Multi-sensor data fusion and online reconstruction method based on planetary topology structure

Abstract

The application discloses a kind of multi-sensor data fusion and online reconfiguration method based on planetary topology structure, including the following steps, construct three-layer multi-sensor data fusion framework based on planetary topology structure, with sensor as outer layer satellite, the subsystem of two two combinations of sensor as middle layer planet, inner layer is main filter;Two-stage fault detection unit is set in satellite and planet layer, sensor fault information is provided by sensor, the effective probability of subsystem is obtained by substituting chi-square test value into fuzzy membership function by subsystem;Filter system reconfiguration table is designed, when fault occurs, the influence brought by fault is reduced online when system reconfiguration is carried out to multi-sensor data fusion system;Finally, global state estimation is completed in inner layer main filter based on the fault information of sensor and subsystem.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of signal and information processing, and particularly relates to a multi-sensor data fusion and online reconstruction method based on a planetary topology. BACKGROUND

[0002] Multi-sensor data fusion technology is an information processing technology that uses computer technology to automatically detect, analyze, combine and estimate measurement information, and can obtain comprehensive information that a single sensor cannot obtain. Data fusion is proposed to meet the multi-source correlation requirements of C 3 I (Command, Control, Communication and Intelligent) military systems, and has become a research hotspot. The purpose of multi-sensor data fusion is to obtain more information to achieve optimal estimation of the system state and improve estimation accuracy. However, as the number of sensors increases, the probability of failure inevitably increases, and false information will pollute the global state estimation and affect the estimation accuracy. Therefore, it is required to build an efficient and flexible data fusion structure, design a fault detection unit, and cooperate with an online fault-tolerant algorithm to effectively isolate fault information and ensure the accuracy of the final global state estimation.

[0003] In order to effectively realize multi-sensor data fusion, domestic and foreign scholars have carried out some research. According to the fusion structure, it is currently divided into centralized fusion structure and distributed fusion structure. The centralized structure lacks fault tolerance capability, and false data can easily cause fusion failure. The distributed fusion structure is mainly a tower filter structure, in which the sensor data is combined and preprocessed by a sub-filter, and then the global state estimation is realized in the main filter. However, this method has a layer-by-layer data progression, and the fault-tolerant algorithm is too dependent on the fault detection module, which limits the flexibility of the filter and the efficiency of the utilization of sensor data. SUMMARY

[0004] The purpose of the present application is to provide a multi-sensor data fusion and online reconstruction method based on a planetary topology.

[0005] The technical solution of the present application is as follows:

[0006] A multi-sensor data fusion and online reconstruction method based on a planetary topology, comprising the following steps:

[0007] Step 1, constructing a three-layer multi-sensor data fusion framework based on a planetary topology, in which each sensor forms an outer satellite ring, each sub-filter system forms a middle planetary ring, and the inner ring is a main filter;

[0008] Step 2, design a two-stage fault detection module, the first stage is the sensor fault information extraction of the satellite ring, and the sensor fault value is obtained; the second stage is the state chi-square detection of the planetary ring subsystem filter system, and the subsystem effective probability is obtained;

[0009] Step 3, design a fault reconstruction table for the data fusion system to realize fault-tolerant reconstruction of the planetary structure fault state system;

[0010] Step 4, based on the fault reconstruction table, combined with the sensor fault value and the subsystem effective probability, complete the global state data fusion in the main filter.

[0011] Specifically, the three-layer multi-sensor data fusion framework based on the planetary topology structure constructed in step 1 specifically includes:

[0012] The design goal of the multi-sensor data fusion system is to improve the performance and robustness of the system through multi-source information fusion, and to overcome the defects and vulnerability of the single system. The data fusion framework is the basic skeleton supporting the multi-sensor data fusion system. Compared with the traditional tower federal filter structure, this patent proposes a planetary fault-tolerant filter structure. The planetary topology structure is composed of 3 layers, the outermost satellite ring is composed of each sensor, the middle layer planetary layer is composed of each sensor combined into a subsystem, and the inner layer is the main filter to realize global system state estimation.

[0013] The fault-tolerant design idea of the planetary topology structure is to divide the fault level into two categories: sensor level and sub-filter level, use corresponding fault detection means to monitor its running state, and formulate a system reconstruction mechanism table for different faults to realize system reconstruction under different fault states.

