ES2536586A1 - Method of location in interior spaces based on detection and pairing of light points (Machine-translation by Google Translate, not legally binding) - Google Patents

Abstract

Method of localization in interior spaces based on detection and pairing of points of light that includes obtaining a map of points of light of an enclosure; detect the movement of a user; detect the passage under points of light arranged in the room, by means of a light sensor provided in a mobile device; where in addition the method comprises: establishing hypothesis of the user's initial location with respect to the map of points of light; update the hypothesis of the user's location according to the information relative to the movement of the user and each time a point of light is detected; assign a probability to the location hypotheses, y; determine the specific location of the user as designated by the location hypothesis assigned the highest probability. (Machine-translation by Google Translate, not legally binding)

Description

Indoor location method based on detection and pairing of light points

DESCRIPTION

 5

Object of the invention

The present invention relates to a method of location in interior spaces based on the detection of conventionally existing points of light in the ceilings of corridors or rooms of a building or an enclosure for lighting the 10 different environments.

The method foresees to make a pairing or comparison of the points of light detected with the points of light that appear in a map of points of light of said enclosure, obtained previously. fifteen

It has application in the industry dedicated to the design, manufacture and marketing of positioning and navigation systems, and more specifically in the industry dedicated to the design and development of location and navigation applications on tablets or mobile phones of the SmartPhone type. twenty

Technical problem to be solved and Background of the invention

In today's world, knowing the exact or approximate location of a person or object in its displacement, both for exterior and interior environments, is essential for many purposes.

In medium or long distance journeys of people and goods, it is extremely useful for the driver of a vehicle to know its location at all times, and this can be difficult without the support of positioning systems that provide the driver with information about his location, when he moves to places unknown to him and that at times, may lack adequate geographical signs on the road.

The satellite positioning systems (GNSS), such as the American Global Positioning System (GPS), the Russian Glonass or the Galileo European positioning system, based on the reception of signals emitted by different satellites to know the geographical coordinates of a device, have become a technology in common use today. 5

However, today there are no sufficiently reliable and economical systems that provide accurate information about the location of people or objects, when they move inside enclosures or buildings, where the GPS signal is not available. 10

It is true that positioning systems have been developed in indoor spaces based on the emission / reception of radio-frequency (RF) signals or WiFi signals. Such systems consist in that a mobile device carried by a person or by an autonomous mobile element receives signals emitted by different fixed devices or 15 beacons arranged ad-hoc in fixed locations of an interior space. In these cases, the fixed devices provide the mobile device with accurate information about its position in an indoor enclosure.

The main drawback of the mentioned systems is that they need to make a specific installation for said system of signal emitters throughout the entire interior of a building.

Likewise, and as a particular case of the aforementioned systems, interior positioning systems based on the collection of 25 by a mobile light device emitted by various fixed light sources and installed specifically for said systems in different locations within have been developed of an interior space.

As an example of the above, the invention described in the article "Indoor 30 positioning and Navigation using Light sensor", M. Ambur, International Journal of Research in Computer and Communication Technology (IJRCCT), vol. 2 no. 1, pp. 20-23, 2013.

As can be seen, in said invention, fluorescent light sources expressly installed for the positioning system emit light signals at a certain frequency, thus carrying accurate information about their location, to be received by a mobile device. This system has the same disadvantage of needing a specific installation, in addition to the facilities of any property.

There are other systems that, by means of a camera provided in a mobile device, capture the light component of a certain frequency emitted by fluorescent light sources. An example of these systems is described in the article “Indoor 10 Navigation using a diverse set of cheap, wearable sensors”, A. R. Golding, Third Int. Symposium on Wearable Computers, (San Francisco), pp. 29-36, 1999.

The drawback of such systems is that they are only valid for a specific type of light source, and for light of a certain frequency. fifteen

The focus of inventions such as that described in the aforementioned article is the attainment of a “footprint” of the lighting of certain rooms or areas. However, the location accuracy obtained by this type of system is low, and they have the added disadvantage of the granularity and the need to carry out 20 frequent calibrations to accommodate changes in the surrounding lighting.

Also, the article “A Corridors Lights based Navigation System including Path Definition using a Topologically Corrected Map for Indoor Mobile Robots”, F. Launay, et al., Proceedings 2002 IEEE, International Conference on 25 Robotics and Automation (Cat. No. 02CH37292), vol. 4, no. May, pp. 3918-3923, 2002.

This article describes an interior navigation system designed for the positioning of robots that move through interior spaces. In this system, the robot is equipped with a conventional camera that points at all times towards the ceiling of an interior enclosure to capture the light coming from points of light of the conventional light installation of a building.

This system has the advantage of not requiring a specific installation and not requiring a special type of light source of a certain frequency. However, 35

The need for processing the images captured by a conventional camera provided in the robot confers some complexity added to said system, to discriminate which are the points of light within the image.

Likewise, it is necessary to have information for each building, about 5 of the place where the different points of light are located in each of the rooms and corridors, so that the robot can compare the detected lights with a map of points of light. This problem is solved in said system by a path previously made by the robot, in which the points of light are captured along a path. 10

The aforementioned system comprises a specific error correction algorithm, to close “cycles” or routes closed by corridors or rooms, thus avoiding that due to the inaccuracy of the displacement detectors, different points of light are identified as what actually It is the same point of light at the beginning and end of these cycles.

The aforementioned displacement detectors for robots are usually based on “odometry” or accumulation of the number of turns on the wheels and the rotation of the direction of the robot. There are other displacement detectors based on inertial measurement units 20 (also known as IMU) that include accelerometers that use the technology known as PDR (Pedestrian Dead-Reckoning), to measure the number of steps given by a user and the orientation in their movement.

