FIELD OF INVENTION
The present invention relates to a device and a method for maintaining the attitude of a carrier using gyroscopes.
STATE OF THE ART
To compute and maintain the attitude of a wearer, it is known to use three primary gyros arranged to measure rotational speeds of the following three different axes during movement of the wearer.
Are calculated from the measurements acquired by the three gyros, a yaw angle, a pitch angle and a roll angle of the wearer, which together define the attitude of the carrier in space.
It is also known to use a secondary gyroscope to correct scaling factor of errors which undermine the measurements acquired by the three primary gyros.
However, use of such a fourth gyroscope is not effective to correct errors of excesses that plague the measurements provided by the gyros.
DISCLOSURE OF INVENTION
The invention therefore proposes to maintain the attitude of a carrier with an accuracy improved compared with the solutions proposed in the state of the art.
It is therefore proposed, according to a first aspect, an attitude of a wearer maintenance device, the device comprising:
• three primary gyros arranged to measure primary rotational speeds of a wearer around the three primary axes,
• A secondary gyrometer arranged to measure a second rotational speed of the carrier around a different minor axis of each of the primary axes,
• a camera having a different optical axis of each of the primary axes,
• a data processing module configured to
o estimate the scale factor error and drift that plague primary speeds using data from the speed of secondary rotation of images acquired by the camera, and
o correct primary rotational speeds with said estimated errors in the present invention, the camera and the secondary gyro combine synergistically. Indeed, the combined operation of the speed provided by the gyroscope and secondary images acquired by the camera can correct not only the scale factor errors which undermine the speeds measured by the primary gyros, but also correct errors drifts.
The speeds measured by the primary gyros are thus specifically corrected with a device with four gyros.
The device according to the first aspect of the invention can be completed with the following optional characteristics taken alone or in combination where technically possible.
• The data processing module comprises a hybridization filter such as a Kalman filter configured to estimate the scale factor errors and drift from the data from the secondary rotational speed and image acquired by the camera.
• The optical axis of the camera is parallel to the secondary axis.
• Secondary gyrometer is selected of a type adapted to generate scale factor errors and / or drift errors in the speed of secondary rotation which are lower than those generated by primary gyros in primary rotational speeds.
• The secondary is the vibrating gyro hemispherical and / or primary gyros are of MEMS.
· The three primary axes are orthogonal two by two, and the minor axis forms an angle of 54.7 degrees with each primary axis.
It is also proposed, according to a second aspect, an attitude maintenance method of a wearer, comprising the steps of:
• masurement of three carrier rotational speeds about three primary axes intersect pairwise using three primary gyrometers,
• measuring a speed of rotation about a secondary axis intersecting with each of the primary axes using a secondary gyro,
the method being characterized by the steps of:
• acquisition of images using a camera having a different optical axis of each of the primary axes,
• estimate of scale factor error and drift that plague primary speeds, using data from the secondary rotation speed and from images acquired by the camera,
• correction of primary speeds using said estimated errors.
The process according to the second aspect of the invention can be completed with the following optional characteristics taken alone or in combination where technically possible.

· Estimation of scale factor errors and drift is implemented by a hybridization filter such as a Kalman filter, by means of observation data depending on a difference δ between a speed rotation depending on the speed of secondary rotation, and a speed of rotation of the carrier about the secondary axis deduced primary rotational speeds.

"The dependent rotational speed of the secondary rotational speed is a speed resulting from correction of the speed of secondary rotation by means of errors arising from the secondary gyro previously estimated by the hybridization filter.

• The method comprises estimating a rotational speed, said rotational speed camera, and is obtained by analyzing the images acquired by the camera, correcting scale factor errors and drift depending on this speed estimated rotation camera.

• The estimated scale factor error and drift is implemented by means of observation data dependent on a deviation e between a dependent speed camera rotation speed, and rotation speed carrier view the deduced primary camera rotation speeds.

• The method further comprises steps of estimating a focal error brought about by the camera, and correcting, by using the focal error of the camera rotational speed obtained by analyzing the images acquired by the camera, so as to produce a corrected rotational speed camera, wherein the rotation speed depending on the rotation speed estimated camera used to calculate the deviation e is the rotation speed corrected camera.

