A federated filtering based multi-redundancy guided information fusion method

Abstract

The application belongs to the technical field of sensor information fusion, and relates to a multi-redundancy guidance information fusion method based on federated filtering. Firstly, measurement data of an airborne inertial navigation system is collected and input into an inertial navigation information solving unit to solve aircraft parameters. Then, measurement data of radar, photoelectric and differential satellite guidance sensors is collected and input into respective guidance information solving units to solve guidance information. The guidance information of the radar, photoelectric and differential satellite guidance sensors is respectively solved in a loosely coupled manner with the inertial navigation information, and a Kalman filter is used to estimate errors of the inertial navigation system to establish a combined guidance model and form a filtering system of respective sub-filters. After time registration and time updating of measurement information, global optimal cooperation is realized. The method can meet the accuracy requirement of mobile target registration. In view of the complex motion model of a mobile target in practice, the method can also achieve good matching effect and fully guarantee the guidance information fusion accuracy.

Description

Technical Field

[0001] This application belongs to the field of sensor information fusion technology, and specifically relates to a redundancy-guided information fusion method based on federated filtering. Background Technology

[0002] Multi-sensor information fusion is a prerequisite for achieving redundant guidance. Relying solely on a single sensor for guidance is susceptible to various factors such as weather, distance, and signal transmission environment, leading to varying degrees of inapplicability and compromised guidance accuracy. For example, photoelectric and radar systems have strong anti-jamming capabilities and are unaffected by weather conditions, but their accuracy is low and significantly affected by distance, making them suitable for short-range guidance. Differential satellite guidance, on the other hand, is unaffected by weather conditions, offers the highest accuracy, and is unaffected by distance, but is easily interfered with, making it suitable for long-range guidance. Therefore, by simultaneously using multiple sensors for guidance and employing redundant guidance information fusion technology based on federated filtering, the stability and reliability of the guidance process can be effectively achieved.

[0003] Currently used information fusion methods have significant limitations, leading to the accumulation of guidance errors. This is mainly due to the following drawbacks:

[0004] a. The timestamps of each sensor were not calibrated. Due to the different installation locations, communication links, and algorithm processing flows of each sensor, the guidance information from each sensor arrived at the fusion center at different times, which had a significant impact on the accuracy of the guidance information fusion results.

[0005] b. The sensor's guidance information was not preprocessed, and interference from outliers was not effectively detected and eliminated. At the same time, there was a lack of handling for occasional sensor failures, resulting in impaired fusion accuracy.

[0006] c. Using a tight combination method, due to the low accuracy of inertial navigation equipment, the error of strapdown calculation accumulates very quickly. When the combined sensor signal fails for a short time, the error of strapdown calculation will accumulate rapidly, resulting in combination failure. At the same time, a lot of calculation and delay compensation are required, which reduces the performance of real-time guidance.

[0007] Therefore, how to efficiently fuse information from different sensors is a problem that needs to be solved. Summary of the Invention

[0008] The purpose of this application is to provide a redundant guided information fusion method based on federated filtering to solve the problem that commonly used information fusion methods have strong limitations and can lead to the accumulation of guidance errors.

[0009] The technical solution of this application is: a redundancy-guided information fusion method based on federated filtering, comprising:

[0010] The measurement data of the airborne inertial navigation system is collected and input into the inertial navigation information calculation unit. The position, velocity and attitude of the aircraft are calculated through inertial navigation mechanics programming to obtain inertial navigation information.

[0011] The measurement data collected from radar, photoelectric, and differential satellite guidance sensors are input into each guidance information calculation unit and the guidance information is calculated to obtain the aircraft's position, velocity, and attitude information;

[0012] The guidance information from radar, photoelectric, and differential satellite navigation sensors is loosely coupled with the inertial navigation information for guidance calculation. Kalman filters are used to estimate various errors of the inertial navigation, and a combined guidance model is established to form a filtering system for each sub-filter.

