WO2012139929A1

WO2012139929A1 - Method for the computer-assisted learning of a reference map on the basis of measurements of features of a wireless network

Info

Publication number: WO2012139929A1
Application number: PCT/EP2012/056041
Authority: WO
Inventors: Joachim Bamberger; Marian Grigoras; Andrei Szabo; Tobias WEIHERER
Original assignee: Siemens Aktiengesellschaft
Priority date: 2011-04-15
Filing date: 2012-04-03
Publication date: 2012-10-18
Also published as: DE102011007486A1; DE102011007486B4

Abstract

The invention relates to a method for the computer-assisted learning of a reference map based on measurements of features of a wireless network, wherein the measurements were carried out such that respective feature vectors (C_meas) of the wireless network are measured for a mobile object (O), which communicates via the wireless network with a plurality of base stations (BS1, BS2, BS3) of the wireless network, in a plurality of unknown object positions (OP) of the mobile object (O), and thus a series of measurements is obtained from a plurality of chronologically consecutive feature vectors (C_meas) for the respective object positions (OP) at the respective measurement times. The reference map contains a plurality of spatial reference positions, wherein the respective feature vectors (C_RM) of the wireless network at the respective reference positions (RP) of the reference map are learned on the basis of the series of measurements. Within the scope of the learning process, estimated object positions are determined on the basis of pattern matching and are subsequently used to determine improved estimated object positions on the basis of an optimization of a cost function. A predetermined movement model, which defines limitations for the object movement using the chronological order of the measurement times of the series of measurements, is taken into consideration in this optimization. Updated feature vectors of the wireless network are then determined at the reference positions of the reference map by means of the object positions determined from the optimization. The method according to the invention has the advantage that an exact localization of mobile objects is made possible with the learned reference map without prior calibration.

Description

description

A method for computer-aided learning of a reference map based on measurements of features of a radio network

The invention relates to a method for computer-aided learning of a reference map based on measurements of features of a radio network and a corresponding device. Furthermore, the invention relates to a method for the computer-aided localization of a mobile object and to a corresponding device. Moreover, the invention relates to a computer program product.

The invention relates to the technical field of localization of mobile objects using features of a radio network. From the prior art, various An ^¬ sets are known as mobile objects can be located via the detection of characteristics of a radio network. In this case, the radio network is formed by a multiplicity of base stations which can transmit and / or receive radio signals. The base stations communicate with a transmitting and / or receiving unit in the mobile object to be located. From the characteristics of the exchanged radio signals, the position of the mobile object can be determined.

Known methods use for localization the absolute transit time or running time differences of radio signals between the respective base stations and the mobile object. Similarly, methods exist, given its. Round-Trip-Time USAGE ^¬, ie the propagation time of a radio signal from a mobile object toward the base station and back to the mobile object or from a base station to a mobile object and back to the base station. In another class of methods, the arrival angle of radio signals at the receiver and the transmission angle of radio signals are used in the transmission, thereby locating objects. Furthermore, there are methods in which a localization of mobile objects takes place via the signal strength of radio signals. Known localization methods are further distinguished as to whether the localization is based on a so-called finger printing or by the evaluation of geometric properties between the mobile object and the base stations. The evaluation of geometric properties is based on tri- or multilateration. In localization based on finger printing, calibration data in the form of features of the radio network from various known reference positions are generally combined in one

Training phase collected to then based on a pattern matching (English, pattern matching) by means of a comparison of corresponding measured features of the radio network with the features at the reference positions to locate an object. The individual reference positions form a so-called reference map, which is also referred to as a radio map. At ^¬ instead of determining this reference card in advance as part of a calibration, there are in the prior art further approaches in which the reference card is initially initialized and is learned online in the context of localization by newly added measurement data. Such a method is described in the publication WO 2007/118518 AI. In contrast to methods in which the localization takes place via geometric properties, finger printing methods have a higher accuracy. However, they require time-consuming manual calibration phase in which a mobile object on ei ^¬ ner plurality of predetermined spatial locations to be located at these positions and the characteristics of the radio network to be measured.

The object of the invention is to provide a method for computer-aided learning of a reference card based on measurements of features of a radio network, which is easy to implement without complex calibration and provides good localization results. This object is achieved by the independent claims. Further developments of the invention are defined in the dependent claims.

The inventive method is used for computer-aided learning a reference card based on measurements of features of a radio network, wherein the radio network in a particularly preferred embodiment form a local radio network and in particular a wireless network. The measurements used for learning are performed such that for a mobile object, which communicates via the radio network with a plurality of base stations of the radio network, ektpositionen at a plurality of non ^¬ knew whether the mobile object respective feature vectors of the radio network are measured, and thereby a Measurement series is obtained from a plurality of temporally successive feature vectors for respective object positions at respective measurement times. The learning process can be carried out online after completion of the measurements or in parallel during the measurement. The to-learn Refe rence ^¬ card includes a plurality of spatial reference ^¬ Posi tions, the respective feature vectors of the radio network to be learned at the respective reference positions of the reference map based on the series of measurements. That is to say that those feature vectors are suitably estimated, which result when the mobile object is arranged at a corresponding reference position. The reference card is thus distinguished by corresponding feature vectors at the individual reference positions of the card.