• The method further comprises steps of calculating a confidence index of the measurement of an optical flow in the image acquired by the camera, and comparing the index with a predetermined threshold, the step of correcting the primary speeds being implemented or not depending on the result of this comparison.

• primary rotational speeds Rbc are corrected by applying the following formula:

febcl 0 0 dbcl'

0 febc2 0 . Rbc - dbc2

0 0 febc3 -dbc3.

where febcl, febc2, febci are scale factor errors and dbcl, DBC2 and dbc3 are drift errors.

• The method further comprises a step of integrating producing attitude data carrier from the primary rotational speeds corrected.

"The method further comprises computing said three rotational speeds blended from the following speeds: primary correction speed corrected using the scale factor error and estimated drift rate resulting from the correction the speed of secondary rotation, one of the blended rotational speed being a speed of rotation around the secondary axis, and wherein the wearer's attitude data is generated by integrating the blended speeds.

• The method further comprises the steps of estimating the carrier attitude errors, and correcting attitude data produced by means of the attitude errors estimated.

DESCRIPTION OF FIGURES

Other features, objects and advantages of the invention will become apparent from the following description, which is purely illustrative and not exhaustive, and should be read in conjunction with the accompanying drawings wherein:

· The Figure 1 schematically illustrates an attitude maintenance device according to an embodiment of the invention;

• Figure 2 shows the orientation in space of individual axes of the Figure 1 device;

• Figure 3 is a flowchart of steps of an attitude measurement method implemented by the device of Figure 1;

• Figures 4, 5 and 6 are flowcharts treatment step made on the measurements acquired by the device of Figure 1, in three different embodiments of the invention.

Of all the figures, similar elements bear identical references.

DETAILED DESCRIPTION OF THE INVENTION

A) An attitude maintenance

Referring to Figure 1, an attitude maintenance device of a wearer comprises three primary gyrometers 1, 2, 3, a secondary gyro 4, a camera 5 and a 6 Data processing module.

The three primary gyrometers 1, 2, 3 are arranged to measure the primary rotational speed of the carrier on three different primary axes.

The three axes are, for example intersecting in pairs. In other words, one of the axes is not included in the plane defined by the other two axes.

It was considered in the following non-limiting example, primary gyrometers 1, 2, 3, oriented such that the three primary axes are substantially orthogonal two by two.

Primary gyrometers 1, 2, 3 together conventionally a "triad" of gyrometers.

The secondary 4 gyrometer is arranged to measure a second rotational speed of the carrier around a different minor axis of each of the primary axes.

The secondary gyro 4 is for example oriented so that its minor axis form the same angle with each of the primary axes, or about 54.7 °. It is also said that the axis of the gyroscope is oriented on a diagonal of the triad of primary gyrometers.

Furthermore, the camera 5 has a different optical axis of each of the primary axes.

The data processing module 6 is configured to generate attitude data carrier from the measured speed by the gyrometers 1, 2, 3, 4 and the images acquired by the camera 5.

The data processing module 6 includes at least one processor configured to implement program code instructions.

In particular, the processor 6 processing module is configured to perform a content analysis algorithm of the images captured by the camera 5, for example an optical flow measurement algorithm in these images.

The processor 6 processing module is further configured to implement a hybridization filter on the basis of data provided by the gyros and the analysis algorithm.

For example, the hybridization filter, whose operation will be described later, may be a Kalman filter. Preferably, the Kalman filter is used in the extended type.

The output data of the device are the attitude data generated by the processing module 6.

The optical axis is preferably parallel to the secondary axis, which has the advantage of along the optical axis - which is the axis of weak performance triad "vision" - the best gyro measured by hybridization gyro measures (this hybridization will be detailed later). In fact the secondary axis is preferred since the three primary axes bring him an information redundancy while conversely, only the minor axis brings a redundant information according to each of the three primary axes (orthogonal to recall them).

Of course, the sensors errors primary gyros have an impact on the attitudes of data produced by the data processing module.