[0013] The global optimal estimation information is allocated among the sub-filters based on the principle of information conservation.

[0014] Each sub-filter acquires the time information of the main filter and updates the time accordingly;

[0015] The filter system measurement information of each sub-filter is updated using the same time point;

[0016] The output information of each sub-filter is sent to the main filter. The main filter performs a weighted average of the output information of each sub-filter to obtain the final error estimate. Then, the obtained error estimate is input into each sub-filter and iteratively solved with the inertial navigation measurement data, and the error estimate in each sub-filter is repeatedly updated.

[0017] Preferably, the inertial navigation information calculation unit is equipped with an inertial navigation health status monitoring module. The inertial navigation health status monitoring module sets different first health status thresholds for different data. The inertial navigation health monitoring module compares the collected measurement data with the corresponding first health status threshold. If the first health status threshold is exceeded, the inertial navigation information calculation unit is judged to be faulty, and the operation of the inertial navigation information calculation unit is stopped.

[0018] Preferably, the guidance information calculation module includes a device health status monitoring module, which sets a second health status threshold for each sensor. The device health status monitoring module compares the guidance information collected by each sensor with the corresponding second health status threshold. If the threshold is exceeded, it indicates that the corresponding sensor is faulty. If all sensors are faulty or not connected to the Internet, the operation of the guidance information calculation module is stopped.

[0019] Preferably, the time update method is as follows: receiving the timestamp information of the main filter, and then collecting the timestamp information of each sensor and comparing it with the timestamp information of the main filter. If the timestamp of the current sensor is inconsistent with the timestamp of the main filter, the timestamp of the corresponding sensor is compensated according to the difference. If they are consistent, the timestamp of the main filter is updated.

[0020] Preferably, each sub-filter is equipped with an outlier detection module. After updating the measurement information of each sub-filter, the innovation value obtained by Kalman filtering, i.e., the measurement value and the one-step prediction value, is used as the statistical innovation. The residual characteristic detection method and residual outlier are used to detect faults and outliers in the measurement information of each sub-filter. If a fault exists, the corresponding sensor is determined to be abnormal, and the sensor is handled accordingly. If an outlier exists, the outlier data is removed.

[0021] This application presents a redundant guidance information fusion method based on federated filtering. First, it collects measurement data from the airborne inertial navigation system (INS) and inputs it into the INS information calculation unit to calculate aircraft parameters. Then, it collects measurement data from radar, electro-optical, and differential satellite navigation sensors and inputs them into their respective guidance information calculation units for guidance information calculation. The guidance information from the radar, electro-optical, and differential satellite navigation sensors is loosely coupled with the INS information for guidance calculation. A Kalman filter is used to estimate various errors in the INS, establishing a combined guidance model and forming sub-filter systems. After time registration and updating the measurement information over time, global optimal matching is achieved. This method can meet the accuracy requirements for maneuvering target registration. It also achieves good matching results for complex motion models of maneuvering targets in reality, fully ensuring the accuracy of guidance information fusion. It is applicable to the entire redundant guidance process and will not lead to local divergence. Attached Figure Description

[0022] To more clearly illustrate the technical solutions provided in this application, the accompanying drawings will be briefly described below. Obviously, the drawings described below are merely some embodiments of this application.

[0023] Figure 1 This is a schematic diagram of the overall process of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be described in more detail below with reference to the accompanying drawings.

[0025] A redundancy-guided information fusion method based on federated filtering, such as Figure 1 As shown, it includes the following steps:

[0026] Step S100: Collect the measurement data of the airborne inertial navigation system and input it into the inertial navigation information calculation unit. Calculate the body position, velocity, and attitude through inertial navigation mechanics programming to obtain inertial navigation information.