 25

In addition, in the aforementioned system, it is assumed that there should only be one point of light in the field of vision of the robot's camera (§ 1, Section 3, p. 3919), so that the detection of points of light In addition to requiring a processing of the images captured by a conventional camera, it can be not very robust and lead to errors if several points of light appear in the field of view of the robot's camera 30.

Additionally, in the system mentioned in F. Launay's article, given an estimate of the location of the robot, the position of the nearest light point of the set of light points in the building is used to update the estimate of position 35

of the robot when a point of light is detected in the field of view of the camera (Section 5.1, p.3921). In this way, the method of F. Launay's article assumes that the uncertainty in the estimation of the location is unimodal, and therefore does not contemplate the coexistence of multiple hypotheses, assuming also that the exact location of the robot at the beginning is known a priori of a route, both aspects being 5 quite limiting.

By means of the invention described in the present application, it is intended to provide a method of location in interior spaces based on inertial measurements and the detection of points of light, of which they are forming part of the conventional light installation 10 of a building, which It can be used by anyone, for example, by carrying a SmartPhone mobile phone that incorporates a light sensor and an inertial motion sensor. By means of the implementation of a particle filter and the pairing of the detected points of light with the points of light that appear in a map of points of previously obtained light, it is possible to converge towards a single hypothesis of location of a person carrying said SmartPhone

The present method does not assume that the exact location of a user carrying said mobile device is known a priori. The method allows the coexistence of 20 multiple hypotheses or particles that, through a particle filter methodology, converge towards a single permissible hypothesis of user location within an enclosure.

As important advantages over the state of the art, the present method avoids the need for the processing of images captured by a conventional camera, since for the detection of points of light the light sensor with which mobile phones are conventionally equipped are used SmartPhone type.

Likewise, it is a dynamic method, which can be implemented in SmartPhone 30 mobile phones without the need for additional hardware, and for the comparison of the detected light points it is not necessary to make a previous tour of a building to experimentally obtain a map, which would also require a specific error correction algorithm, but a downloadable file in situ or

via the internet with the exact location of all the points of light in the interior space in question.

Description of the invention

 5

The present invention relates to a method of localization in indoor spaces based on detection and pairing of points of light comprising:

 obtain a map of points of light of an enclosure, said map of points of light comprising information relative to the location of all the points of light 10 of the enclosure that can be detected by means of a light sensor;

 detect the movement of a user, in its journey through the enclosure, by means of a motion sensor provided in a mobile device;

 detect the passage under points of light arranged in the enclosure, by means of a luminous sensor provided in a mobile device; fifteen

The location method object of the present invention further comprises:

a) estimate the initial location of the user based on the points of light detected by the luminous sensor of the mobile device, establishing hypothesis of the user's location with respect to the map of points of light, where the non-detection of a point of light implies that any hypothesis about the user's initial location with respect to the map of points of light is admissible;

b) update the hypothesis of the user's location with respect to the map of 25 light points, according to the information related to the movement of the user, provided by the motion sensor of the mobile device, where the hypothesis of the user's location is displaced with respect to to the map of points of light based on said movement related information;

 30

c) update the user location assumptions with respect to the map of points of light each time a point of light is detected by means of the luminous sensor of the mobile device, where the location assumptions are concentrated

around points of light that appear on the map of points of light each time a point of light is detected;

d) assign a probability to the location assumptions each time a light point is detected by the light sensor of the mobile device;

e) determine the user's specific location as the one designated by the location hypothesis that has the highest probability assigned.

 10

According to the method of localization in interior spaces based on detection and pairing of points of light object of the present invention, to determine the location of the user with respect to the map of points of light, a particle filter type technique, PF, is used. where a location hypothesis is made up of at least one particle positioned on the map of points of light, and where each particle is assigned a state that defines its position and orientation on the map of points of light.

According to a preferred embodiment of the present method of location in 20 interior spaces based on detection and matching of light points, the map of light points comprises information regarding the size and orientation of all the light points of the enclosure.

The following mathematical representation is used to define a position and orientation of each particle:

which incorporates the three-dimensional position of the particle in space, and the orientation "" of the particle on the horizontal plane.

The user is located when the particles that constitute each hypothesis of location are concentrated within a single and small region of the enclosure that appears on the map of points of light, said particles comprising similar positions and orientations.

The size of said small region of the enclosure represents the accuracy of the location method.

According to a preferred embodiment of the present method of location in indoor spaces based on detection and pairing of light points, for each step that the user takes, an estimation of the change of position and orientation is performed that is represented mathematically in the following way:

  [] [] [] [] [],

 10

where "j" is an ordinal index referred to the number of steps taken by the user and accounted for by the motion sensor of the mobile device, and the increments "refer to a temporary reference system of the user of the mobile device in the previous step, where the instants of time in which step "j-1" and its successive occur, are indicated by this notation: []. fifteen

The method of localization in indoor spaces based on detection and pairing of points of light preferably comprises using a previously established movement model, determined by a covariance matrix that represents the uncertainty in the estimation of the change of position and orientation of the user in each Passage of your tour of the enclosure.

This covariance matrix takes values that depend on the motion sensor and the user's walking mode.

 25

According to a preferred embodiment of the present location method, a particular state is attributed to each particle that is part of a location hypothesis, according to the following mathematical formula:

     [] [] [] []

 30

where "it symbolizes the state of a particle," i ", said state being given by its position and orientation estimated by the motion sensor in the previous step, and

by a nonlinear function "f" of the estimated change of orientation and position in the last step taken by the user, and of the given number of steps, where the function f is represented mathematically as follows:

  ([] []) [([]) ([]) ([]) ([])] ([[] [] [] []] √ []) 5

where

- "[]" is the orientation of each particle, "i", with respect to a local reference, and corresponding to step (j); 10

- "" is a random vector of dimension 4x1, which has a Gaussian distribution of zero mean, and which serves to disperse the particles according to the movement model used in the location method, and;

- “P [j]” is the covariance matrix that determines the movement model. fifteen

The assignment of a probability to each user location hypothesis is carried out by assigning a weight to each particle that constitutes said location hypothesis, depending on the separation distance between each of the particles that constitute each location hypothesis and the points of light listed on the map of 20 points of light.