While gyro sensors is subject to errors which undermine the speeds it provides. These sensors errors include errors of scale factors and drift errors. Thus, the gyros are conventionnement classified according to their precision, that is to say according to the value of scale factor error and / or drift that may be encountered in the measurements acquired with these gyros.

In the attitude maintenance device, the secondary gyro 4 is selected of a type adapted to generate scale factor errors and / or lower drift errors to those generated by primary gyrometers 1, 2, 3 .

Is taken in the following Example primary gyrometers example MEMS type, that is to say gyrometers micromechanical vibrating structure.

The type gyros "MEMS" cause of scale factor error and substantial abuses. In particular, the scale factor error of a current typical MEMS gyroscope is a Class 3000 to 10,000 ppm. Thus, an error scale factor of OOOOppm causes a heading error of 1 .8 ° to 180 ° excursion. The stability of drift (technical low end of the Allan variance) is the class 10 ° / h.

However, MEMS gyroscopes have the advantage of being inexpensive and very compact. Furthermore, despite the relatively high value of their errors of scale factors, they can be effectively corrected by the secondary gyro when it is chosen greater accuracy.

The secondary gyro is for example a hemispherical vibrating gyro (HRG).

B) General principle for correcting scale factor errors primary gyros with the secondary gyrometer

Is presented below the theoretical principle of operation of attitude maintenance device.

The four gyroscopes 1, 2, 3, 4 together define a heterogeneous inertial architecture.

Secondary 4 gyro is configured to observe a mixture of errors which undermine the measurements acquired by primary gyros 1, 2, 3. This concept provides access among others to a linear combination of scale factors raised by primary gyros 1 , 2, 3, 4 when the carrier [p] is in rotational motion, that is to say when the error scale factors are predominant.

Not limited to but to simplify the presentation of the general principle of general correction of scale factor errors in the measurements made by the gyroscope, it is considered that the rotational speed measurement measured by the secondary gyro 4, denoted Rhp , is vitiated by a negligible error.

The trihedral formed by the three primary gyrometers 1, 2, 3 axes {GBCL, GBC2, Gbc3}

[Rbcll

rotates speed measurement Rbc = RBC2

Rbc3

Defining a virtual origin orthonormal reference axes {GO and Gl, G2, G3} whose cube diagonal is exactly the Ghp axis. The rotational speeds true to the three

great

axes that perfect trihedron are denoted R = R2

R3

On a :

[1 1 1].R

Rhp =

f (1 )

Furthermore, the three axes {GBCL, GBC2, Gbc3} do not coincide with the axes of the coordinate system {Gl, G2, G3}. The three gyros 1, 2, 3 facing along the axes {GBCL, GBC2, Gbc3} make measurements vitiated by a scale factor error and drift error.

Note Kbc 3x3 matrix containing on its diagonal the correction of errors of scale factors and the off-diagonal correction terms axes of the setting errors

{GBCL, GBC2, Gbc3} relative to the axes {Gl, G2, G3}.

Γ1 + febcl cbcl2 cbcl3

Krbc = cbc21 1 + febc2 cbc23

cbc31 cbc32 1 + febc3J

dbcl

Note dbc = DBC2 the column vector containing the terms of drift of the three dbc3

primary gyrometers 1, 2, 3.

We have the following equation relating the real rotations R to the measured rotations Rbc: Kbc. Rbc - dbc = R (2)

Otherwise define δ as a gap between the Rhp measurement performed by the secondary gyro 4, and similar information deduced from the measurements provided by the triad of gyros primary 1, 2, 3

c „, [1 1 l].Rbc [1 1 1].R [1 1 l].Rbc

o = Rhp— (3)

Hence, by injecting (2) into (3):

[1 1 l]. [Kbc. Rbc - dbc] [l 1 l]. Rbc

d =

V3 V3

Is finally obtained

febcl + kbc21 + kbc31

δ = Rbc T . febc2 + kbcl2 + kbc32 [1 1 l]. dBc

febc3 + kbcl3 + kbc23.