Within the scope of the inventive method, the respective object positions at which the feature vectors of the radio network were measured are estimated in a step a) based on a pattern matching which compares the respective feature vectors of the measurement series with the feature vectors at the reference positions of the reference map. It is assumed that an appropriate initialization of the reference card is provided at the beginning of the process, which is initially used in the pattern matching. encryption Driving for pattern matching are well known in the art and are commonly referred to as pattern matching. They are based on the fact that similarities or similarities between predetermined patterns, which according to the invention are present as feature vectors of a reference ^map , and measured feature vectors are identified and the object position is determined therefrom.

In a step b) of the inventive method are determined based on optimization of a cost function optimized estimated Whether ektpositionen, wherein one or more boundary conditions are taken into account in the optimization, which are defined by a predetermined motion model for the mobile object, wherein the predetermined Be ^¬ wegungsmodell using the temporal order of the points of measurement of the measurement series sets of one or more Beschränkun ^¬ gene for the object motion. Finally, in a step c), updated feature vectors of the radio network are determined at the reference positions of the reference card by means of the optimized estimated object positions.

The inventive method is characterized in that in the context of learning the reference map and the zeitli ^¬ che order of the measurement time points is taken into account, and flows in a ge ^¬ suitable motion model. The motion ^¬ model may take into account useful in the measurement series both restrictions with regard to the last movement of the mobile object and with regard to the future movement of the mobile object. Which improved from the optimization resulting estimated object positions then provide a good basis for Approximati ^¬ on the corresponding feature vectors at the reference positions of the reference card.

In the method according to the invention, an initial calibration of the reference card can be dispensed with. Rather, a suitable initialization of the reference card is sufficient. and by repeating the learning steps several times, an increasingly accurate reference map can be learned.

The method according to the invention can be used for time-based measurements, in which the feature vectors of the radio network include entries as runtime-based variables, each of which depends on a runtime of one or more radio signals of the radio network. The duration of the radio signal or signals can be the one-way transit time of a radio signal between a respective base station and the mobile object. Also, the term may be a two-way time (also referred to as round-trip time hereinafter) be a Funksig ^¬ Nals of the mobile object to a respective base station and from there back to the respective base station or the two-way travel time of a radio signal from a respective Ba ^¬ sisstation to the mobile object and from there back to the respective base station. In a particularly preferred embodiment of the invention, however, the runtime-based variables depend on a transit time in the form of a transit time difference. The transit time difference is defined for respective pairs of base stations and represents the difference between a first and a second runtime, wherein the first runtime is the transit time of a radio signal between the mobile object and the one base station of the pair and the second runtime the transit time of a radio signal between represents the mobile object and the other base station of the pair.

In a particularly preferred embodiment, the runtime-based variables in the entries of the feature vectors of the radio network are so-called channel impulse responses. These can be detected for the above-described transit times or transit time differences and represent the amplitude of the radio signal detected in dependence on the transit time or transit time difference in response to a transmit pulse. In a particularly preferred embodiment, the channel impulse response is in a respective entry of Feature vector the Kanalimpulsant ^¬ word for the duration difference for a respective pair of base stations. That is, over the corresponding amplitudes the radio signals detected at the base stations for different propagation time differences, the channel impulse response is detected. In a particularly preferred embodiment in which the run-time-based variables are the channel impulse responses for respective pairs of base stations, the reference map is initialized at the beginning of the process such that for each ^¬ stays awhile reference position, the channel impulse response for a JE weiliges pair of base stations by a normal distribution whose mean value is a transit time difference based on estimated distances between the respective reference position and the base stations of the respective pair. The normal distribution is preferably standardized in such a way that the integral over the normal distribution yields 1, so that the impulse responses can be understood as probabilities for the occurrence of certain transit time differences.

In a further embodiment of the method according to the invention, in the pattern matching in step a), a number of feature vectors are determined at reference positions with the greatest agreement with the measured feature vector at a respective object position, based on the reference position (s) for the number of feature vectors, in particular by averaging the reference positions, the respective object position is estimated. This method is based on the known search for the k nearest neighbors, where k corresponds to the number of feature vectors.

In a further, particularly preferred disclosed embodiment of the method according to the invention, the feature vectors of the reference map are refreshes ^¬ alisiert within the pattern matching in parallel by the feature vectors of the reference map for reference positions in a predetermined environment can be corrected by an estimated using the pattern matching object position in each case with a correction term , wherein the correction term of the difference between the feature vector at the estimated via the pattern matching estimated ektposition and the respective feature vector of the reference map to be corrected. Preferably, in the document WO

2007/118518 AI described method used.