In the preferred case that the map of points of light also includes information regarding the size and orientation of the points of light of the enclosure, the assignment of a probability to each hypothesis of location of the user, is performed by assigning a weight to each particle which constitutes said location hypothesis, also taking into account the size and orientation of the points of light that appear on the map of points of light.

The following mathematical formula is preferably used for the attribution of a weight to each constituent particle of each hypothesis (3) of location:

     [] [] [] | ̂ []

where "[] is the weight assigned to the particle," i ", for the instant detection" k "of a point of light; where "k" refers to an instant in which a luminosity measurement is produced by the light sensor, said instant "k" referred to a particular temporal reference system of the light sensor, "[]", and where the Weight “[]” is attributed based on the following parameters:

- the weight "[]", assigned to the particle, "i", in the previous instant, "k-1";

- an estimated state, "̂ []", of the particle, "i", at the instant "k" referred to the particular temporal reference system of the light sensor, "[]";

- the probability, "p", that, given an instant "k" in the particular temporal reference system of the light sensor, "[]", and an estimated state "̂ []" for the particle, "i", there is a detection of a point of light by the light sensor, "[]", where the variable z [k] takes only two values, 0 or 1, depending on, respectively, at the moment "k" whether or not a light point is detected by the light sensor;

- a coefficient “”, whose function is to normalize the weights attributed to each particle so that the sum of all of them is equal to 1.

 twenty

The estimated state, "̂ []", of the particle, "i", at the instant "k" referred to the particular temporal reference system of the light sensor, "[]" is preferably calculated by interpolating between two instants "j-1 "And" j "of a local temporal reference system of the motion sensor, [], said instants being" j-1 "and" j "respectively before and after the instant" k "in which a measurement of 25 brightness by the light sensor.

To perform the interpolation mentioned in the previous paragraph, the following mathematical formula is preferably used:

 30 ̂ [] [] [] []

being [] [] [] [], an interpolation coefficient.

The probability, "p", that, given an instant "k" in the particular temporal reference system of the light sensor, "[]", and an estimated state "̂ []" for the particle, 5 "i", a detection of a light point by the light sensor, "[]", is preferably calculated by the following mathematical expression:

  ([] | ̂ []) ∑ √ | | {̂ [] ̂ []}

where: 10

- "̂ []", is an estimated position of the particle, "i", at the instant "k" referred to the particular temporal reference system of the light sensor, "[]", and;

- L represents the total number of points of light on the map of points of light;

 fifteen

this probability being therefore the sum of multiple Gaussians located with center in each point of light and with a shape defined by the size and orientation of each point of light, represented by “.

The variable "[]" is typically evaluated by the following operations: 20

- low pass filtering;

- calculation of the derivative of the luminosity;

- peak detection based on finding a zero crossing surrounded, before and after, by minimum values of the luminosity derivative. 25

Finally, the most likely location of a user after step "j" and the measure "k" is calculated using the following mathematical expression:

         ∑ [] []

where

- N is the total number of particles, "i", which constitute the location assumptions; 5

- [] {[] [] [] []}, is the state of each particle after the step “j” given by the user, and;

- [] are the weights of each particle after the “k” measurement taken by the light sensor.

 10

According to a possible embodiment of the method of location in indoor spaces based on detection and pairing of light points, the map of light points is obtained by downloading from an online server on a mobile device.

According to another possible embodiment of the present method of location in 15 interior spaces based on detection and pairing of light points, the map of light points is obtained, by downloading via bluetooth or other method of local communication on the mobile device , in each enclosure.

Brief description of figures 20

The present invention will be better understood in view of the following figures, which are shown in an explanatory and non-limiting manner.

Figure 1: Shows a representation of a user carrying a mobile device with a light sensor to detect points of light when traveling through an enclosure.

Figures 2a - 2f: They show by way of scheme, an example of the operation of the method of localization in interior spaces object of the present invention, in successive instants of time along the route of a user through an enclosure. 30

Figure 3: Shows a block diagram with the constituent elements of the method of localization in interior spaces object of the present invention, according to a possible embodiment of the present invention.

Below is a list of the numerical references used in the figures.

1. Point of light. 5

2. Particle.

3. Hypothesis.

4. Enclosure.

5. Travel.

6. Mobile device. 10

7. User.

100. First block.

101. Motion sensor.

102. Step detection module.

103. Movement model. fifteen

200. Second block.

201. Luminous sensor.

202. Light bulb detector.

203. Measurement model.

204. Map of points of light. twenty

300. Third block.

301. First stage.

302. Second stage.

303. Location block.

304. Additional stage. 25

400. Other sensors or movement restrictions.

1000. First information.

2000. Second information.

3000. User location.

4000. Additional sources of information. 30

Detailed description of an embodiment of the invention

The present invention relates, as already described, to a method of localization in indoor spaces based on the detection of light points (1).

conventionally provided in the light installation of an interior or real estate (4), and in the pairing or collation of said points of light (1) detected with the points of light (1) contained in a map of points of light (204) ) of said enclosure (4), previously obtained.

 5

The method object of the present invention is typically implemented in mobile devices (6), generally in SmartPhone mobile phones, which are provided with a light sensor (201) and a motion sensor (101), typically consisting of a unit inertial measurement, also known as IMU, which typically comprises three accelerometers and three gyroscopes, arranged in three orthogonal directions.

However, the present method of location in indoor spaces is capable of being implemented in any mobile device (6), such as tablets or cameras, which have a light sensor (201) and a motion sensor 15 (101). ).

