METHOD FOR CALIBRATING A GYROMETER BUILT INTO AN OBJECT

Description

GENERAL TECHNICAL FIELD

[0001] The present invention relates to the field of navigation without GNSS.

[0002] More specifically, it concerns a method of calibrating a gyrometer attached to magnetometers. STATE OF THE ART

[0003] It is now common to track the position of a vehicle by GNSS (Global Navigation Satellite System, for example GPS) or by using a wireless communication network (triangulation using transmitting terminals, Wi-Fi network or others).

[0004] These methods are very limited because they do not ensure the availability and accuracy of information, both of which are affected by possible masking between the sources and the receiver. They also depend on external technologies such as satellites for GNSS, which may be unavailable or even deliberately jammed.

[0005] Alternatively, there are also known "autonomous" methods for tracking the relative movement of a vehicle in any environment, for example using an inertial or magneto-inertial unit. By relative movement, we mean the trajectory of the vehicle in space relative to a point and a reference frame given at initialization. In addition to the trajectory, these methods also make it possible to obtain the orientation of the vehicle relative to the same initial reference frame.

[0006] A navigation-class inertial unit consists of at least three accelerometers and three gyrometers arranged in a triaxial arrangement. Typically, the gyrometers "maintain" a reference frame, in which a double time integration of the accelerometer measurements makes it possible to estimate the movement.

[0007] It is well known that in order to use classical inertial navigation methods, as implemented in heavy applications such as the navigation of fighter or airliner aircraft, submarines, ships, etc., it is necessary to use very high precision sensors. Indeed, the double time integration of an acceleration measurement means that a constant acceleration error creates a position error that increases proportionally to the square of the time.

[0008] Alternatively, it is known that if the velocity vector information in the carrier's frame of reference can be provided by an external source (odometry on a car, log on a ship, Pitot tube on an airplane), it is possible to obtain the carrier's trajectory as soon as we are able to know its orientation and in particular its heading.

[0009] In a number of cases, the initial heading is known (initial "alignment" of an inertial device) and can be obtained by sensors other than inertial (e.g. magnetic) or deduced directly from inertial sensors when these allow the Earth's rotation to be measured precisely.

[0010] The difficulty when working with low-cost sensors is to "increase" their performance by continuously calibrating their error parameters, so as to best maintain the orientation information of a carrier, in particular the heading information on a land vehicle.

[0011] It was therefore proposed in this context to calibrate a gyrometer using magnetic measurements.

[0012] Application US2012 / 0116716 proposes to examine whether the evolution of a magnetic field between two measurement points corresponds to the change in attitude predicted by the integration of gyrometers. However, it assumes that the magnetometric / accelerometric measurements are reliable, which is difficult. Indeed, we observe in practice that: the ambient magnetic field is not always homogeneous and stationary; metallic elements (for example a body) or nearby magnets have an impact on the magnetic measurement (soft iron and hard iron type effects), see also application US2004 / 0123474. To correctly calibrate a magnetometer, it is necessary to place it in a maximum spatial orientation so as to "excite the error model" to have observability. This problem is easily soluble for equipment that can be carried by hand and thus turned in all directions, see for example application FR1757082, before using it. If, on the contrary, the magnetometer is fitted to a vehicle, we are constrained: ∘ by the fact that during a typical journey of a vehicle, roll and pitch deviate from zero only slightly.∘ by the fact that the body is magnetized, or that the vehicle has been loaded, which means that the calibration may no longer be relevant; which will require regular recalibrations.

[0013] Application US2011 / 0178707 relates more specifically to smartphone-type devices comprising a magnetic compass and a gyrometer, and addresses the problem of magnetic disturbances. This application is aimed more at calibrating the magnetic compass from gyrometric data, but also proposes the reverse (i.e. calibrating a gyrometer from magnetic data), taking a quaternion as its starting point. q magneto-accelerometric and obtaining the angular velocity of recalibration from this quaternion ω= 2 q -1< q̇.However, it is still assumed that the magnetometric / accelerometric measurements are reliable, and if not it is simply proposed either not to update the gyrometer bias while this situation lasts, or to use a default quaternion stored in a memory.

[0014] Document US 2012 / 259572 A1 discloses a method for calibrating a gyrometer which measures an angular velocity of an object moving in an ambient magnetic field, the method using magnetic measuring means secured to said object to measure at least two components of the magnetic field and determining by data processing means values ​​of at least one calibration parameter of the gyrometer minimizing an expression defined by an estimated angular velocity of the object and a magnetic equation on the components of the magnetic field, the method detecting a condition of stationarity and uniformity of the magnetic field to calibrate the gyrometer.

[0015] It would be desirable to have a new method for calibrating an object's gyrometer in order to estimate the movement of this object which allows excellent quality of result and is not restrictive. PRESENTATION OF THE INVENTION

[0016] The present invention thus relates, according to a first aspect, to a method for calibrating a gyrometer equipping an object moving in an ambient magnetic field, comprising steps a), b), c) and d) according to independent claim 1.