[0014] The improvement of the traditional tower structure is:

[0015] 1) Each layer of the planetary structure has a direct channel to the main filter;

[0016] 2) The "satellite" and "planet" on each layer of the planetary structure are provided with fault detection units;

[0017] 3) The main filter of the planetary structure uses a fault reconstruction table for system reconstruction under fault state.

[0018] Compared with the traditional structure, this design has better timeliness, robustness and flexibility, which is specifically reflected in:

[0019] 1) Multi-type sensor faults can be directly detected at the sensor level without the need for detection at the sub-filter level, and the planetary structure has better timeliness for such faults;

[0020] 2) The sensor data alone can still be used to achieve state estimation. For a three-element system, the traditional structure can only cope with single filter failure, while the planetary structure can cope with double sensor failure;

[0021] 3) The planetary structure can fully consider the working state of the system and its dependence on each other through fault reconstruction table for system reconstruction.

[0022] Specifically, the two-stage fault detection module constructed in step 2 specifically includes:

[0023] As described in step 1, each node on each orbit of the planetary topology structure is provided with a fault detection unit. For the sensor fault information of the satellite ring, the sensor itself provides the sensor fault information, and for the fault information of the planetary ring filter, the chi-square detection is used to obtain the fault information;

[0024] Step 2-1, for a single sensor output as a non-model system, there is no reference output information for comparison, so its fault information can only be obtained through its sensor internal acquisition, and the fault detection unit can obtain the sensor hard fault information, and the soft fault cannot be identified. The system state output by the i-th sensor is defined as n is the number of sensors, P ii,k|k is defined as the covariance of the system state r ii is defined as the fault detection value of the i-th sensor:

[0025]

[0026] Step 2-2, for the sub-filter composed of two sensors, the system state x ij,k|k obeys Gaussian distribution. The prior estimation of the i-th and j-th sensor combined sub-filter at time k is defined as n is the number of sensors, P ij,k|k is defined as the covariance of the state The posterior estimation of the i-th and j-th sensor combined sub-filter is defined as P ij,k|k-1 is defined as the covariance of the state The difference β ij,k between the prior estimation and the posterior estimation of the system state obeys the following zero-mean normal distribution:

[0027] β ij,k ~ N (0, B ij,k )

[0028] wherein,

[0029]

[0030] B ij,k = P ij,k|k-P ij,k|k-1

[0031] According to the orthogonality principle β ij,k The Mahalanobis distance obeys chi-square distribution, and the chi-square test statistic M ij,k :

[0032] M ij,k =(β ij,k ) T (B ij,k ) -1 β ij,k ~χ 2 (m)

[0033] Where m is the dimension of the system state. When the sub-filter fails, β ij,k cannot maintain a Gaussian distribution, so the chi-square detection value M ij,k can be used to detect the sensor. Since the binary logic of the traditional method is difficult to select the fault detection threshold T ij , the fault detection threshold T ij is fuzzified, and instead of accurately determining whether the sub-filter is faulty, the degree of the sub-filter between the faulty and non-faulty states is calculated. This degree is defined as the effective probability, which is obtained by setting the fuzzy logic function. The effective probability of the ith and jth sensor combination sub-filter is defined as μ ij , and the upper and lower bounds of the detection threshold are set as T max and T min , which represent the chi-square test values corresponding to the confidence probabilities of 99% and 90%, respectively, which are obtained by querying the chi-square distribution table. The fuzzy logic membership function is represented as follows:

[0034]

[0035] Specifically, the fault reconstruction table of the data fusion system designed in step 3 specifically includes:

[0036] For a planetary structure multi-sensor data fusion system, this project uses reconstruction logic to reconstruct the fault state system. Since the sensor level fault detection module can only detect hard faults, the isolation strategy should be adopted, and the sub-filter level fault detection can effectively deal with soft faults and hard faults through chi-square detection, and the response strategy is to adjust the weight according to the effective probability μ ij of the sub-filter (when the weight is 0, it means isolation, which means that the output of this sensor does not perform fusion operation).

