METHOD AND SYSTEM FOR LOCALIZING A MOBILE NODE BY RADIO MEASUREMENTS USING A VIEWING CARD

Description

Technical domain

[0001] The invention relates to a method and system for locating a mobile node in a given area, using known positions of reference nodes.

[0002] The technical field of the invention is that of localization using radio signals. More specifically, it concerns methods using the time of flight of the electromagnetic wave.

[0003] As illustrated by the figure 1 Existing localization methods consist of measuring the time of flight of the signal between reference nodes (A1, A2, A3), whose position is generally known, and a mobile node TM, whose position we seek to determine.

[0004] For example, in a Low Power Wide Area Network (LPWAN), the reference node is an anchor, meaning a device with a known and fixed position, and the mobile node is a tag. The tag is typically attached to an object to transmit telemetry data. In a cellular network, the reference node is a base station, and the mobile node can be a cell phone. In a satellite navigation system, the reference node is a satellite (its position is known), and the mobile node can be a navigation receiver.

[0005] Knowing the speed of wave propagation, which is equal to that of light, it is possible to determine the distance traveled by the wave and therefore, the distance separating the fixed nodes from the moving node. This is referred to as a "ranging" measurement, which corresponds to the measurement of the time of flight between two devices, the measurement of the corresponding distance, or even a pseudo-distance.

[0006] Locating a moving node is therefore a classic trilateration problem, as illustrated in the figure 1 whose unknowns (in this two-dimensional example: x,y) satisfy the following equations: d 1 = x − x a 1 2 + y − y a 1 2 d 2 = x − x a 2 2 + y − y a 2 2 d 3 = x − x a 3 2 + y − y a 3 2

[0007] Accurate measurement of flight time, however, requires that the clocks of the fixed nodes and the moving node be perfectly synchronized (with an accuracy typically between 1 ns and 100 ns), a synchronization error of one nanosecond leading to a distance measurement error of approximately 30 cm.

[0008] However, this precise synchronization is generally not possible, leading to a description of the measurements mi that includes an additional unknown Δi reflecting this synchronization defect. Furthermore, these measurements can be affected by other errors εi caused, for example, by measurement noise and wave reflections. In the general case, the system of equations [Math1] becomes: m 1 = d 1 + Δ 1 + ε 1 m 2 = d 2 + Δ 2 + ε 2 m 3 = d 3 + Δ 3 + ε 3

[0009] There are various techniques to circumvent the synchronization problem and to resolve the unknowns Δ i that reflect this synchronization defect.

[0010] Depending on the environment through which the wave propagates, it can be reflected (this is called multipath propagation) and / or obstructed. We speak of Line Of Sight (LOS) when the wave propagates along a direct path, without reflection, between the transmitter and the receiver, and of Non-Line Of Sight (NLOS) in the opposite case.

[0011] This occurs, for example, in urban environments where buildings can act as both obstacles and / or reflectors for electromagnetic waves.

[0012] There figure 2 illustrates these different situations.

[0013] Reference node A1 is in a line of sight (LOS) without multipath propagation relative to the moving node TM because it only receives the direct path, without any reflected paths. Reference node A2 is in a line of sight with multipath propagation relative to the moving node TM because it receives the direct path t2A as well as a reflected path t2B. Reference node A3 is in a non-line of sight (NLOS) with multipath propagation relative to the moving node TM because it does not receive the direct path t3A, which is obstructed by building B1, but it does receive a reflected path t3B from building B2.

[0014] An NLOS situation without multipath leads to non-reception of the signal since it has no path to reach its destination.

[0015] In the representation of the figure 2The measurement performed by the reference node A1 can be considered the most reliable because it is not subject to any disturbances in its propagation. The associated error (ε1) is in this case mainly determined by the measurement noise, which constitutes a minimal error.

[0016] For example, for a LoRa type waveform, the error caused by noise alone is typically on the order of a hundred meters (between 10 m and 300 m).

[0017] The measurement performed by the fixed terminal A2 can be affected by the presence of multipath propagation. The resulting error depends on the ability of the reference node A2 to distinguish the direct path from the reflected path (which will arrive later than the direct path), as well as on the relative power of the reflected path compared to the direct path. This is therefore an intermediate situation.

[0018] For example, for a LoRa waveform, the error caused by noise and multipath propagation varies greatly depending on the characteristics of the reflected paths. Errors can range from a few tens of meters (similar to noise alone) to several kilometers.

[0019] The measurement taken by the reference node A3 is, however, highly biased. Indeed, the only signal reaching this reference node has traveled a distance corresponding to path t3B, significantly greater than the direct path t3A; the time of flight therefore reflects this distance and not that of the direct path. This measurement is generally considered an artifact (or outlier) because it exhibits an error that can be several orders of magnitude greater than the error due to noise alone.