[0027] Preferably, the inertial navigation information processing unit includes an inertial navigation health status monitoring module. This module sets different first health status thresholds for different data types. The module compares the collected measurement data with the corresponding first health status threshold. If the threshold is exceeded, the inertial navigation information processing unit is deemed faulty and its operation is terminated. This allows for timely detection and handling of abnormal inertial navigation data. The first health status threshold is the standard value for the corresponding data.

[0028] Step S200: Collect measurement data from radar, photoelectric, and differential satellite guidance sensors, input them into each guidance information calculation unit, and perform guidance information calculation to obtain the aircraft's position, velocity, and attitude information;

[0029] Preferably, the guidance information calculation module includes a device health status monitoring module. This module sets a second health status threshold for each sensor. The monitoring module compares the guidance information collected by each sensor with the corresponding second health status threshold. If the threshold is exceeded, it indicates a sensor malfunction. If all sensors are malfunctioning or not connected to the internet, the operation of the guidance information calculation module is terminated. This allows for timely detection and handling of sensor malfunctions. The second health status threshold serves as a standard value for the corresponding data.

[0030] In step S300, the guidance information from radar, electro-optical, and differential satellite navigation sensors is loosely coupled with the inertial navigation information for guidance calculation. A Kalman filter is used to estimate various errors of the inertial navigation (position, velocity, attitude, angular rate, and acceleration) to establish a combined guidance model and form a sub-filter system. Each sub-filter system can estimate the motion state of the aircraft and can adapt to the complex motion model during the aircraft landing process.

[0031] The sub-filters are GPS / INS filter, radar / INS filter, and electro-optical / INS filter, with the radar / INS filter containing different errors. Therefore, the estimates of aircraft position, velocity, and attitude obtained after calculation using the Kalman filter will also differ. Setting different sub-filters also helps avoid data cross-contamination.

[0032] Step S400: The global optimal estimation information is allocated among the sub-filters according to the principle of information conservation; that is, when the error ratio between a certain data on the sensor and the measurement data corresponding to the inertial navigation is the smallest, the error ratio is allocated to other data collected by the sensor to improve data accuracy.

[0033] In step S500, each sub-filter acquires the time information of the main filter and updates the time accordingly;

[0034] Preferably, the time update method is as follows: receiving the timestamp information of the main filter, and then collecting the timestamp information of each sensor and comparing it with the timestamp information of the main filter. If the timestamp of the current sensor is inconsistent with the timestamp of the main filter, the timestamp of the corresponding sensor is compensated according to the difference; if they are consistent, the timestamp of the main filter is updated. This allows for time registration of the measurement values ​​collected by each sensor.

[0035] Step S600: Update the filter system measurement information of each sub-filter using the same time node;

[0036] Preferably, each sub-filter is equipped with an outlier detection module. After updating the measurement information of each sub-filter, the innovation value obtained through Kalman filtering, i.e., the measured value and the one-step prediction value, is used as the statistical innovation. The residual characteristic detection method and residual outlier are used to detect faults and outliers in the measurement information of each sub-filter. If a fault is found, the corresponding sensor is determined to be abnormal, and appropriate special handling is performed on the sensor. If an outlier is found, the outlier data is removed. This avoids outlier data from entering the fusion center and affecting the information fusion accuracy, and can solve the problem of discontinuities and divergences in the data collected from the sensors.

[0037] In step S700, the output information of each sub-filter is sent to the main filter. The main filter performs a weighted average of the output information of each sub-filter to obtain the final error estimate, thereby achieving global optimal coordination.

[0038] The obtained error estimate is then input into each sub-filter and iteratively calculated with the inertial navigation measurement data, repeatedly updating the error estimate in each sub-filter. Through weighted averaging, a relatively accurate error estimate can be obtained. This error estimate is then sent to each filter, and newly acquired data is calculated according to the new error estimate. Through repeated iterations, the accuracy of error calculation can be continuously improved.