In a particularly preferred variant of the embodiment just described, the size of the correction term decreases with increasing distance of the respective reference position in the predetermined environment from the object position estimated via the pattern matching. In this case, the correction term preferably depends on a Gaussian function with center at the position estimated via the pattern matching.

In a further, particularly preferred embodiment, guide the process is dimensionally performed iteratively for learning of the reference map, in which after a number of updates of the feature vectors of the reference map in step a) the steps b) and c) are carried out and then step a) ^¬ is returned, if a predetermined abort criterion, such as a predetermined number of iterations, is not yet met.

The predetermined movement model used in step b) is suitably adapted to the mobile object under consideration. In a preferred embodiment, the motion model sets a maximum velocity for the movement of the mobile object as a constraint. This maximum speed can then be suitably taken into account within the measurement series for successive measurement times in such a way that two estimated object positions at consecutive measurement times must not be further apart than is possible according to the maximum velocity of the mobile object. In a further embodiment of the method according to the invention, the cost function used in step b) is a sum of Gaussian functions, each of which is based on the difference in magnitude between the respective data ^item and the sample item. Equally estimated is the ect position and the optimized estimated ect position to be determined.

In a further embodiment of the method according to the invention, in step c) the updated feature vectors of the radio network at the reference positions of the reference card are determined such that the optimized estimated object position having the smallest distance to a respective reference position of the reference card is determined, and the feature vector at the respective reference position is replaced by the measured feature vector at the optimized estimated object position with the smallest distance to the respective reference position of the reference map. As a result, a particularly simple update of the feature vectors in the reference map is achieved.

In a further disclosed embodiment of the inventive method in step c) the updated feature vectors of the radio network at the reference positions of the reference card is determined such that, for a respective reference Posi ^¬ tion a weighted sum of measured feature vectors is determined on optimized estimated object positions, wherein the weighted sum represents the updated feature vector at the respective reference position and wherein the respective weights of the summands in the sum are the smaller the farther an estimated optimized object position is from the respective reference position. Preferably, the weights of the weighted sum are modeled by Gaussian functions, wherein a respective Gaussian function depends on the relative distance between a respective reference position and the respective optimized estimated object position.

In order to stabilize the learning according to the invention and thus to avoid rotation or displacement of the reference card at certain reference positions, in a preferred embodiment, in the context of performing step a) and / or step b) fixed points with fixed spatial positions are used. NEN and feature vectors taken into account, wherein, in the measurement ^¬ row one or more Whether ektpositionen are pre-assigned fixed points and / or be associated with the measured feature vectors of the series of measurements fixed points, eg by a feature vector whose accordance with a feature vector of a respective fixed point over-writing a predetermined threshold ^¬ tet, is equated with the respective fixed point. The spatial positions of fixed points in step a) are treated as estimated object positions or treated as optimized estimated object positions in step b). That is, the spatial positions of the fixed points are considered given and not determined by means of an estimate in step a). Likewise, the fixed points in the optimization in step b) are not considered as free parameters.

In a particularly preferred disclosed embodiment of the invention the measured feature vectors of the series of measurements being an example of a sol ^¬ chen heuristic in the detailed description is set forth are assigned to fixed points on a heuristic. In a further embodiment, the spatial positions of the fixed points correspond to the spatial positions of the base stations.

In addition to the above-described method for computer-aided learning of a reference map, the invention further comprises a method for the computer-aided localization of a mobile object based on such a learned reference map. Here, a feature vector of the radio network at the site to Loka ^¬ lisierenden object position is measured. Based on a pattern matching which compares the measured feature vector with feature vectors at the reference positions of the learned reference card, the object position is subsequently determined. The invention further comprises a device for computer-aided learning of a reference map based on measurements of features of a radio network, wherein the measurements are made such that for a mobile object, which via the Radio network communicates with a plurality of base stations of the radio network, at a plurality of unknown ob ektpositionen of the mobile object respective feature vectors of the radio network are measured, and thereby a series of measurements from a plurality of temporally successive feature vectors for respective object positions is obtained at respective measurement times. The reference card in this case comprises a plurality of spatial reference positions and the apparatus includes a processing unit, with the operation of the device based on the measurement series, the respective feature vectors of the radio network at the respective reference positions, the reference card are learned with the inventive method described above.

The invention further relates to a device for the computer-aided localization of a mobile object, based on a reference map, which is learned with the learning method described above. In this case, the device comprises a measuring and processing device with which a feature vector of the radio network is measured at the object position to be located and with which based on a pattern matching, which compares the measured feature vector with the feature vectors at the reference positions of the learned reference map, the object position is determined ,

The invention further comprises a Computerprogrammpro ^¬ domestic product with a program stored on a machine-readable carrier, the program code for performing the method for learning a reference map or for the localization of a mobile object described above, according to the invention, when the program runs on a computer.

