EP2366116A1

EP2366116A1 - Hybridization device with segregated kalman filters

Info

Publication number: EP2366116A1
Application number: EP09771565A
Authority: EP
Inventors: Sébastien VOURC'H; Victor Mauger
Original assignee: Sagem Defense Securite SA
Current assignee: Safran Electronics and Defense SAS
Priority date: 2008-12-17
Filing date: 2009-12-16
Publication date: 2011-09-21
Also published as: US8625696B2; US20110243194A1; FR2939901B1; WO2010070011A1; FR2939901A1

Abstract

The invention relates according to a first aspect to a hybridization device (1) comprising a virtual platform (2) calculating inertial measurements (PPVI), a bank (3) of Kalman filters each formulating a hybrid navigation solution on the basis of the inertial measurements of the virtual platform (2) and of measurements of signals emitted by a constellation of satellites, which measurements are delivered by a satellite positioning system (GNSS), characterized in that it comprises, for each filter of the bank, a module for correcting the satellite measurements (4) delivering to the filter the measurements of the satellite positioning system (GNSS) which are used by the filter after correction with the aid of the hybrid navigation solution formulated by the filter.

Description

KALMAN SEGREGES FILTER HYBRIDIZATION DEVICE

The field of the invention is that of the carriers using information provided both by an inertial unit and by a satellite navigation system, such as a GPS system.

The invention relates to a closed-loop hybridization device, and relates more particularly to performing hybridization with protection in the sense of integrity of the output data.

Carriers such as aircraft or boats have many navigation systems. These systems include a hybrid INS / GNSS (Inertial Navigation System and Global Navigation System).

An inertial unit provides noisy and accurate information in the short term. However, in the long term, the localization performance of an inertial unit degrades (more or less quickly depending on the quality of the sensors, accelerometers or gyroscopes for example, and treatments used by the plant). While information acquired from a satellite navigation system is very unlikely to drift in the long term, it is often noisy and variable in accuracy. In addition, inertial measurements are always available when GPS information is not available or is likely to be deceived and scrambled.

Hybridization consists of combining the information provided by the inertial unit with the measurements provided by the satellite navigation system to obtain hybrid position and speed information by taking advantage of both systems. Thus, the accuracy of the measurements provided by the GNSS receiver makes it possible to control the inertial drift and the low noise inertial measurements make it possible to filter the noise on the measurements of the GNSS receiver. Modern navigation systems calculate protection radii around the calculated position that can contain the error from true position to a given integrity risk, that's what defines the integrity of a system.

According to the state of the art, these protection radii can be calculated by means of a Kalman filter bank which makes it possible to protect against the appearance of a single failure.

These filters hybridize information from the satellite navigation system to that from the inertial unit. One of the filter bank filters, referred to as the main filter, uses all GNSS measurements consisting of pseudo-measurements and quality information. The other filters, called secondary filters, use only a part of the available GNSS measurements. If a failure occurs at a satellite measurement, it will not be seen by the filter not receiving this measurement: this filter will remain unpolluted. In the state of the art, integrity is based on the fact that in the event of a satellite failure, one of the bank's filters is not affected by the failure.

According to a first architecture described for example as an earlier art in EP1801539 A (see FIG. 1 and corresponding discussion), N closed-loop filters are used on N virtual platforms. This architecture has the advantages of being insensitive to inertial drift and to present high-performance outputs at high frequency. However, it is expensive in terms of computing load due to the integration of N virtual platforms.

According to a second architecture which is the subject of document EP1801539 A (see FIG. 2), N closed-loop filters are used on a virtual platform. The advantages are the low cost of computing due to the integration of a single virtual platform, while the disadvantages are the sensitivity to inertial drift and the fact that the outputs are performing at the frequency of the filters. As indicated in this document EP1801539 A, a satellite failure detection must be implemented in order to select the Kalman filter whose output (state vector) will be applied (as stabilization vector) to the inertial measurements from the virtual platform to recalibrate it.

Thus, this document provides for selecting the main Kalman filter when no failure of one of the satellites is detected, or, when a failure is detected, to select the secondary Kalman filter not affected by the failure. Insofar as the corrections thus come from a single filter (the stabilization vector of the inertial measurements thus being a copy of the state vector estimated by the selected Kalman filter) and where this filter is not affected by a failure. satellite, one does not come to apply to the inertial measurements of erroneous corrections by propagation of information corrupted by a satellite failure.