Generally, any mobile device (6) in which the installation of a GPS system is provided, is equipped with a motion sensor (101), necessary to improve, through sensory fusion algorithms, the operation of the GPS system. twenty

The light sensor (201) of the mobile device (6), detects the light coming from the lamps or points of light (1) provided in the light installation of an enclosure (4) when a user (7) moves through the interior of said enclosure (4) with the mobile device (6) in the hand or simply, with the mobile device (6) arranged in such a way that the luminous sensor (201) of the latter can receive the light coming from said points of light ( one).

For the correct operation of the method object of the present invention, the mobile device (6) must be arranged such that its light sensor (201) 30 is pointing towards the ceiling, or in general, pointing towards the wall, ceiling, or floor where the points of light (1) that illuminate the different environments of the enclosure (4) are provided.

Generally, said points of light (1) are arranged on the ceiling, and the light sensor (201) of the mobile device (6) is typically arranged on that face of the

mobile device (6) that is provided with a screen. So when a user (7) moves looking at the screen of his mobile device (6), said mobile device (6) is usually placed face up, whereby the light sensor (201) points towards the ceiling, where they are arranged the points of light (1) to be detected.

 5

The present method of indoor location based on detection and pairing of light points (1) makes use of said light sensor (201) of the mobile device (6) and of the motion sensor (101) of the mobile device (6).

The technique for detecting displacement using the motion sensor (101) 10 is typically known conventionally as PDR (Pedestrian Dead-Reckoning).

The present method of location in indoor spaces provides for comparing or comparing each time a light spot (1) is detected by the light sensor (201) of the mobile device (6), and by a particle filter type methodology (also known as PF, by its acronym in English, the distances of each particle (2) to each point of light (1), between the points of light (1) that appear on a map of points of light (204) of the enclosure (4) in which the exact location, orientation and dimensions of said points of light (1) are indicated. The map of points of light (204) is typically obtained by downloading a file through the internet, by means of a wireless connection (eg WiFi or telephony) or wired, or via Bluetooth or other method of field communication local upon arrival at the site (4).

Although there is numerous bibliography referring to methodologies of the “25-particle filter” type, PF, it is enough to clarify that with the term “particle” (2), in the present description, reference is made to a possible location and specific orientation of the user (7 ) in the enclosure (4). A set of particles (2) located very close to each other, constitute what is called "hypothesis" (3) of user location (7) within the enclosure (4). 30

A particle filter (PF) is used, which implies the use of a set of numerous particles (2), to represent among them all the hypotheses (3) or areas of possible location of a user (7) within the enclosure (4 ). A hypothesis (3), as just mentioned, is constituted by a group of particles (2) that by their similarity in 35

Position and orientation are considered to form a single entity and therefore support a single hypothesis (3) of specific location. In English terminology a hypothesis (3) is formed by a “cluster”, or group of similar particles (2).

A user (7) carrying the mobile device (6) that detects the light points (1) 5 while making a certain route (5), can benefit from the present invention to have a continuous update of the hypotheses (3) that define Your location or location. A particle filter type (PF) methodology is responsible for purifying, discarding and regenerating the particles (2) that define the possible location hypotheses (3) until they are concentrated spatially within the enclosure (4), minimizing the possible regions where the user (7) can be located, or in other words, limiting the number and size of the possible admissible hypotheses (3), eventually converging towards a single admissible hypothesis (3), which corresponds to a concentration of the particles (2) on a single and small region of the enclosure (4), the size of which represents the precision of the location method.

The present method of location in interior spaces has the added advantage that it is not excessively relevant that any of the lamps or points of light (1) of an enclosure (4) is turned off and without emitting light, so that it cannot 20 be detected by the light sensor (201) of the mobile device (6), although according to the combination of information provided by the motion sensor (101) and by the map of points of light (204), it is inferred that it should be under a point of light (1) that appears on the map of points of light (204).

 25

Thanks to the map of points of light (204), the motion sensor (101), and the particle filter methodology, PF, which is used by this method of location in indoor spaces, when points of light are detected ( 1) along a route (5) through an enclosure (4), the most probable location of the user (7) carrying the mobile device (6) with which the 30 point of detection has been detected is determined at all times light (1), based on a calculation of probabilities, which discards possible hypotheses (3) of user location (7) as the path (5) travels through the enclosure (4) and according to the light sensor (201 ) is detecting points of light (1).

Initially, the particles (2) are scattered throughout the enclosure (4) contained in the map of points of light (204), so that all hypotheses (3) of the user's location are considered admissible, given that no light spot detection has yet occurred (1).

 5

At the moment when the detection of the first point of light (1) occurs, the particles (2) are grouped around each and every one of the points of light (1) that appear on the map of points of light (204), so that each grouping of particles (2), comprising particles (2) oriented in all possible directions of the horizontal plane, is considered to be a hypothesis (3) of location of the user (7) within the enclosure (4).

After this first detection of a point of light (1), all particles (2) are assigned the same weight or probability, so that all assumptions (3) of user location (7) are considered equally admissible. fifteen

From that moment, the particles (2) move along the map of points of light (204), in all directions and in as much distance and as much angle of rotation as indicated by the motion sensor (101) of the mobile device ( 6).

 twenty

When a second detection of a point of light (1) occurs, the distance of each particle (2) to each and every one of the points of light (1) contained in the map of points of light (204) is calculated. , and it is considered that the most probable location of each particle (2) is that determined by the shortest possible distance to one of the points of light (1) of the map of points of light (204). 25

At this time the weight of each particle (2) is updated, giving more weight to the particles (2) that are closest to one of the points of light (1) that appear on the map of points of light (204) , ensuring that the sum of weights of all particles (2) is equal to 1. 30

The method object of the present invention converges towards an isolated grouping of particles (2), that is, towards a hypothesis (3) of unique location, as the path (5) is advanced along the enclosure (4) and according to more light points (1) are detected, or put another way, as more information about the 35

location of the user (7) carrying the mobile device (6), which makes it possible to rule out false hypotheses (3) of the location of the mobile device (6) with respect to the enclosure (4).

