Determination of a heading using a field measured by magnetic sensors

Description

GENERAL TECHNICAL FIELD AND PRIOR ART

[0001] The present invention relates generally to magneto-inertial techniques.

[0002] More specifically, it concerns the determination of heading by magnetometers.

[0003] It is particularly useful for measurements in urban or indoor environments, i.e. inside buildings.

[0004] Magnetometers are traditionally used for heading calculation in embedded systems.

[0005] In this case, we assume that the magnetic field measured by the sensor is the Earth's magnetic field, which points, for its horizontal component, towards magnetic north. The difference between the direction of magnetic north and geographic north (called magnetic declination) is known and tabulated. Without loss of generality, we will therefore consider in the following that magnetic north and geographic north are the same, and that the magnetometers therefore indicate what we will call north.

[0006] Typically, the determination methods used for calculating heading from magnetometer measurements are based on: a model which links the measurement provided by the magnetometer and information relating to the heading of the system, a characterization of the relevance of this relationship for the measurements carried out.

[0007] In a classical approach, the calculation implemented can be a Kalman-type filtering with the magnetic field as a measure, allowing the heading contained in the state to be recalibrated, with measurement noise depending on the characterization of the relevance.

[0008] The modeling consists of writing that the measured magnetic field M contains heading information, for example by the equation M = R ψ θ φ T M TERRE where R is the rotation matrix allowing the transition from the object's frame of reference to the Earth's inertial frame of reference. ψ , θ, φ the Euler angles and M EARTH the Earth's magnetic field.

[0009] The relevance of this equation is characterized by a measurement variance, that is, we assume that the error of this equality is a Gaussian random variable of zero expectation.

[0010] This variance is used to automatically calculate the Kalman gain which weights the recalibration to take into account the different noises (dynamic noises linked to the external environment and measurement noises from the magnetometers).

[0011] In yet another approach, a calculation is implemented by linear filtering using the same type of modeling and characterization.

[0012] In this approach, it is the relative adjustment of the amplitude of the equation's gain that characterizes its relevance. In general, this gain is adjusted manually. Those skilled in the art will then know how to weight it according to the relevance of the modeling.

[0013] Thus, in these two approaches, the parameters characterizing the relevance of the model (variance of the Gaussian noise in the case of the Kalman filter, gain in the case of linear filtering) are generally constant parameters, independent of the measured magnetic field.

[0014] It has recently been proposed, for example in the publications: WT Faulkner, R. Alwood, WT David, and J. Bohlin, "Gps-denied pedestrian tracking in indoor environments using an imu and magnetic compass," in Proceedings of the 2010 International Technical Meeting of The Institute of Navigation, (San Diego, CA), pp. 198 - 204, January 2010, MH Afzal, V. Renaudin, and G. Lachapelle, "Magnetic field based heading estimation for pedestrian navigation environments," in 2011 International Conference on Indoor Positioning and Indoor Navigation, (Guimaraes, Portugal), Sep. 2011, characterizations of the relevance of the modeling taking into account the measured magnetic field by comparing it to a geomagnetic model of the Earth's magnetic field.

[0015] The characterizations thus carried out are not, however, fully satisfactory, firstly because they are based on an a priori model of the magnetic field, which requires having a model at least as precise as the desired precision and secondly, because they do not allow all disturbances to be rejected.

[0016] The paper "Unscented Filtering for Spacecraft Attitude Estimation" (J.L. Crassidis and F. Landis Markley) describes a model for estimating a craft's attitude. This model is based on the application of a Kalman filter and does not take into account any correlation. However, this model also has biases that make it unreliable.

[0017] It is also known from document EP 2 264 485 to propose a simplified model, in which the estimated magnetic field is equal to the sum of the value of the measured magnetic field, plus the value of the magnetic disturbance, plus the value of the measurement noise. However, this model of the evolution of the disturbance only takes into account a temporal correlation of the disturbance, which makes it unreliable.

