FR2704665A1 - Process and system for locating a land vehicle - Google Patents

Abstract

The process combines a process of radio location using for example the "GPS" system and a process of dead reckoning using measurements of speed of rotation (Wd, Wg) of the wheels of the vehicle. The dead reckoning calculates predicted values of position (x, y) by virtue of a model (M1) whose parameters are bound to the geometry of the vehicle. In order to improve accuracy, the parameters are corrected by means of a parameter fitting algorithm (AAP). Use in particular to produce a navigation system.

Description

- 1-

LOCATION METHOD AND SYSTEM

OF A TERRESTRIAL VEHICLE

  The invention relates to the problem of locating a land vehicle, in particular for the constitution of

navigation system.

  Precise knowledge of the location of a vehicle is basic information necessary for any navigation system. In the automotive field, the need for a location system appears in particular for the following applications: - navigation aid via a location and guidance system which makes it possible to advise a driver wishing to reach a given destination on possible routes and guide him step by step along this route; - vehicle fleet management (truck, taxi, ...) by equipping each vehicle with a localization system as well as a communication system allowing it to transmit its position to a central office and receive it

the instructions.

  For fleet management, the permanent availability of position information is the most important quality. Depending on the case, such an application may be satisfied with a precision of

  a few hundred meters to a few kilometers.

  For a navigation aid system on a motor vehicle, information that is both permanent and very precise is sought. For guidance on urban sites, accuracy better than a hundred meters

may be necessary.

  There are three main categories of navigation systems: autonomous systems, infrastructure-based systems and radiolocation systems. Autonomous navigation systems are based on the use of endogenous information, collected from sensors fitted to the vehicle and data

  on-board, stored in a mass memory.

  However, the sensors always tainting the information they deliver with a certain error, the calculation which results from the exploitation of this leads to a "dead reckoning" navigation which drifts as time goes by. Because of this, the dead reckoning of the position is not sufficient and needs to be readjusted using on-board cartographic data. The advantages of this type of system are therefore the permanence of information and complete autonomy while a disadvantage is the need for manual intervention at initialization and each time the system is lost. Their use being limited to a mapped area corresponding to the on-board data, another drawback is the need to update the on-board database. Infrastructure-based systems use a network of beacons covering a certain geographic area. Each time the vehicle passes in front of a beacon, it is possible for it to exchange 3- information with it by a short range communication. The beacon can thus transmit its absolute position to it with great precision (a few meters). Between two tags, the system can extrapolate its position using a dead reckoning navigation system. The advantage of this system lies in the possibility of offering additional services helping or supplementing the location service: information on road traffic, location of service stations, phone booths, parking lots, restaurants, hotels, etc. Its major drawback is the need to set up

heavy infrastructure.

  Radiolocation systems allow localization by triangulation by measuring the propagation time of signals emitted by terrestrial or space transmitters. From the reception of three transmitters, it is possible to locate a vehicle in absolute terms in a reference landmark. Among the radiolocation systems, there may be mentioned the "LORAN C" and the "GPS" ("Global Positioning System"). The "LORAN C" is a maritime system, which limits its use to coastal regions. "GPS" is a global positioning satellite system which allows better accuracy

than a hundred meters in 95% of cases.

  The major drawback of radiolocation systems is that they rely on the reception of at least three transmitters, reception which can be made difficult or even impossible, depending on the environment of the vehicle (building, tree, road sign, tunnel). So they only release their information intermittently. It is for this reason that it is desirable to couple them with a dead reckoning navigation system in order to be able to provide the position of the vehicle at all times. These systems however have the essential advantage of delivering absolute position information, the error of which remains

time bound.

  Since dead reckoning and localization systems have complementary qualities, it is interesting to design hybrid systems combining these two types of system. For example, French patent application No. 91 14544, filed on 11/26/1991, describes a method of assisting the navigation of a land vehicle consisting in combining navigation information with dead reckoning, making it possible to calculate point by point, the heading and the position of the vehicle relative to a starting position, with location information received by radio enabling the absolute position of the vehicle to be determined directly. Radio information is used in order to readjust course and position in real time on all

navigation points.