However, these two types of architecture are not completely satisfactory insofar as they do not intrinsically guarantee that one of the filters of the bank will not be polluted by the failure. Indeed, if we do not detect a failure or if we exclude the bad satellite, the position calculated by the GNSS receiver is polluted by the failure. The correction models applied to the pseudo-measurements are then polluted which causes a degradation of all the corrected pseudo-measurements and therefore of the filter which does not use the pseudo-measure where the failure is actually present.

The object of the invention is to overcome these drawbacks, and proposes for this purpose according to a first aspect, a hybridization device comprising a virtual platform calculating inertial measurements, a Kalman filter bank each developing a hybrid navigation solution to from the inertial measurements of the virtual platform and measurements of signals emitted by a constellation of satellites delivered by a satellite positioning system, characterized in that it comprises, for each filter of the bank, a satellite measurements correction module delivering to the filter the measurements of the satellite positioning system which are used by the filter after correction using the hybrid navigation solution developed by the filter. Some preferred, but not limiting, aspects of this device are:

the satellite measurement correction module receives as input the hybrid navigation solutions developed by the filters and the measurements corresponding to the entire constellation delivered by satellite positioning system;

the measurements of the satellite positioning system which are used by a filter are pseudo-measurements;

the device generates a hybrid output corresponding to the inertial measurements calculated by the virtual platform corrected by a stabilization vector;

the stabilization vector corresponds to a state vector estimated by a filter of the bank not affected by a satellite failure.

each component of the stabilization vector is developed according to the corresponding components of the set of state vectors estimated by the filters of the bank;

the device comprises a module for generating the stabilization vector configured, for each component of the stabilization vector (dC [state]), so as: o to analyze the sign of the set of corresponding components of the state vectors; when the set of these corresponding components are not of the same sign, to develop a component of zero value for the stabilization vector; when the set of these corresponding components are of the same sign, to develop a non-zero value component for the stabilization vector, determined according to the value of each of these components of the state vectors.

the non-zero value of the component of the stabilization vector corresponds to the minimum of the set of corresponding components of the state vectors when all of these components corresponding are positive, and corresponds to the maximum of the set of corresponding components of the state vectors when all of these corresponding components are negative;

the non-zero value of the component of the stabilization vector corresponds to the average of the P plus, corresponding small components of the state vectors, taken in absolute value;

the stabilization vector is applied to the input of all the filters of the filter bank;

- the hybrid output is looped back to the input of the virtual platform; the Kalman filter bank comprises a main Kalman filter receiving the measurements of the signals emitted by n satellites corrected with the main hybrid navigation solution that it develops, and n secondary Kalman filters each receiving the measurements of the signals emitted by the n satellites excluding a satellite corrected using the secondary hybrid navigation solution that it develops.

According to a second aspect, the invention proposes an INS / GNSS hybridization method implementing a bank of Kalman filters each developing a hybrid navigation solution from the inertial measurements calculated by a virtual platform and from measurements of signals transmitted by a satellite constellation delivered by a satellite positioning system, characterized in that the satellite measurements used by each filter of the bank are pre-corrected using the hybrid navigation solution developed by the filter.

Other aspects, objects and advantages of the present invention will appear better on reading the following detailed description of preferred embodiments thereof, given by way of non-limiting example, and with reference to the appended drawings in which: Figure 1 is a diagram illustrating a possible embodiment of a device according to the first aspect of the invention. With reference to FIG. 1, there is shown a hybridization device 1 according to a possible embodiment of the first aspect of the invention, intended to be embedded in a carrier such as an aircraft. The hybridization device 1 uses information provided by an UMI inertial measurement unit and by a GNSS satellite navigation system. The device 1 comprises a single virtual platform 2 and a bank 3 of Kalman filters in parallel.

The virtual platform 2 receives inertial increments from the sensors (gyroscopes, accelerometers) of a unit of inertial measurements. The inertial increments correspond in particular to angular increments and to increments of speed. Inertial navigation information (such as the orientation, speed or position of the carrier) is calculated by the virtual platform from these increments. This inertial navigation information is designated PPVI inertial measurements thereafter.