[0020] For example, for a LoRa type radio, the error in an NLOS situation is typically several hundred meters to several kilometers.

[0021] The difficulty in such a situation is that it is very difficult, if not impossible, to determine the reliability of each of the measures.

[0022] In general, we can differentiate between sequential algorithms (which exploit previous solutions), and instantaneous approaches ("single epoch", "snapshot") which rely solely on measurements received at time t.

[0023] In a sequential approach, the system can predict the possible position at time t based on the solutions found at previous times t - 1, t - 2, with an acceptable margin of error (this notion of acceptable margin of error is assessed with regard to the precision of the measurement given by the noise for example).

[0024] From the predicted position, a "predicted measurement" can be deduced, along with its associated margin of error, which can then be compared to the actual measurement. If the discrepancy between the prediction and the measurement exceeds the calculated margin of error (possibly increased by an arbitrary tolerance), the faulty measurement can be excluded from the solution calculation, as proposed in the article "NLOS Mitigation Method for TDOA Measurement," H. Jiang, J. Xu, and Z. Li, 2010 Sixth International Conference on Intelligent Information Hiding and Multimedia Signal Processing, 2010, pp. 196-199.

[0025] The sequential approach is only feasible if the measurement refresh rate is sufficiently high relative to the speed of movement of the object to be located, in order to extract usable information from a previous estimate, as the information is diluted over time.

[0026] However, increasing the measurement refresh rate leads to an increase in the mobile terminal's power consumption, which goes against the conditions of use of low-power networks such as LPWAN networks.

[0027] The most widespread non-sequential (“single epoch”) approaches generally rely on a redundancy of measurements, that is to say that the number of reference nodes (Na) is significantly greater than the number of unknowns (n+1): Na > n+1. In this case, it is possible to construct subsets of measurements, each of which allows a solution to be calculated.

[0028] The solutions obtained can be compared to determine which subsets lead to consistent solutions and which lead to marginal solutions. From this, it is generally possible to deduce, by analyzing the composition of the inconsistent subsets, the measure or measures that may be faulty.

[0029] These techniques work well when redundancy is very high (Na >> n+1), but they then lead to a very large number of subsets to analyze, which requires high computational capacity.

[0030] Alternatively, the algorithm can rely on knowledge of the environment through a 3D map, in order to predict, for a given position, which measurements are most likely to be biased.

[0031] Such maps are now easily accessible and available for a large number of cities (for example, the OSM Buildings database). Each building is identified by the two-dimensional position of its footprint and by its height.

[0032] Some approaches exploit maps to simulate wave propagation between each possible position of the moving node and the various reference nodes using so-called "ray tracing" techniques to determine a map of the biases that will affect each measurement for each of the possible positions.

[0033] This mapping can then be taken into account by algorithms to correct the measurements and improve the estimation of positions; this is called "NLOS-mitigation" (see for example the article "Mobile Station Localization Emitter in Urban NLoS using Multipath Ray Tracing Fingerprints and Machine Learning", MN De Sousa and RS Thomä, 2018 8th International Conference on Localization and GNSS (ICL-GNSS), 2018, pp. 1-6).

[0034] However, these approaches are on the one hand extremely computationally expensive because bias mapping requires simulating a "complete" propagation (considering possible signal reflections), for each position on the map and with respect to each reference node, and on the other hand very unreliable because these simulations, however complex they may be, struggle to predict the real propagation which is much more complex at the scale of a city.

[0035] Other approaches use the map to predict what measures will be taken in a NLOS situation.

[0036] In particular, patent application WO 2022 / 051615 A1 describes, in the context of a connected vehicle, the use of a map to detect routes in NLOS situations. The map includes the position of all reflectors in the environment of the user terminal. The map makes it possible to determine at least one reference positioning signal (PRS), which propagates along at least one NLOS route.

[0037] The user terminal's position is determined from the PRS reference positioning signal. Patent application WO 2022 / 051615 A1 does not describe how the geolocation algorithm uses the knowledge of the LOS / NLOS condition of each point on the map.

[0038] Patent application WO 2009 / 035952 A1 discloses a method for determining the position of a mobile terminal, using one of the fixed nodes as a reference node (a TDOA type approach, which therefore requires a particular measurement serving as a pivot).

[0039] In this document, an initial position for the mobile node is determined, without considering the map. From this initial position, the map, and the positions of the fixed terminals, the LOS / NLOS paths are evaluated. The reference node is selected from among the LOS nodes. The position of the mobile node is then calculated, taking into account the LOS / NLOS conditions of the measurements.