Embodiments of the invention are described below in detail with reference to the accompanying drawings.

Show it: Fig. 1 is a schematic representation of a physical environment, is explained with the aid of learning ^¬ Re ference card for reference positions in the spatial environment of the present invention;

FIG. 2 is a diagram illustrating by way of example the course of the channel impulse response used in an embodiment of the invention as an entry of feature vectors; FIG.

Fig. 3 is a schematic representation showing the iterative

Sequence of an embodiment of the method according to the invention; and

Comparing Fig. 4 is a diagram showing results of the invention ^shown SEN process with other processes.

Embodiments of the method according to the invention based on feature vectors of a radio network will now be described, which contain as entries so-called channel impulse responses for a transit time difference of radio signals between base stations of corresponding pairs of base stations of the radio network. In a preferred variant, the radio network is a WLAN network in which the base stations represent corresponding access points. In this case, a mobile Whether ^¬ ject to the corresponding standard of the radio network emits Funksig ^¬ dimensional whose channel impulse response as a function of the transit time difference for respective pairs of base stations is measured. Based on this, the expected feature vectors are then determined for a multiplicity of spatial reference positions of a reference map, wherein the reference ^{map can} subsequently be used, for example, to localize a mobile object by so-called pattern matching. As part of this pattern matching, the corresponding Merkmalsvek ^¬ tor for a mobile object at an unknown location is measured, followed by comparison of the measured feature vector with the respective feature vectors of the reference Positions is determined in a suitable manner the ektposition.

Fig. 1 illustrates generally the operation of the invention ^shown SEN method for learning a reference map. In this case, at ^¬ way of example, a spatial surroundings R is shown in the shape of a right ^¬ square space in which a plurality of regularly spaced and known reference positions RP are included a reference map. The corresponding feature vectors at these reference positions are to be determined in the course of learning the reference map. The reference positions are indicated in FIG. 1 by corresponding crosses, with only two of the reference positions being denoted by reference symbol RP for reasons of ^clarity . The feature vector present at these positions is denoted by C _RM . By way of example, three base stations BS1, BS2 and BS3 of the radio network are indicated in FIG. 1, which communicate with a corresponding mobile object 0 at an object position OP, the feature vector C _{meas being} measured for the object position OP.

The method according to the invention is based on measurements within a predetermined measurement period in which the object 0 moves in the spatial environment R and the feature vectors C _meas at the corresponding object positions OP are detected at a plurality of measurement times. These object positions are not known, ie the measurement is not a calibration process in the actual sense in which corresponding feature vectors are detected at known positions. A measured feature vector includes an entry for each possible pair of two Basissta ^¬ functions of the radio network, that is, in the scenario of FIG. 1, an entry for the pair of base stations BSl and BS2, for the pair of base stations BSl and BS3 and for the pair of Base stations BS2 and BS3. By way of example, FIG. 1 illustrates a corresponding measurement of a feature vector for the pair of base stations BS1 and BS2. The measurement is based on the transit time difference LD of radio signals between a first transit time LI from the object 0 to the base station. on BS1 and a second transit time L2 from the object 0 to the base station BS2. Here, the sum of the respective amplitudes of the radio signals, which are detected by the respective Basissta ^¬ functions BS1 and BS2 is detected as a function of a distance and the corresponding propagation time difference between the base stations. The resultant curve in this case represents the so-called. Channel impulse response, which is illustrated in ^¬ way of example in Fig. 2.

In the diagram of Fig. 2, the distance d is indicated along the abscissa, the different (multiplied by the Signalgeschwin speed) propagation time differences of the radio signals for a pair of base stations reproduces. Along the Ordi nate the amplitude AP is plotted, which is composed of the amplitudes of the radio signals, which are to be received at each Ba sisstation of the pair for the respective distance. The channel impulse response is represented as a solid curve K1. The maximum value of the channel impulse response ent ^¬ usually speaks of the time difference for a clear line of sight connection between the base stations and the mobile object. The other maxima of the curve are usually caused by Mehrwe ^¬ geausbreitungen due to reflections. In Fig. 2, a further dashed curve K2 is also given ^¬ again. This curve is used as an initialization for the channel impulse responses at a reference position of the reference map, and their calculation will be described later.