[0017] According to other advantageous and non-limiting characteristics: the estimated angular velocity ω gyro estimation of the object is related to the measured angular velocity ω gyro mesure by a model ω gyro estimation = D ⋅ ω gyro mesure + b gyro , Or D And b gyroare the calibration parameters of the gyrometer; the method comprises a step (e) of re-estimating said parameter representative of an error on the calibration parameters; said parameter representative of an error on the calibration parameters is a function of the value of said first expression, or where appropriate of said second expression, calculated for the determined values ​​of the calibration parameters; step (c), and where appropriate step (e), comprises determining an accumulated error interval as a function of the value of said parameter representative of an error on the calibration parameters and a duration since the last calibration of the gyrometer, said predetermined threshold depending on said accumulated error interval;the components of the magnetic field and / or of an i-th derivative of the magnetic field acquired in step (a) are measured components, and those used by said first and second magnetic equations are components estimated as a function of the measured components and of calibration parameters of the magnetic measuring means, step (b) and, where appropriate, step (d), also comprising the determination of values ​​of at least one calibration parameter of the magnetic measuring means; the magnetic measuring means are triaxial magnetometers, the estimated components; M ( estimate )< of the magnetic field being linked to the measured components M (measure)< by a model M (measure) < = A · M ( estimate )< + b magneto , Or A And b magneto are the calibration parameters of the magnetic measuring means; the first magnetic equation(s) are of the form ∇ n M ⋅ = f n ω × ∇ n M, and where appropriate the second magnetic equation(s) are of the form ∇ n M ⋅ = f n ω × ∇ n M + g n ∇ n + 1 M . V where ∇ n M is an n-th derivative of the magnetic field, ω angular velocity, V linear speed, and fn And gn predetermined functions; the first expression is a function of ∇ n M estιmation ⋅ − f n ω gyro estimation × ∇ n M ( estimate )< | and the second expression is a function of ∇ n M estιmation ⋅ − f n ω gyro estimation × ∇ n M estimation − g n ∇ n + 1 M estimation . V . said parameter representative of an error on the calibration parameters is either the average of ∇ n M estιmation ⋅ − f n ω gyro estimation × ∇ n M estimation 2 Or ∇ M <none / > estιmation <mprescripts / > n <none / > ⋅ − f n ω gyro estimation × ∇ n M estimation − g n ∇ n + 1 M estimation . V 2 over a given time interval, either all or part of a spatial gradient of said components of the magnetic field and / or of an i-th derivative of the magnetic field; the method comprises a step (f) of estimation by the data processing means of the movement of said object (1) as a function of the measured angular velocity of the object, of the components of the magnetic field and / or of an i-th derivative of the magnetic field, and / or of the possible measured linear velocity of the object, and of the values ​​of the calibration parameters, in which an orientation of the vehicle is estimated in step (f) solely as a function of the measured angular velocity of the vehicle and of the values ​​of the calibration parameters.

[0018] According to a second aspect, there is provided an object moving in an ambient magnetic field, comprising a gyrometer configured to acquire a measured angular velocity of the object, according to independent claim 13.

[0019] According to other advantageous and non-limiting characteristics: • The object is a wheeled vehicle, the additional acquisition means being at least two odometers fitted to the vehicle's wheels.

[0020] According to a third aspect, there is provided a computer program product, according to claim 14, comprising code instructions for executing a method according to the first aspect of the invention for calibrating a PRESENTATION OF FIGURES

[0021] Other characteristics and advantages of the present invention will appear on reading the following description of a preferred embodiment. This description will be given with reference to the appended drawings in which: there figure 1 represents an example of vehicle architecture for implementing the method according to the invention; the figure 2is a diagram representing the steps of a preferred embodiment of the method according to the invention. DETAILED DESCRIPTION Architecture

[0022] In reference to the figure 1 , the present method allows the calibration of a gyrometer 11 of an object 1 evolving in an ambient magnetic field (typically the Earth's magnetic field, possibly altered by nearby metallic objects), noted M . As already explained, the magnetic field is a vector field in this case three-dimensional, that is to say associating a vector of dimension three with each three-dimensional point in which the object 1 is mobile.

[0023] This object 1 can be any mobile object whose position knowledge is desired, for example a vehicle, in particular a wheeled vehicle, a drone, etc., but also a person or a part of this person's body (their hands, their head, etc.).

[0024] The object 1 is equipped with the gyrometer 11 and magnetic measuring means 20. The means 20 and the gyrometer 11 are integral with the object 1, for example integral with the bodywork in the case of a vehicle, and generally fixed in the reference frame of the object 1.

[0025] The magnetic measuring means 20 are typically magnetometers and / or magnetic gradiometers. In the latter case, instead of measuring the value of the components of the magnetic field M, the gradiometers directly measure the value of the components of the gradient of the magnetic field. M , i.e. the value of the spatial derivatives (possibly in addition to the magnetic field). Such magnetic gradiometers 20 are known to those skilled in the art. According to certain embodiments of the invention, it will be possible to use gradiometers directly measuring the value of the second derivatives (gradient of order two), and generally of the i-th derivatives (gradient of order i ).

[0026] In the remainder of the description we will take the example of magnetometers but those skilled in the art will be able to transpose this to gradiometers.

[0027] The magnetometer(s) 20 are preferably “axial”, i.e. capable of measuring a component of said magnetic field, i.e. the projection of said magnetic field vector. M along their axis (or, in the case of a gradiometer, a component of a spatial derivative of order i of said magnetic field, i.e. the variation of a component of ∇ i -1< M at the level of the gradiometer along its axis).

[0028] The magnetometers 20 are at least two in number, and preferably at least three in number, so as to be able to acquire respectively at least two or three components of the magnetic field. It will be understood in fact that two magnetometers may suffice, if they are arranged orthogonally to each other in a plane orthogonal to the vertical axis of rotation of the vehicle, i.e. a horizontal plane. Nevertheless, using three magnetometers 20 and acquiring three components makes it possible to overcome the need for this particular arrangement.

[0029] Advantageously, the magnetometers 20 are even greater than three in number, such as six or even nine, advantageously organized in groups of three in "triaxes", i.e. a triplet of two-by-two orthogonal magnetometers 20 associated with the same spatial position and measuring the magnetic field along the three axes (in the case of six magnetometers 20 there are two triaxes and in the case of nine magnetometers 20 there are three triaxes). Two triaxes make it possible to measure the magnetic field as well as part of its gradient (see below), and from eight magnetometers 20 which have been suitably positioned and oriented, all the components of the gradient can be determined. This is for example the case for three triaxes positioned at the corners of an isosceles rectangular triangle.

[0030] Preferably, the orthonormal reference frame associated with the object is chosen by convention (and for ease for the remainder of this description) such that the triaxes are advantageously oriented in accordance with said orthonormal reference frame, so as to further facilitate the calculations.