[0040] However, with this approach, the evaluation of LOS / NLOS conditions based on the map is unreliable because the position of the considered mobile node is determined by an initial estimation that does not utilize knowledge of the LOS / NLOS conditions; it can therefore be highly inaccurate. Consequently, the evaluation of propagation conditions is not based on correct positions.

[0041] For example, if the position considered is wrong, even by only ten meters, an obstacle may or may not be in the path of the wave, making the evaluation erroneous.

[0042] The document "3D building model-based pedestrian positioning method using GPS / GLONASS / QZSS and its reliability calculation" (Hsu et al.) describes a positioning method using a satellite geolocation system, in which the effects of multipathing in an urban environment are mitigated. This is achieved by using a particle filter to distribute the possible positions.

[0043] US document 2022 / 050211 A1 (Bennington et al.) describes a method for providing predictions of accuracy reduction in GNSS navigation for vehicle route calculations.

[0044] There is therefore a need to provide a method for locating a mobile node in a given area, using known positions of reference nodes, which has a high level of accuracy, without however impacting the energy consumption of the mobile node. Summary of the Invention

[0045] The invention aims to remedy the aforementioned drawbacks by proposing a flexible and easy-to-use method.

[0046] The invention relates to a method for locating a mobile node in a given area, as defined in claims 1 to 11.

[0047] The invention also relates to a system for locating a mobile node in a given area, as defined in claims 12 to 14. Figure description

[0048] Other features, details and advantages of the invention will become apparent from the description made with reference to the attached drawings given by way of example. There figure 1 The already described example illustrates the principle of localization by trilateration. figure 2 The already described example illustrates situations of multipath and obstruction. figure 3 is a flowchart of the different steps of the process according to the invention. The figure 4illustrates an example of a geographic area for which the visibility map is established. figure 5 illustrates an example of a visibility map. The figure 6 illustrates an example of a consistency score map. Detailed Description

[0049] A flowchart of the different stages of the process is illustrated by the figure 3 .

[0050] The first step of the process consists of establishing a plurality of visibility maps (step a) of the process). The visibility map corresponds to a predefined 2D or 3D space, which is discretized to obtain a regular or irregular mesh of cells, which correspond to as many possible positions of the moving node whose position we seek to determine.

[0051] The method according to the invention uses as many visibility cards C j that there are reference nodes N j usable for ranging measurements. The position of each reference node N jis known in the visibility map C j .

[0052] The area defined by the visibility map is identical for all visibility maps. It could be, for example, a square several kilometers on each side. The position of a cell can be defined by two coordinates (x, y in 2D) or three coordinates (x, y, z) in 3D. These positions can be chosen according to a grid, for example, where they are uniformly distributed and separated from each other by a predefined interval, such as ten meters in both the x and y directions (or even in all three directions).

[0053] For each cell P and of the visibility map C j a visibility indicator l ij corresponding to radio propagation in direct line of sight between the cell P and and the reference node N j is calculated.

[0054] There figure 4illustrates an example of a geographic area for which the visibility map is established, using several reference nodes N 1 ... N 9 .

[0055] The visibility indicator l ij provides a "Line Of Sight" (LOS) or "Non Line Of Sight" (NLOS) type indication between the cell P and and the reference node N j .

[0056] The visibility indicator l ij This can be calculated through simulation, for example using Matlab software, by loading certain terrain mapping data. The data can come, for example, from the collaborative project "OpenStreetMap," possibly supplemented by data indicating the height of each building. This data can be obtained from the "OSM Building" library.

[0057] Matlab software includes a "RayTracing" function, which allows you to establish a propagation model between two given cells, specifying that there should be no reflections in the propagation path, in order to obtain LOS / NLOS information. Therefore, it is not necessary to simulate a "complete" propagation, considering all possible reflections of the signal.

[0058] Any other simulation software that allows obtaining LOS / NLOS type information between two cells of a map could also be used.

[0059] The visibility map can also be determined by field surveys carried out in another context, rather than by simulations.

[0060] According to one embodiment, the visibility indicator l ijhas a binary value indicating a situation of radio propagation within line of sight or a situation of no radio propagation within line of sight. For example, l ij = 1 in LOS situation between the cell P and and the reference node N j , And l ij = 0 otherwise. Using binary values ​​greatly simplifies subsequent calculations to locate the mobile node.

[0061] Alternatively, the visibility indicator can take a continuous value between 0 and 1, allowing for nuanced obstruction levels and improved accuracy of the localization process.