In the following, the individual steps of an embodiment of the inventive method for learning a reference map based on channel impulse responses for time differences for pairs of base stations will be explained in detail. As already mentioned, the learning of the reference map is based on a measurement series from a plurality of feature vectors which have been measured for different measurement positions. For a respective measurement position or a corresponding measurement time i, the following feature vector results: ^ meas l>meas; fiSl, 2 ^■ > · · · ^■ > ^ sas BSl, N '* - meas; BS2,3 ^■ > · · · ^■ > msas BS2, N ^■ > · · · ^■ > msas BSQi - X, NJ

The individual entries of the feature vector C ^{(^} _Eils are the channel impulse responses in the time domain, which are approximated by a 5 inverse Fourier transform of a channel estimation in the frequency domain. Each entry of the above vector containing the channel impulse response in the form of L Amp ^¬ litudenwerten to corresponding propagation time differences 1-T, 2-T, LT, which in this sense represent support points of the continuous channel impulse response for corresponding propagation time differences and distances respectively Mathematically, a corresponding entry is made for a pair of base stations BSm and BSn follows:

1 ^ 5 c * - ⁽ m ^,) eas; = \ Lα "m ^ω eas; BS>", n; lr ' ^{- - -} ' α "m ^ω eas; BS", n; ir VJ

The discrete time interval T of the channel impulse response depends as ^¬ in from the measurement setup. In order to reduce the influence of irregular measured values (so-called outliers), the amplitude values of several measurements can be averaged around the actual position before the learning described below.

In order to start the learning of the reference card, it is necessary for a suitable initialization of the reference card to be present, ie a meaningful set of K reference positions with corresponding initialized feature vectors is given. At the beginning of the method, there exists a position vector P _RM = [pj ^, ..., pj ^] ^r with corresponding k reference positions of the reference card. In the present case 30, two-dimensional spatial positions are considered wel ^¬ che describes the location of the object with x and y coordinate within ei ^¬ nes space without regard to its height. Gege ^¬ appropriate, is also possible to extend the method to three Dimensio ^¬ nen.

The k-th reference position in the reference map is thus described by the following position vector: 'RM [RM' RM -

As indicated in FIG. 1, the individual positions are preferably the vertices of an equidistant grid. To loading 5 are beginning for each of the K reference positions initialized feature vectors Cj ^ _j of the reference map RM (RM of the method

= Radio Map), so that the reference map is described by the following matrix of initialized feature vectors:

In the embodiment described here, a special initialization of the individual feature vectors based on a Gaussian estimation is used. Alternatively it is

It is possible for a certain number of calibration measurements to be used for initialization, wherein feature vectors for spatial positions between calibration measurements are determined by linear interpolation. The Gaussian estimate used in the embodiment described herein

20 rests on the geometric arrangement of the respective reference positions with respect to the individual base stations. The coordinates of all base stations are known. These are described by the following position vector P _BS with corresponding two-dimensional positions ρ 'B _Β S ₈ :

25

P ^A BS = LFBS '-' - 'I'B ^(A S ° J

N denotes the number of base stations. The 30 actual distance difference d for a pair of Basissta ^¬ functions n, m and a corresponding reference position is calculated k based on the following formula: d = ^ ^(X S ^~ RM) + 0 s' ^~~ Yiu ^~ ^ ^(X S ^~ RM) + (BS ^~~ ^ RM)

35

Based on the assumption that the base stations and the respective reference position are in line of sight, and assuming Gaussian measurement errors, and a propagation v (usually the Lichtge ^¬ speed) can be approximated the amplitudes of the respective channel impulse responses for propagation time differences VLT by the formula: a ^ -BS _m, NIT = aN (YLT; d, a)

This amplitude is reproduced by way of example for a channel impulse response as curve K2 in FIG. N designates the normal distribution with a suitable standard deviation σ and a normalization factor a. The normalization factor is chosen such that the integral over the Normalvertei ^¬ system is 1, and thus the individual amplitude values can be interpreted as true ^¬ probabilities for a certain distance difference between two base stations.

As part of learning the reference card according to the invention, corresponding estimated object positions are now determined for all measured channel impulse responses with a pattern matching known per se (pattern matching). The quadratic difference between a measured feature vector C ^ _eas and all entries Cj ^ of the reference map is determined as follows: y Nl NL

c) _C (k) _{) =} yy ₍ ((*) γ

\ ^u mensBSm, nl-T RM; BSm, n; lT)

m-1 n-m + 1-l

As part of a known "k nearest neighbors" algorithm then a predetermined number of entries of the reference card with the shortest distances to the current measured feature vector is selected. The final estimated position P ge ^¬ ⁼ [^ PM PM] ^for the measurement i is determined by averaging the respective reference positions in accordance with the predetermined number of entries of the reference map.