Thereafter, assuming the correction of scales febcl factors febc2, febc3 are in 3000ppm class, and calibrations that are known to 0.5 mrad, and that the terms of drifts are in Class 20 ° / h is 0005 ° / s, when one or more of Rbc rotations exceed class 10 ° / s may make the approximation:

febcl

d Rbc = T . febc2 (4)

febc3

Thereafter, if stacked column several achievements, we get an overdetermined system:

febcl

D = A. febc2

febc3

Where Δ is a stack column of several δ achievements and A stacking columns of several embodiments of the vector line Rbc T .

The foregoing show that it is therefore possible to identify errors and drifts axes timings near neglected in equation (4), the scale factors of gyros primary 1, 2, 3 through the gyrometer S4.

C) General principle of error correction of aberrations primary gyrometers using the camera

In addition to the four gyroscopes 1, 2, 3, 4, the addition of the camera 5 can increase the performance of attitude maintenance device.

The camera 5 is configured to be used as an ego-motion sensor, by using a solution algorithm of the global optical flow of images captured by the camera 5.

It will be seen hereinafter that the camera 5 provided Rvi vision information which are combined with the measurements provided by the gyros 1, 2, 3, 4.

The measurement algorithm of global optical flow used by the camera 5 provides a confidence index of the extent to modulate its influence on the attitude solution based on its availability: usable content in images, likelihood of the result . Concretely, analyzes the amount of usable pixels in the image, and the value of the image formed by subtracting the two images repositioned therebetween.

In one embodiment, the observation vision Rvi may or may not be included depending on the comparison to a predefined threshold of the confidence index of the optical flow measured.

It is considered that the camera 5 takes a measurement free of drift error: the measurement is performed between pairs of images; one of the two images is called "reference" and one "test". The "test" image is the current image.

The "reference" image is changed only when needed, and typically remains unchanged in a scene from observation phase with small movements (less than a quarter of the field of view). In practice we will consider the "Record" changes rarely enough so that we can neglect the drift in the measurement of rotational movement by the camera 5.

Consider an original orthonormal reference axes {VO and V1, V2, V3}, VO being the optical center, the optical axis is VI. Note that this axis is colinear with the high performance gyro axis Ghp defined above to a translation nearly equal to the difference between the origin GO of the mark and the origin mark of VO.

It is considered that the orthonormal {V0, V1, V2, V3} is the mark attached to the carrier

rc

[P]. As part of an interview of the attitude of said device, the heading, pitch and roll

are calculated and allow a change of reference between the mark on the carrier [p] and the local geographical reference frame [g]. Tgp We denote the change of reference matrix

[G] to [p] which is obtained directly from that and we do not need

detail here.

The trihedron formed by the minor axis and the camera {Gvil, Gvi2, Gvi3} performs

rRvill

rotation speed measurement Rvi = Rvi2

Rvi3

Rvi here is a rotational velocities vector comprising three components, a component of which is a speed of rotation around the optical axis.

Note Kvi 3x3 matrix containing on its diagonal errors corrections of scale factors and the off-diagonal terms of correction of timing errors of the axes vision {Gvil, Gvi2, Gvi3} relative to the axes {V1, V2, V3 }.

Kvi

Note that we consider the pixel array as perfectly square and known, implying that the error factors ' scale according to VI and V2 axes are identical and equal to the ignorance of the focal length.

Note that the camera 5 measurement rotations relative to the landmark and not in relation to the inertial reference frame as gyroscopes. Let Rt the Earth rotation vector expressed - and known - in the landmark [g].

We have the following equation relating the true inertial V rotations in the coordinate system {VO, VI, V2, V3} the measured rotations Rvi:

Kvi. Hrivi = V - Tgp T K Rit (5)

The triad {V1, V2, V3} is obtained by applying three rotations trihedron {G1, G2, G3} virtual defined above:

• A 45 ° rotation according G3 to form a trihedral which will be called

{Glbis, G2bis, G3bis)

• Turning - atan (^) "35.26 ° according G2bis, to form the triad

{Glter, G2ter, G3ter)

• then a rotation of 165 ° according Glter, to form the triad {V1, V2, V3}

The relationship between the inertial true V rotations in the virtual marker vision {V0, V1, V2, V3} and real rotations R in the virtual inertial frame {G0, G1, G2, G3} is:

R = Tgyvi. V (6)

0.5744 0.5744 0.5744

Where Tgyvi = 0.5744 -0.7887 0.2113

0.5744 0.2113 -0.7887

Rvil

Maintenant définissons ε comme talented between the Net RVI = Rvi2 effectuée by the

Rvi3

triad vision, and the equivalent information reconstructed by the triad of primary gyros:

= Ε Tgyvi T K Arbisi- Tgp T K Hrit- Hrivi

(TgyviT(Kbc.Rbc-dbc)-TgpT.Rt)

ε = TgyviT. Rbc— TgpT. Rt■

Kvi

fevil cvil2 cvil3

Considering now that Kvi = Id + cvi21 fevi2 cvi23 = Id + kvi and cvi31 cvi32 fevi3

febcl cbcl2 cbcl3

Kbc = Id + cbc21 febc2 cbc23 = Id + KBC, and development limité- Id = - kvi cbc31 cbc32 febc3

ε = Tgyvi T . Rbc - Tgp T . Rt - (Id - kvi). (Tgyvi T ((Id + KBC) Rbc -. Dbc) - Tgp T Rt.) (7) If we now consider a use phase of the device with low attitudes movements, typically an observation phase a fixed point of view, and the summation for example ε 10s. Consider that in 10s the change of attitude does not exceed 0.1 °. This means that ε on the full terms Rbc do not exceed the class 50 ° / h, which means that the terms of the second order products between the terms of matrices and KBC kvi not exceeding 0.01, and Rbc terms, do not exceed 0.5 ° / h. Similarly, the Earth's rotation of 15 ° / h multiplied by kvi terms are Class 0.15 ° / h. These values are to be compared with values drift gyros low cost, class 20 ° / h. Then one can in equation (7) neglecting the terms of the second and third order Tgyvi T . KBC. Rbc, kvi. Tgyvi T . Rbc, kvi. Tgyvi T . KBC. Rbc,, kvi. Tgyvi T . dbc, and finally kvi. Tgp T . Rt, for:

ε = Tgyvi T . dbc (8) In other words, comparing the measurement of vision and equivalent information reconstructed by primary gyros can observe the terms of primary drift gyros, neglecting in equation (8) the terms of the second dependent order of rotation speeds and timings errors trihedral.

D) A method for attitude maintenance of a wearer

The general principles outlined in Sections B) and C) above can be used to implement an attitude maintenance process implemented by the maintenance device shown schematically in Figures 1 and 2.

Shows the attitude of the carrier [p] as the vector comprising rotations

I-CT

heading, pitch and roll

Referring to Figure 3, in a step 100, the three primary gyrometers 1, 2, 3, measure the three rotational speeds RBC1, RBC2, RBC3 the wearer around the three primary axes.

Rbcl

These velocities are grouped into the vector Rbc = RBC2

Rbc3

Furthermore, the secondary 4 gyro measurement in a step 102 a Rhp rotational speed of the carrier about its minor axis which intersects each of the primary axes.

In a step 104, the camera 5 acquires images.

In a step 106, an analysis of images captured by the camera 5 is implemented so as to produce an estimate of the carrier rotational speed in the space under rRvill

the shape of the vector Rvi = Rvi2.

Rvi3.

Image analysis can be implemented using a measurement algorithm of an optical flow in these images, as previously stated in Section C).

The following rotation data is transmitted to the processing module 6:

• Rbc (provided by primary gyros 1, 2, 3)

• Rhp (provided by the secondary gyro 4)

• Rvi (from the analysis of images captured by the camera 5).

These data are processed by the processing module 6 for generating a measure of the attitude of the wearer.

In a first embodiment of processing illustrated in Figure 4, the processing unit 6 applies the following treatments.

Processing unit 6 estimates 200 the scale factor error and drift that plague primary speeds, using the secondary speed Rhp and Rvi data from images captured by the camera 5.

The estimate 200 of the scale factor errors and drift is implemented by the hybridization filter at least one executed by the processor 6 processing module.

The hybridization filter further provides an attitude error estimate: a heading error, a roll error dr, dt and a pitch error.