[0037] When the number of sensors in the satellite ring is 3, let the three sensors in the satellite ring be A, B, and C, and the sub-filter system in the planet ring be <ab> 、 <bc> 、 <ac>; then the fault reconstruction table is specifically as follows:

[0038] Table 1: Fault reconstruction table of the planetary structure filtering system

[0039] System state Reconstruction structure No fault: BC <bc>, fault: A, failure: <ab> <ac> < / ac> < / ab> < / bc> BC <bc> < / bc> No fault: AC <ac>, fault: B, failure: <ab> <bc> < / bc> < / ab> < / ac> AC <ac> < / ac> No fault: AB <ab>, fault: C, failure: <ac> <bc> < / bc> < / ac> < / ab> AB <ab> < / ab> No fault: C, fault: AB, failure: <ab> <ac> <bc> < / bc> < / ac> < / ab> C No fault: A, fault: BC, failure: <ab> <ac> <bc> < / bc> < / ac> < / ab> A No fault: B, fault: AC, failure: <ab> <ac> <bc> < / bc> < / ac> < / ab> B No fault: ABC <bc> <ac>, fault: <ab> < / ab> < / ac> < / bc> ABC <bc> <ac>, down-weighting: <ab> < / ab> < / ac> < / bc> No fault: ABC <ab> <bc>, fault: <ac> < / ac> < / bc> < / ab> ABC <ab> <bc>, down-weighting: <ac> < / ac> < / bc> < / ab> No fault: ABC <ab> <ac>, fault: <bc> < / bc> < / ac> < / ab> ABC <ab> <ac>, down-weighting: <bc> < / bc> < / ac> < / ab> No fault: ABC <bc>, fault: <ab> <ac> < / ac> < / ab> < / bc> ABC <bc>, down-weighting: <ab> <ac> < / ac> < / ab> < / bc> No fault: ABC <ab>, fault: <ac> <bc> < / bc> < / ac> < / ab> ABC <ab>, down-weighting: <ac> <bc> < / bc> < / ac> < / ab> No fault: ABC <ac>, fault: <ab> <bc> < / bc> < / ab> < / ac> ABC <ac>, down-weighting: <ab> <bc> < / bc> < / ab> < / ac> No fault: ABC, Fault: <ab> <ac> <bc> < / bc> < / ac> < / ab> ABC, down-weighted: <ab> <ac> <bc> < / bc> < / ac> < / ab> No fault: BC, Fault: A <bc>Failure: <ab> <ac> < / ac> < / ab> < / bc> BC, down-weighting: <bc> < / bc> No fault: AC, Fault: B <ac>Failure: <ab> <bc> < / bc> < / ab> < / ac> AC, down-weighting: <ac> < / ac> No fault: AB, Fault: C <ab>Failure: <ac> <bc> < / bc> < / ac> < / ab> AB, down-weighting: <ab> < / ab> No fault: ABC <ab> <ac> <bc> < / bc> < / ac> < / ab> ABC <ab> <ac> <bc> < / bc> < / ac> < / ab>

[0040] Wherein, the failure refers to sensor self output failure or sub-filter lacking input information and not working.

[0041] Specifically, the global state estimation in the main filter is realized according to the system fault reconstruction table in step 4, and specifically includes:

[0042] In the planetary structure, both the "satellite" and the "planet" are designed to estimate the state of the overall system The "satellite" sensor will obtain the system state and its covariance P ii,k|k of the sensor, and the "planet" sub-filter will obtain the system state and its covariance P ij,k|k of the sub-filter through filtering after combining the sensors two by two. In addition, the sensor fault detection value r ii and the sub-filter effective probability μ ij are input into the main filter for fusion. The essence of the information fusion process of the main filter is weighted summation of multiple independent local state estimates, and the fusion algorithm of the main filter is globally optimal in the least square sense, and the expression is:

[0043]

[0044]

[0045] Wherein, and P k are the estimation and covariance matrix of the global state respectively. Thus, the multi-sensor data fusion is completed, and the global state estimation is obtained.

[0046] Compared with the prior art, the present application has the following advantages: 1) the sensor data and the sub-filter output data are directly connected to the main filter channel, the data utilization rate is higher, and the combination mode is more flexible; 2) the planetary structure adopted sets a fault detection unit on each orbit node, so that the fault can be found at the sensor level without the need for detection at the sub-filter level, thus having better timeliness; 3) the state chi-square detection at the sub-filter level uses fuzzy logic, which reduces the influence of threshold and improves the data utilization rate; 4) for a ternary system, the traditional structure can only cope with but sensor failure, while the planetary structure can better utilize the remaining sensor information; 5) the planetary structure can fully consider the working state of each sensor and the dependence relationship between them through system reconstruction by the reconstruction table.

