The present invention relates to a method of aligning an inertial unit located on a static or quasi-static carrier.
It also relates to an inertial system implementing such an alignment process.
STATE OF THE ART
Initialization of an inertial location is to develop initial values ​​of attitude, speed and position of the inertial locating algorithm. This step corresponds to a mode of inertial called "Alignment".
Initialization of an inertial location includes many variants, depending on the applications, and classified into two categories:
· Static alignment in the case of a holder not moving,
• alignment movement in the case of a movable carrier on Earth.
The wearer's movements during a static alignment degrade the accuracy of the initialization of navigation. Most applications include a motion detection function to check the immobility of the carrier and secure alignment of inertial navigation. Various algorithms exploit the inertial measurements to detect movement of the wearer. They can operate in addition to type information "Weight on Wheels (WoW)" parking brake (aircraft or ground vehicle), power on the rotor (helicopter), collective (helicopter), etc.
The detection power of motion detection algorithms depends on the type of movement. In the case of low-frequency oscillations, current movement of the tests, the limited precision inertial sensors used by the inertial localization, are not enough to ensure the robustness of the static alignment algorithm.
Recall below some known results on regressors least squares as they are a way to estimate the initialization of an inertial location by observing the inertial speeds or observational inertial movement maintained by integration of inertial speeds.
On static or quasi-static holder and after a coarse alignment phase, the horizontal velocity errors of an inertial location can be modeled by polynomials which are a function of the alignment and length which are each of order 2 horizontal axes. The coefficients of these two polynomials are used to estimate attitude errors and speed the inertial positioning and thus to achieve alignment. Similarly, errors of horizontal displacement of an inertial location can be modeled by polynomials function of the alignment and length of order 3 for each of the horizontal axes. The coefficients of these two polynomials are used to estimate attitude errors and speed the inertial positioning and thus to achieve the alignment. Using a regressor operator inertial movement is well suited to the case of oscillatory movements carriers.
Moreover, it is well known that the estimated heading error of the inertial location is very sensitive to the wearer movements because it is based on the Earth's rotation is a physical signal much more difficult to measure than gravity earthly. This heading estimate is derived from the curvature for a regressor observing the inertial speeds or polynomial coefficient of order 3 for a regressor observing the inertial movement.
Kalman filters are also a classic way to estimate initializing an inertial location. One advantage known regressors least squares with respect to Kalman filters is that you simply merely know the effect on the alignment of a parasite movement under assumption that this movement is known and limited amplitude.
The table below summarizes the maximum impact on the polynomial cmax or dmax of the highest order coefficient for order regressors 2 and 3 who observe a limited amplitude signal V D in speed or movement. We know that the maximum impact is achieved for a signal that switches between its terminals (+ or - V if observing speed errors, + or - D if observations of trips) exactly 2 times for an order regressor 2 or 3 times for a serial regressor 3.
Regressor least regressor least
squares of order 2 squares of order 3
2 switching times 3 switching times that are rated u1 and u2: that are rated U1, U2 and U3:
VI5
moments
u1 = 1 /2 - 10
switching u1 = 1/2 - 6
u2 = 1/2
VI5
. 1 = U2 / 2 + . 6
u3 = 1/2 + 10
maximum term
20 -V
the order plus 9 \ D

élevé CMAX " S - T2
where T represents the duration of the temporal regression and u represents the normalized time u = t / T
Figure 1 shows a scenario for the 2nd order model (there is a symmetric second scenario). In abscissa, the normalized time u and ordinate, normalized error Terror equal to / V.
In this figure, the curve MP represents the movement corresponding to the worst case. R curve is of order 2 regressor.
On the curve, the movement of the carrier switches 2 times between its terminals. This generates maximum curvature and Cmax Terror on the polynomial estimate of order 2.
It is known that the maximum inertial drift induced by the wearer's movement is thus equal to:
Maximum inertial drift =
where g is gravity.
Figure 2 shows in turn a scenario for the model of order 3. On the horizontal axis, the normalized time u, y-axis, normalized Terror equal to error / D.

The movement of the carrier switches 3 times (curve PM) between its terminals (+ or - D). This thus generates the maximum dmax error on the order coefficient 3 on the order of polynomial estimation 3 (curve R). It is known that the maximum inertial drift induced by the wearer's movement is thus equal to
. , 6 - d 273 D
_inertielle_maximale = = drift -
gg - t
Whatever regressor, the maximum inertial drift induced by the wearer's movements must remain small compared with the precision inertial sensors. An order of magnitude which is usually considered is the drift induced by the movements must be less than one third of the drift of the inertial gyros used by location as 1/3 in short RMS generates about 10% of degradation.