In a step 202, the processing unit 6 calculates a second rotational speed of the carrier about the secondary axis from the primary rotational speeds Rbc, and this calculated velocity component redundant information with respect to the rotational speed provided Rhp by the secondary gyro 4.

In the case where the primary axes are orthogonal in pairs and where the minor axis is oriented diagonally to the primary axes, this redundant speed is written

Γΐ_ _ _1 Rbc

L ^ -, as indicated above in Section B.

The processing module 6 calculates also a rotational speed vector of the camera 5, from the primary rotational speeds Rbc, forming a redundant information with respect to the speed Rvi, obtained by analyzing the images captured by the camera 5.

In a step 204, the processing unit 6 calculates the difference δ between the secondary rotation speed measured by the secondary gyro 4 and the estimated secondary rotational speed constituting a redundant information.

In a step 206, the processing unit 6 calculates the difference e between the speed of rotation around the optical axis of the camera 5, deduced from the images captured by the latter, and redundant rotational speed about the same axis optical, estimated using RBC primary velocity measurements.

In the case where the primary axes are orthogonal in pairs and where the minor axis is oriented diagonally with respect to the primary axis, the deviation e can be calculated according to formula (7) previously mentioned in the section C).

The speed differences δ and e are temporally integrated in steps 208 and 210 and the J and J O e results of these time integrations are used by the hybridization filter 200 to estimate the scale factor errors and mistakes drift that plague primary speed measurements.

As indicated above, the hybridization filter may be a Kalman filter, which is a single estimator to implement.

In a particular embodiment, the Kalman filter implements the following equations of evolution and observation:

(X = F X). X

{Y = H(X). X

Where X is a state vector made up of three attitudinal errors, dt, dr, six errors sensors (that is to say, the three errors of FEBC scale factor and the three errors dbc drift) and the time integral of the two gaps δ and ε defined above.

of

dt

dr

febcl

febc2

febc3

X = dbcl

dbc2

dbc3

t d

and the

J e2

J e3

The evolution matrix F (X) is:

The terms here ~ noted the evolution matrix F (X) are readily determinable by the skilled person. These words contain the dependencies of changes of attitudes vis-a-vis the attitude errors errors themselves and sensor errors.

The observation vector Y of the Kalman filter directly contains the full redundancy observations formed by the gap δ and e:

Finally the matrix H of observation state of passage written

After the judgment step 200 implemented using the filter hybridization, the processing module 6 212 corrects the speed data Rbc using the scale factor error estimates and drift produced by this hybridization filter.

This correction 212 is typically implemented by applying the following formula:

\ Fe

The result of this formula is a primary corrected speed vector with respect to the primary speed Rbc provided by the triad of primary gyrometers.

In one embodiment, the calculations redundant information leading to gaps δ and e are not implemented on the basis of Rbc speed data provided directly by primary gyros (and thus tainted factor sensor errors scale and excesses), but based on the corrected primary speeds.

In other words, the hybridization filter operates using a sensor error loop primary gyrometers 1, 2, 3.

In a step 214, the processing module integrates the corrected rotational speeds

I-CT obtained via the correction step 212, so as to obtain a first estimated

the attitude of the wearer.

In a step 216, the processing unit 6 applies a correction processing of the first estimated attitude using the attitude errors, dr, dt produced by the hybridization filter. In other words, is implemented by the Kalman filter looping attitudinal errors.

The correction processing of step 216 product and an attitude of the filtered carrier, more accurate than the first estimated of this attitude.

The Kalman filter performs a loopback error sensors and integration attitude errors, at a slower frequency than the attitude integration frequency. During a loopback, the terms of the state vector X are set to zeros.

Note that the confidence index of the extent of vision Rvi is not shown in Figure 4. A simple implementation may be to ignore the Rvi vision measurement in the process when the measurement confidence index optical flow crosses a predetermined threshold. In this case, no observation Rvi vision available and the Kalman filter "expects" the next available observation for corrections of errors and attitude errors dependent sensors of vision.