[0031] But the person skilled in the art will in any case be able to transpose this to any spatial arrangement of magnetometers / gradiometers.

[0032] By gyrometer 11 is meant inertial measurement means capable of measuring the angular velocity of the object 1 according to a system of three orthogonal axes, which define the reference frame of the object 1, i.e. measuring the three components of an angular velocity vector noted ω. It is thus understood that the gyrometer 11 can in fact designate a set of three gyrometers associated with one of the three axes, in particular in triaxis (i.e. each capable of measuring one of the three components of the angular velocity vector noted ω).

[0033] In the preferred embodiment where the object 1 is a vehicle, this reference frame preferably comprises a vertical axis and two horizontal axes, in particular a longitudinal axis and a transverse axis. The rotation around the vertical axis of the vehicle is described by the angle at which the driver acts by turning the steering wheel. On generally flat ground, the changes in direction of the vehicle are in the horizontal plane, i.e. also along said vertical axis of the yaw. In reality, non-zero values ​​for roll (rotation along the longitudinal axis of the vehicle) and pitch (rotation along the transverse axis of the vehicle) may be the result, for example, of a sloping road but are typically low.

[0034] The method which is the subject of the present invention makes it possible to calibrate the three components of the angular velocity vector ω.

[0035] The object 1 is further equipped with “additional” acquisition means 10 for a measured linear speed of the object 1 (denoted V), i.e. the displacement. These means 10 can directly or indirectly make it possible to obtain the linear speed, and thus be of numerous types, for example inertial measurement means. Thus the gyrometer 11 can be supplemented by one or more accelerometers, or even the object 1 can comprise a “6-axis” inertial unit. The accelerometers are sensitive to external forces other than gravitational forces applied to the sensor, and make it possible to measure a specific acceleration denoted γ.

[0036] Alternatively, the means 10 may consist, if the object 1 is a wheeled vehicle, of at least two odometers each for one wheel of the vehicle, for example the two rear wheels, as shown in the example of the figure 1. Note that a set of odometers is a simple and reliable way to obtain a simple linear speed.

[0037] An odometer is a device capable of measuring the speed of a wheel by counting the revolutions ("rev counter"). Generally, odometers have a part fixed to the wheel (for example a magnet), and detect each passage of this fixed part (called "top") so as to count the number of revolutions per unit of time, which is the rotation frequency. Other techniques are known, for example the optical detection of a mark on the wheel, or the magnetometer of patent FR2939514 which detects the rotation of a metallic object such as a wheel.

[0038] Here the "speed" of a wheel is a scalar, i.e. the norm of the speed of the wheel in the terrestrial frame of reference (assuming no skidding). If the radius r of the wheel is known, the measurement of the frequency fof rotation allows to estimate this speed standard: v =2 πrf.

[0039] Alternatively, the additional means 10 may be a GNSS receiver. It will be understood that the person skilled in the art is not limited to any particular technique for measuring speed.

[0040] The object 1 further comprises, as explained, processing means 21 (typically a processor) for implementing, in particular in real time, the processing of the present method, for example an on-board computer of a vehicle, and possibly a memory 22 and an interface 23 for restoring information on the movement of the object 1 (an instantaneous speed value, a heading, a position on a map, etc.), and / or sending commands to the object 1. The object 1 may in this respect be in particular an autonomous vehicle, and the processing means 21 configured to implement the autonomous navigation of the vehicle. Thus, said commands are sent to the control members of the vehicle (engine, steering wheel actuators, etc.) so as to simulate driving by the driver.

[0041] Note that the processing means 21 may be external to the object 1, and for example connected to the latter by a wireless network. Alternatively, the gyrometer 11 and the magnetic measuring means 20 may be connected to the data processing means 21 in a wired manner, for example via Ethernet. Process

[0042] The method which is the subject of the present invention is a method for calibrating at least the gyrometer 11. By calibration, we mean the determination of one or more calibration parameters, a list of which will be seen later. In particular, certain calibration parameters can be considered reliable and predetermined, and others to be determined. With regard to those to be determined, it can be provided that they have “current” values ​​(in other words that a calibration has already taken place), and that these values ​​will, if necessary, be modified (in the event of a new calibration).

[0043] In a particularly preferred embodiment, the method may further be a method for calibrating the magnetic measuring means 20, i.e. the gyrometer 11 and the magnetometers 20 may be calibrated simultaneously. This is an extremely advantageous mode, since as will be seen there is no longer even any need to assume that one of the gyrometry (the gyrometric data) and the magnetometry (the magnetic data) is reliable to use it as a reference to calibrate the other: the two automatically calibrate each other. Alternatively, it will of course be possible to consider the magnetometers 20 as well calibrated and to calibrate the gyrometer 11 accordingly, which makes it possible for example to calibrate more parameters of the gyrometer 11.

[0044] As will be seen, in an advantageous embodiment the present method is even a method for estimating the movement of the object 1, i.e. it comprises, following the calibration, the use of the measurements to reliably deduce one or more components of the movement.

[0045] In a first step (a), the method comprises the acquisition by the gyrometer 11 of a measured angular velocity of the object 1, noted ω gyro mesure , and by the magnetic measuring means 20, of at least two components of the magnetic field and / or of an i-th derivative of the magnetic field. These components are more precisely so-called measured components, forming a vector noted M (measure)< in the case of the magnetic field (in the case of an i-th derivative, we can note ∇ i M ( measure )< ) .Preferably, three components of the angular velocity and at least six or even eight magnetic components are acquired (advantageously two or even three magnetometric triaxes positioned at the corners of an isosceles rectangular triangle), so as to be able to deduce the field and at least part of its gradient (i.e. the derivatives to the following order along the three axes).

[0046] These quantities are advantageously measured with dt sampling (i.e. every “dt” seconds) with dt very small compared to the characteristic time of the movements of object 1, typically 40 ms.

[0047] In a step (b), the data processing means 21 determine the values ​​of at least one calibration parameter of the gyrometer 11 minimizing a first expression defined by an estimated angular speed of the object 1 and at least a first magnetic equation on the components of the magnetic field and / or an i-th derivative of the magnetic field.