[0018] Another model for estimating the orientation of a craft is known from document US 2014 / 122015.

[0019] Document FR 2 999 703 also discloses a method for determining the bias of a magnetometer, with a view to recalibrating this magnetometer and thus correcting the magnetic field measurements provided by the magnetometer. GENERAL PRESENTATION OF THE INVENTION

[0020] A general aim of the invention is to propose a solution allowing a better characterization of the relevance of the modeling used, particularly in highly disturbed environments.

[0021] In particular, the invention provides a method for determining heading as defined in claim 1.

[0022] In this way, the processing takes into account the evolution of the magnetic heading disturbance between two sampling instants and takes into account the possible spatial correlation of the disturbance due to the environment in the case of successive sampling instants.

[0023] The resulting recalibration is more reliable than in the prior art.

[0024] Preferred features of this heading determination method are set out in claims 2 to 9. The invention further relates to: a heading determination device as defined in claim 10, a magneto-inertial navigation system comprising at least one such heading measuring device, a computer program product as defined in claim 12, and a storage means as defined in claim 13. PRESENTATION OF FIGURES

[0025] Other characteristics and advantages of the invention will emerge from the following description, which is purely illustrative and non-limiting, and must be read in conjunction with the appended figures in which: there figure 1 is a diagram of equipment for implementing the method according to the invention; the figure 2 represents in more detail an example of a box for implementing the method according to the invention; the figure 3 illustrates the main steps of a method in accordance with an embodiment of the invention. DESCRIPTION OF ONE OR MORE METHODS OF IMPLEMENTATION AND REALIZATION General - Measuring device

[0026] In reference to the figure 1 , the proposed measuring device is for example used for estimating the movement of an object 1 moving in an ambient magnetic field (typically the Earth's magnetic field, possibly slightly altered by nearby metallic objects), noted B As we know, the magnetic field is a three-dimensional vector field, that is, associating a three-dimensional vector with each three-dimensional point in which the object is moving.

[0027] This object 1 can be any mobile object whose position knowledge is desired, for example a wheeled vehicle, a drone, etc., but also a pedestrian.

[0028] The object 1 comprises in a housing 2 (support) a plurality of magnetic measurement sensors 20, i.e. axial magnetometers 20. By axial magnetometer, we mean an element capable of measuring a component of said magnetic field, i.e. the projection of said magnetic field vector B at the level of said magnetometer 20 along its axis.

[0029] More precisely, the magnetometers 20 are integral with the housing 2. They present a movement substantially identical to the housing 2 and to the object 1 in the terrestrial reference frame.

[0030] Preferably, the reference frame of the object 1 is provided with an orthonormal Cartesian reference frame in which the magnetometers 20 have a predetermined position in this reference frame.

[0031] On the figure 2, the box 2 is fixed to the object 1 (for example a limb of the pedestrian) by attachment means 23. These attachment means 23 consist for example of a bracelet, for example with a self-gripping strip which encloses the limb and allows the integral connection.

[0032] Of course, the invention is not limited to estimating the movement of a pedestrian, but it is particularly advantageous in such use because it allows for a very reduced footprint, which is necessary for the box to be ergonomically portable by a human.

[0033] The box 2 may comprise calculation means 21 (typically a processor) for the direct implementation in real time of the processing of the present method, or the measurements may be transmitted via communication means 25 to an external device such as a mobile terminal (smartphone) 3, or even a remote server 4, or the measurements may be recorded in local data storage means 22 (a memory, for example of the flash type) local memory for subsequent processing, for example on the server 4.

[0034] The communication means 25 may implement short-range wireless communication, for example Bluetooth or Wifi (in particular in an embodiment with a mobile terminal 3) or even be means of connection to a mobile network (typically UMTS / LTE) for long-distance communication. It should be noted that the communication means 25 may be, for example, a wired connection (typically USB) for transferring data from the local data storage means 22 to those of a mobile terminal 3 or a server 4.