[0047] The application is further described in detail with the following structural drawings. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 A flow chart of the method of multi-sensor data fusion and online reconstruction based on planetary topology structure of the application;

[0049] Figure 2 A comparison diagram of traditional and multi-sensor data fusion structure of the application;

[0050] Figure 3 A state chi-square detection block diagram of two sensors combined in the embodiment;

[0051] Figure 4 A state chi-square detection block diagram of two sensors combined in the embodiment; DETAILED DESCRIPTION

[0052] In order to make the purpose, technical scheme and advantages of the application more clear, the application is further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the application and not to limit the application.

[0053] In one embodiment, in combination Figure 1 , a method of multi-sensor data fusion and online reconstruction based on planetary topology structure is provided. The method comprises the following steps:

[0054] Step 1, constructing a three-layer multi-sensor data fusion framework based on planetary topology structure, with each sensor forming an outer ring satellite, each sub-filter system forming a middle ring planet, and the inner ring being a main filter;

[0055] Step 2, designing a two-stage fault detection module, the first stage being sensor fault information extraction of the satellite ring, and the second stage being state chi-square detection of the sub-filter system of the planet ring;

[0056] Step 3, designing a data fusion system fault reconstruction table to realize fault-tolerant reconstruction of the planetary structure fault state system;

[0057] Step 4, based on the fault reconstruction table, combining the sensor fault value and the effective probability of the subsystem, completing global state data fusion in the main filter.

[0058] Further, in one of the embodiments, the three-layer multi-sensor data fusion framework based on planetary topology structure constructed in step 1 specifically comprises:

[0059] The design goal of multi-sensor data fusion system is to improve the performance and robustness of the system through multi-source information fusion, and to overcome the defects and vulnerability of single system. The data fusion framework is the basic skeleton supporting the multi-sensor data fusion system. Compared with the traditional tower federated filtering structure (such as Figure 2 (a) shown), the patent proposes a planetary fault-tolerant filtering structure, as shown in Figure 2 (b). The planetary topology structure is composed of 3 layers. The outermost layer of satellite ring is composed of each sensor, the middle layer of planet layer is composed of each sensor two two combined subsystem, and the inner layer is the main filter to realize the global system state estimation.

[0060] The fault-tolerant design idea of planetary topology structure is: the fault level is divided into two categories: sensor level and sub-filter level, the corresponding fault detection means is used to monitor its running state, and the system reconstruction mechanism table is formulated according to different faults, to realize the reconstruction of the system under different fault states.

[0061] The improvement of the traditional tower structure is:

[0062] 1) Each layer track of the planetary structure has a direct channel to the main filter;

[0063] 2) The "satellite" and "planet" on each layer track of the planetary structure are provided with fault detection units;

[0064] 3) The main filter of the planetary structure uses the fault reconstruction table to reconstruct the system under the fault state.

[0065] Compared with the traditional structure, this design has better timeliness, robustness and flexibility, which is specifically embodied in:

[0066] 1) Multi-type sensor faults can be directly detected at the sensor level without the need for detection at the sub-filter level. The planetary structure has better timeliness for such faults;

[0067] 2) The data of the individual sensor can still be used to realize state estimation. For a three-element system, the traditional structure can only cope with single filter failure, while the planetary structure can cope with double sensor failure;

[0068] 3) The planetary structure reconstructs the system through the fault reconstruction table, which can fully consider the working state of the system and the dependence relationship between them.

[0069] Further, in one embodiment, the two-level fault detection module constructed in step 2 specifically includes:

[0070] As described in step 1, each node on each orbital layer of the planetary topology is equipped with a fault detection unit. For the satellite orbit, sensor fault information is provided by the sensors themselves; for the planetary orbit filter, fault information is obtained through chi-square detection.