  Another example of a hybrid system can be illustrated by American patent US-A-4758959 which also describes a complete navigation system fulfilling the functions of localization and guidance. The proposed location function is based on a dead reckoning navigation system, which can be paired either with a cartographic database or with a satellite location system. The dead reckoning position is calculated from the measurement of the transverse acceleration and the vehicle speed. The system -5 implements a mechanism which makes it possible to reduce the measurement errors of the accelerometer and to improve the accuracy of the assembly. Various solutions are also provided for adjusting dead reckoning using geographic data from the road network stored in

a mass memory.

  Although the previous solutions provide an answer to the problem of the permanence of location information, they only answered

  partially to the problem of its accuracy.

  In fact, these solutions are essentially based on the use of a radiolocation system, the intrinsic precision of which is generally sufficient for the service concerned. However, in the phases where the radiolocation data are unavailable, these systems always face an accuracy which in addition to decreasing over time is strongly conditioned

by the terms of use.

  Thus, according to French application No. 91 14544, the dead reckoning is calculated from the speeds of the rear wheels of the vehicle, which makes it possible to get rid of a compass whose use is difficult to

  by the disturbances to which it may be subjected.

  The radio positions are used to evaluate the absolute heading of the vehicle. However, no remedy is proposed to attenuate the influence of the conditions of use: speed, state of charge, state of the tires, etc. In addition, its precision is also linked to the value of coefficients involved in the estimated calculation of the position and corresponding to "manufacturer" data which generally have a fairly large dispersion. If by chance, on a given vehicle, these values can lead to a low deadweight navigation drift, in the worst case the deadlock obtained may prove to be unusable. In fact, it is in practice essential to calibrate the "vehicle" parameters to obtain an acceptable accuracy in the absence of radio position, which supposes at least manual intervention in the

development of such solutions.

  The solution proposed in US-A-4,758,959 mainly focused on solving the problem of precision. It essentially seeks to best develop a location error from map data or data received from satellites, which it then uses to correct the measurement from the accelerometer. The need to take into account the tilt angle of the vehicle is also emphasized to correct the impact of terrestrial gravity on the acceleration measurement carried out. However, none

  solution to measure this cant is written.

  In order to overcome the drawbacks of the systems and methods described above, the invention provides a hybrid system ensuring that the global navigation system operates continuously at its highest degree of precision, especially during the phases of

  non-reception of radiolocation data.

  To this end, the subject of the invention is a method for locating a land vehicle combining a radiolocation method and a dead reckoning method which uses measurements of the speed of rotation of the right and left wheels of the same 7 - axle of the vehicle, characterized in that the dead reckoning method consists in calculating predicted values of the position and the heading of said vehicle thanks to a mathematical model which establishes by means of parameters notably related to the geometry of the vehicle , relationships between the variations of the heading and of the position and said speeds of rotation, in that, by means of the radiolocation method are calculated correction values of said predicted values and in that the values of at least one of said parameters are adjusted according to said correction values at

  using a parametric adaptation algorithm.

  For a certain level of disturbance, this solution makes it possible to keep the system above a certain precision threshold, a threshold which is really linked to the quality of the instruments used and no longer to the conditions of use of the vehicle. It also has the advantage of avoiding any manual intervention linked directly or indirectly to the calibration of the system such as a reset of the system after a change of

pneumatic for example.

  There are many possibilities for choosing the geometric parameters, depending in particular on the type of vehicle. It is however advantageous to choose these parameters so as to minimize the number and type of sensors necessary for the implementation of the

  dead reckoning process.

  Also, according to an additional characteristic of the invention, the geometric parameters are functions of the circumferences of the wheels and of the axle track. In this case, it suffices to provide a sensor -8 of rotation speed per wheel or else to use existing sensors for example if the vehicle is equipped with a braking system provided with an anti-lock system.

wheels (ABS).