Based on the initialized reference map and using the estimated ob ect positions, borrowed the learning of the reference card by iterative Aktualisie ^¬ tion of its entries, where a series of measurements of Merkmalsvekto ^¬ ren at I time points or measuring positions is taken into account, which is described by the following matrix: meas L * - ^ meas'"""'meas'"""'meas J

Each individual measurement is assigned a corresponding time stamp with the measurement time, which can be described by the following vector: t meas = r Lt m ⁽¹ e ⁾ as '''''t me ⁾ as''''' tm ^(I e ⁾ as l J

In the embodiment described here disclosed embodiment of the invention, an iterative learning is used which so-called. Selbstorgani ^¬ sierende cards (engl. Seif Organizing Maps) is used. The ^¬ ses learning is referred to hereinafter as SOM learning and provides an adaptation is of the per se known Kohonen Seif Organizing Maps. In addition, an optimization is used in Inventive ^¬ contemporary learning, which is a motion model of the mobile object in a suitable manner for taken into account the complete measurement series. This optimization is referred to below as "Complete Track Optimization" and abbreviated CTO. The CTO-process represents a Wesent ^¬ Liche component of the inventive method, which is guide combined form as described herein from the additional SOM learning. Particularly good results are obtained by combining these two learning methods.

Fig. 3 shows an example of a flow chart, which represents the ith ^¬ rative learning based on a combination of the SOM learning and CTO-learning. It designates the rationsschritt Ite ^¬. At the beginning of the process (it = 0) a complete update of the corresponding feature vectors of the reference card is first Runaway ^¬ performs, based on the SOM learning. After each iteration step (it = it + l), a check step A checks whether a predetermined number of iteration steps it_cto is reached. If this is not the case, the reference card is updated again based on the SOM learning (branch B2). If the check A, where the predetermined number is reached to 5 iterations, a CTO-learning step is Runaway ^¬ leads (branch Bl). Subsequently, the SOM learning is again performed for a predetermined number of iterations it_cto. The iteration is finally terminated upon reaching a predetermined total number n of iteration steps (branch 10 B3).

The following section first describes SOM learning. As already mentioned, the learning rule is adapted from so-called Kohonen Seif Organizational Maps. A variant of the SOM learning has already been described in document WO 2007/118518 A1 and will now be suitably adapted within the scope of the invention to term-based feature vectors.

As explained above, by pattern matching, for every 20 measured feature vector C ⁽ ^ _AS, an estimated octet position

P _PI J _J determined. This position now functions as a center for an update process of the corresponding feature vectors of the reference map with reference positions in the vicinity of the corresponding pattern-matching determined object position. In this case, the so-called neighborhood error is calculated as a difference between the currently measured characteristic vector C ^ _EAS and the individual feature vectors

Reference card defined as follows: η ΛΡ ^ '_ / "" ⁽ ' ⁾ _ f ^(k)

^{JU ~~} ^ meas ^ RM

Based on the distance r ^(k) = ^ (x ^ - - _M ) ^{2 +} ( _PM ^~ ^ _RM ) ² between the predetermined position, which was determined for the corresponding measurement by pattern matching, and the previously known reference positions , the influence of the above-calculated error AC ^(k) becomes appropriate depending on the distance between object position and reference position adapted. This is done by applying a weighting radio ^¬ tion f is used, which is described by a Gaussian κ surface in combination with an additional learning rate parameter, wherein said function f is as described herein from guide defined form as follows:

Based on this function, all entries of the reference map are updated via the so-called Kohonen update rule as follows:

C & = C & ₊ / (r «). AC <"It is apparent that thus the corresponding correction of the watch ^¬ times vector is the smaller of the reference card, the further away is the reference position of the estimated object position. In a particularly preferred embodiment, for the computationally efficient implementation of the method, the actual update is only considered within a range of 3σ around the estimated object position, because the Gaussian bell can be considered virtually zero if it exceeds three times the standard deviation of its Center is removed.

The updating described above can be iteratively repeated several times, wherein in each iteration, preferably in random order, all corresponding measuring times or measured feature vectors are considered. Each individual iteration always uses the newly calculated feature vectors of the reference card.

As already mentioned above, after a predetermined number of iteration steps, the CTO method is performed, which will be explained below. The sequenced ^¬ tielle chronological sequence of measurements which t from the above-described measurement time point vector is _meas shows in suitable for formulating an optimization problem using a predetermined motion model for the mobile object. The goal is to improve the individual ^¬ nen entries of the reference card in the context of the entire measurement series ^¬ and the associated tracks of the mobile object.

The starting point of the CTO optimization is again the position vector of the object positions estimated via the pattern matching, which reads as follows:

As part of the optimization of these positions can be approximated, thus resulting in improved estimated object positions preserver ^¬ th are being considered in the approximation that predetermined movement of the mobile object constraints are not violated. These movement restrictions are defined within a predetermined movement model. The motion model can be defined depending on the mobile object ge ^¬ suitable. In the embodiment described herein the motion model based on the assumption that the mobile object can move a maximum with a maximum Ge ^¬ speed v _MM of, for example 2 m / s. Un ^¬ ter this assumption is a Kos ^¬ tenfunktion g used to optimize and set a (1-1) dimensional limitation radio ^¬ tion h. These functions depend on possible positions of the mobile object at the individual measurement times. These positions are defined by the following vector:

P = [p> ⁽¹⁾ , .., Ρ ⁽⁰ , .., Ρ ^(/) ί

The cost function g for this position vector P to be optimized with the corresponding restriction functions hi for successive measurement times i is as follows: = ι 2πσ

h (P) = [^ (P), ... ^ (P),. , , , Ä _i _ ₁ (P)]

The aim of the optimization is to injure the effect minimizing the above cost function g (J *) without the constraints of the motion model that the above Beschränkungsfunkti on ^¬ z, (P) is zero becomes smaller. In this way into account ^¬ Untitled that a mobile object between two successive measurement time points can not move faster than the above maximum speed v _MM.