These inertial measurements PPVI are transmitted to a device for calculating pseudo-distances estimated a priori (not shown in Figure 1) which also receives data on the position of the satellites. On the one hand, inertial measurements and on the other hand by data on the position of the satellites, the device for calculating pseudo-distances estimated a priori calculates the pseudo-distances a priori between the carrier and the different visible satellites of the carrier. .

The hybridization device 1 also receives from the GNSS satellite navigation system the pseudo-measurements between the carrier and the different visible satellites. The discrepancies (called observations) are then classically calculated between the pseudo-measurements estimated a priori and the pseudo-measurements delivered by the GNSS system.

The hybridization device 1 further comprises a Kalman filter bank 3 which hybridises between the inertial information coming from the inertial unit and the information from the satellite navigation system. In addition to a function of providing statistical information on the output measurements, the role of the filters is to maintain the virtual platform 2 in a linear operating area image of that modeled in the Kalman filter each estimating a state vector dXO-dXn (generally having the order of 30 components).

In a manner conventionally known per se, the filterbank 3 comprises several Kalman filters in parallel. One of the filters is called the main Kalman filter 8: it takes into account all the observations (and receives all the measurements from the GNSS system) and elaborates a main hybrid navigation solution.

The other filters 9i, 9n are called secondary filters: they take into account only a part of the observations, for example (n-1) observations among the n observations relating to the n visible satellites so that the i-th Kalman filter Secondary 6i receives from the GNSS system the measurements of all satellites except the i-th, and each develop a secondary hybrid navigation solution.

It should be noted that the observation development process described above is not common to all the filters of the bank 3, but is carried out for each of the filters. Thus, the calculation of the pseudo-distances a priori and the calculation of the observations which are mentioned above are not common to all the filters of the bank, but the hybridization device 1 according to the invention performs these calculations for each bank filter. In the same way, FIG. 1 shows only a single module for correcting the satellite measurements 4. It will be understood, however, that the hybridization device 1 according to the first aspect of the invention comprises a satellite measurements correction module. 4 by bank filter.

In the context of the invention, it is thus advantageously provided to associate with each filter of the bank a satellite measurements correction module 4 delivering to the filter the measurements (typically the pseudo-measurements) of the satellite positioning system (GNSS). which are used by the filter after correction using the hybrid navigation solution developed by the filter. In order to develop the hybrid navigation solutions, the device 1 comprises a bench of summators 10, where each summator is positioned at the output of the filter bank to add to the state vector dXO-dXn elaborated by a filter the hybrid output SH which will be presented in more detail later.

It will be noted that the reference navigation (hybrid output SH) produced by the device 1 is only used internally. It is thus the information delivered at the output of the summator bank 7 that provides the optimal navigation solutions (main navigation solution NAV INS / GPS 0 from the main Kalman filter, secondary navigation solution NAV INS / GPS i "derived from the secondary Kalman filter of index i).

The raw measurement correction module 4 associated with a filter of the bank receives as input the satellite measurements (typically pseudomesures) used by the filter corresponding to all (main Kalman filter) or part (secondary Kalman filter) of the all of the constellation delivered by GNSS satellite positioning system, and outputs for each of the filters of the bank 3 said satellite measurements used by the filter after implementation of a deterministic error correction model developed from the hybrid navigation solution estimated by the filter. The calculation of the observations is then made on the basis of these satellite measurements corrected from the information of the filter, and not from the information of the GNSS system as is conventionally the case.

Classically, the pseudo-measurements performed by a GNSS receiver are corrected inside this same receiver. Indeed deterministic errors can be corrected largely by using models that require positioning information to be calculated. It is proposed in the context of the invention to make this correction not in the receiver but for each filter of the bank, on the basis of the position estimated by the filter. Specifically, the errors that affect GNSS raw measurements are modeled. The modeling of atmospheric errors is dependent on the estimated navigation parameters. By adding (positively or negatively according to the errors and the type of measurement) the value of the error calculated by model to the gross measurement, we correct in part the error which is modeled. The error residue is then modeled statistically and characterized by a standard deviation calculated from the navigation parameters.