In the embodiment illustrated in Figure 4, the hybridization filter used is relatively simple. However, a more sophisticated hybridization filter may be constructed with a state vector X containing all errors of all the sensors used (gyro 1, 2, 3, 4 and camera 5) drifts factors scales and axes timings, and by constructing the evolution matrix F (X) with all the interdependencies

corresponding. A correction of sensor errors and comprehensive, coherent and optimal attitude is well developed. In particular, it may be relevant in an embodiment to estimate also the error of focal length of the camera 5.

In application of this principle is illustrated in Figure 5 the steps of a method wherein the vector of the Kalman filter state includes not only the scale factor errors and drift of the gyrometers primary, but also the errors scale factor and drift secondary gyro.
The method according to this second embodiment differs from that illustrated in Figure 4 by the following additional steps.
In a step 203, the processing module corrects the secondary Rhp speed measurement supplied by the secondary gyro using the scale factor errors and drift on this secondary gyrometer which have been previously estimated by the filter 'hybridization. The result of this correction stage is a corrected speed of secondary rotation.
The gap δ calculated in step 204 by the processing module is the difference between the corrected velocity and the secondary rotational speed of secondary rotation redundant.
In other words, the speed of rotation measured by the secondary gyro 4 used in the first embodiment illustrated in Figure 4 to calculate the deviation δ is replaced in the second embodiment illustrated in Figure 5 by the speed corrected secondary rotation.
Similarly, in a step 205, the processing unit 6 corrects the rotational speed measurement Rvi (which comprises, as a reminder, three components including a rotational component around the optical axis) from the camera 5 at the using the errors arising from the camera 5 (e.g., an error of focal length of the camera 5) which have been previously estimated 200 by the hybridization filter. The result of this correction stage is a corrected rotational speed camera.
The deviation e calculated by the processing module in step 206 is the difference between the corrected speed and the speed of secondary rotation redundant.
In other words, the camera speed, used in the first embodiment illustrated in Figure 4 to calculate the deviation e is replaced in the second embodiment illustrated in Figure 5 by a camera speed corrected.
In a third embodiment illustrated in Figure 6, additional processing 213 is implemented between the correction steps 102 of scale factor errors and drift which sully the primary rotation speed measurements, and step integration 214.
This additional processing 213 is a reconstruction of an inertial coordinate system from the corrected speed of secondary rotation via the correcting step 212.
This inertial trihedron is composed of the minor axis and two axes orthogonal to the secondary axis.
The recomposition processor 213 produces a vector of three rotation speeds Rhpbc, which are the subject of the integration 214 in place of the primary rotational speed Rbc corrected to obtain the attitude data c, t, r.
In this way, the qualities of the secondary axis Ghp are used directly in the integration of attitude 214 and thus improves the accuracy of the attitude solution, especially in the trajectories for which this axis is predominantly excited .
The table below summarizes the advantages and disadvantages of various elements of attitude maintenance device of the invention, and how these elements synergize with each other when combined in accordance with the invention.
Each element of attitude maintenance device is thus more or less efficient with regard to:
• its ability to produce low marred by scale factor error data;
• its ability to produce low marred by drift errors data;
• availability, that is to say its ability to produce continuous data (high availability) or only intermittently (limited availability).
This table shows that the combination of gyros primary, secondary gyroscope and camera offers effective protection against the scale factor error and drift, while being capable of providing continuous attitude estimation.

CLAIMS

1. Device for attitude maintenance of a wearer, the device comprising:

• three primary gyros (1, 2, 3) arranged to measure primary rotational speeds (Rbc) of a wearer around the three primary axes,

• A secondary gyro (4) arranged to measure a speed of secondary rotation (Rhp) of the carrier around a different minor axis of each of the primary axes,

the device being characterized in that it further comprises:

· A camera (5) having a different optical axis of each of the primary axes,

• a data processing module (6) configured to

o estimate the scale factor error and drift that plague primary rotational speeds (Rbc) using data from the secondary rotation speed (Rhp) and images acquired by the camera (5) and o correct primary rotational speeds with said estimated errors.

2. Device according to the preceding claim, wherein the data processing module (6) comprises a hybridization filter such as a Kalman filter configured to estimate the scale factor errors and drift from data from the speed of secondary rotation (Rhp) and images acquired by the camera (5).