The optimization problem just described can be solved with known from the prior art algorithms, such as the Active Set algorithm. By manually supplying gradients of both the cost function and the constraint function, the computation time for the optimization can be significantly reduced. In a preferred embodiment, the above positions P are initialized by means of a per se known particle filter to thereby improve the results. The just described Opti ^¬ optimization can be regarded as a kind of tracking algorithm in which the complete history and future of data ^¬ set of the series of measurements is taken into account. After performing the above optimization, as a result, the position _vector P _{opt is} obtained, which contains as entries the corresponding optimized estimated positions for the individual measurement times. Using this vector in a preferred variant of the invention, for each re ference will be position in the reference map the closest opti mized ^¬ position from the vector P _opt is determined as follows:

(I)

k fu _i = arg min ^ (χ ^ - x®) ² + (x ^ - x {opt / Finally, the corresponding feature vector of the reference card is replaced by the obtained closest measurement as follows:

C ^ _jy is therefore the updated feature vector for Refe ^¬ ence position k of the reference card. In a variant, the determination of an updated feature vector of Refe rence ^¬ map is not performed by replacing the feature vector with the closest measured feature vector but using a weighted sum, which is represented by the following equation:

Where a denotes a normalization factor. The a posteriori probability in the above equation is directly available when using particle filters, but can also be modeled by Gaussian curves around the positions P _opt as follows:

Studies conducted by the inventors simulations have shown that updating the reference card with the feature vector ^¬ ektposition according to the nearest Whether slightly better results than the just described Aktualisie ^¬ tion.

In order to improve the stability of the learning process described above, fixed points are so-called guide form in a preferred. Taken into account, which represent spatial positions in the reference map Be ^¬ rich and predetermined their assignments describe between measured feature vectors and their actual positions. The fixed points can be manually set in advance in the context of measurements of the feature vectors thereby, that in this case, corresponding Mes ^¬ solutions predetermined spatial positions of fixed points are assigned. These spatial positions are then no longer changed in the subsequent optimizations. Likewise corresponding fixed points can also be automatically determined from the Messda ^¬ th. For this purpose, for example, a suitable heuristic ^¬ tik used for automatic detection of these fixed points. An example of such a heuristic will be described below.

Both in the manual definition and in the automatic determination of fixed points, these fixed points are described by a spatial position vector P _FIX with F entries for the respective spatial positions of the individual fixed points and the associated feature vectors in the form of the matrix C _FLX . It thus applies:

P = TD ⁽¹⁾ D ^(/) D ⁽⁾

In the context of automatic detection of fixed points, a corresponding measured feature vector is assigned to a fixed point during learning if the quadratic difference between the feature vector of the measurement and the predetermined measurement at the fixed point is below a predetermined threshold d _fix , ie if: r (cL _s , c ^) <^ _x o

Measurements which are assigned to fixed points are not estimated in the context of the method described above, but are regarded as predetermined. In particular, in the SOM learning described above, the pattern matching is performed. averaged positions replaced by the fixed point positions pj ', which in the context of SOM learning leads in most cases to a correct position of the center of the update ^¬ surfaces. This replacement is also done as part of CTO optimization. In this case, the corresponding fixed-point position is no longer a free parameter of the optimization, so that after the optimization, this fixed-point position is not changed and is equivalent to an estimated optimized position ektposition.

In a further modification of the process according to the invention fixed points are determined based on a heuristic ge ^¬ Mäss of such measured feature vectors are detected, which are close to a corresponding base station. In the context of this heuristic, is determined as follows for a base station n is an appropriate measurement position, that is closest to the base station n ^¬:

For each measurement (ie, for each measurement time i), it is determined how long the runtime difference is for each pair of base stations that the base station n contains. The transit time difference is then since ^¬ by derived from the channel impulse response that the maximum value of the channel impulse response is used as the transit time difference. The channel impulse responses are thus considered as probabilities for corresponding propagation time differences. For each measurement, the runtime differences are summed for each base station pair containing the base station n. In this case, a sign is taken into account in a suitable manner as to whether in the corresponding base station pair the considered base station n is the first and the second base station. As a result, the closer the point of the corresponding measurement is to the base station n currently under consideration, the greater the sum becomes. From these summed values, the measurement with the largest sum is selected. The ent ^¬ speaking measurement position is then set equal to the position of the base station n. Based on this heuristic, such measurements can be calculated in a computationally efficient way which are located at base stations. These measurements are considered as fixed points and equated with the spatial positio ^¬ nen of the base stations. The inventive method has been tested by the inventors basie ^¬ rend on simulations. 4 shows a diagram which compares the results of the method according to the invention with methods known per se. In this case, a reference map learned with the method according to the invention was used for the localization of a mobile object and compared with localization methods known per se. The abscissa of the diagram of Fig. 4 shows the localization error LE and along the ordinate the cumulative probability Pcum for the corresponding localization error. That is, the steeper the corresponding curves extend upwards, the better the corresponding localization.