Conventionally, the addition (positive or negative) of errors calculated by model to gross measurements is performed by the GNSS receiver, so that the entire filter bank receives corrected measurements and error residual modeling. It follows that the calculation of the error models and standard deviations of the residuals is then dependent on the position calculated by the GNSS receiver.

Taking the example of pseudo-distance measurements:

Raw PR = true distance + errors errors modeled = f (X j

PR corrected = true distance + errors - modeled errors σ = h (x) The invention proposes to deport this calculation of the error models and the standard deviations of the residues, as well as the application of the models to the raw measurements in the module correction of the satellite measurements 4, and to perform the calculations for each filter from the navigation parameters specific to each filter. In this way, the corrected measurements received by a filter are no longer dependent on the navigation parameters estimated by the GNSS receiver. Therefore, in the event of a satellite failure, the corrected measurements received by the filter that does not use the failed satellite are not affected.

Thus, the main Kalman filter receives from module 4 all the pseudo-measures, corrected with the aid of the main navigation solution that it develops. The secondary Kalman filter of index i 9i receives meanwhile from module 4 all the pseudo-measurements with the exception of that corresponding to the satellite of index i, corrected using the secondary navigation solution that it develops.

Since the correction of the raw measurements is specific to each filter, the filter which does not use the pseudo-measure polluted by the appearance of a failure is not affected by the failure (its solution of unpolluted navigation the failure makes it possible to correct the pseudomesures that it uses), and consequently it remains unpolluted.

The hybridization device 1 according to the first aspect of the invention produces a hybrid output SH corresponding to the PPVI inertial measurements calculated by the virtual platform 2 and corrected by a stabilization vector dC.

According to one possible implementation of the invention, the corrections to be applied to the inertial measurements come from a single Kalman filter. Thus, the stabilization vector is equal to the correction vector estimated by the selected Kalman filter. The selection is effected for example according to EP1801539 A by detecting a possible satellite failure.

According to an advantageous implementation of the invention, the stabilization vector is elaborated component by component, using for each component the set of state vectors estimated by the Kalman filters.

The device 1 according to the invention comprises for this purpose a development module of the correction 5 configured to develop each of the components dC [state] of the stabilization vector dC according to the set of corresponding components dXO [state] -dXn [state] of the correction vectors dXO-dXn.

According to one possible embodiment of the invention, the correction elaboration module 5 is configured, for each component dC [state] of the stabilization vector dC, so as to: analyzing the sign of the set of corresponding components dXO [state] -dXn [state] of the correction vectors estimated by the Kalman filters; and

- when all these corresponding components are not of the same sign, to develop a zero value component (dC [state]

= 0) for the stabilization vector;

when all these components dXO [state] -dXn [state] are of the same sign, to develop a non-zero value component for the stabilization vector, determined according to the value of each of these components dXO [state ] -dXn [state].

For example, the correction elaboration module 5 is configured in such a way that the non-zero value of the component of the stabilization vector dC [state] corresponds to the minimum of the set of corresponding components dXO [state] -dXn [ state] of the correction vectors when these components dXO [state] -dXn [state] are all positive, and corresponds to the maximum of the set of components dXO [state] - dXn [state] of the correction vectors when these components dXO [ state] - dXn [state] are all negative.

As a variant, the correction elaboration module 5 can be configured in such a way that the non-zero value of the component of the stabilization vector dC [state] corresponds to the average of the smaller Ps (P being, for example, equal to 2) corresponding components dXO [state] - dXn [state] of the correction vectors, taken in absolute value.

The stabilization vector developed in accordance with this possible embodiment of the invention makes it possible to minimize the estimated errors for all the filters.

This development of the stabilization vector proves judicious insofar as it is not constrained by a fault detection and exclusion mechanism (FDE mechanism according to the English terminology "Fault Detection and Exclusion"), and where the validity of the protection radii is not constrained by an EDF. In the context of this variant, the virtual platform and the filter bank will be polluted by a satellite failure, but not the filter that excludes the failed satellite.

The stabilization vector dC thus produced by the module 5 or simply selected from the state vectors of the filters makes it possible to correct, with a delay 6, the PPVI inertial measurements calculated by the virtual platform, using conventionally known per se a subtractor 7.