[0062] Thus, the visibility indicator l ij between the cell P and and the reference node N j can be defined by: l ij = 1 − b int b max si b int < b max 0 sinon

[0063] Considering the NLOS segment that connects the cell P and and the reference node N j the length b intcorresponds to the length of the segment that prevents direct line of sight between the cell P and and the reference node N j . For example, the obstruction can be caused by passing through a building, or by the ground in the case of an intersection with the terrain in the case of uneven ground. A maximum length b max is predefined, for example equal to ten meters. If the obstruction portion of the segment is greater than the maximum length b max it is considered that l ij = 0. This allows us to take into account only small obstacles, for example the edges or borders of buildings, and thus to qualify the LOS / NLOS type information.

[0064] For example, if the segment does not intersect any building (b int = 0), the visibility indicator l ijis equal to 1 as before; if it cuts a building over five meters (b int = 5), the indicator is equal to 0.5, and if the length is greater than or equal to ten meters (b max) then the indicator is equal to 0.

[0065] The visibility indicators l ij are calculated for all cells P and of the visibility map C j, and there are as many visibility maps C j as there are reference nodes N j in the area under consideration. It is possible to consider only a portion of the reference nodes. N j of the area, in particular to reduce the computational cost of the process.

[0066] There figure 5 illustrates a detailed view of the visibility map C 2 associated with the reference node N 2. The grayed-out cells are cells for which the visibility indicator l ij is non-zero (LOS situation). The white cells are cells for which the visibility indicator l ij is zero (NLOS situation).

[0067] Visibility maps can be calculated "offline", that is, before the mobile node location step. They can be calculated once and for all; however, it is preferable that they be calculated periodically, and / or when there are changes to reference nodes (activation of a new reference node or deactivation of a reference node), or when there are modifications in the given area that may influence the calculations of visibility indicators (for example, construction of a building, or the movement, addition or removal of a reference node).

[0068] Creating visibility maps over a large area can lead to particularly long and intensive simulations.

[0069] To shorten the time required to create visibility maps, it is proposed to subdivide each visibility map into multiple zones. Step a) is implemented for each zone independently. Only buildings and other landmarks belonging to the zone in question are loaded for processing.

[0070] The partial establishment of visibility maps can, depending on the implementation methods, be carried out sequentially (each zone after the other), or in parallel (all zones at the same time).

[0071] An overlap between two adjacent areas can be implemented to improve the reliability of the visibility indicator calculation. Indeed, when the cell P andIf the area is at the edge of the zone, some NLOS situations may not be detected because buildings outside the zone are not represented. Adding an overlap (or margin) ensures that the immediate environment of each point on the map is correctly modeled in all directions.

[0072] Advantageously, an overlap of around 25% can be implemented: for an area defined by a 2 km x 2 km square, an overlap of 500 m can be used, which makes it possible to represent all buildings within a neighborhood of at least 500 m.

[0073] The method according to the invention then comprises several steps which are performed "online", as they are implemented each time it is necessary to locate a mobile node. The first of these steps (after the establishment of the visibility maps) consists of providing a set of ranging measurements from signals emitted or received by a plurality of reference nodes.

[0074] The invention is not limited to a particular ranging measurement. Several techniques typically exist for locating a mobile terminal using ranging measurements.

[0075] The first technique, called the "two-way ranging" (or TWR) technique, assumes that the reference nodes and the moving node can act as both transmitters and receivers. In this case, a transmission in one direction (e.g., from the reference node to the moving node) is followed by a transmission in the other direction (from the moving node to the reference node). It can be demonstrated that by combining the different measured start and finish times, the synchronization error can be eliminated for each measurement, thus allowing the distance to be determined.

[0076] A second technique, based on the Time Difference of Arrival (TDOA), allows for the localization of a mobile node by performing unidirectional transmissions. This means that, for example, only the mobile node transmits while the reference nodes receive only, or only the reference nodes transmit while the mobile node receives only. It also assumes that the reference nodes are precisely synchronized with each other; that is, that the unknowns Δ1, Δ2, and Δ3 are identical and reduce to a single value Δ representing the synchronization error between all the reference nodes and the mobile node. In this case, calculating "secondary measures" constructed as the difference of the measurements eliminates the unknown Δ and leads to the following system of equations: m 1 − m 3 = x − x a 1 2 + y − y a 1 2 − x − x a 3 2 + y − y a 3 2 + ε 1 − ε 3 m 2 − m 3 = x − x a 2 2 + y − y a 2 2 − x − x a 3 2 + y − y a 3 2 + ε 2 − ε 3

[0077] It can be noted that, for this approach, a measure (here m 3 ) plays a special role and serves as a kind of pivot.