[0035] If it is a mobile terminal 3 (respectively a server 4) which hosts the “intelligence”, it comprises calculation means 31 (respectively 41) such as a processor for implementing the processing of the present method which will be described. When the calculation means used are those 21 of the box 2, the latter can also include communication means 25 for transmitting the estimated position. For example, the position of the wearer can be sent to the mobile terminal 3 to display the position in an interface of navigation software.

[0036] The data calculation means 21, 31, 41 respectively of the box 2, of a smartphone 3 and of a remote server 4 can indifferently and depending on the applications carry out all or part of the steps of the method.

[0037] For this purpose, they each include storage means in which all or part of the sequences of code instructions for the execution of the method are stored. Prediction and recalibration

[0038] The calculation means implement ( figure 3 ) a filtering 100 calculating on the one hand an estimate of the heading value by prediction (step 101) and on the other hand implementing a recalibration according to an error estimate (step 102).

[0039] In particular, knowing the angular velocity ω given for example by a gyrometer, step 101 calculates the heading ψ k +1 at time k+1 as equal to ψ k + 1 = ψ k + ω . Δ t Or ψ k is the heading at the previous instant k and where Δ t is the duration separating these two sampling instants.

[0040] The recalibration 102 takes into account the measurements made by the magnetometers 20.

[0041] In the following, the measurement of the magnetic heading (from the magnetometer measurements) for a given instant k is noted z ψ [ k ], with z ψ k = ψ k

[0042] Typically, the magnetic heading is given by z ψ k = atan My k Mx k where My and Mx are two horizontal components of the magnetic field in a terrestrial frame of reference, these two components being calculated as a function of the attitude determined for object 1 by an inertial unit of the latter.

[0043] The calculation means calculate the recalibrated heading ψ k + 1 recale corresponding to time k+1 as equal to the sum of the estimated heading ψ k +1 at this moment and a recalibration which is advantageously a function of a gain K k calculated or adjusted relative to the previous instant k by an error estimate Err ( ψ k +1 , z ψ ) between the predicted heading ψ k +1 and the measure z ψ

[0044] Thus, the calculation means use the magnetic heading resulting from the measurement by the magnetometers to recalibrate the state, and in particular the heading, by calculating ψ k + 1 recale = ψ k + 1 + K k . Err ψ k + 1 z ψ

[0045] Typically, the error Err ( ψ k +1 , z ψ ) can be a simple difference between the predicted heading, estimated at time k+ 1 (before recalibration), and the magnetic heading z ψ from magnetometer measurements.

[0046] Other error functions are nevertheless possible, particularly in the case of non-linear filtering.

[0047] In particular, the recalibration K k . Err ( ψ k +1 , z ψ), and more particularly the gain K k , can advantageously depend on the gradient of the magnetic field measured by the magnetometers 20.

[0048] In this way, the recalibration correction is all the more important as the magnetic field varies strongly and is therefore likely to induce significant errors in the heading measurement.

[0049] The recalibrated heading value thus obtained is stored by said calculation means 21, 31, 41 and / or used by said means for the continuation of the processing and for the calculations of magneto-inertial navigation information (linear speed, angular speed, position, heading, etc.).

[0050] It can also be transmitted by the computing means to interface means, for example on the telephone to be displayed on the telephone screen. Model of the evolution of environmental disturbances

[0051] The following details an example of a possible calculation of the recalibration in which the magnetic heading measurement z ψ is corrected by an estimate of the magnetic field disturbance linked to the environment.

[0052] The magnetic cape z ψ from the measurement by magnetometers can in fact be considered as being broken down as follows: z ψ = ψ + ψ d + v z ψ Or: ψcorresponds to the actual heading that we are trying to determine, v z ψ corresponds to a Gaussian measurement error and ψ ( d )< (“d” for “disturbance” in English) corresponds to a disturbance of magnetic heading linked to the environment (typically, disturbances linked to metallic infrastructures and electrical cables in urban environments or in buildings).