[0048] The idea is to separately estimate gyrometric data and magnetic data which are theoretically linked by a magnetic equation. Thus, thanks to the equation, we can express a quantity which ideally should be zero (i.e. the gyrometric and magnetic data exactly verify the magnetic equation), otherwise the calibration needs to be perfected.

[0049] The estimated angular velocity of object 1, denoted ω gyro estimation , is a function of the measured angular velocity and calibration parameters of the gyrometer 11, i.e. in general ω gyro estimation = g ω gyro mesure with g a function (application) that is determined by the calibration parameters. In the simplest case, g is affine and ω gyro estimation is related to the measured angular velocity ω gyro mesure by a formula of the type ω gyro estimation = D ⋅ ω gyro mesure − b gyro , Or D And b gyro represent the calibration parameters of the gyrometer 11.

[0050] In the case of a three-dimensional attitude, b gyro is a bias vector and D is a 3x3 matrix = (orthogonal matrix for transition to the correct reference frame) × (upper triangular matrix containing the scale factors and the adjustments). In a simplified manner, we can for example consider that D is predetermined (in practice it varies very slowly) and that the only calibration parameter to be determined for the gyrometer 11 is b gyro , which actually tends to vary over time (we speak of gyrometer drift 11).

[0051] As explained, the components of the magnetic field and / or of an i-th derivative of the magnetic field acquired in step (a) are preferentially measured components (i.e. potentially biased), and those used by said first magnetic equation are estimated components (the corresponding vector is noted M ( estimate)< in the case of the magnetic field - in the case of an i-th derivative, we can note ∇ i M ( estimate )< ), depending on the measured components and calibration parameters of the magnetic measuring means 20.

[0052] In the case where the magnetic measuring means 20 are magnetometers (i.e. we work on the components of the magnetic field and not a derivative), the estimated components M ( estimate )< of the magnetic field are advantageously linked to the measured components M ( measure )< by a model of the form M ( measure )< = A · M ( estimate )< + b magneto , where the 3x3 matrix A and the vector b magneto are the calibration parameters of the magnetic measuring means 20. To the extent that it is rather the estimated component that we wish to express as a function of the measured component, we can write M ( estimate )< = A -1< · ( M ( measure)< - b magneto ) .

[0053] b magneto is generally representative of “hard iron type” effects, and matrix A is representative of “soft iron type” effects. These effects correspond to hard / soft iron effects strictly speaking but can also include phenomena with the same impact but a different cause (e.g. effects due to the electronics / physics of magnetometer 20).

[0054] As with the gyrometer, more generally, we can look at a general error model : M ( estimate )< = h ( M ( measure )< ) with h a function (application) which is not necessarily affine.

[0055] In the affine case ( M ( measure )< = A · M ( estimate )< + b magneto ) , A is a 3×3 matrix, which can be written as A=R·Ã with R orthogonal and  upper triangular.

[0056] Note that it is assumed that the magnetic measuring means 20 themselves are a priori perfectly calibrated, it is only their direct environment (bodywork, etc.) which must be corrected in calibration in order to be able to estimate the Earth's magnetic field which would be measured in the absence of object 1. The matrix R is thus important and to be determined so that the object magnetic reference remains stable.

[0057] Classical calibration methods with respect to spheres do not determine R than a rotation around an axis (see the document E. Dorveaux, D. Vissière, AP Martin, N. Petit, "Iterative calibration method for inertial and magnetic sensors", in Proc. of the 48th IEEE Conf. on Decision and Control 2009 ). The sensor axes do not remain fixed after applying these methods.

[0058] In practice, we can determine eight coefficients of Ain calibration (a coefficient, the overall scale factor that relates the measurement to the physical unit in Tesla (or Gauss), cannot be determined, but at the same time, it is not necessary to know it to calibrate the gyrometer 11).

[0059] Thus, in the particularly preferred embodiment of simultaneous dual calibration, there are only three calibration parameters to be determined: b gyro , A And b magneto , This limited number of parameters makes such a double calibration possible as soon as object 1 presents a “varied” trajectory (directions / speeds which present a great diversity of values ​​as a function of time). Magnetic equation

[0060] The evolution of the magnetic field in the frame of object 1 (called body ) over time is described by the equation M ˙ = − ω × M + ∇ M ⋅ V + ∂ M ∂ t .

[0061] The first term -ω × Mdescribes the change in the field in the frame of object 1 which is due to the rotation of this frame relative to the fixed frame. The second term ∇ M · V describes the change in the measured field resulting from motion in a region with an inhomogeneous field. Finally, the third term ∂ M ∂ t takes into account the instationarities of the magnetic field (e.g. periodic currents such as 50 Hz in Europe, movements of a magnet / steel object near said object, etc.).

[0062] If there is no "mobile" magnetic disturbance, i.e. the magnetic field has a stationary character, ∂ M ∂ t ≈ 0 , and we can reduce the equation to Ṁ = - ω × M + VM · V.

[0063] If there is no magnetic disturbance at all, i.e. the magnetic field has a uniform character in addition to being stationary, ∇ M ≈ 0 and we can further reduce the equation to Ṁ = - ω × M.

[0064] In the case of using magnetic sensors to determine a heading, this double hypothesis of uniformity and stationarity is most often verified or assumed (as when using a compass). Indeed, for the Earth's magnetic field, uniformity is locally observed in very good approximation in our latitudes, except in cases where there is, for example, a metallic or reinforced concrete structure (typically a bridge) in the vicinity. Unsteadiness due to the movement of nearby objects is even rarer because the sources of disturbance would have to move, and possible unsteadiness due to periodic currents are on average zero and can also be treated with estimation techniques, see for example document C.-I. Chesneau, M. Hillion, and C. Prieur, Motion estimation of a Rigid Body with an EKF using Magneto-Inertial Measurements, 7th Conf.on Indoor Positioning and Indoor Navigation (IPIN'16), Madrid, Spain, 2016.