[0078] This also implies that there must be a number of measurements greater than or equal to n+1, n being the number of unknowns of the position (two in 2D, three in 3D).

[0079] One drawback of this technique is that a single measure plays a special role, since it appears in several of the secondary measures. Indeed, if this measure contains a significant error, this error will be reflected in all the secondary measures. Furthermore, when the number of measures exceeds the minimum required, it is possible, through arbitrary choices, to construct a large number of subsets by selecting different "pivot measures."

[0080] A third technique, called TOA (or Time Of Arrival), is very similar to the TDOA technique, except that, rather than combining measurements to "mathematically" eliminate the synchronization unknown Δi, it is added to the list of unknowns to be determined. This technique requires at least n+1 measurements, but without the drawback of having to arbitrarily determine subgroups of secondary measurements or define pivot measurements. This approach is used by all satellite radionavigation receivers (the receiver plays the role of the moving node, and the satellites that of the reference nodes).

[0081] The invention is presented within the framework of the TOA technique, with a mobile transmitting node and receiving reference nodes. Those skilled in the art will be able to adapt the invention to other localization techniques and to other transmission / reception configurations between the mobile node and the reference node.

[0082] In the first step, it is therefore assumed, according to one embodiment, that a set of M measurements corresponding to the arrival times (ATOs) of the same transmission measured by M reference nodes are received. These arrival times typically have a resolution of a few nanoseconds and are delivered as the decimal part of the current second. Indeed, the arrival times measured by the different reference nodes are sufficiently close (differences << 1 ms) that they belong to the same second.

[0083] These measurements can be transformed into pseudo-distances by multiplying time by the speed of light in order to convert the times into meters, and reduce them to the following equations: m 1 = c . TOA 1 = d 1 + Δ + ε 1 m 2 = c . TOA 2 = d 2 + Δ + ε 2 m 3 = c . TOA 3 = d 3 + Δ + ε 3

[0084] It can be noted that the measurements can be constructed according to the different techniques described above, and can be directly in the form of distance if the measurement uses the round-trip technique TWR.

[0085] The number of measurements received may be less than the number of reference nodes deployed, especially if some reference nodes do not pick up the signal emitted by the mobile node, or if they pick it up but with insufficient power.

[0086] Once ranging measurements have been obtained (or at least part of the ranging measurements), a score will be assigned to each cell, the score indicating the likelihood that the moving node is located in the cell.

[0087] To this end, according to a first sub-step c1) (cf. figure 3 ), a confidence index σ ij is calculated from the visibility indicator l ij between the cell P and and the reference node N j identified by the ranging measurement mj In the data frame of the ranging measurement mj , or in the set of corresponding frames, the identity of the reference node, whose position is known, can be easily identified.

[0088] Considering a cell P and on the visibility map C j associated with the reference node N j the confidence index σ ij associated with a measure mj supposed to be in the cell P and uses the visibility indicator l ij Depending on the embodiment in which the visibility indicator l ij has a binary value, the confidence index σ ij can be defined as follows: σ ij = σ A si l ij = 1 σ B si l ij = 0

[0089] More generally, depending on the implementation method where the visibility indicator l ij is not necessarily a binary value, and can take a continuous value between 0 and 1, the confidence index σ ij is calculated according to the formula: σ ij = σ 0 + 1 − l ij σ 1

[0090] σ 0 and σ 1 correspond to predetermined values, identical for all cells and for all reference nodes.

[0091] For example, it is possible to use a ratio of one to twenty between σ 0 and σ 1, in particular the following values: σ 0 = 100 m and σ 1 = 2000 m .

[0092] The confidence level can also depend on the altitude of the reference node. For example, when the altitude of a reference node is above a certain predefined altitude, the confidence level is a constant value regardless of the visibility indicator. The reference node is indeed visible regardless of the cell's position and its surroundings. This allows for a more accurate representation of reality (the calculated NLOS of the map is not effective, and propagation, in practice, more closely resembles LOS conditions), thus improving the accuracy of the process.

[0093] In another embodiment, if no visibility indicator is associated with the cell in question, the available visibility indicator of the nearest available cell can be used. Thus, for the calculation of the confidence index, it is not essential to have the visibility indicator of all the cells on the map.

[0094] The second substep consists of calculating a normalized residual δ ij N , for each cell P and and for each ranging measure mj , with the following formula: δ ij N = m j − d ij σ ij

[0095] is a corresponds to the theoretical distance between the cell P and and the reference node N j .

[0096] mj corresponds to the ranging measurement of the reference node N j . If the ranging measurement is an arrival time, or a difference in arrival time, it can be transformed into a pseudo-distance by multiplying by the speed of light.