[0071] Step 2-1: For a single sensor output, which is a model-less system with no reference output information for comparison, its fault information can only be obtained from within the sensor itself. This fault detection unit can obtain hard fault information from the sensor but cannot identify soft faults. The system state output by the i-th sensor is defined as... n is the number of sensors, P ii,k|k Defined as system state The covariance. Define r. ii For the fault detection value of the i-th sensor:

[0072]

[0073] Step 2-2, for the sub-filter formed by combining two sensors in pairs, the system state x ij,k|k It follows a Gaussian distribution. (Combined with...) Figure 3 Define the prior estimates of the i-th and j-th sensor combination sub-filters at time k. n is the number of sensors, P ij,k|k Defined as state The covariance of the i-th and j-th sensor combined sub-filters is defined as follows. P ij,k|k-1 Defined as state The covariance of the system state. Then the difference β between the prior and posterior estimates of the system state. ij,k It follows a zero-mean normal distribution:

[0074] β ij,k ~N(0,B ij,k )

[0075] in,

[0076]

[0077] B ij,k =P ij,k|k -P ij,k|k-1

[0078] According to the orthogonality principle β ij,k Mahalanobis distance follows a chi-square distribution; construct the chi-square test statistic M at time k. ij,k :

[0079] M ij,k =(β) ij,k ) T (B ij,k ) -1 β ij,k ~ χ 2 (m)

[0080] where m is the dimension of the system state. When the sub-filter is faulty, then β ij,k cannot maintain a Gaussian distribution, so the chi-square detection value M ij,k can be used to detect the sensor. Since the binary logic of the traditional method is difficult to select the fault detection threshold T ij , the fault detection threshold T ij is fuzzy, and instead of accurately determining whether the sub-filter is faulty, the degree to which the sub-filter is between the two states of being faulty and not being faulty is calculated. This degree is defined as the effective probability, which is obtained by setting the fuzzy logic function. The effective probability of the ith and jth sensor combination sub-filter is defined as μ ij , and the upper and lower bounds of the detection threshold are set to T max and T min , which represent the chi-square test values corresponding to confidence probabilities of 99% and 90%, respectively, which are obtained by consulting the chi-square distribution table. The fuzzy logic membership function is represented as follows:

[0081]

[0082] Further, in one of the embodiments, the fault reconstruction table of the data fusion system designed in step 3 specifically includes:

[0083] For a multi-sensor data fusion system with a planetary structure, the reconstruction logic shown in the following table is used to reconstruct the fault state system. Since the sensor level fault detection module can only detect hard faults, the isolation strategy should be used, and the sub-filter level fault detection can effectively deal with soft faults and hard faults through chi-square detection, and the strategy is to adjust the weight according to the effective probability μ ij of the sub-filter (when the weight is 0, it is isolated). When the number of sensors in the satellite ring = 3, let the three sensors in the satellite ring be A, B, and C, and the sub-filter system in the planet ring be <ab> 、 <bc> 、 <ac>; combined Figure 2 (b) and Figure 4 The fault reconstruction table is as follows:

[0084] Table 1: Fault reconstruction table of planetary structure filtering system

[0085] System state Reconstruction structure No fault: BC <bc>, fault: A, failure: <ab> <ac> < / ac> < / ab> < / bc> BC <bc> < / bc> No fault: AC <ac>, fault: B, failure: <ab> <bc> < / bc> < / ab> < / ac> AC <ac> < / ac> No fault: AB <ab>, fault: C, failure: <ac> <bc> < / bc> < / ac> < / ab> AB <ab> < / ab> No fault: C, fault: AB, failure: <ab> <ac> <bc> < / bc> < / ac> < / ab> C No fault: A, Fault: BC, Failure: <ab> <ac> <bc> < / bc> < / ac> < / ab> A No fault: B, fault: AC, failure: <ab> <ac> <bc> < / bc> < / ac> < / ab> B No fault: ABC <bc> <ac>, fault: <ab> < / ab> < / ac> < / bc> ABC <bc> <ac>, down-weighting: <ab> < / ab> < / ac> < / bc> No fault: ABC <ab> <bc>, fault: <ac> < / ac> < / bc> < / ab> ABC <ab> <bc>, down-weighting: <ac> < / ac> < / bc> < / ab> No fault: ABC <ab> <ac>, fault: <bc> < / bc> < / ac> < / ab> ABC <ab> <ac>, down-weighting: <bc> < / bc> < / ac> < / ab> No fault: ABC <bc>, fault: <ab> <ac> < / ac> < / ab> < / bc> ABC <bc>, down-weighting: <ab> <ac> < / ac> < / ab> < / bc> No fault: ABC <ab>, fault: <ac> <bc> < / bc> < / ac> < / ab> ABC <ab>, down-weighting: <ac> <bc> < / bc> < / ac> < / ab> No fault: ABC <ac>, fault: <ab> <bc> < / bc> < / ab> < / ac> ABC <ac>, down-weighting: <ab> <bc> < / bc> < / ab> < / ac> No fault: ABC, Fault: <ab> <ac> <bc> < / bc> < / ac> < / ab> ABC, down-weighted: <ab> <ac> <bc> < / bc> < / ac> < / ab> No fault: BC, Fault: A <bc>Failure: <ab> <ac> < / ac> < / ab> < / bc> BC, down-weighting: <bc> < / bc> No fault: AC, Fault: B <ac>Failure: <ab> <bc> < / bc> < / ab> < / ac> AC, down-weighting: <ac> < / ac> No fault: AB, Fault: C <ab>Failure: <ac> <bc> < / bc> < / ac> < / ab> AB, down-weighting: <ab> < / ab> No fault: ABC <ab> <ac> <bc> < / bc> < / ac> < / ab> ABC <ab> <ac> <bc> < / bc> < / ac> < / ab>