[0039] This application first collects measurement data from the airborne inertial navigation system (INS) and inputs it into the INS information calculation unit to calculate aircraft parameters. Then, it collects measurement data from radar, electro-optical, and differential satellite navigation sensors and inputs them into each guidance information calculation unit for guidance information calculation. The guidance information from the radar, electro-optical, and differential satellite navigation sensors is loosely coupled with the INS information for guidance calculation. A Kalman filter is used to estimate various errors in the INS, establishing a combined guidance model and forming sub-filter systems. After time registration and updating the measurement information over time, global optimal matching is achieved. This can meet the accuracy requirements for maneuvering target registration. It also achieves good matching results for complex motion models of maneuvering targets in reality, fully ensuring the accuracy of guidance information fusion. It is applicable to the entire process of redundant guidance and will not lead to local divergence.

[0040] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A redundancy-guided information fusion method based on federated filtering, characterized in that, include: The measurement data of the airborne inertial navigation system is collected and input into the inertial navigation information calculation unit. The position, velocity and attitude of the aircraft are calculated through inertial navigation mechanics programming to obtain inertial navigation information. The measurement data collected from radar, photoelectric, and differential satellite guidance sensors are input into each guidance information calculation unit and the guidance information is calculated to obtain the aircraft's position, velocity, and attitude information; The guidance information from radar, photoelectric, and differential satellite navigation sensors is loosely coupled with the inertial navigation information for guidance calculation. Kalman filters are used to estimate various errors of the inertial navigation, and a combined guidance model is established to form a filtering system for each sub-filter. The global optimal estimation information is allocated among the sub-filters based on the principle of information conservation. Each sub-filter acquires the time information of the main filter and updates the time accordingly; The filter system measurement information of each sub-filter is updated using the same time point; The output information of each sub-filter is sent to the main filter. The main filter performs a weighted average of the output information of each sub-filter to obtain the final error estimate. Then, the obtained error estimate is input into each sub-filter and iteratively solved with the inertial navigation measurement data, and the error estimate in each sub-filter is repeatedly updated.

2. The redundancy-guided information fusion prevention method based on federated filtering as described in claim 1, characterized in that: The inertial navigation information calculation unit is equipped with an inertial navigation health status monitoring module. The inertial navigation health status monitoring module sets different first health status thresholds for different data. The inertial navigation health monitoring module compares the collected measurement data with the corresponding first health status threshold. If the first health status threshold is exceeded, the inertial navigation information calculation unit is judged to be faulty and the operation of the inertial navigation information calculation unit is stopped.

3. The redundancy-guided information fusion prevention method based on federated filtering as described in claim 1, characterized in that: The guidance information calculation unit is equipped with a device health status monitoring module. The device health status monitoring module is equipped with a second health status threshold corresponding to each sensor. The device health status monitoring module compares the guidance information collected by each sensor with the corresponding second health status threshold. If the threshold is exceeded, it indicates that the corresponding sensor is faulty. If all sensors are faulty or not connected to the Internet, the operation of the guidance information calculation unit is stopped.

4. The redundancy guidance information fusion prevention method based on federated filtering as described in claim 1, characterized in that, The time update method is as follows: receive the timestamp information of the main filter, and then collect the timestamp information of each sensor and compare it with the timestamp information of the main filter. If the timestamp of the current sensor is inconsistent with the timestamp of the main filter, the timestamp of the corresponding sensor is compensated according to the difference. If they are consistent, the timestamp of the main filter is updated.

5. The redundancy guidance information fusion prevention method based on federated filtering as described in claim 1, characterized in that: Each sub-filter is equipped with an outlier detection module. After updating the measurement information of each sub-filter, the new information obtained by Kalman filtering, namely the measurement value and the one-step prediction value, is used as the statistical new information. The residual characteristic detection method and residual outlier are used to detect faults and outliers in the measurement information of each sub-filter. If a fault exists, the corresponding sensor is judged to be abnormal, and the sensor is handled accordingly. If outliers exist, remove those outlier data.

Images