  In a dead reckoning system, the calculation of the heading is the operation which requires maximum precision because the errors on the heading are those which have the most important consequences on the drift of the

  calculated position in relation to the actual position.

  Also, according to another characteristic of the invention, one of the adjustable parameters is the relative difference between the circumferences of the wheels because the precision of the evaluation of this difference conditions the precision on

the cap.

  According to yet other aspects of the invention, other adjustable parameters are the average value of the wheel circumferences and the cornering gain defined as the ratio between the average circumference and the

axle track.

  In general, the geometric parameters of the vehicle vary according to its dynamic behavior. For example, the circumference of the wheels varies according to the load applied due to the deformation of the tires. Consequently, the circumferences of the wheels will be a direct function of the transverse and longitudinal accelerations of the vehicle which create load transfers. These variations generate position errors which are normally automatically corrected by means of the position measurement provided by the radiolocation system. However, in certain circumstances such as passing over long curves at high speed, in the absence of radiolocation information, such a correction can no longer take place, and the predicted values of

  position may be marred by significant errors.

  To solve this problem and according to another aspect of the invention, the method is characterized in that other adjustable parameters of the model are dynamic parameters expressing the variations of the geometric parameters as a function of the vector.

vehicle acceleration.

  The invention also relates to a system for implementing the method according to the invention. This system is characterized in that it includes a programmable processing unit connected to right and left wheel speed sensors on the same axle of the vehicle, to a satellite radiolocation receiver and to a control and display, said processing unit being programmed to

  execute said parametric adaptation algorithm.

  Other aspects and details of realization of

  the invention will appear in the following description

with reference to the figures.

  - Figure 1 shows schematically the main geometric elements of a vehicle to define a model according to the invention; - Figure 2 shows a hybrid location system according to the state of the art;

- 10 -

  - Figure 3 shows a location system according to the invention; - Figure 4 shows an alternative embodiment of the location system of Figure 3; - Figure 5 illustrates the relationships between the geometric parameters and the transverse acceleration of the vehicle; - Figure 6 shows an embodiment of a

  digital location system according to the invention.

  Figure 1 is a schematic representation of a vehicle 1 o appear the main geometric parameters and position variables which are

  involved in the model according to the invention.

  The rear left and right wheels 2g, 2d have a rotation speed respectively Wg, wd corresponding respectively to the linear speeds vg, vd. The wheels 2g, 2d are spaced apart by a distance V which is usually called the axle track. The tangential speed of the vehicle is defined as the mean value v of the left and right linear speeds vg, vd of the wheels. The difference between the speeds vg and vd corresponds to a rotation of the vehicle around a center of curvature C with a radius of curvature R. The middle of the rear axle then travels the curve L with a yaw speed e 'equal to the derivative with respect to the time of / dt of the angle of rotation e of the vehicle. We can also say that the heading undergoes a variation of of between the instants t and t + dt corresponding respectively to the positions P and P + dP of the center of the axle.

- 11 -

  If we call Dd and Dg respectively the circumferences of the right wheels 2d and left 2g, we can define the average circumference D of the wheels: (1) D = (Dd + Dg) / 2 We can also define the relative deviation e between the circumferences: (2) e = (Dd - Dg) / D, as well as the gain at turn G:

(3) G = D / V

  We deduce a geometric model defined by the two equations: (4) e '= G [(1 + E) .wd - (1 - e) .Wg] (5) v = D [(1 + e) .wd + (1 - e) .Wg] / 2 This model therefore makes it possible to continuously calculate the yaw speed e 'and the tangential speed v of the vehicle based solely on the measurements of the rotational speeds Wg and wd of the left and right wheels. We then obtain the course e by a simple integration with respect to time: u (6) e (u) = eO + J e '(z) .dz o The course 6 then makes it possible to obtain the position x, y:

- 12 -

  t (7) x (t) = xO + v (u) .cos [e (u)]. du ot (8) y (t) = Yo + / v (u) .sin [8 (u)]. of Similar geometric models have already been used in mixed art location systems

  prior. Such a system is shown in Figure 2.

