measurement unit, and fuel management system. Feature sensing module: Deployed in non-ablation-critical parts of the rocket body (grid fin mounting end, tail fin fixed section, and propellant tank top end), it consists of a high-temperature resistant millimeter-wave radar sensor, a high-dynamic inertial measurement unit, and a laser profile sensor. The millimeter-wave radar has a range resolution ≤0.1cm and an angular resolution ≤0.01°. The laser profile sensor has a profile acquisition accuracy ≤0.05cm, a sampling frequency ≥1kHz, a vibration resistance level ≥20g, and an operating temperature range of -50℃ to 800℃. It is used to collect real-time structural features of the rocket itself (grid fin profile / deflection angle, tail fin profile / deployment state, and propellant tank feature profile) and attitude features (rocket body pitch angle, yaw angle, roll angle, and angular velocity). The acquired signals are only for the rocket's own components and do not include any external environmental background information. Feature-targeted matching module: It has a built-in standard feature library and dynamic update unit for the entire rocket flight phase. The standard feature library pre-stores the standard structural feature parameters, attitude feature reference ranges and feature association models of the rocket in each phase of takeoff, reentry, subsonic, and landing. The dynamic update unit corrects the standard feature library parameters in real time based on the mass change of rocket fuel consumption and the aerodynamic deformation data of the rocket body generated in real time by the rocket airborne aerodynamic simulation module based on real-time flight speed, altitude and angle of attack. This module adopts an improved SIFT key point matching algorithm combined with contour feature matching, with a matching calculation delay of ≤10ms. It accurately matches the real-time features collected by the feature perception module with the corrected standard features and outputs the current position deviation of the rocket body (three-dimensional spatial coordinate deviation), attitude deviation (pitch / yaw / roll angle deviation, angular velocity deviation) and deviation change rate. The guidance and control module has a built-in deviation-control command mapping model (based on the active disturbance rejection PID control algorithm). It presets the control constraints of the rocket attitude control engine and the main engine (maximum deflection angle of the grid fin ±30°, main engine thrust adjustment rate 0~5% / ms, and attitude control engine jet angle ±15°). Based on the deviation and deviation change rate output by the feature target matching module, combined with the real-time mass and aerodynamic parameters of the rocket, it generates attitude adjustment commands (grid fin deflection command, attitude control engine jet angle / duration command) and thrust control commands (main engine thrust magnitude, start / stop timing command) to drive the rocket to execute trajectory and attitude correction. Closed-loop verification module: It has built-in landing accuracy thresholds (positioning deviation ≤ 5cm, attitude deviation ≤ 0.5°) and a secondary adjustment trigger mechanism. At key nodes such as rocket reentry phase, subsonic phase, and 1000m / 100m / 10m before landing, it collects feature data from the feature perception module and execution data from the guidance and control module in real time to verify the deviation between the actual trajectory / attitude of the rocket and the target trajectory / attitude. If the deviation exceeds the threshold, a secondary correction command is immediately triggered. If the deviation still exceeds the threshold 10m before landing, an emergency landing procedure is initiated. This module also links with the rocket fuel management system to incorporate the remaining fuel amount into the calculation of correction commands.

[0006] 2. Methods and Steps A feature-targeted rocket recovery guidance method, with no GPS / satellite navigation signal access, no external environmental feature input, and no data interaction between the rocket and the ground after liftoff, includes the following steps: Benchmark calibration and feature library construction: Before rocket launch, the inertial measurement unit is initially zero-calibrated using a ground reference station. The ground reference station is only used for initial calibration before launch and there is no data interaction throughout the process. A standard feature library for the entire flight phase of the rocket is pre-established, including standard structural feature parameters of grid fins, tail fins, and propellant tanks for each phase, reference ranges for pitch / yaw / roll angles, and feature-attitude correlation models under different fuel masses. At the same time, dynamic update rules for the feature library, guidance and control constraints, closed-loop verification thresholds, and secondary adjustment triggering mechanisms are preset. Real-time feature acquisition: From liftoff to landing, the rocket’s own structural and attitude features are continuously acquired through the feature perception module. The acquired signals are only for the rocket’s own components, filtering out all external environmental background information. During the reentry into the atmospheric plasma sheath stage, features are acquired without obstruction through millimeter-wave radar sensors. Dynamic updates to the feature library: Based on the fuel consumption data transmitted in real time by the rocket fuel management system, the real-time mass change of the rocket body is calculated. Combined with the aerodynamic deformation prediction data of the rocket body generated in real time by the rocket airborne aerodynamic simulation module based on real-time flight speed, altitude, and angle of attack, the feature parameters and associated models in the standard feature library are dynamically corrected to ensure that the benchmark features match the actual state of the rocket. Feature-targeted precise matching: The real-time features collected in step 2 are compared with the standard features corrected in step 3 by combining the improved SIFT key point matching algorithm with contour feature matching. The current three-dimensional spatial position deviation, pitch / yaw / roll angle deviation, angular velocity deviation and the rate of change of each deviation are calculated. The matching calculation delay is ≤10ms. Guided control command generation: Based on the deviation and deviation change rate in step 4, the deviation-control command mapping model of the active disturbance rejection PID control algorithm is used to generate attitude adjustment commands (grid fin deflection angle, attitude control engine jet parameters) and thrust control commands (main engine thrust magnitude and timing) by combining the rocket's real-time mass, aerodynamic parameters and preset control constraints. Trajectory and attitude correction execution: Guidance and control commands are sent to the rocket attitude control engine, main engine, and grid fin actuator to drive the rocket body to complete attitude adjustment and trajectory correction, and achieve real-time deviation compensation; Multi-node closed-loop verification and secondary correction: At key nodes such as rocket reentry phase, subsonic phase, and 1000m / 100m / 10m before landing, the deviation between the actual trajectory / attitude and the target value is verified through the closed-loop verification module. If the deviation exceeds the preset threshold, steps 2-6 are repeated to trigger secondary correction; if the deviation meets the threshold requirement 10m before landing, the command is maintained; if it exceeds the threshold, the emergency landing procedure is initiated. High-precision landing completed: At the moment of rocket touchdown, the feature perception module collects the final landing attitude and position data, and the closed-loop verification module records the final deviation, completing the high-precision recovery guidance closed loop of the rocket without satellite navigation dependence.