[0086] Where, failure refers to the failure of the sub-filter caused by sensor failure.

[0087] The design logic of the fault reconstruction table is:

[0088] Sensor fault detection value r ii Obtained from the state of the sensor itself.

[0089] Sub-filter effective probability μ ij There are two cases:

[0090] The first kind: the i-th and j-th sensors have faults, according to the fault reconstruction table, the sub-filter is defined as a failed filter, and isolation operation is performed on it, so: μ ij = 0.

[0091] The second kind: the i-th and j-th sensors have no faults, then the effective probability μ ij is calculated by combining the chi-square value obtained by the state chi-square detection algorithm with the fuzzy logic membership function. For the sub-filter whose chi-square detection is a fault, according to the fault reconstruction table, weight reduction processing is performed, and the weight size is the sub-effective probability μ ij calculated in claim 3.

[0092] In the embodiment, n = 3 is taken as an example, and the specific fault reconstruction table form is given. Based on this, those skilled in the art can deduce the fault reconstruction table when n > 3 according to the above design logic, which will not be repeated here.

[0093] Further, in one of the embodiments, based on the fault reconstruction table, the sensor fault value and the subsystem effective probability are combined in step 4 to complete the global state data fusion in the main filter, which specifically includes:

[0094] In the planetary structure, both the "satellite" and the "planet" are designed to estimate the state of the overall system The "satellite" sensor will obtain the system state and its covariance P ii,k|k of the sensor, and the "planet" sub-filter will obtain the system state and its covariance P ij,k|k of the sub-filter by filtering after combining the sensors two by two. In addition, the sensor fault detection value r ii and the sub-filter effective probability μ ij The input is fused into the main filter. The essence of the main filter information fusion process is the weighted summation of multiple independent local state estimates, and the fusion algorithm of the main filter is globally optimal in the least square sense, and the expression is:

[0095]

[0096]

[0097] wherein, and P k are the global state estimate and covariance matrix respectively. At this point, the multi-sensor data fusion is completed, and the global state estimate is obtained.

[0098] The present application is based on a multi-sensor data fusion and online reconstruction method based on a planetary topology structure, adopts a planetary multi-sensor data fusion structure, constructs a sensor level and a sub-filter level fault detection unit, introduces fuzzy logic to define the effective probability of the sub-filter, and designs a data fusion system fault reconstruction table, combines the sensor fault value and the sub-system effective probability, and completes the global state estimation in the main filter. Compared with the traditional tower topology data fusion method, this method not only can realize high-precision data fusion, but also can realize online system reconstruction to complete fault tolerance, improve the utilization rate of sensor data, has the advantages of high efficiency and flexibility, and can realize the system state estimation of a single sensor through the system reconstruction table, effectively improves the reliability and precision of multi-sensor data fusion.