Obviously, the fewer points of light (1) are turned off, and therefore, more points of light (1) can be detected by the light sensor (201) of the mobile device (6), the more information is provided to the method, being able to improve the comparison of the points of light (1) detected with the map of points of light (204) and more quickly discard false hypotheses (3) about the location, and therefore, more rapidly converging the method towards a hypothesis (3) unique.

 10

The method of the present invention involves the merging of at least two types of information:

1) the detection of the presence of points of light (1) or lamps lit on the ceiling of an enclosure (4), and; fifteen

2) obtaining information about the advance and rotation of a user (7) carrying a mobile device (6) while he is walking (pedometric information, Dead-Reckoning or simply PDR).

The method is also called "light pairing", LM (for the 20 acronym in Light Matching), because it relates the lights or points of light (1) detected by the light sensor (201) of the mobile device (6) with the points of light (1) indicated on the map of points of light (204), so that, as the user (7) moves through the enclosure (4), the hypotheses ( 3) that do not match or match the detected light points (1) with the 25 points of light (1) that appear on the map of points of light (204); in this way, only those hypotheses that make the detected light points (1) and their relative arrangement compatible with the light points (1) indicated on the map of light points (1) persist as valid or admissible hypotheses. 204).

 30

The method seeks as a final objective that only a set of particles (2) grouped around the same point of light (1) persists, that is, that only a single hypothesis (3) of admissible location persists that makes compatible the points of light (1) detected and estimated movements, with the points of light (1) indicated in the

map of points of light (204), at which time the user (7) is located and can know its exact location within the enclosure (4).

From the moment in which the user (7) is located and there is only one hypothesis (3) of admissible location, the method continues to operate in the same way, trying to collect enough information from the motion sensor (101) and the light sensor (201), to try to make the hypothesis (3) remain unique and the user (7) continues to be located continuously.

When the user (7) is located, the method works as a system for locating a browser that receives its signal from the light points (1) detected by the light sensor (201) of the mobile device (6) and informs the user (7) in real time of its precise location along its route (5) through the enclosure (4).

 fifteen

The operating principle of this method is summarized below. If it is assumed that there is a map of points of light (204) where exactly the arrangement of the lamps or points of light (1) appears on the ceiling of an enclosure (4) or building (position, size and orientation) and it is also assumed that there is a method to detect if there is a light point (1) on a user (7) while a user makes a tour (5) through the enclosure (4), then it can be deduced that, at the moment when the presence of a light is detected, the user (7) is under one of the light points (1) that appear on the map of light points (204) of the enclosure (4).

 25

For the general case of a user (7) with unknown position and initial orientation, a detection of a point of light (1) implies the creation of as many hypotheses (3) as the number of points of light (1) exist in the building or enclosure (4). These multiple hypotheses (3), representing possible locations of the user (7) within the enclosure (4), can be propagated (moved and rotated), as the user (7) travels along its path (5) through the enclosure (4), according to a motion model (103) and experimental measurements of the motion sensor (101), so that the hypotheses (3) are updated with the information received by said motion sensor (101).

 35

Likewise, the particles (2), which represent the hypothesis (3) of location, can be “weighed” (update their probability) according to each of the light spot detections (1) that are produced by the light sensor (201). After successive detection of points of light (1), only particles (2) that coincide under a point of light (1) persist and the rest of particles (2) disappear, or at least, their probability is minimized. After a period of repetitive detections of points of light (1), of repeating the probabilities of each particle (2), and of propagation of the particles (2) by personal inertial estimation (PDR - Pedestrian Dead-Reckoning), it is possible finally reduce the spatial dispersion and orientation of the particles (2), that is, gradually reduce the number of hypotheses (3) of the user's location (7) 10 and even converge to a single admissible hypothesis (3), the which has the particles (2) concentrated locally and with a predominant orientation, which reliably determines the location of the user (7) within the enclosure (4).

The process of convergence of multiple hypotheses (3) to a single one can be facilitated 15 using optionally other additional sources of information (4000) that come from other sensors or movement restrictions (400), such as a magnetic compass, beacon signals external, etc.

An illustrative example of the operation of the present method is shown in the Figures 2a-2f.

In Figure 2a, two points of light (1) are observed in an enclosure (4) the initial position of the user (7) being unknown; therefore, the hypothesis (3) of initial location is very dispersed, consisting of particles (2) distributed in multiple locations and 25 orientations. The actual travel (5) of the user (7), and its position for different instants of time, is indicated by a straight vertical line.

When the user (7) passes under one of the points of light (1), at time t = k, it is detected that said point of light (1) is on, and the particles (2) representing 30 possible hypotheses (3 ) of location of the user (7), are reduced to those under the points of light (1) of the enclosure (4), as can be seen in Figure 2b.

After advancing the user (7) a distance "d" the particles (2) are distributed as shown in Figure 2c (instant t = k + 1). 35

After advancing another distance "d" (instant t = k + 2 in Figure 2d) another point of light (1) is detected and the particles (2) are updated as shown in Figure 2e.

Figure 2e shows that the two main location hypotheses (3), which correspond to the two main groupings of particles (2), include two defined orientations (as opposed to Figure 2b where they had dispersed orientation), that is, oriented upwards in the upper group, and oriented downwards in the lower group. After further advancing another distance "d" each group of particles (2) moves in their respective directions as seen in Figure 2f. 10

The correct hypothesis (3), of the two remnants, is the superior one, which may finally predominate and be the only one persistent if more light points (1) are still detected, or additional sources of information (4000) from other sensors or movement restrictions (400), such as a magnetic compass 15, external beacon signals, etc.