[0007] (III) Beneficial Effects Completely autonomous and without external dependence: The rocket's own structure / attitude characteristics serve as the sole target benchmark. There is no GPS / satellite navigation signal access or external environmental characteristic input throughout the entire process. The ground reference station only completes the initial calibration before takeoff. After the rocket takes off, there is no data interaction with the ground end. It completely eliminates dependence on space-based and ground-based external infrastructure, significantly improves anti-interference and anti-denial capabilities, and adapts to the rocket recovery needs in extreme environments. Continuous and reliable guidance throughout all stages: The feature perception module adopts a combination of high-temperature and vibration resistant millimeter-wave radar, laser profile sensor and high dynamic inertial measurement unit, which clearly defines the core detection accuracy of the sensor. It can realize the unobstructed acquisition of rocket's own features in scenarios such as plasma sheath, complex electromagnetic, extreme weather, etc., covering the entire stage of rocket reentry, gliding and landing, without guidance scene gaps; It combines high precision and high real-time performance: It adopts an improved SIFT key point matching algorithm combined with contour feature matching, with a matching delay of ≤10ms. Combined with the active disturbance rejection PID control algorithm, it achieves centimeter-level landing accuracy with positioning deviation ≤5cm and attitude deviation ≤0.5°, which meets the high-precision control requirements of rocket vertical recovery. Moreover, the deviation correction response speed is fast and can compensate for dynamic deviations during high-speed rocket flight. Strong robustness and wide adaptability: It sets up a feature library dynamic update unit, corrects the benchmark features based on rocket fuel consumption and aerodynamic deformation data calculated in real time by the airborne aerodynamic simulation module, and incorporates rocket body mass and remaining fuel into the control command calculation, effectively compensating for changes in the rocket's state during flight. It can be adapted to different models and different loads of launch vehicle recovery scenarios. Deployment can be completed by updating the standard feature library only, with low R&D and modification costs. Closed-loop controllability and high safety: The design incorporates a closed-loop verification mechanism and secondary correction triggering rules for multiple key nodes, combined with detailed emergency landing procedures, to achieve real-time monitoring, correction, and safety control of deviations throughout the entire rocket recovery process. This effectively avoids recovery failures caused by single-stage malfunctions, thereby improving the safety and success rate of rocket recovery. Detailed Implementation

[0008] The following uses the vertical recovery of the first stage of a medium-sized launch vehicle as an example to describe the technical solution of the present invention in detail. The test conditions of this embodiment are: the rocket takeoff mass is 300t, the target landing point of the first stage recovery is a marine recovery platform, the preset landing positioning deviation is ≤5cm, the attitude deviation is ≤0.5°, there is no satellite navigation signal access throughout the process, and there is no data interaction between the rocket and the ground after takeoff.