[0097] The normalized residue δ ij N therefore reports the residue δ ij = mj - d ij with respect to the expected precision defined by σ ij , which contains LOS or NLOS information.

[0098] According to one embodiment, the residues σ ijtoo large, and, by transposition, the normalized residuals δ ij N Measurements that exceed a predetermined threshold can be removed from the rest of the process. These measurements are, in fact, highly biased.

[0099] The normalized residue δ ij N is therefore obtained by assuming that the mobile node is located in the cell P and Then, by calculating the theoretical measurements (for example, the geometric distance) between the cell's position and the various reference nodes. The residuals are finally formed by calculating the difference between these theoretical measurements and the actual measurements. Thus, the closer the considered position on the map is to the actual position, the lower the residual error, and vice versa.

[0100] The implementation of the process may involve first calculating the confidence index. σ ij, then the residue mj - d ij , or vice versa. Substeps c1) and c2) can thus be executed simultaneously.

[0101] If ranging measurements are obtained using the round-trip technique, the reference nodes are synchronized with each other, so the only synchronization error to identify is the synchronization error between the moving node and the reference node. This synchronization error is eliminated during the measurement difference calculations, as previously described.

[0102] If ranging measurements are obtained using other techniques, such as TOA or TDOA, as used in the LoRa system, the reference nodes are not synchronized. The mj measurements are therefore pseudo-distances, which have a high value (because the mj transmission time is unknown). It is therefore necessary to account for a synchronization error between all the reference nodes. N j for the cell P and assuming that the mobile node is located in the cell P and Thus, the synchronization defect is taken into account in the calculation of the normalized residual: δ ij N = m j − d ij − Δ i σ ij

[0103] For each cell P and The synchronization error, which corresponds to the moment the measurement was taken, can be calculated by performing a weighted average of the residuals: Δ i = 1 ∑ j = 1 M σ ij 2 ∑ j = 1 M m j − d ij σ ij 2 where M corresponds to the total number of reference nodes N j considered for the location of the mobile node.

[0104] Thus, the process according to the invention advantageously exploits the residue mj - is a and it is assumed that the synchronization error is equal to the weighted average of the residuals. Each residual is weighted by the confidence index. σ ij This gives us the true distance by subtracting the synchronization error value Δ i at the residual mj - d ij .

[0105] According to one embodiment, the synchronization fault Δ i may have resulted from one or more previous estimates. Indeed, the synchronization defect Δ i may evolve over time, but it can be considered that the temporal drift of the synchronization fault Δ i The synchronization error, related to the system's electronics, is relatively slow and is not linked to the movement of the mobile node. This embodiment is particularly applicable for satellite positioning. The determination of the synchronization error Δ i from previous estimates can be made using for example a moving average, a low-pass filter or a Kalman filter.

[0106] In another embodiment, the visibility map can include a predetermined bias value for certain cells. The bias value corresponds to a systematic error, i.e., a constant residual observed over time. The bias value can be subtracted during the calculation of the residual and the normalized residual. This significantly improves the accuracy of the localization. This embodiment is particularly suitable for localizing a moving node along a known path, for example, along a railway line or on a production line.

[0107] In a third sub-step c3), a consistency score if is calculated for each cell P and It indicates the consistency between the cell's position and the observed measurements. This indicator can be calculated using the following formula: s i = K . exp − 1 2 ∑ j = 1 M δ ij N 2

[0108] This formula corresponds to the calculation of a likelihood, with K an optional normalization constant such that, for example, the sum of all scores equals 1. Calculating such a score for each position on the map leads to a "mapping" of the scores, modeled on visibility maps, and illustrated as an example on the figure 6 .

[0109] There figure 6 therefore illustrates the value of the indicator if for each cell P and on a scale of 0 to 100%. The x and y coordinates correspond, in meters, to the distance from the origin of the graph. The probability of the moving node being located in a cell is related to its score. In this case, the moving node is more likely to be located in a cell with a high score, for example, between 90% and 100%.

[0110] Advantageously, the consistency score ifat a given time t can be calculated based on the consistency score if at a time t - 1, according to the following formula: s i t = s i t − 1 ∗ K t . exp − 1 2 ∑ j = 1 M δ ij N 2

[0111] This approach transforms the instantaneous algorithm into a sequential one that incorporates historical measurements and previous solutions. The measurement refresh rate is determined by the speed of movement of the mobile node, particularly if it is mounted on a moving platform. For example, if the mobile node is intended to be carried by a pedestrian, the score measurement refresh rate will be lower than if the node is intended to be carried by a vehicle. For objects designed to be static, such as home automation sensors or outdoor sensors for measuring environmental data, a daily refresh rate may be sufficient.