The method of the present invention is implemented using a type of Bayesian techniques that allow the use and treatment of multiple hypotheses (3), as explained above, using a particle filter (PF), whose mathematical details of implementation. Preferred are given below.

Each particle (2) is assigned a state, which incorporates the three-dimensional position of the particle (2) in space, being its orientation on the horizontal plane. 25

The method, as already mentioned, can be implemented in a standard mobile device (6) (eg smartphone, tablet, laptop, etc.) or other measuring equipment, which incorporates as a fundamental sensor a light sensor ( 201) and a motion sensor (101), typically consisting of an inertial measuring unit 30, IMU, with accelerometers and gyroscopes, and having a suitable processor for calculating position estimates.

Figure 3 shows a block diagram of a possible embodiment of the invention, using a particle filter, PF. The diagram consists of three fundamental blocks (100, 200, 300).

The first block (100) is the PDR block, with inertial pedometry, which in turn consists of a motion sensor (101), a step detection module (102) and integration of the advance made by the user (7 ) when walking, and of a movement model (103) that represents the uncertainty in the estimation of the movement by means of a covariance matrix P [j] associated with each step “j” given by the user (7) and detected by the module step detection (102). This covariance matrix 10 P [j] is diagonal, has a dimension of 4x4 and represents the uncertainty after each step in the four dimensions (forward, lateral, height and turn).

The second block (200) corresponds to the pairing of lights, LM, which in turn consists of a light sensor (201), a light bulb detector (202), of a database or map of points of light (204), with the position, orientation and size of all the light points (1) of the enclosure (4), and finally a measurement model (203) that indicates the probability that, given a state or position / orientation , X, a light point (1) can be detected on, LD (Light Detection), that is, P (LD | X). twenty

A third block (300) corresponds to the particle filter, PF, which is used to merge a first information (1000), from the first block (100), PDR block, and a second information (2000), from the second block (200), light pairing block, LM. Said third block (300) collects the first and the second information (1000, 2000) provided by the PDR (100) and LM (200) blocks, and processes it to finally filter and obtain the user's location (3000), which It is the ultimate goal of the method.

The third block (300) or particle filter block, PF, consists of a first stage 30 (301) of prediction or propagation of the particles (2) that constitute the different hypotheses (3) of the user's location (7), and of a second stage (302) of updating the weights of the particles (2) according to the points of light (1) detected. A location block (303) uses the state of the different particles (2) and their weight, to generate the most likely user location (3000). 35

An additional step (304) is included, as in any conventional particle filter, to re-sample particles (2) in case of degeneration, that is, in case there are a few particles (2) that have all the weight and The vast majority have no weight. The particles (2) are re-sampled by placing many new particles (2) close to the particles (2) that had a greater weight, and the majority of the particles (2) that had little weight were eliminated, leaving the number of particles (2) totals constant.

Finally, the method also includes the possibility of including as many additional sources of information (4000) as desired from other sensors or 10 movement restrictions (400).

Detailing the first block (100), PDR block, the inertial measurement unit, IMU, which constitutes the motion sensor (101), must contain at least one accelerometer and a gyroscope with the sensitive axes aligned vertically, and thus being able to 15 determine the user's progress (7) with respect to the position he occupied in the previous step and his change of orientation. Preferably, three accelerometers and three orthogonal gyros should be used with each other (as is usual in commercial IMU’s); Next, there is a module for the detection of steps (102) and integration of the advance, in order to estimate the travel (5) of a user (7) in the enclosure (4) as accurately as possible.

The first information (1000) related to the change of position and orientation estimated in the step detection module (102) and progress integration, is represented as follows: [] [] [] [] [], for the step number "j" 25 given by the user (7), and referred to the temporal reference system of the motion sensor (101) in the previous step, ie [j-1].

The instants of time in which step j-1 occurs, and its successive j, j + 1, etc., are indicated by this notation: [], which denotes the local time reference system of the motion sensor (101 ).

The movement model (103), modeled with the matrix P, whose value depends on the type of motion sensor and the way of walking, represents the uncertainty in the

position change estimation, [] [] [] [] [], and is a 4x4 dimension matrix.

The third block (300), of particle filter, PF, propagates the state, X, of each of the particles (2), "i", which represent different hypotheses (3) of user location 5 (7), i being an index ranging from 1 to the number N of particles (2) that are desired to be used (typically more than one thousand and less than one million). This propagation is done by this expression:

    [] [] [] [] (ec. 1) 10

Where the function f is a nonlinear transformation given by:

 ([] []) [([]) ([]) ([]) ([])] ([[] [] [] []] √ []) (ec. 2)

 fifteen

Where "[]" is the orientation of each particle (2), "i", with respect to a local reference, and corresponding to step (j). On the other hand, "" is a random vector of dimension 4x1, which has a Gaussian distribution of zero mean, and which serves to disperse the particles (2), "i", according to the movement model, P (eg a matrix whose diagonal is [0.01m, 0.01m, 0.01m, 0.005 rad]). twenty

The second stage (302), of updating the weights of the third block (300), or particle filter block, PF, reallocates new weight values [] for the measurement k, based on the previous weight [] and the model LM measurement (203). That is: 25

    [] [] [] | ̂ [] (ec. 3)

This second stage (302), of updating weights for each elementary hypothesis (3) or particle (2), "i", represented in the previous generic equation, is particularized for this invention, where the variable z [k] ( the measure), is a Boolean variable that refers to each light detection, LD, that is, it takes an output value of the type

all / nothing, 0 or 1, generated by the light bulb detector (202), to indicate when a light spot (1) has been detected.