[0099] The basic principles, main features and advantages of the present application are shown and described above. It should be understood by those skilled in the art that the present application is not limited by the above examples, and the above examples and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.< / ac> < / bc> < / ab> < / ac> < / bc> < / ab>

Claims

1. A multi-sensor data fusion and online reconfiguration method based on a planetary topology, characterized in that, Comprising the following steps: Step 1, a three-layer multi-sensor data fusion framework based on planetary topology is constructed, each sensor forms an outer satellite circle, each sub-filter system forms a middle planetary circle, and the inner circle is a main filter; Step 2, design two-stage fault detection module, the first stage is satellite ring sensor fault information extraction, obtain sensor fault value r ii ; the first stage is the state chi-square detection of the planetary ring filter system, and the effective probability μ of the subsystem is obtained ij ; The two-stage fault detection module constructed in step 2 specifically includes: Each node on each layer orbit of the planetary topology is provided with a fault detection unit; the fault information of the sensors of the satellite circle is provided by the sensors themselves, and the fault information of the sub-filter of the planetary circle is obtained by chi-square detection; Step 2-1, for single sensor output, no model system, no reference output information comparison, the system state output by the i th sensor is defined as i = 1, 2, …, n, n is the number of sensors, P ii,k|k is defined as the covariance of the system state r ii is defined as the i th sensor fault detection value: Step 2-2, for the sub-filter composed of two sensors, the system state x ij,k|k obeys Gaussian distribution; define the priori estimation of the sub-filter combined by the ith and jth sensors at time k as i,j = 1, 2, …, n; j > i, n is the number of sensors, P ij,k|k is defined as the covariance of state ; define the posteriori estimation of the sub-filter combined by the ith and jth sensors as i,j = 1, 2, …, n; j > i, P ij,k|k-1 is defined as the covariance of state ; then the difference β between the priori estimation and the posteriori estimation of the system state is ij,k obeys the following zero-mean normal distribution: β ij,k ~ N(0, B ij,k ) Wherein, B ij,k = P ij,k|k - P ij,k|k-1 Wherein, ~ indicates that the variable is subject to the probability distribution behind; Then according to the orthogonality principle β ij,k The Mahalanobis distance obeys chi-square distribution, and the chi-square test statistic M is constructed at k moment ij,k : M ij,k = (β ij,k ) T (B ij,k ) -1 β ij,k ~χ 2 (m) where χ 2 (m) denotes the m-dimensional chi-square distribution, m is the dimension of the system state, β ij,k When the Gaussian distribution cannot be maintained, it indicates that the sub-filter is faulty; the effective probability of the ith and jth sensor combination sub-filter is defined as μ ij , which is the subsystem effective probability; the upper and lower bounds of the detection threshold are set as T max and T min , which are the chi-square test values corresponding to the confidence probabilities of 99% and 90%, respectively, and are obtained by consulting the chi-square distribution table; the fuzzy logic membership function is shown as follows: wherein μ ij is 0 means isolation; Step 3, a fault reconstruction table of the data fusion system is designed to realize fault-tolerant reconstruction of the planetary structure fault state system. Step 4, Reconfiguration based on fault table, combined with sensor fault values r ii and subsystem effective probability μ ij Global state data fusion is done in the main filter.

2. The multi-sensor data fusion and online reconfiguration method based on planetary topology according to claim 1, characterized in that, The three-layer multi-sensor data fusion framework based on the planetary topology constructed in step 1 adopts a planetary fault-tolerant filtering structure, and each layer orbit of the planetary structure has a direct channel to the main filter.

3. The multi-sensor data fusion and online reconfiguration method based on planetary topology according to claim 1, characterized in that, The data fusion system fault reconfiguration table designed in step 3, when the number of satellite ring sensors = 3, let the three sensors of the satellite ring be A, B, and C, and the sub-filter systems of the planet ring be <ab> 、 <bc> 、 <ac>; The fault reconstruction table is as follows:< / ac> < / bc> < / ab> Fault reconstruction table wherein the failure refers to a sub-filter failure caused by a sensor failure, and the down-weighting refers to a down-weighting according to a subsystem validity probability μ ij adjusting the weight.

4. The multi-sensor data fusion and online reconfiguration method based on planetary topology of claim 1, wherein, In step 4, the global state estimation is realized in the main filter according to the system fault reconstruction table, which specifically includes: In the planetary structure, both the "satellite" and the "planet" are designed to estimate the state of the overall system The "satellite" sensor will obtain the system state and its covariance P ii,k|k of the sensor ij,k|k The "planet" sub-filter filters through the combination of two sensors and obtains the system state and its covariance P ii,k|k of the sub-filter ij,k|k The sensor fault detection value r ii and the sub-filter effective probability μ ij are input into the main filter for fusion.

5. The multi-sensor data fusion and online reconfiguration method based on planetary topology according to claim 4, characterized in that, The expression of the information fusion of the main filter is: where, and P k are the estimate and covariance matrix of the global state, respectively.

Images