[0112] The fourth step of the process (see step d), figure 3 ) consists of determining the position of the mobile node NM based on the consistency score if of each cell P and The estimation of the mobile node's position can be done by choosing, for example, the position with the highest score ( P i = arg max here ). Cells with a high score have a high probability that the moving node is located there.

[0113] Other embodiments can be considered to determine the position of the mobile node NM.

[0114] In particular, the score map, which delimits the same area as the visibility maps, can be subdivided into a plurality of zones, which may or may not be of equal size. The zones are not defined a priori and are formed based on scores calculated using clustering techniques. An average coherence score is calculated for each zone, based on the average of the coherence scores of the cells within that zone. Finally, the position of the moving node corresponds to the cell with the highest coherence score, within the zone having the highest average coherence score. This implementation resolves ambiguities when cells with high scores are very far apart, or if they are located in disjoint zones.

[0115] Instead of using the average consistency score, it would also be possible to use the median score.

[0116] Alternatively, only cells with a score above a predetermined threshold, and / or only areas with an average score above a predetermined threshold, are taken into consideration to determine the location of the mobile node.

[0117] The maximum a posteriori method can also be used to estimate the position of the moving node from the score mapping.

[0118] The invention has been described, so far, by implementing the localization method over the entire sector corresponding to the visibility maps, at all possible positions on the map. According to one embodiment, if the part of the map in which the mobile node can a priori be located is known (for example, from a previous iteration of the method, or through prior knowledge of the area in which the mobile node is located), the search can be limited to the cells of the considered part of the map, thus reducing the computational cost.

[0119] In another embodiment, the calculation is performed only on a subset of cells, for example, randomly selected cells or cells selected according to a predefined resolution (e.g., one cell out of ten, or one cell out of one hundred). The results, i.e., the scores, obtained from this first subset can then be refined by performing a second calculation on a second subset of cells, chosen, for example, around the cells in the first subset with the highest scores. This embodiment therefore consists of performing a first calculation with a coarse sample and then a second calculation with a finer sample around the cells with high scores. This embodiment reduces the number of scores to be calculated.

[0120] The invention is particularly suited to time-of-flight measurements over relatively large distances (typically greater than 100 meters), which concerns LPWAN (Low Power Wide Area Network) type networks, such as the LoRaWAN network, or satellite radionavigation systems (GPS, Galileo) or 4G / 5G type cellular networks.

[0121] The invention also relates to a system capable of implementing the predefined method. The system may, for example, be a satellite positioning system: the mobile node is a radio navigation receiver, and the reference nodes are satellites. The invention may also be implemented in a positioning system in which the reference nodes (anchors) operate in transmit mode, and the mobile nodes (tags) operate in receive mode. According to another configuration, the reference node may be a receiving anchor of a telemetry system, and the mobile node may be a transmitting terminal.

[0122] For cellular localization (e.g., according to 4G or 5G standards), base stations are the reference nodes operating in transmission, and phones are the mobile nodes operating in reception.

[0123] The different stages of the process can be executed by a computing device integrated into one of the reference nodes or the moving node, or by an external computing device.

Claims

1. Method for locating a mobile node (NM) in a given area, using reference nodes, the positions of which are known in said area, comprising at least one iteration of the steps of: a) creating, for each reference node (Nj), a visibility map (Cj) comprising, for a plurality of cells (Pi) of the visibility map (Cj), a visibility indicator (lij) corresponding to a line-of-sight radio propagation between the cell (Pi) and the reference node (Nj); b) providing a set of ranging measurements (mj) based on signals transmitted or received by a plurality of reference nodes (Nj); c) computing, for each cell (Pi) of the plurality of cells, a coherence score (si) between the position of the cell (Pi) and the set of ranging measurements (mj), the coherence score (si) being computed, for each reference node (Nj), on the basis of the visibility indicator (lij) between the cell (Pi) and the reference node (Nj) identified by the ranging measurement (mj); d) determining the position of the mobile node (NM) as a function of the coherence score (si) of each cell (Pi), characterized in that step c) comprises: c1) computing, for each ranging measurement (mj) and each cell (Pi), a confidence index (σij) of the ranging measurement (mj) at the cell (Pi) on the basis of the visibility indicator (lij) between the cell (Pi) and the reference node (Nj) identified by the ranging measurement (mj); c2) computing, for each ranging measurement (mj), a normalized residual ( δ ij N ) corresponding to a difference between the ranging measurement (mj) and the distance (dij) between the cell (Pi) and the reference node (Nj), related to the confidence index (σij); c3) computing the coherence score (si) between the position of the cell (Pi) and the set of ranging measurements (mj), on the basis of the normalized residuals ( δ ij N ) associated with each of the reference nodes (Nj).