  The kinematic quantities Wd, Wg of the vehicle are applied to the input of the geometric model M which provides calculated values of position x, y and of heading e. A module M4 called "estimator" receives the calculated values x, y, e as well as position values X, Y supplied by a radiolocation system. The M4 module calculates correction values dxe, dY, d by appropriate filtering. This filtering is advantageously a statistical filtering such as the filtering of KALMAN-BUCY in its extended form: "EKF" ("Extended Kalman Filter"). This estimation technique is described for example in the book "ADAPTIVE FILTERING PREDICTION AND CONTROL" by Graham C. Goodwin and Kwai Sang Sin, 1984, Prentice-Hall, Information and System

Sciences Series.

  The correction values dxe, dye, dee of the predicted values x, y, e provided by the model M make it possible to obtain the corrected values xe, Ye, Be- These correction quantities are also used for the registration of the model M, it is that is to say, to evaluate the initial values e0, x0, yo used for the subsequent calculation of e,

  x, y in accordance with equations (7) and (8).

- 13 -

  Although the system which has just been described satisfies a first objective which is the precise evaluation of the drift of dead reckoning, it does not make it possible to correct the modeling errors nor to reduce their effects. In reality, the values of the parameters e, G, D used in the model are values ci, Gi, Di generally different from the real values Er, Gr, Dr. By showing the absolute error Se on e and the relative errors kG and kD on respectively G, D, we will have: (9) Ei = Er + Se (10) Gi = (1 + kG) Gr (11) Di = (1 + kD) Dr These errors on the parameters e, G, D will therefore cause errors 68ee and 6ve respectively on the yaw speed e 'and the vehicle speed v in accordance with the equations: (12) 6e'e = e'- e'e = kG.e'e + 6Se. (2 / Ve) .Ve (13) 6ve = v - ve = Se. (Ve / 2) e'e + kD.Ve These equations result from a calculation limited to the first order, o e'e, Ve, Ve are respectively estimated values of the real values of e ', v, V. From these linear equations in parameter, it is then possible to identify the errors Se, kG, kD using an appropriate method such as the method of least squares or derivatives. Advantageously, we

- 14 -

  will be able to choose the method of the "stochastic gradient" presenting the advantages of simplicity and its ability to pursue nonstationarity. This calculation can be carried out using a parametric adaptation algorithm which can operate recurrently, which provides great efficiency. Such methods are

  described in the book cited above.

  The location method according to the invention is shown in more detail in Figure 3. We find the estimator M4 receiving the position data X, Y of the radiolocation system such as "GPS". It also receives the predicted values of position x, y and heading e calculated by the module M in

  function of the rotational speeds Wd, Wg of wheels.

  The estimator M4 (assumed here to include means of addition) delivers the estimated values Xe, Ye and ee calculated as a function of 8, x, y, X, Y. It also provides the values e'e and ve to an algorithm

parametric adaptation of AAP.

  The module M comprises a mathematical model M1 which calculates the angular velocity e 'and linear v as a function of the rotational speeds Wd, Wg and the geometric parameters e, D, G in accordance with equations (4) and (5). The predicted value of heading e is calculated by an integrator I as a function of the angular velocity e 'and of an initial value of heading e0 whose value is equal to the estimated value of heading ee at the start of the integration period. A kinematic module M2 calculates the predicted position x, y, as a function of the predicted values of the heading e and the speed v in accordance with equations (7) and (8). The initial position values x0, yo are the estimated values

- 15 -

  Xe, Ye at the start of the integration period. The module M finally comprises a module for correcting the parameters M3 receiving initial values of the parameters Ei, Di, Gi as well as correction values for these parameters Se, kD, kG. These correction values are provided by the parametric adaptation algorithm AAP as a function of the predicted values e ', v and estimated e'e, ve of the yaw rate and the tangential speed. The M3 module continuously calculates the corrected values e,

  G, D according to equations (9) to (11).