In the context of a closed-loop hybridization device, the hybrid output SH is looped back to the input of the virtual platform. Moreover, as shown in FIG. 1, the stabilization vector dC can be applied to the input of all the filters of the filter bank. In this way, the Kalman filters adjust by subtracting from their estimate (correction vector dX) the correction dC, and are thus kept consistent with the virtual platform. The architecture proposed by the invention has the following advantages.

- it only requires the integration of a single virtual platform;

the Kalman filter which does not use the diseased satellite is not polluted by the failure;

- the filters of the bank are completely segregated; the calculation of the stabilization control of the platform is not constrained by a fault detection method.

The invention is also not limited to a hybridization device according to its first aspect, but also extends to an INS / GNSS hybridization method implementing a bank of Kalman filters each developing a navigation solution. hybrid based on the inertial measurements calculated by a virtual platform and measurements of signals emitted by a constellation of satellites delivered by a satellite positioning system, characterized in that the satellite measurements used by each filter of the bank are corrected beforehand. help of the hybrid navigation solution developed by the filter.

Claims

1. Hybridization device (1) comprising a virtual platform (2) calculating inertial measurements (PPVI), a bank (3) of Kalman filters each developing a hybrid navigation solution from the inertial measurements of the virtual platform ( 2) and measurements of signals emitted by a constellation of satellites delivered by a satellite positioning system (GNSS), characterized in that it comprises, for each filter of the bank, a satellite measurements correction module (4) delivering to the filter the measurements of the satellite positioning system (GNSS) that are used by the filter after correction using the hybrid navigation solution developed by the filter.

2. Device according to claim 1, wherein the satellite measurements correction module (4) receives as input the hybrid navigation solutions developed by the filters and the measurements corresponding to the entire constellation delivered by satellite positioning system. (GNSS).

3. Device according to one of claims 1 to 2, wherein the measurements of the satellite positioning system (GNSS) which are used by a filter are pseudo-measurements.

4. Device according to one of claims 1 to 3, developing a hybrid output (SH) corresponding to the inertial measurements (PPVI) calculated by the virtual platform (2) and corrected by a stabilization vector (dC).

5. Device according to claim 4, wherein the stabilization vector corresponds to a state vector estimated by a bank filter unaffected by a satellite failure.

6. Device according to claim 4, wherein each component of the stabilization vector (dC [state]) is developed according to the corresponding components of the set of state vectors estimated by the filters of the bank.

7. Device according to claim 6, comprising a module for generating the stabilization vector configured, for each component of the stabilization vector (dC [state]), so as to:

analyzing the sign of the set of corresponding components (dXO [state] -dXn [state]) of the state vectors;

when the set of these corresponding components are not of the same sign, to develop a component of zero value (dC [state] = 0) for the stabilization vector;

when the set of these corresponding components are of the same sign, to develop a non-zero value component for the stabilization vector, determined as a function of the value of each of these components of the state vectors.

8. Device according to claim 7, wherein the non-zero value of the component of the stabilization vector corresponds to the minimum of all the corresponding components of the state vectors when all of these corresponding components are positive, and corresponds to the maximum of all the corresponding components of the state vectors when all of these corresponding components are negative.

9. Device according to claim 7, wherein the non-zero value of the component of the stabilization vector corresponds to the average of P plus, small corresponding components of the state vectors, taken in absolute value.

10. Device according to one of claims 1 to 9, wherein the stabilization vector is applied to the input of all filters in the filter bank.

11. Device according to one of claims 1 to 10, wherein the hybrid output is looped back to the input of the virtual platform.

12. Device according to one of claims 1 to 11, wherein the Kalman filter bank (3) comprises a main Kalman filter (8) receiving the measurements of the signals emitted by n satellites corrected by the main hybrid navigation solution. that it elaborates, and n secondary Kalman filters (9i, 9n) each receiving the measurements of the signals emitted by the n satellites excluding a satellite corrected with the aid of the secondary hybrid navigation solution that it develops.

13. INS / GNSS hybridization method implementing a bank of Kalman filters each developing a hybrid navigation solution from the inertial measurements calculated by a virtual platform (2) and measurements of signals emitted by a constellation of satellites delivered by a satellite positioning system (GNSS), characterized in that the satellite measurements used by each filter of the bank are previously corrected using the hybrid navigation solution developed by the filter.