[0009] Example 1: Vertical Recovery Guidance of Rocket First Stage Benchmark Calibration and Feature Library Construction: Before rocket launch, the inertial measurement unit of the rocket body is initially zero-calibrated using a marine ground reference station. The ground reference station only completes this calibration, and there is no data exchange after rocket launch. A standard feature library for this type of rocket is pre-established, including the standard profile parameters of the grid fins / tail fins / propellant tanks for the launch phase, reentry phase (Mach 5~10), subsonic phase (Mach < 1), and landing phase; the reference ranges for pitch / yaw / roll angles (pitch angle 0°~15°, yaw angle ±5°, roll angle ±3°); and the feature-attitude correlation model for fuel mass from 150t to 0t. The maximum grid fin deflection angle is preset to ±30°, the main engine thrust adjustment rate is 0~5% / ms, and the closed-loop verification thresholds are positioning deviation ≤5cm and attitude deviation ≤0.5°. The critical verification nodes are 1000m / 100m / 10m before landing. Real-time feature acquisition: After rocket liftoff, high-temperature resistant millimeter-wave radar sensors (range resolution ≤0.1cm, angular resolution ≤0.01°), laser profile sensors (profile acquisition accuracy ≤0.05cm), and high-dynamic inertial measurement units (sampling frequency 1kHz) deployed at the grid fin mounting end and tail fin fixed section continuously acquire grid fin profile / deflection angle, tail fin deployment status, rocket body pitch / yaw / roll angle and angular velocity, filtering external sky and ocean background information; during atmospheric reentry, the plasma sheath shields optical signals, enabling unobstructed continuous feature acquisition through millimeter-wave radar sensors; Feature library dynamic updates: The rocket fuel management system transmits fuel consumption data in real time. As the rocket body mass gradually decreases from 300t to 150t, the feature target matching module, based on the mass change data and combined with the aerodynamic deformation data calculated by the airborne aerodynamic simulation module based on real-time flight speed, altitude, and angle of attack, corrects the tank contour parameters and feature-attitude correlation model in the standard feature library in real time to adapt to the changes in rocket body state. Feature-targeted precise matching: The real-time collected grid fin, tail fin contour and attitude data are matched with the corrected standard features using improved SIFT key point + contour feature matching. The matching delay is 8ms. The calculation results show that the yaw angle deviation of the rocket body in the reentry phase is 1.5° and the position deviation is 80m, while the pitch angle deviation in the subsonic phase is 0.8° and the position deviation is 10m. Guidance control command generation: Based on the deviation data, and using the active disturbance rejection PID control model combined with the real-time mass of the rocket body (180t), guidance control commands are generated as follows: grid rudder deflection +8°, ​​attitude control engine jet angle +5°, and main engine thrust increase by 3%. Trajectory and attitude correction execution: After receiving the command, the actuator drives the grid rudder to deflect, the attitude control engine jet to adjust the thrust of the main engine, and the yaw angle, pitch angle and position deviation of the rocket body are quickly compensated. The deviation in the subsonic range is reduced to 0.3° yaw angle, 0.2° pitch angle and 3m position deviation. Multi-node closed-loop verification and secondary correction: At the 1000m / 100m nodes before landing, the verification deviations of the closed-loop verification module all meet the threshold requirements, and no secondary correction is required; at the 10m node before landing, the positioning deviation is verified to be 3cm and the attitude deviation is 0.2°, which meets the high-precision guidance requirements and maintains command execution. High-precision landing completed: The rocket's first stage vertically touched down on the sea recovery platform, with a final landing positioning deviation of 4cm and an attitude deviation of 0.3°, meeting the preset accuracy requirements. The closed-loop verification module recorded the deviation data throughout the process, completing the recovery guidance.

[0010] This embodiment verifies that the system and method of the present invention can achieve high-precision vertical recovery of the first stage of a rocket without satellite navigation dependence and without data interaction between the rocket and the ground after takeoff. The entire process guidance is continuous and reliable, and the deviation correction response is fast, meeting the needs of engineering applications.

Claims

1. A rocket recovery guidance system based on feature targeting, characterized in that, Completely exclude the integration of satellite navigation modules and external environment feature perception modules, including airborne feature perception modules, airborne feature target matching modules, airborne guidance and control modules, and airborne closed-loop verification modules that are sequentially electrically connected to the rocket attitude control engine, main engine, inertial measurement unit, and fuel management system, among which: The feature perception module, deployed in the non-ablation-critical parts of the rocket body, consists of a high-temperature resistant millimeter-wave radar sensor, a high-dynamic inertial measurement unit, and a laser profile sensor. The millimeter-wave radar has a range resolution of ≤0.1cm and an angular resolution of ≤0.01°. The laser profile sensor has a profile acquisition accuracy of ≤0.05cm and a sampling frequency of ≥1kHz. It is used to acquire the rocket's own structural and attitude features in real time, and the acquired signals are only for the rocket's own components and do not contain any external environmental background information. The feature-targeted matching module has a built-in standard feature library and dynamic update unit for the entire flight phase of the rocket. The correction basis of the dynamic update unit includes the mass change of rocket fuel consumption and the predicted aerodynamic deformation data of the rocket body generated in real time by the rocket airborne aerodynamic simulation module based on real-time flight speed, altitude and angle of attack. This module adopts an improved SIFT key point matching algorithm combined with contour feature matching, with a matching calculation delay of ≤10ms. It is used to match the real-time acquired features with the dynamically corrected standard features and output the rocket body position deviation, attitude deviation and deviation change rate. The guidance control module has a built-in deviation-control command mapping model based on the active disturbance rejection PID control algorithm, and presets the control constraints of the engine and actuators. It is used to generate attitude adjustment commands and thrust control commands based on the deviation and the rate of change of deviation. The closed-loop verification module has a built-in landing accuracy threshold and a multi-node secondary adjustment triggering mechanism. It is used to verify deviations at key nodes of rocket recovery. If the deviation exceeds the threshold, a secondary correction is triggered. It is also linked with the rocket fuel management system.