The multiplicative factor "" is just a normalization constant to make the sum of all the probabilities or weights of the particles (2) equal to 1. 5

The variable ̂ [] in equation 3 represents an estimate of the position of the particles (2), "i", at the instant of time, [], in which the measurement is produced, which in general is different from the At which point each step occurs, [], and can be calculated as an interpolation between the positions of the particles (2) in step 10 and j-1:

 ̂ [] [] [] [] (ec. 4)

being [] [] [] [], the interpolation coefficient. fifteen

To finish calculating the weights of the particles (2), as indicated in equation 3, the measurement model (203) used in the second block (200), LM, is as follows:

 twenty

 ([] | ̂ []) ∑ √ | | {̂ [] ̂ []} (ec. 5)

This equation indicates that the probability of detecting a point of light (1) when the user (7) is in the position indicated by the particle (2), "i", is equal to the sum of a series of Gaussian distributions, so many such as the number, L, of points of light (1) 25 that are on the map of points of light (204). Each of these Gaussian distributions represents the illumination area of each point of light (1) on the horizontal plane of the enclosure (4), and therefore these Gaussians are two-dimensional. Each Gaussian "l" (for l from 1 to L) is centered on the position of the point of light (1), "l", and the shape of said Gaussian depends on the size and orientation of the point of light 30 (1). The shape of each Gaussian "l" is indicated in ec. 5 using the 2x2 (two-dimensional) matrix, “, which can be worth [1 1; 1 1] in the case of a round spotlight without preferential orientation, up to other more generic values [a b; c d] that can be

calculate following algebraic techniques known in the state of the art based on the eigenvalues and eigenvectors associated with the main axes defined by the dimensions and orientation of each point of light (1).

With respect to the light bulb detector block (202), which gives as an output measure z 5 [k] = [], this is typically based on the scalar measurement of a specific light sensor (201), which is preferred over sensors matrix images or images of conventional cameras, since these require more processing. The detection principle of the light bulb detector (202) can be the following: 1) Low pass filtering, 2) Calculation of the derivative of the luminosity, 3) Peak detection based on 10 finding a zero crossing surrounded, before and later, by minimum values of the derivative of luminosity.

Finally, the most probable location (3000) of the user (7) after step "j" and the measure "k" can be calculated using this expression:

        ∑ [] [] (ec. 6)

where [] {[] [] [] []}, is the state of each particle (2), "i", after the step "j" given by the user (7), calculated in equation 1, and [ ] are the weights 20 of each particle (2), "i", after the measurement "k" taken by the light sensor (201), calculated in equation 3.

The present invention is defined by the following claims.

 25

Claims (15)