Thus, for a given amplitude of the wearer's movements, alignment is even less disturbed it is set for an observation period T high. This finding also applies to other types of estimators especially Kalman filter base.
PRESENTATION GENERAL DE INVENTION
A general object of the invention to provide an alignment method that does not have the disadvantages of the prior art and that improves the robustness to unwanted movements.

In particular, an object of the invention is to provide an alignment method for a maximum inertial drift induced by the movements of the lower carrier to the accuracy of the inertial sensors.

The present invention proposes to simultaneously perform several alignments (by least-squares regression, Kalman filtering or by using any other estimator errors of an inertial location) dimensioned for several amplitudes of oscillations of low frequency and select alignment of minimum duration whose consistency has been checked by an inter-alignments consolidation.

There is thus provided a method of aligning an inertial unit carried by a carrier static or quasi-static, wherein:

is implemented simultaneously more alignment treatment dimensioned with different alignment observation times,

determining the minimum observation time for which the consistency of alignment information obtained is checked by testing the consistency of alignment information obtained for the smallest observation time and optionally also the consistency of alignment information obtained for higher observation times,

Alignment information is determined based on alignment information determined for the minimum observation time.

Such a process is advantageously completed by the following different characteristics taken alone or in combination:

the selected alignment information is the last alignment information determined for the minimum observation period;

the selected alignment information is an average of the last alignment information determined for the minimum duration of observation and alignment information determined just before by the same treatment and the same minimum observation period;

the selected alignment information is alignment information determined with a treatment:

• which is sized to the same amplitude of carrier movements the alignment processing implemented with the minimum selected observation period, and

• which is implemented for a longer observation period.

The invention also relates to an inertial system, characterized in that it comprises processing means adapted to implement the aforesaid method.

PRESENTATION DES FIGURES

Other features and advantages of the invention still will match the following description, which is purely illustrative and non-limiting and should be read with reference to the appended figures in which:

- Figures 1 and 2 each schematically show, firstly a curve of the critical motion of the carrier, and secondly a curve giving the error of maximum curvature in the case of a polynomial estimation of order 2 and a polynomial estimation of order 3;

Figures 3a and 3b illustrate two possible embodiments of an inertial system in accordance with various embodiments of the invention;

Figure 4 is a schematic diagram giving the various stages of a possible treatment for the implementation of the invention;

Figures 5 to 7 illustrate three possible implementation modes for the invention.

DETAILED DESCRIPTION OF THE INVENTION

General information on the inertial system

In a first embodiment, with reference to Figure 3a, a system

(Inertial) comprises a processing unit 1 and an inertial navigation 2.

The one data processing unit typically includes at least one processor to implement a computer program. This computer program comprises program code instructions configured to implement the improved alignment device 10. This processing unit may share one of the inertial unit 2 mounted on a static or animated carrier parasitic movements such that a vehicle, a boat or an aircraft. The two inertial unit comprises a navigation unit 20 and the inertial sensor 22. The inertial unit 2 may also comprise a user interface.

The inertial sensors 22 comprise for example accelerometers providing measurements of Sa and gyroscopes providing Sg measures.

The navigation unit 20, known in itself, is particularly configured to implement an inertial location and a coarse alignment which performs

the estimation of attitudes within a few degrees in accordance with a method known to the art. The navigation unit 20 for example includes at least one processor configured to execute a maintenance algorithm of the inertial location. Furthermore, the improved alignment device 10 and the navigation unit 20 may be self physical components or grouped together in a single housing.

The processing unit 1 further comprises a memory 12 for storing data including received measurements of the inertial unit 2.

Alternatively (Figure 3b), the system comprises a processing unit 1 and an Inertial Measurement Unit 2a also called UMI.

Examples of alignment treatment

The system shown in Figures 3a or 3b for example implements the steps shown in Figure 4.

In a first step, the alignment stage is started and an initial alignment information is performed (step 1). The initial values ​​of attitude, speed and position and optionally used are, for example raw acquisition data provided by various sensors of the carrier on which the system is placed. Many other techniques are also possible for this initialization.

Following this step 1, the system runs in parallel alignment different treatments (specifically referenced from 1 to n in the figure, with no finite integer).

Each of these treatments is to observation periods (D1, ..., Dn) different.

They are also dimensioned with different amplitudes of carrier movement.