2. Method according to claim 1, wherein the visibility indicator (lij) has a binary value indicating a situation involving a line-of-sight radio propagation or a situation not involving a line-of-sight radio propagation.

3. Method according to claim 1, wherein the visibility indicator lij between the cell Pi and the reference node Nj has a continuous value, defined by: l ij = 1 − b int b max if b int < b max 0 otherwise wherein bint corresponds to an obstruction length of a segment between the cell Pi and the reference node Nj, and bmax corresponds to a predefined maximum length.

4. Method according to any one of claims 1 to 3, wherein the confidence index σij is computed according to the formula: σij = σ0+ (1-lij) σ1, wherein lij corresponds to the visibility indicator between the cell Pi and the reference node Nj, and σ0 and σ1 correspond to predetermined values.

5. Method according to any one of claims 1 to 3, wherein the confidence index, for a reference node (Nj), is equal to a constant value independently of the visibility indicator if the altitude of a reference node (Nj) is higher than a predefined altitude.

6. Method according to any one of claims 1 to 5, wherein, if a visibility indicator is not available for a cell (Pi) of the visibility map (Cj), the available visibility indicator (lij) of the nearest cell is used to compute the confidence index (σij).

7. Method according to any one of the preceding claims, wherein, in the absence of synchronization between all the reference nodes Nj, the normalized residual δ ij N is computed according to the formula: δ ij N = m j − d ij − Δ i σ ij where mj corresponds to the ranging measurement, dij corresponds to the distance between the cell Pi and the reference node Nj, σij corresponds to the confidence index associated with the measurement mj at the position of the cell Pi, and Δi corresponds to a synchronization error value between all the reference nodes Nj for the cell Pi.

8. Method according to any one of claims 1 to 7, wherein a bias value is associated with at least one cell, the bias value corresponding, for said cell, to a normalized residual ( δ ij N ) that is constant and greater than a predetermined threshold observed for a plurality of previous iterations of the method, and wherein the bias value is subtracted from the ranging measurement (mj) during the sub-step c2).

9. Method according to any one of claims 1 to 8, wherein the coherence score si is computed according to the formula: s i = K exp − 1 2 Σ M j = 1 δ ij N 2 where K is a normalization constant.

10. Method according to any of the preceding claims, wherein the coherence score si at a time t is also computed as a function of the coherence score si at a time t - 1.

11. Method according to any of the preceding claims, wherein - the visibility map (Cj) is subdivided into a plurality of areas; - an average coherence score is computed for each area, based on the average of the coherence scores of the cells in the area; - the position of the mobile node (NM) corresponds to the cell (Pi) for which the coherence score (si) is maximum, in the area with the maximum average coherence score.

12. System for locating a mobile node (NM) in a given area, using reference nodes, the positions of which are known in said area, the system being configured for: a) creating, for each reference node (Nj), a visibility map (Cj) comprising, for a plurality of cells (Pi) of the visibility map (Cj), a visibility indicator (lij) corresponding to a line-of-sight radio propagation between the cell (Pi) and the reference node (Nj); b) providing a set of ranging measurements (mj) based on signals transmitted or received by a plurality of reference nodes (Nj); c) computing, for each cell (Pi) of the plurality of cells, a coherence score (si) between the position of the cell (Pi) and the set of ranging measurements (mj), the coherence score (si) being computed, for each reference node (Nj), on the basis of the visibility indicator (lij) between the cell (Pi) and the reference node (Nj) identified by the ranging measurement (mj); d) determining the position of the mobile node (NM) as a function of the coherence score (si) of each cell (Pi); characterized in that the system is further configured to: c1) computing, for each ranging measurement (mj) and each cell (Pi), a confidence index (σij) of the ranging measurement (mj) at the cell (Pi) on the basis of the visibility indicator (lij) between the cell (Pi) and the reference node (Nj) identified by the ranging measurement (mj); c2) computing, for each ranging measurement (mj), a normalized residual ( δ ij N ) corresponding to a difference between the ranging measurement (mj) and the distance (dij) between the cell (Pi) and the reference node (Nj), related to the confidence index (σij); c3) computing the coherence score (si) between the position of the cell (Pi) and the set of ranging measurements (mj), on the basis of the normalized residuals δ ij N ) associated with each of the reference nodes (Nj).

13. System according to claim 12, wherein the mobile node (NM) is a transmitter tag, and the reference nodes (Nj) are anchors.

14. System according to claim 12, wherein the mobile node (NM) is a radionavigation receiver, and the reference nodes (Nj) are satellites.