  With a view to implementing the method by means of a synchronous digital processing system, it is simpler to use the variations of, dL of course e and distance traveled L. Equations (12) and ( 13) then become: (14) de - dee = kG.dee + 6th. (2 / Ve) .dLe (15) dL - dLe = Se. (Ve / 2) .dee + kD.dLe, where the magnitudes of, dee, dL, dLe are the variations respectively of e, ee, L, Le between two periods

system sampling.

  The process is then that shown in FIG. 4 o the integrator I consists of a simple adder

  synchronous receiving the successive variations of the heading.

  The kinematic module M2 first calculates the position variations dx, dy according to the equations: (16) dx = dL.cose (17) dy = dL.sine

- 16 -

  The integrals appearing in equations (7) and (8) are then calculated by means of adders performing the accumulation of the variations dx, dy at each

  new synchronization period.

  The M4 estimator essentially consists of a Kalman filter FK which calculates the correction values dxe, dye d8e, dLe as a function of the predicted positions x, y and of the positions X, Y provided by the radiolocation system. The correction values dxe, dye, dee are used to correct the predicted values x, y, e to provide the estimated values Xe, Ye, ee. The correction values dee, dLe are used for the algorithm

parametric adaptation of AAP.

  As we have already pointed out previously, certain geometric parameters of the vehicle vary according to its dynamic behavior. This is the case in particular of the circumference of the wheels which varies as a function of the longitudinal or transverse acceleration due to the deformation of the tires. Thus, during acceleration, the rear tires undergo a crushing which has the effect of reducing their apparent circumference. Similarly, during turns, the right and left tires undergo unequal transverse deformations which modify both their circumferences and the relative distance between their circumferences. This phenomenon is illustrated in Figure 5 o are shown the two left and right wheels 2g, 2d under the influence of inertial forces during a

  left turn with transverse acceleration rt.

  Under these circumstances, there is a decrease in the circumference of the right wheel, an increase in that of the left wheel as well as a decrease in the

- 17 -

  lane V relative to its value V0 when the vehicle is at rest or in a straight line. There is also a phenomenon called "tire drift" which means that the linear speed of the vehicle, taken at the rear axle, is no longer in the longitudinal axis thereof. Taking into account the preceding considerations, the M1 model can be improved by showing dynamic parameters expressing the variations of the geometric parameters as a function for example of the transverse acceleration. Thus, by limiting ourselves to the first order, we can write: (19) e = co - Ke.rt (20) G = Go (1 + KG. Irtl) where eo0 is the relative deviation of the wheel circumferences when the vehicle is at rest or in a straight line, Ke and KG are correction coefficients to hold

  account for the dynamic effects on the vehicle.

  As for the geometrical parameters, one will define the relative or absolute errors between the dynamic parameters used in the model and the real dynamic parameters. These errors will then appear in error equations similar to equations (12) and (13) and we can again use a parametric adaptation algorithm to best adjust the

  values of these dynamic parameters.

  More precisely, when using the coefficients Ke and KG defined above, we can ask:

- 18 -

  (21) (KG) i = (KG) r + 6KG (22) (KE) i = (Ke) r + 6KE (23) Gi = Gr + 6G (24) Di = Dr + ED (25) Ci = Er + e (26) A1 = [er- Gr.Vr - e'r.-Dr / 2] / [(1 - Er) Gr] (27) A2 = (2 Gr.Vr - Dr.Er.e'r ) / [(l - Er) .Dr], O the indices r and i respectively indicate a

  real size and an uncorrected quantity.

  We then obtain: (28) se'e = SG.e'e / Ge + SKG.I (rt) el.O'e / [l + (KG) e] + 6e.Ale - SKE.A1e (29) 6ve = 6D.ve/De - Se.A2e - 6Ke.A2e,

  o the index e indicates an estimated quantity.