Thus, for a first period of observation D1:

the system calculates a first alignment information S1 c_1 set by implementing an alignment processing (step 21) dimensioned to amplitude S1 of the wearer's movements; this calculation is made from the initial data provided in step 1 and the speed error and / or horizontal movements of the inertial location acquired in step 8; the game information S1 c_1 obtained is stored (step 21 a);

the calculation of step 21 is then iterated one or more times, using each time the game information obtained with the previous alignment processing as input data (the alignment information of game S1 c_1 to calculate the alignment information of game c_2 S1 in the case of step 31); the treatment is in each case identical to that implemented in step 21 (same dimensioning S1 amplitude and the same duration D1);

at the end of the iterations, a consistency check is carried out between the alignment information obtained (step 41). For example, in the case of Figure 4 where there is an iteration, it checks the consistency of the last two alignment information sets S1 S1 c_1 and c_2 obtained;

when the result of the checking step is that the information is considered consistent, alignment is considered acceptable and as finished (step 51); a set of alignment values ​​is then determined; this alignment game may correspond to the previous game S1 of c_2 obtained alignment information for this treatment or game recalculated based on the game information S1 c_2 and one or more other sets of information determined in parallel for higher observation times (step 61) (e.g., a precision alignment treatment S1 p_1 implemented with a sizing S1 for the movement of carriers and a longer duration than D1 (e.g. 2 times the duration D1) - see below).
As will be understood, the system implements parallel steps similar to one or more other all distinct periods of observation of each other with different alignment treatment sized for different amplitudes.

In particular, for Dn observation period and amplitude Sn sizing, it implements the following steps in parallel with steps 21 to 61:

2n steps and 2na: calculating a first set of alignment Snc_1 and storing information;
3n step: calculation iteration and in particular a Snc_2 alignment;
4n step: consistency check;
5n step: end of the alignment and output of a set of estimated alignment information,
6n step of: alignment processing implemented in parallel for a duration greater than Dn (e.g. 2 times Dn) and for the same dimensioning
Sn d'amplitude.
In the example in FIG 4, when the result of the check carried out in step 41 is that the alignment information obtained are not consistent, it triggers the checking for the upper right observation period and this is iterated verification process by changing successively time to find an observation period for which consistency is verified. Of course, these verification processes can be conducted in parallel.
If none of the treatments conducted for various periods D1, Dn do to verify consistency, there is failure of alignment. A message to that effect is provided (step 7 output Sn coherence test of the highest Dn duration).
The different durations D1, Dn are selected for example according to a geometric sequence.
The number n of selected durations may typically be between 3 and 5.
For example: D1 = 2 min 30 s, D2 = 5 min, 10 min = D3, D4 = 15 min, D5 = 30 min. In the case illustrated in Figure 5, the alignment treatment uses three times alignment chosen according to a geometric progression of ratio 2.

At time Ta, one can verify the consistency of information estimated by the alignments sized for S1 amplitude of carrier movements No. S1 S1 c_1 and c_2.
In case of consistency, alignment is complete and it may provide the information of the last alignment c_2 S1 or the average information of two alignments S1 S1 c_1 and c_2.
In case of inconsistency, two choices are possible:
setting up a ratchet effect no longer operator alignments sized for S1 amplitude of carrier movements.
we continue to exploit the alignments designed for a range of S1 bearer movements since the wearer's movements are not necessarily stationary. The description of this example is with this choice.
The alignment is then extended by exploiting alignments sized for S2 amplitude of wearer movements.
At time Tb, one can check the consistency of the information estimated by S2c_1 S2c_2 and alignments.

In case of consistency, alignment is complete and can provide the final alignment S2c_2 information or the average information of two alignments S2c_1 and S2c_2.
In case of inconsistency, the alignment extends operator alignments dimensioned for an S3 amplitude of carrier movements.
At time Te, one can check the consistency of information estimated by S3c_1 S3c_2 and alignments.
In case of consistency, alignment is complete and can provide the final alignment S3c_2 information or the average information of two alignments S3c_1 and S3c_2.

The example illustrated in Figure 6 is a variation of the first example by making the assumption that the S3 amplitude is always larger than the amplitude of carrier movements.
In this case, it is unnecessary to check the consistency of alignment dimensioned for an S3 amplitude movements of the holders. No checks are then implemented to this amplitude.
In another variant (Figure 7), it can also improve this first example by adding dedicated alignment accuracy rated S1 and S2p p * * (* represents the alignment of embodiment number). Alignments dedicated to consistency are noted S1 c_ * and * S2c_.

Alignments type S1 p * are set in the same way as the type of alignments S1 c_ * except that they continue to observe a period of 2 twice as large: if consistency 2 alignments c_1 S1 and S1 c_2, alignment is complete and

is the estimate given by S1 p_1. In case of inconsistency of these alignments, we test the coherence of two alignments S2c_1 and S2c_2.

If these alignments are consistent then the alignment is completed and corresponds to the estimate given by S2p_1.
Consistency check
The consistency check between 2 alignments in different examples uses tests on the absolute values ​​of deviations of the estimated speeds on horizontal errors and 3 errors vis-à-vis attitudes details of the estimated 2 alignments . Thus it is possible to use the following treatments to achieve the consistency check.