[0065] We note that we know the possibility of generalizing the magnetic equations (reduced or not), so as to link not the spatial and temporal variations of the magnetic field evaluated using magnetometer measurements, but the spatial and temporal variations of the magnetic gradients (i.e. the spatial variations of the magnetic field) evaluated in particular using gradiometer measurements. More precisely, the equation Ṁ = - ω × M + VM · V derives into ∇ n M ⋅ = f n ω . ∇ n M + g n ∇ n + 1 M . V , with n + 1 ≥ i ≥ 1, where ∇ n M is an nth derivative of the magnetic field, and fn And gn predetermined functions, see application FR1756675.

[0066] In the remainder of this description, we will use for convenience the equations of order 0 (i.e. based on Ṁ = - ω× M + ∇ M · V ), but the person skilled in the art will know how to transpose the process to higher orders.

[0067] In reference to the figure 2 , the present calibration method cleverly proposes a calibration at several levels of hypotheses.

[0068] More precisely, in step (b) we initially attempt to calibrate the gyrometer by making the double hypothesis of uniformity and stationarity of the field, then in a possible step (c) of estimation of a parameter representative of an error on the calibration parameters we verify the absence of disturbances, i.e. the relevance of the hypotheses.

[0069] According to the invention, if this verification fails (said parameter representative of an error is greater than a predetermined threshold), the calibration is retried in a step (d) by only making the hypothesis of stationarity (and no longer that of uniformity). The relevance of this hypothesis can then be rechecked in a step (e). Alternatively, if the verification of step (c) fails (said parameter representative of an error is greater than a predetermined threshold), the process can be stopped and the current calibration kept.

[0070] It is understood that the presence of step (d) makes it possible to carry out the calibration in any cases where the magnetic field is not uniform, provided that the additional acquisition means 10 mentioned above are available. Nevertheless, according to an aspect not belonging to the present invention, it is possible to be satisfied with having only the gyrometer 11 and the magnetic measurement means 20, and to implement only steps (a), (b) and possibly (c).

[0071] Step (d) thus comprises the acquisition by the additional acquisition means 10 of a measured linear speed of the object 1, noted V(in practice this acquisition can be implemented concomitantly with the other acquisitions of step (a), in particular at each time step), then the re-determination by the data processing means 21 of values ​​of the at least one calibration parameter of the gyrometer 11 minimizing a second expression defined by the estimated angular speed of the object 1 and at least one second magnetic equation on the linear speed and the components of the magnetic field and / or an i-th derivative of the magnetic field, the at least one second magnetic equation assuming only a stationary character of the magnetic field at the level of the magnetic measuring means 20.

[0072] Here we will understand "acquisition" in the broad sense, i.e. direct or indirect. In practice, the means 10 will most often measure a first quantity which is not the linear speed (accelerations, the rotation frequency of a wheel, etc.), and the so-called measured linear speed is in fact estimated from these quantities, possibly combined. Those skilled in the art will know how to obtain a linear speed of the object 1 in many ways, for example by magnetic gradiometry.

[0073] The at least one second equation assuming only a stationary character of the magnetic field at the level of the magnetic measuring means 20 used in step (d) is thus preferably of the form ∇ n M ⋅ = f n ω × ∇ n M + gn (∇ n+1 M ) .V, with ω the theoretical angular velocity, and as explained Ṁ = - ω × M + VB · Vin the preferred embodiment in which a component of the magnetic field is directly measured.

[0074] It is therefore understood that in a hypothesis of stationarity of the magnetic field at the level of the magnetic measuring means 20, the quantity ∇ n M estιmatιon ⋅ − f n ω gyro estimation × ∇ n M estimation − g n ∇ n + 1 M estimation . V (and in particular its version at order 0 M estιmation . + ω gyro estimation × M estimation − ∇M ( estimate )< . V ) is equal to zero up to calibration errors.

[0075] Similarly, the at least one first magnetic equation assuming a uniform and stationary character of the magnetic field at the level of the magnetic measuring means 20 used in step (b) is a simplified version of the at least one second equation, and is preferably of the form ∇ n M ⋅ = fn ( ω ) × ∇ n M , with as explained Ṁ = - ω × Min the preferred embodiment in which components of the magnetic field are measured directly. It is therefore understood that under a double assumption of uniformity and stationarity of the magnetic field at the level of the magnetic measuring means 20, the quantity ∇ n M estιmatιon ⋅ − f n ω gyro estimation × ∇ n M estimation (and in particular its version at order 0 M estιmatιon ⋅ + ω gyro estimation × M estimation ) is equal to zero up to calibration errors.

[0076] We see that the first calibration is easy to implement because it does not require knowledge of the linear speed V .

[0077] Advantageously, the first expression (minimized in step (b)) is a function of M estιmatιon ⋅ + ω gyro estimation × M estimation , and preferably of M estιmatιon ⋅ + ω gyro estimation × M estimation or of M estιmatιon ⋅ + ω gyro estimation × M (estimate )< ) 2< ; and the second expression (minimized in step (d)) is a function of M estιmatιon ⋅ + ω gyro estimation × − ∇ M estimation . V , and preferably of M estιmatιon ⋅ + ω gyro estimation × − ∇ M estimation . V Or M estimation ⋅ + ω gyro estimation × - ∇ M ( estimate )< . V ) 2< .Under generic conditions, minimizing these expressions will give the unambiguous parameters for a generic trajectory. It will be understood, however, that alternatively, the person skilled in the art may use as the first / second expression any other functional sensitive to deviations from the relationship defined by the first / second equation (norm L 2< , L ∞< , etc.).

[0078] To implement this minimization, the data processing means 21 can work over time on an interval of a given length. For this purpose, in a known manner a recursive filter (RLS methods, recursive least squares, etc.) or an optimization (least squares method, etc.) can be used.