[0053] The magnetic heading disturbance related to the environment is strongly spatially correlated. The magnetic field related to environmental disturbances is indeed a continuous vector field and the magnetic fields at two given points A and B are all the closer the closer these two points A and B are in space.

[0054] The course disturbance ψ ( d)< linked to the environment is estimated by the calculation means using a formulation taking into account the evolution correlations that can be expected for magnetic field disturbances between two successive sampling instants (temporal, spatial or more complex correlation).

[0055] This estimate is constructed to allow a Markovian model (which can be used in a recursive filter) for the evolution of ψ d And allows to take into account a model of the evolution of the magnetic heading disturbance (temporal, spatial or more complex correlation), allows to construct a filtering model for which the heading is observable (otherwise, we lose all hope of constructing a heading estimator).

[0056] In the proposed estimation, the amplitude of the recalibration is estimated based on a model of the evolution of magnetic disturbances between two sampling times and an a priori estimate of the amplitude of the disturbances.

[0057] The inventors found (and mathematically verified) that in the case of spatial correlation (case of disturbances in an urban or "indoor" environment), an adapted estimate of the evolution of the disturbance ψ ( d )< between two sampling instants separated by a time step is as follows: ψ d k + 1 = α k ψ d k + û k + 1 − α k û k 2 + v ψ d k Or α k = 1 − σ u k 2 2 a k 2 a [ k ] being a parameter which represents the a priori amplitude of the magnetic disturbance, û And σ u being the expectation and variance of a random variable which is a model (in the form of a Gaussian random variable) of the variation of the magnetic heading disturbance between 2 sampling instants, v ψ (d) is a Gaussian random variable with variance a[k] 2< (1 - α[k] 2< ).

[0058] This estimate is subtracted from z ψ for the error calculation, which can be directly calculated between ψ k +1 and ( z ψ - ψ ( d )< [k + 1]), or entrusted to a Kalman type filter which has both ψ ( d )< and ψ in his condition. Determination of the parameter a [ k ]

[0059] The parameter a[k] represents the a priori amplitude of the magnetic heading disturbance. It characterizes the heading and is, for example, calculated as a linear function of the norm of the magnetic field gradient.

[0060] For example, a k = a 0 + a 1 N k with N k = ∇ B ^ k a 0 and a 1 being two parameters previously fixed before implementing the filtering treatment, B̂ [ k ] being the magnetic field at time k.

[0061] In this way, as indicated above, the term of recalibration K k . Err ( ψ k +1 , z ψ ) is a function of the gradient of the magnetic field measured by the magnetometers 20, which allows an effective correction.

[0062] Other methods are possible to determine the parameter a [ k ].

[0063] Notably, a [ k ] can also be determined by comparison with a model of the Earth's magnetic field, for example by implementing the techniques as proposed in the publication: "Assessment of Indoor Magnetic Field Anomalies using Multiple Magnetometers Assessment of Indoor" - MH Afzal, V. Renaudin, G. Lachapelle - Proceedings of the 23rd International Technical Meeting of The Satellite Division of the Institute of Navigation (ION GNSS 2010) September 21 - 24, 2010.

[0064] The solutions in which the parametera [ k ] is a function of the magnetic field gradient, however, allows for better corrections and does not require models of the Earth's magnetic field. Determination of the parameter û

[0065] The parameter û[k] represents the most probable value for the evolution of the disturbance.

[0066] This parameter is calculated, for example, using gyrometers that are assumed to be good in the short term. The most probable evolution of the disturbance is then given by the difference between the evolution of the magnetic heading measurement and the evolution of the gyrometric heading: u ^ k = E ψ d k + 1 − ψ d k = E ψ d k + 1 + ψ k + 1 ︸ cap magn é tique mesur éà l ′ instant k + 1 − ψ d k + ψ k ︸ cap magn é tique mesur éà l ′ instant k − ψ k + 1 − ψ k ︸ é volution de cap = z ψ k + 1 − z ψ k − ω k dt where: E[.] is the mathematical expectation, ω [ k ] is the gyrometric rotation speed and dt is the sampling step.