  FIG. 6 represents an exemplary embodiment of a digital system allowing the implementation of the invention. The system comprises a programmable processing unit 3 connected by an interface 15 to a control and display device 6. The unit 3 is also connected by an interface 13 and a serial link 14 to a receiver 5 for radiolocation by satellite. The unit 3 is finally connected by an interface 10 to two digital counters themselves connected to sensors 4d, 4g of speed of rotation of the right and left wheels by means of a shaping circuit 12. The unit 3 is organized around a bus B on which the interfaces 13 and 15, the interface 10, a microprocessor 7, a program memory 8, a working memory 9 and a non-volatile memory 16 are connected. together is synchronized by a

clock not shown.

- 19 -

  All the operations described with reference to FIGS. 3 or 4 are carried out by the execution of a localization calculation program contained in the memory 8. The working memory 9 is used to store the intermediate calculation results while the non-volatile memory 16 stores the last calculated position as well as the last values of the calculated geometric and dynamic parameters. This last provision allows to save all the parameters of the

  model when the vehicle is put at rest.

  The system described in FIG. 6 being a microprocessor system of conventional type, its implementation is within the reach of those skilled in the art, taking into account the

  explanations given above.

- 20 -

Claims (10)

  1. Method for locating a land vehicle (1) combining a radiolocation method and a dead reckoning method which uses rotational speed measurements (wd, Wg) of the right and left wheels (2d, 2g) of the same axle of the vehicle, characterized in that the dead reckoning method consists in calculating predicted values of the position (x, y) and of the heading (8) of said vehicle using a mathematical model (MI) which establishes by means of parameters (e, D, G) notably linked to the geometry of the vehicle, relations between the variations (el 'de, v, dx, dy) of the heading (e) and of the position (x, y ) and said rotational speeds (Wd, Wg), in that by means of the radiolocation method are calculated correction values (dxe, dye) of said predicted values (x, y) and in that the values of at at least one of said parameters (e, D, G) is adjusted as a function of said correction values (dxe, dye) by means of a   parametric adaptation algorithm (AAP).   2. Method according to claim 1, characterized in that said geometric parameters (e, D, G) are functions of the circumferences (Dd, Dg) of said wheels   (2d, 2g) and track (V) of said axle.   3. Method according to claim 2, characterized in that one of said parameters adjustable by means of said algorithm (AAP) is the relative difference (e) between the   circumferences (Dd, Dg) of said wheels (2d, 2g). - 21 -   4. Method according to claim 3, characterized in that other adjustable parameters are the average value (D) of the circumferences (Dd, Dg) of the wheels and the cornering gain (G) defined as the ratio between said average circumference ( D) and the track (V) of said axle. 5. Method according to claim 4, characterized in that other adjustable parameters of the model are dynamic parameters (Ke, KG) expressing the variations of said geometric parameters (e, D, G) as a function of   vehicle acceleration vector (F).   6. Method according to claim 5, characterized in that one of said adjustable dynamic parameters (KE) is proportional to the ratio between the variation of said relative deviation (e) due to a transverse acceleration (rt) of the vehicle and said transverse acceleration (Ft). -   7. Method according to claim 6, characterized in that one of said adjustable dynamic parameters (KG) is proportional to the ratio between the relative variation of the cornering gain (G) due to said transverse acceleration (Ft) and said transverse acceleration (Ft).   8. Method according to one of claims 1 to 7,   characterized in that said adaptation algorithm   parametric is the stochastic gradient algorithm.   9. Method according to claim 8, characterized in that said correction value (dXe, dye) is determined by the use of a Kalman filter using said predicted values (x, y) and - 22 -   position data (X, Y) provided by said radiolocation method. 10. Land vehicle location system for implementing the method according to one of   Claims 1 to 9, characterized in that it comprises   a programmable processing unit (3) connected to right and left wheel rotation speed (2d, 2g) sensors (4d, 4g) of the same axle of the vehicle, to a satellite radiolocation receiver (5) and to a control and display device (6), said processing unit (3) being programmed to execute   said parametric adaptation algorithm.