SI (alignment # 1 has failed)

OR (alignment # 2 failed)

OR abs (X axis velocity error of aln 1 - speed error X axis of aln # 2)> S_VX
OR abs (error rate Y of # 1 aln - speed error of the Y axis aln # 2)> S_VY
OR abs (X axis attitude error of aln 1 - X axis attitude error of aln 2)> S_AX
OR abs (Y attitude error of aln 1 - attitude error Y of aln 2)> S_AY
OR abs (Z axis attitude error of aln 1 - Z axis attitude error of aln 2)> S_AZ
SO
Consistency = KO
IF NOT
Consistency = OK
END YES
The S_VX thresholds S_VY, S_AX, S_AY S_AZ and are calculated as follows:
S_VX = K * RacineCarrée ((speed error uncertainty on X axis of aln No. 1) 2
+ (Speed error uncertainty on X axis of aln # 2) 2 ) = K * S_VY RacineCarrée ((speed error uncertainty on Y axis of aln No. 1) 2
+ (Speed error uncertainty on Y axis of aln # 2) 2 )
S_AX = K * RacineCarrée ((uncertainty on the attitude error of axis X aln No. 1) 2
+ (Uncertainty on the attitude error of axis X aln # 2) 2 )
S_AY = K * RacineCarrée ((uncertainty on the attitude error of the Y axis aln No. 1) 2
+ (Uncertainty on the attitude error of the Y axis aln # 2) 2 ) = K * S_AZ RacineCarrée ((uncertainty on the attitude error of axis Z aln No. 1) 2
+ (Uncertainty on the attitude error of axis Z aln # 2) 2 )
where K is a fixed constant chosen according to the desired false alarm probability Pfa that is to say depending on the probability of declaring two wrongly incoherent alignments. Usually the false alarm probability is calculated from the quantile of a Gaussian distribution. Similarly, the minimal bias detected with a probability of not detecting Pnd is calculated from the risk of the second kind to a Gaussian distribution. The following table provides some possible values ​​for the constant K and adjustment of the value of the minimum detectable through K2 for Pnd chosen equal to 0.1%.
The lowest detectable values ​​on speed and attitude errors of the inertial location are calculated as follows:
VX_détect = K2 * RacineCarrée ((speed error uncertainty on X axis of aln No. 1) 2
+ (Speed error uncertainty on X axis of aln # 2) 2 )
VY_détect = K2 * RacineCarrée ((speed error uncertainty on Y axis of aln No. 1) 2
+ (Speed error uncertainty on Y axis of aln # 2) 2 ) = K2 * AX_détect RacineCarrée ((uncertainty on the attitude error of axis X aln No. 1) 2
+ (Uncertainty on the attitude error of axis X aln # 2) 2 )
AY_détect = K2 * RacineCarrée ((Uncertainty attitude Terror Y Tain # 1) 2
+ (Uncertainty Terror attitude Y Tain # 2) 2 )
AZ_détect = K2 * RacineCarrée ((Uncertainty attitude Terror Z axis Tain # 1) 2
+ (Uncertainty Terror attitude axis Z Tain # 2) 2 )

WE CLAIMS

1. A method of aligning an inertial unit carried by a carrier static or quasi-static, wherein:
implement is brought simultaneously more alignment treatment dimensioned for several wearer's movements with amplitudes separate alignment observation times,
a minimum observation period is determined corresponding to the alignment observation period for which the consistency of the alignment obtained through treatment dimensioned alignment information for a given amplitude of the wearer's movement is verified, it is determined the alignment information based on the alignment information determined for the minimum observation time.
2. The method of claim 1, wherein the selected alignment information is the last alignment information determined for the minimum observation time.
3. The method of claim 1, wherein the selected alignment information is an average of the last alignment information determined for the minimum duration of observation and alignment information determined previously by the same treatment and the same minimum observation time.
4. The method of claim 1, wherein the selected alignment information is alignment information determined with a treatment: that is sized the amplitude of carrier movements of the alignment treatment implemented with the minimum of selected observation, and
which is implemented for a longer observation period.
5. The method of claim 4, wherein the longer observation period is equal to twice the minimum observation period.
6. Method according to one of the preceding claims wherein observation of the durations of the various alignment treatments are chosen according to a geometric sequence.
7. The method of claim 6, wherein the geometric progression of common ratio 2.
8. Method according to one of the preceding claims, wherein the alignment information includes attitude values ​​and / or speed and / or position.
9. inertial system, characterized in that it comprises processing means adapted to implement the method according to one of the preceding claims.