[0067] So the parameter û[k]is calculated by the computing means as the difference between the magnetic headings determined directly from the magnetometer outputs for times k+1 and k, from which the predicted rotation is subtracted ω [ k ] dt. Determination of the parameters σ u

[0068] The parameter σ u represents the image of the correlation between the disturbance at step k and that at step k+1.

[0069] It is advantageous to choose to index it on the speed of movement (or on the movement between two successive sampling instants, which is similar) so as to account for the spatial correlation of the magnetic disturbances.

[0070] For example, we can take σ u = c . v where c is an adjustment coefficient.

Claims

1. A method for determining heading by magnetic sensors, in which a magnetic field is measured by magnetometers and calculation means implement a recursive processing calculating for a given sampling time: - an estimation of a heading prediction which is a function of the heading determined at the preceding sampling time, - a resetting of the thus estimated predicted heading as a function of a magnetic heading determined from magnetometer measurements, characterised in that during said resetting, the calculation means estimate the amplitude of the resetting as a function: - of a model of a change in magnetic disturbances between two sampling times determined beforehand to take into account a spatial correlation of the changes in disturbances, and - of an a priori estimation of the amplitude of the disturbances. and in that, for a given sampling time k+1, the calculation means estimate the heading disturbance linked to the environment by calculating α k ψ d k + û k + 1 − α k û k 2 + v ψ d k where α k = 1 − σ u k 2 2 a k 2 and vψ(d) is a random Gaussian variable of variance a[k]2(1 - α[k]2), - k being the preceding sampling time, - a[k] being a parameter which represents the a priori amplitude of the magnetic disturbance, - û and σu being two parameters calculated by the calculation means as an estimation of the expectation and the variance in the change in disturbance.

2. The method according to claim 1, in which said resetting is a function of the measured magnetic field gradient.

3. The method according to claim 2, in which the calculation means estimate the parameter a[k] as a linear function of the norm of the magnetic field gradient.

4. The method according to claim 1, in which the heading prediction is also a function of a predicted change in heading and the parameter û[k] is calculated by the calculation means as the difference between the magnetic headings determined directly from the magnetometer outputs for the times k+1 and k, from which is subtracted the predicted change in heading.

5. The method according to claim 1, in which the parameter σu is estimated as a function of the velocity of displacement or the displacement between two successive sampling times.

6. The method according to claim 1, in which the processing implements a Kalman filtering of which the state has at least as parameter the real heading and the magnetic heading disturbance (ψ, ψ(d)).

7. The method according to one of the preceding claims, in which the heading prediction is determined as a function of measurements of one or more sensors of an inertial unit.

8. The method according to claim 1, in which the change model is determined beforehand in order to take account of a temporal correlation of the changes in disturbances.

9. The method according to one of the preceding claims, in which a magneto-inertial navigation system is used in a urban environment or inside buildings, the system comprising at least one device for determining heading by means of magnetic sensors, the device comprising magnetometers and calculation means for calculating the heading from the magnetic field measured by said magnetometers, the calculation means implementing, for different successive sampling instants, the processing of the method according to one of the preceding claims.

10. A device for determining heading by magnetic sensors, comprising magnetometers and calculation means for the calculation of the heading from the magnetic field measured by said magnetometers, characterised in that the calculation means implement, for different successive sampling times, the processing of the method according to one of claims 1 to 9.

11. A magneto-inertial navigation system comprising at least one device for determining heading according to claim 10.

12. Computer programme product including code instructions for the execution by a device for determining heading according to claim 10 of a magnetic field measurement method according to one of claims 1 to 9, when said programme is executed on a computer.

13. Storage means readable by a computer equipment on which a computer programme product includes code instructions for the execution by a device for determining heading according to claim 10 of a magnetic field measurement method according to one of claims 1 to 9.