[0079] For example, assuming that the calibration parameters of the magnetometers 20 are determined simultaneously, i.e. the parameters A And b magneto ,we can implement a principle called differential calibration in which we minimize ∫ M estimation ⋅ + ω gyro estimation × M estimation 2 dt Or ∫ M estimation ⋅ + ω gyro estimation × M estimation − ∇ M estimation . V 2 dt , with ω gyro estimation = D ⋅ ω gyro mesure + b And M ( estimate )< = A -1< · ( M ( measure )< - b magneto ) to least squares. Error characterization

[0080] As explained, the method advantageously comprises a step (c) of estimating a parameter representative of an error in the calibration parameters, and even more advantageously (if steps (c) and (d) have taken place) a step (e) of new estimation of said parameter representative of an error in the calibration parameters.

[0081] The idea is to estimate the quality of the information provided by the magnetic measurement means 20 to validate or not the levels of hypotheses on the ambient magnetic field. More precisely, and as we see on the figure 2 : if the first estimate (step (c)) of the parameter representing an error is lower than said predetermined threshold, the double hypothesis of uniformity and stationarity of the ambient magnetic field is validated and the result of the calibration of step (b) is accepted.Step (c) then comprises the effective calibration of the gyrometer 11 and / or the magnetic measuring means 20 with the values ​​determined in step (b) of the calibration parameters, and there is no need to implement steps (d) and (e); if the first estimate (step (c)) of the parameter representative of an error is greater than said predetermined threshold, the double hypothesis of uniformity and stationarity of the ambient magnetic field is rejected and the result of the calibration of step (b) is not accepted, one can either stop or implement the recalibration of step (d): ∘ if the second estimate (step (e)) of the parameter representative of an error is this time less than said predetermined threshold, the hypothesis of only stationarity of the ambient magnetic field is validated and the result of the calibration of step (d) is accepted.Step (e) then comprises the effective calibration of the gyrometer 11 and / or the magnetic measuring means 20 with the values ​​determined in step (d) of the calibration parameters; ∘ if the second estimate (step (e)) of the parameter representing an error is still greater than said predetermined threshold, even the hypothesis of stationarity of the ambient magnetic field is rejected (there are disturbances) and the result of the calibration of step (d) is not accepted. The parameters of the old calibration are then kept as calibration. We can then either attempt to use the complete magnetic equation (with the term . ∂ M ∂ t ), or temporarily discard magnetic measurements.

[0082] It should be noted that determined values ​​of the calibration parameters during an occurrence of step (b) and / or (d), but not used for the actual calibration, may be stored on the data storage means 12, and used during a future occurrence of step (c) and / or (e). For example, it may be provided that as long as the parameter representative of an error is above the threshold, the determined calibration parameters are stored, and when it falls below the threshold, the actual calibration also takes into account the stored values. Furthermore, the result of steps (c) and (e) may depend on the age of the parameters of the old calibration (for example via the threshold value).

[0083] Note that an error interval accumulated after a duration t since the last good calibration can be estimated as a function of this duration t and the estimated error in this calibration, for example by adding to this error an increment which is proportional to the duration t. The predicted error interval can be used to characterize the result of a future calibration, and in particular to define the value of said predetermined threshold: for example, the larger the error interval, the more urgent it is to recalibrate the gyrometer, and thus the more "tolerant" one will be with the value of the parameter representing an error (i.e. the predetermined threshold will be higher). Furthermore, to the extent that the error interval increases proportionally faster when the estimated error is high, this makes it possible to keep "good" calibrations for a long time with a very low estimated error, while forcing "less good" calibrations to be recalibrated more quickly.

[0084] Generally, said parameter representing an error on the calibration parameters is a function of the value of the first / second expression (depending on whether it is the first or second occurrence of estimation of the parameter), calculated for the determined values ​​of the calibration parameters.

[0085] A first embodiment of step (c) / (e) is said to be intrinsic since it only uses quantities available in step (b) / (d). Preferably, the estimation residuals of step (b) / (d) are used, i.e. said parameter representative of an error is in particular the norm (e.g. L 2< or L ∞< ) of the value of the first / second expression over a given time interval, typically ∫ M estimation ⋅ + ω gyro estimation × M estimation 2 dt Or ∫ M estimation ⋅ + ω gyro estimation × M ( estimate )< - ∇M ( estimate )< . V | 2< dt.In such a case, step (c) / (e) can be implemented concomitantly with step (b) / (d). In the case of a recursive filter, one can use the norm (e.g. L 2< or L ∞< ) of the filter innovation over a given period of time.

[0086] Alternatively, one can quantify the deviation from a normal distribution, i.e. look at whether the statistic of M estimation ⋅ + ω gyro estimation . M estimation (Or M estimation ⋅ + ω gyro estimation . M estimation − ∇ M estimation . V ) is similar to a Gaussian or non-Gaussian noise statistic, for example by calculating moments of order greater than 2.

[0087] In a second embodiment of step (c) / (e), called extrinsic, quantities of another type are used, so as to calculate on the basis of one or other of these quantities a so-called reference angular speed, so as to compare it with the estimated angular speed. And those skilled in the art will know how to calculate such a reference angular speed from various types of quantities, and in particular it will be possible to use for example: a steering angle (if object 1 is a vehicle), GNSS data, e.g. GPS, (if / when available), odometric data.

[0088] In the latter case, if the additional acquisition means 10 of a linear speed consist of at least two odometers (always if the object 1 is a vehicle), said theoretical angular speed can in fact be obtained from the measured wheel speeds. In particular, in the case where the rear wheels are each equipped with an odometer, assuming the Ackermann differential geometry, a second estimated angular speed ω odo estimation of object 1 is related to the measured speeds v L , v R respectively for the left rear wheel and the right rear wheel by a formula of the type ω odo estimation = v R − v L d , with d the distance between the wheels, and said parameter representing an error is then in particular the norm (e.g. L 2< or L ∞< ) of ω ↑ , gyro estimation − ω odo estimation 2 over the given time interval, where ω ↑ , gyro estimation is the angular velocity for rotation alone around the vertical axis of object 1.