1. Method of location in indoor spaces based on detection and pairing of light points (1) comprising:  obtaining a map of points of light (204) of an enclosure (4), said map of points of light (204) comprising information relating to the location of all 5 points of light (1) of the enclosure (4) capable of be detected by means of a light sensor (201);  detect the movement of a user (7), in its path (5) through the enclosure (4), by means of a motion sensor (101) provided in a mobile device (6); 10  detect the passage under points of light (1) arranged in the enclosure (4), by means of a light sensor (201) provided in a mobile device (6); characterized in that it also includes: a) estimate the user's initial location (7) based on the light points (1) detected by the light sensor (201) of the mobile device (6), 15 establishing hypotheses (3) of the user's location (7) with with respect to the map of points of light (204), where the non-detection of a point of light (1) implies that any hypothesis (3) about the user's initial location (7) with respect to the map of points of light (1 ) is admissible; b) update the assumptions (3) of the user's location (7) with respect to the map of light points (204), according to the information related to the user's movement (7), provided by the motion sensor (101) of the mobile device (6), where the hypothesis (3) of the user's location (7) are displaced with respect to the map of points of light (204) based on said movement-related information; 25 c) update the hypothesis (3) of user location (7) with respect to the map of points of light (204) each time a point of light (1) is detected by the light sensor (201) of the mobile device (6) ), where the location hypotheses (3) are concentrated around points of light (1) that they appear on the map of points of light (204) each time a point of light (1) is detected; d) assign a probability to the location hypotheses (3) each time a light point (1) is detected by the light sensor (201) of the mobile device (6); 5 e) determine the specific location of the user (7) as that designated by the location hypothesis (3) that has the highest probability assigned; where:  to determine the location of the user (7) with respect to the 10-point light map (204), a particle filter type technique, PF, is used, where a hypothesis (3) of location consists of at least one particle (2) positioned on the map of points of light (204), and where each particle (2) is assigned a state that defines its position and orientation on the map of points of light (204), and; fifteen  the assignment of a probability to each hypothesis (3) of the user's location (7) is done by assigning a weight to each particle (2) that constitutes said hypothesis (3) of location, depending on the separation distance between each one of the particles (2) that 20 constitute each hypothesis (3) of location and the points of light (1) that appear on the map of points of light (204). 2. Method of locating in indoor spaces based on detection and matching of light points (1) according to claim 1, characterized in that the map of light points (204) comprises information regarding the size and orientation of all the points of light (1) of the enclosure (4). 3. Method of location in interior spaces based on detection and pairing of light points (1) according to claim 1, characterized in that a position and orientation of the particle (2) is defined according to the following mathematical representation: which incorporates the three-dimensional position of the particle (2) in space, and the orientation "" of the particle (2) on the horizontal plane. 4. Method of localization in indoor spaces based on detection and matching of light points (1) according to claim 1, characterized in that the user (7) is located when the particles (2) that constitute each 5 hypotheses (3) of location is concentrated within a single and small region of the enclosure (4) contained in the map of points of light (204), said particles (2) comprising similar positions and orientations, and where the size of said small region of the enclosure ( 4) represents the accuracy of the location method. 5. Indoor location method based on detection and pairing of light points (1) according to claim 3, characterized in that for each step taken by the user (7), an estimation of the position and orientation change is made which is represented mathematically as follows:   [ ] [ ] [ ] [ ] [ ] , fifteen where "j" is an ordinal index referred to the number of steps taken by the user (7) and accounted for by the motion sensor (101) of the mobile device (6), and the increments "are referred to a time reference system of the user (7) of the mobile device (6) in the previous step, where the instants of time in which step "j-1" and its successive occur, are indicated by this notation: []. 6. Method of localization in indoor spaces based on detection and pairing of points of light (1) according to claim 5, characterized in that a previously established motion model (103) is used, determined by a covariance matrix representing the uncertainty in the 25 estimate of the change of position and orientation of the user (7) in each step of its route (5) through the enclosure (4). 7. Indoor location method based on detection and pairing of light points (1) according to claim 6, characterized by that a particular state is attributed to each particle (2) that is part of a location hypothesis (3), according to the following mathematical formula: [] [] [] [] where "it symbolizes the state of a particle (2)," i ", said state being given by its position and orientation estimated by the motion sensor (101) in the previous step, and by a non-linear function" f "of the change Estimated 5 orientation and position in the last step taken by the user (7), and of the given number of steps, where the function f is represented mathematically as follows:   ([] []) [([]) ([]) ([]) ([])] ([[] [] [] []] √ []) where 10 - "[]" is the orientation of each particle (2), "i", with respect to a local reference, and corresponding to step (j); - "" is a random vector of dimension 4x1, which has a Gaussian distribution of zero mean, and which serves to disperse the particles (2) according to the movement model (103) used in the location method, and; - “P [j]” is the covariance matrix that determines the movement model (103). 8. Method of localization in indoor spaces based on detection and matching of light points (1) according to claims 1 and 2, 20 characterized in that the assignment of a probability to each hypothesis (3) of user location (7), It is done by assigning a weight to each particle (2) that constitutes said location hypothesis (3), taking into account the size and orientation of the points of light (1) that appear on the map of points of light (204).  25 9. Method of location in indoor spaces based on detection and matching of light points (1) according to claim 1, characterized by that the following mathematical formula is used for the attribution of a weight to each particle (2) constituting each hypothesis (3) of location: [] [] [] | ̂ [] where "[] is the weight assigned to the particle (2)," i ", for the instant detection" k "of a point of light (1); where "k" refers to an instant in which a luminosity measurement is produced by the light sensor 5 (201), said instant "k" referred to a particular temporal reference system of the light sensor (201), "[ ] ”, And where the weight“ [] ”is attributed based on the following parameters: - the weight "[]", assigned to the particle (2), "i", in the previous instant, "k-1"; 10 - an estimated state, "̂ []", of the particle (2), "i", at the instant "k" referred to the particular temporal reference system of the light sensor (201), "[]"; - the probability, "p", that, given an instant "k" in the particular temporal reference system of the light sensor, "[]", and an estimated state "̂ []" for the particle (2), "I", there is a detection of a point of light (1) by the light sensor (201), "[]", where the variable z [k] takes only two values, 0 or 1, depending on whether , respectively, at the moment "k" the detection of a light point (1) by the light sensor (201) occurs or not; twenty - a coefficient “”, whose function is to normalize the weights attributed to each particle (2) so that the sum of all of them is equal to 1. 10. Method of localization in indoor spaces based on detection and matching of light points (1) according to claims 7 and 9, characterized in that the estimated state, "̂ []", of the particle (2), "i" , at the instant "k" referred to the particular temporal reference system of the light sensor (201), "[]" is calculated, interpolating between two instants "j-1" and "j" of a local temporal reference system of the motion sensor (101), [], said instants being "j-1" and "j" respectively before and after instantly "K" in which a luminosity measurement has been made by the light sensor (201), according to the following mathematical formula: ̂ [] [] [] [] being [] [] [] [], an interpolation coefficient. 11. Method of location in indoor spaces based on detection and matching of light points (1) according to claims 9 and 10, 5 characterized in that the probability, "p", that, given a moment "k" in the system of particular temporal reference of the light sensor (201), "[]", and an estimated state "̂ []" for the particle (2), "i", there is a detection of a point of light (1) per part of the light sensor (201), “[]”, is calculated using the following mathematical expression: 10 ([] | ̂ []) ∑ √ | | {̂ [] ̂ []} where: - "̂ []", is an estimated position of the particle (2), "i", at the instant "k" referred to the particular temporal reference system of the light sensor (201), "[]", and; - L represents the total number of points of light (1) on the map of points of light 15 (204); this probability being therefore the sum of multiple Gaussians located with center at each point of light (1) and with a shape defined by the size and orientation of each point of light (1), represented by “. 12. Indoor location method based on detection and pairing of light points (1) according to claim 9, characterized in that the variable "[]" is evaluated by the following operations: - low pass filtering; - calculation of the derivative of the luminosity; - peak detection based on finding a zero crossing surrounded, before and after, by minimum values of the luminosity derivative. 13. Indoor location method based on detection and pairing of light points (1) according to claim 11, characterized in that the most probable location of a user (7) after step "j" and measurement 5 "k ”Is calculated using the following mathematical expression: ∑ [] [] where - N is the total number of particles (2), "i", which constitute the hypothesis (3) of location; - [] {[] [] [] []}, is the state of each particle (2) 10 after the step “j” given by the user (7), and; - [] are the weights of each particle (2) after the “k” measurement taken by the light sensor (201). 14. Indoor location method based on detection and pairing of light points (1) according to claim 1, characterized in that the light point map (204) is obtained by downloading from an online server in a mobile device (6). 15. Indoor location method based on detection and pairing of light points (1) according to claim 1, characterized in that the map of light points (204) is obtained by downloading via bluetooth or other communication method locally in the mobile device (6), in each enclosure (4).