3. Device according to one of the preceding claims, wherein the optical axis of the camera is parallel to the secondary axis.

4. Device according to the preceding claim, wherein the secondary gyro (4) is selected of a type adapted to generate scale factor errors and / or drift errors in the speed of secondary rotation which are lower than those arising from primary gyros (1, 2, 3) in the primary rotation speed.

5. Device according to one of the preceding claims, wherein the secondary gyro (4) is hemispherical vibrating type and / or primary rate gyros (1, 2, 3) are of the MEMS type.

6. Device according to one of the preceding claims, wherein the three primary axes are mutually orthogonal, and the minor axis forms an angle of 54.7 degrees with each primary axis.

7. A method for attitude maintenance of a wearer, comprising the steps of:

• measuring (100) carrying three rotational speeds about three primary axes intersect pairwise using three primary gyrometers,

• measuring (102) a speed of rotation about a secondary axis intersecting with each of the primary axes using a secondary gyro,

the method being characterized by the steps of:

• acquisition (104) of images with a camera having a different optical axis of each of the primary axes,

• estimating (200) scale factor error and drift that plague primary speeds, using data from the secondary rotation speed and from images acquired by the camera,

• correcting (202) of the primary rotational speed by using said estimated errors.

8. Method according to the preceding claim, wherein estimating (200) the scale factor errors and drift is implemented by a hybridization filter such as a Kalman filter, by means of data observation δ dependent on a difference between:

• a speed dependent on the speed of secondary rotation,

• a carrier rotation speed about the secondary axis deduced primary rotational speeds.

9. Method according to the preceding claim, wherein the dependent rotation speed of the secondary rotational speed is a speed resulting from a correction (203) of the speed of secondary rotation by means of errors arising from the secondary gyro previously estimated (200) the hybridization filter.

10. A method according to one of claims 7 to 9, comprising estimating a speed of rotation, said rotational speed camera, and is obtained by analyzing (106) the images acquired by the camera (5), the correction of scale factor error and drift depending on the estimated speed camera.

January 1. Method according to the preceding claim, wherein estimating (200) the scale factor errors and drift is implemented by means of observation data depending on a deviation e between:

• a speed depending on the camera speed, and

· A rotational speed of the carrier seen by the camera (5), which is derived (202) of the primary rotational speed.

12. Method according to the preceding claim, further comprising steps of:
• estimating (200) a focal error triggered by the camera (5)
• correcting (205), using the focal error of the camera rotational speed obtained by analyzing the images acquired by the camera, so as to produce a corrected rotational speed camera,
and wherein the rotational speed depending on the rotation speed estimated camera used to calculate the deviation e is the rotation speed corrected camera.
13. A method according to one of claims 7 to 12, comprising the further steps of:
• calculation of a confidence index on the measurement of an optical flow in the images acquired by the camera (5)
• comparison of the index with a predetermined threshold,
the correcting step (202) of the primary rotational speed is implemented or not depending on the result of this comparison.
14. A method according to one of claims 7 to 13, wherein the primary rotational speeds Rbc are corrected (212) by applying the following formula:
febcl 0 0 dbcl'
0 febc2 0 . Rbc - dbc2
0 0 febc3 -dbci.
where febcl, febc2, febc3 are scale factor errors and dbcl, dbcl and dbc3 are drift errors.
15. A method according to one of Claims 7 to 14, further comprising a step of integrating (214) producing attitude data carrier from the primary rotational speeds corrected.
16. A method according to preceding claim in its attachment to claim 9, further comprising calculating (213) of said three rotational speeds reconstituted from:
• Primary speeds corrected using the scale factor error and estimated drift
• speed resulting from the correction of the speed of secondary rotation
wherein a reconstituted rotational speed is a speed of rotation around the secondary axis, and wherein the wearer's attitude data is generated by integrating the blended speeds.
17. A method according to one of claims 15 and 16 further comprising the steps of:
• estimate (200) of the wearer attitude error,
• correcting (216) attitudinal data produced by the attitude errors estimated.