[0089] According to a third embodiment, with regard to the first occurrence (step (c)) of the estimation of the error parameter, one can simply estimate all or part of a spatial gradient of said component of the magnetic field and / or of an i-th derivative of the magnetic field, i.e. ∇ M or ∇ i+1 M according to the order of the second equation used.

[0090] Indeed, this estimate must be close to zero for the uniform character to be retained. We understand that the estimation of a subset of the components of the gradient can in practice be sufficient for this characterization, because each non-zero component is already a manifestation of an inhomogeneity.

[0091] Alternatively or in addition, learning can be used to improve the estimation of this error parameter and / or develop an approach to identifying favorable conditions (i.e. stationary and uniform) in a more robust manner and with increased availability.

[0092] In particular, learning mechanisms such as neural networks, support vector machines, nearest neighbor methods, decision tree forests, etc. can be implemented. Thus, at each occurrence of steps (b) to (e), a learning base can be enriched in which each set of measurement data is "tagged" with the corresponding value of the parameter representing an error, so as to progressively (as steps (b) to (e) are successively performed) and automatically learn to distinguish acceptable calibrations from unacceptable ones. Thus, the calibration improves itself continuously. Movement estimation

[0093] As explained, the method advantageously comprises a step (f) of estimation by the data processing means 21 of the movement of said object 1 as a function of the measured angular velocity of the object 1 and / or of the at least two components of the magnetic field and / or of an i-th derivative of the magnetic field, and / or of the possible measured linear velocity of the object 1 (by the means 10), and of the values ​​of the calibration parameters, i.e. after recalibration. This step (f), which is not shown on the figure 2 which focuses on calibration, can be implemented continuously in parallel.

[0094] By estimation of the movement, we mean in particular at least the estimation of an orientation of the object 1 (in the horizontal plane, i.e. a heading) and advantageously the estimation of a speed standard. The orientation is typically obtained by integration of the angular speed from a known initial orientation.

[0095] Preferably, the estimation by the data processing means 21 of the heading of said object 1 is made solely as a function of the measured angular velocity of the object 1 and the values ​​of the calibration parameters, i.e. using only the gyrometric data. In summary, the magnetic data are used to calibrate the gyrometer 11, and then only the gyrometric data are used for the heading estimation (since these are always available). Where appropriate, the measured linear velocity (obtained via the means 10) is used in step (f) only to determine an overall velocity of the object 1.

[0096] Note that in the case of an autonomous vehicle, step (f) may comprise the generation of a command for said vehicle 1 as a function of the estimated movement, so as to bring the vehicle 1, for example, to a desired destination, or to stop the vehicle 1 while keeping it in a trajectory free of obstacles. Equipment and object

[0097] According to a second aspect, the invention relates in particular to a set of equipment 11, 20, 21 and possibly 10 for implementing one or other of the embodiments of the method.

[0098] This assembly can be installed as a kit in a “classic” object 1 so as to transform it. Alternatively, in particular in the case where the object 1 is a vehicle, it can be an autonomous vehicle already equipped with data processing means 21 for the navigation of the vehicle, as well as sensors such as the gyrometer 11 and / or odometers serving as additional acquisition means 10.

[0099] In particular, an object 1 is proposed, in particular of the wheeled vehicle type, comprising: a gyrometer 11 configured to acquire a measured angular velocity of the object 1; magnetic measuring means 20 configured to acquire at least two components of the magnetic field and / or an i-th derivative of the magnetic field; where appropriate, additional acquisition means 10 configured to acquire a measured linear velocity of the object 1 (advantageously odometers equipping at least two of the wheels of the vehicle and configured to acquire measured speeds of said two wheels);data processing means 21 configured to ∘ determine values ​​of at least one calibration parameter of the gyrometer 11 minimizing a first expression defined by an estimated angular speed of the object 1 and at least one first magnetic equation on the components of the magnetic field and / or an i-th derivative of the magnetic field, ▪ the estimated angular speed of the object 1 being a function of the measured angular speed and of calibration parameters of the gyrometer 11, and ▪ the at least one first magnetic equation assuming a uniform and stationary character of the magnetic field at the magnetic measurement means 20, the data processing means 21 being configured to: ∘ Estimate a parameter representative of an error on the calibration parameters;and ∘ If said parameter representative of an error is greater than a predetermined threshold, re-determine from the values ​​of the at least one calibration parameter of the gyrometer 11 a second expression defined by the estimated angular speed of the object 1 and at least one second magnetic equation on the measured linear speed and the components of the magnetic field and / or of an i-th derivative of the magnetic field, the at least one second magnetic equation assuming only a stationary character of the magnetic field at the level of the magnetic measuring means 20. ;

[0100] As explained previously, the object 1 may further comprise a memory 22 and an interface 23, as well as other sensors such as a steering wheel angle sensor or a GNSS receiver.

[0101] Furthermore, the data processing means 21 may further be configured to estimate a movement of said object 1 as a function of the measured angular velocity of the object 1, of the components of the magnetic field and / or of an i-th derivative of the magnetic field, and / or of the possible measured linear velocity of the object 1 (where appropriate according to the result of a comparison of said parameter representative of an error with a threshold), and of the values ​​of the calibration parameters. Computer program product

[0102] According to a third and a fourth aspect, the invention relates to a computer program product comprising code instructions for the execution (on the processing means 21) of a method for calibrating a gyrometer 11 according to the first aspect of the invention, as well as storage means readable by computer equipment (for example data storage means 22) on which this computer program product is found.