2. A rocket recovery guidance method based on feature targeting, characterized in that, The entire process involves no GPS / satellite navigation signal access, no external environmental feature input, and no data exchange between the rocket and the ground after liftoff, including the following steps: 1) Before the rocket takes off, the inertial measurement unit is initially zeroed through the ground reference station. The ground reference station is only used for the initial calibration before takeoff and there is no data interaction throughout the process. A standard feature library for the entire flight phase of the rocket is pre-established, and the feature library dynamic update rules, guidance and control constraints, closed-loop verification thresholds and secondary adjustment triggering mechanisms are preset. 2) From liftoff to landing, the rocket’s structural and attitude features are continuously collected by the feature sensing module. The collected signals are only for the rocket’s own components, filtering out all external environmental background information. During the rocket’s reentry into the atmospheric plasma sheath phase, features are collected without obstruction by millimeter-wave radar sensors. 3) Based on the real-time fuel consumption data of the rocket fuel management system, and combined with the predicted aerodynamic deformation data of the rocket body generated in real time by the rocket airborne aerodynamic simulation module based on real-time flight speed, altitude, and angle of attack, the feature parameters and associated models in the standard feature library are dynamically corrected. 4) Accurately match the real-time acquired features with the dynamically corrected standard features, calculate the current position deviation, attitude deviation and the rate of change of each deviation of the rocket body, and the matching calculation delay is ≤10ms; 5) Based on the deviation and the rate of change of deviation, combined with the real-time mass of the rocket, aerodynamic parameters and preset control constraints, generate attitude adjustment commands and thrust control commands; 6) Send control commands to the rocket actuators to drive the rocket body to complete trajectory and attitude corrections; 7) Check the deviations at key nodes 1000m / 100m / 10m before landing during the rocket reentry phase, subsonic phase, and landing phase. If the deviation exceeds the threshold, repeat steps 2-6 to trigger a secondary correction. If the deviation meets the standard 10m before landing, keep the command executed; otherwise, start the emergency landing procedure. 8) After the rocket touches the ground, it collects the final landing data to complete the high-precision recovery guidance closed loop.

3. The system according to claim 1, characterized in that, The feature sensing module has a vibration resistance level of ≥20g, an operating temperature range of -50℃ to 800℃, and is deployed at least one of the following locations: the rocket grid fin mounting end, the tail fin fixing section, and the upper end cap of the propellant tank.

4. The system according to claim 1, characterized in that, The structural features include at least one of the following: grid rudder profile / deflection angle, tail fin profile / deployment state, and tank feature profile; the attitude features include at least one of the following: rocket body pitch angle, yaw angle, roll angle, and angular velocity.

5. The system according to claim 1, characterized in that, The control constraints include at least one of the following: maximum grid rudder deflection angle ±30°, main engine thrust adjustment rate 0~5% / ms, and attitude control engine jet angle ±15°.

6. The system according to claim 1, characterized in that, The landing accuracy threshold of the closed-loop verification module is a positioning deviation ≤ 5cm and an attitude deviation ≤ 0.5°. The secondary adjustment trigger nodes include at least one of the following: rocket reentry phase, subsonic phase, and 1000m / 100m / 10m before landing.

7. The method according to claim 2, characterized in that, The precise matching described in step 4 is an improved SIFT keypoint matching algorithm combined with contour feature matching.

8. The method according to claim 2, characterized in that, The emergency landing procedure described in step 7 includes at least one of the following: the attitude control engine maintaining the vertical attitude of the rocket body at full power, the main engine thrust being reduced to a safe threshold, the power supply to non-critical auxiliary systems of the rocket body being cut off, and the grid fins being returned to zero deflection angle.

9. The method according to claim 2, characterized in that, The attitude adjustment commands in step 5 include grid rudder deflection angle commands and attitude control engine jet parameter commands, and the thrust control commands include main engine thrust magnitude commands and main engine timing commands.

10. The system according to claim 1 or the method according to claim 2, characterized in that, The rocket is a first or second stage launch vehicle, suitable for vertical recovery scenarios such as offshore recovery platforms and land-based recovery sites.