Claims

1. Method for calibrating a gyrometer (11) of an object (1) moving in an ambient magnetic field, the method comprising the steps of: (a) acquisition - by the gyrometer (11), of a measured angular velocity of the object (1), and - by magnetic measuring means (20) which are secured to said object, (1) of at least two components of the magnetic field and / or of an i-th derivative of the magnetic field, around the magnetic measurement means (20); (b) determining, by data processing means (21), values of at least one calibration parameter of the gyrometer (11) minimising a first expression defined by an estimated angular velocity of the object (1) and at least one first magnetic equation on the components of the magnetic field and / or of an i-th derivative of the magnetic field, - the estimated angular velocity of the object (1) being a function of the measured angular velocity and of calibration parameters of the gyrometer (11), and - the at least one first magnetic equation assuming that the magnetic field is uniform and stationary around the magnetic measurement means (20); (c) estimating a parameter representative of an error on the calibration parameters. the method being characterised in that it further comprises the step of: (d) if said parameter representative of an error is greater than a predetermined threshold, acquiring by additional acquisition means (10) a measured linear velocity of the object (1) and re-determining by the data processing means (21) values of the at least one calibration parameter of the gyrometer (11) minimising a second expression defined by the estimated angular velocity of the object (1) and at least one second magnetic equation on the components of the magnetic field and / or of an i-th derivative of the magnetic field and on the measure linear velocity, the at least one second magnetic equation assuming only that the magnetic field is stationary around the magnetic measurement means (20)2. Method according to claim 1, wherein the estimated angular velocity ω gyro estimation of the object (1) is linked to the measured angular velocity ω gyro measurement by a model ω gyro estimation = D ⋅ ω gyro measurement + bgyro), where D and bgyro are the calibration parameters of the gyrometer (11).

3. Method according to claim 1 or 2, wherein, during step (c), said parameter representative of an error on the calibration parameters is a function of the value of said first expression calculated for the determined values of the calibration parameters.

4. Method according to one of claims 1 to 3, comprising a step (e) of a new estimating of said parameter representative of an error on the calibration parameters.

5. Method according to claim 4, wherein, during step (e), said parameter representative of an error on the calibration parameters is a function of the value of said second expression calculated for the determined values of the calibration parameters.

6. Method according to one of claims 1 to 5, wherein the step (c) comprises determining an accumulated error interval according to the value of said parameter representative of an error on the calibration parameters and of a duration since the last calibration of the gyrometer (11), said predetermined threshold depending on said accumulated error interval.

7. Method according to one of claims 1 to 6, wherein the components of the magnetic field and / or of an i-th derivative of the magnetic field acquired in step (a) are measured components, and those used by the first magnetic equation or equations and the second magnetic equation or equations, are estimated components as a function of the measured components and of calibration parameters of the magnetic measuring means (20), the step (b) and the step (d) also comprising the determining of values of at least one calibration parameter of the magnetic measuring means (20).

8. Method according to claim 7, wherein the magnetic measuring means (20) are tri-axes magnetometers, the estimated components M(estimation) of the magnetic field being linked to the measured components M(measurement) by a model M(measurement) = A · M(estimation) + bmagneto, where A and bmagneto are the calibration parameters of the magnetic measuring means (20).

9. Method according to one of claims 1 to 8, wherein the first magnetic equation or equations are of the form ∇ n M ⋅ = f n ω × ∇ n M, and the second magnetic equation or equations are of the form ∇ n M ⋅ = f n ω × ∇ n M + gn(∇n+1M).V, where ∇nM is an n-th derivative of the magnetic field, ω the angular velocity, V the linear velocity, and fn and gn predetermined functions.

10. Method according to claim 9, wherein the first expression is a function of ∇ n M estimation ⋅ − f n ω gyro estimation × ∇ n M estimation and the second expression is a function of ∇ n M estimation ⋅ − f n ω gyro estimation × ∇ n M(estimation) - gn(∇n+1M(estimation)).V.

11. Method according to one of claims 5 and 10 in combination, wherein said parameter representative of an error on the calibration parameters is either the mean of ∇ n M estimation . − f n ω gyro estimation × ∇ n M estimation 2 or ∇ n M estimation ⋅ − f n ω gyro estimation × ∇ n M estimation − g n ∇ n + 1 M estimation . V 2 over a given time interval, or all or a portion of a spatial gradient of said components of the magnetic field and / or of an i-th derivative of the magnetic field.

12. Method according to one of claims 1 to 11, comprising a step (f) of estimating, by the data processing means (21), the movement of said object (1) according to the measured angular velocity of the object (1), the components of the magnetic field and / or of an i-th derivative of the magnetic field, and / or of any measured linear velocity of the object (1), and of values of calibration parameters, wherein an orientation of the object (1) is estimated in step (f) only as a function of the measured angular velocity of the object (1) and of values of the calibration parameters.

13. Object (1) moving in an ambient magnetic field, comprising a gyrometer (11) configured to acquire a measured angular velocity of the object (1), magnetic measuring means (20) configured to acquire at least two components of the magnetic field and / or of an i-th derivative of the magnetic field, additional acquisition means (10) configured to acquire a measured linear velocity of the object (1), the object (1) further comprising data processing means (21) configured to: - determine values of at least one calibration parameter of the gyrometer (11) minimising a first expression defined by an estimated angular velocity of the object (1) and at least one first magnetic equation on the components of the magnetic field and / or of an i-th derivative of the magnetic field, ∘ the estimated angular velocity of the object (1) being a function of the measured angular velocity and of calibration parameters of the gyrometer (11), and ∘ the at least one first magnetic equation assuming that the magnetic field is uniform and stationary around the magnetic measurement means (20); - estimate a parameter representative of an error on the calibration parameters; the object being characterized in that the data processing means (21) are furthermore configured to: - if said parameter representative of an error is greater than a predetermined threshold, re-determine values of at least one calibration parameter of the gyrometer (11) minimising a second expression defined by the estimated angular velocity of the object (1) and at least one second magnetic equation on the measured linear velocity and on the components of the magnetic field and / or of an i-th derivative of the magnetic field, the at least one second magnetic equation assuming only that the magnetic field is stationary around the magnetic measurement means (20).

14. Computer program product comprising code instructions for the execution of a method for calibrating a gyrometer (11) according to one of claims 1 to 12, when said program is executed on a computer.