The curve CU1 shows a localization carried out on the basis of the inventively learned reference card by means of pattern matching. In contrast, the curve CU2 shows a localization based on pattern matching with a manually calibrated reference map. The curve CU3 shows a prior art location without a reference map, in which, for each possible position of the mobile object, the corresponding channel impulse responses of the individual pairs of

Base stations are multiplied and the location with the maximum ^¬ value of the product as the position of the mobile object identi ^¬ ficated. In contrast, the curve CU4 shows a pattern matching, in which the reference card is not learned, but the reference card originally initialized via a Gaussian function is used.

As is apparent from the diagram of Fig. 4, of course, the best results are achieved by manual calibration of the reference map. However, this localization is associated with high calibration effort. However, in comparison to methods which do not use calibration, the method according to the invention is according to the illustrated curve CU1 the best. In particular, the method according to the invention leads to better localization results than the above-described method according to the prior art, which is represented by the curve CU3. As might be expected, without learning a reference map, the worst localization results are achieved, as indicated by the flat curve CU4.

The embodiments of the method according to the invention described above have a number of advantages. In particular, the reference map learned with the method improves runtime-based localization systems without the need for additional calibration phases or other measurement data. Moreover, the method does not involve the disadvantages of the manual calibration must be gemes ^¬ sen in which corresponding with a high effort at predetermined known spatial positions of feature vectors of the radio network. In a preferred form of the method according to the invention, predetermined fixed points can be incorporated within the scope of the learning, which can also be detected automatically, for example based on a heuristic.

Claims

claims

1. A method for computer-aided learning of a reference map based on measurements of features of a radio network, wherein the measurements are made such that for a mobile Ob ^¬ jekt (0), which via the radio network with a plurality of base stations (BS1, BS2, BS3) the radio network communicates, respective feature vectors (C _meas ) of the radio network are measured at a plurality of unknown ob ektpositionen (OP) of the mobile object (0) and thereby a measurement series of a plurality of temporally successive feature vectors (C _MEAS ) for respective object positions ( OP) is obtained to jewei ^¬ time measuring times, wherein the reference card has a plurality of spatial reference positions (RP), wherein based on the series of measurements the respective Merkmalsvek ^¬ factors (C _RM) of the radio network at the respective Referenzpositio ^¬ NEN (RP) of the reference map be learned, where:

a) based on a pattern matching which compares the respective feature vectors (C _MEAS) of the measurement series with the shopping ^¬ times vectors (C _RM) (at the reference positions RP) of Re ^¬ conference map, the respective object position (OP) in which the feature vectors ( C _MEAS ) of the radio network are estimated;

b) based on optimization of a cost function optimized estimated object position (OP) determined who ^¬, wherein one or more boundary conditions are taken into account in the optimization, which are provided by a vorbe ^{¬-determined} motion model for the mobile object (0) ^¬ type, wherein the predetermined motion model determines one or more object movement constraints using the temporal order of the measurement times of the measurement series;

c) using the optimized estimated object positions (OP) updated feature vectors (C _RM ) of the radio network at the reference positions (RP) of the reference card are determined.

2. The method of claim 1, _wherein the feature vectors (C _MEAS , C _RM ) of the radio network as entries runtime-based Grö ^¬ Shen comprise, each of which depends on a duration of one or more radio signals in the radio network.

3. The method of claim 2, wherein the duration of the radio signals or the one-way runtime of a radio signal between ^¬ rule a respective base station (BS1, BS2, BS3) and the mobile object (0) and / or the two-way Run time of a radio signal from the mobile object (0) to a respective one

Base station (BS1, BS2, BS3) and from there back to the respective base station (BS1, BS2, BS3) and / or the two-way transit time of a radio signal from a respective base station (BS1, BS2, BS3) to the mobile object (0 ) and from there back to the respective base station (BS1, BS2, BS3).

4. The method according to claim 2, wherein the propagation time of the radio signal or signals comprises the transit time difference for a respective pair of base stations, wherein the transit time difference is the difference between a first and a second delay and a second transit time (LI, L2), wherein the first transit time is the transit time of a radio signal between the mobile object (0) and the one base station (BS1, BS2, BS3) of the pair and the second transit time the transit time of a radio signal between the mobile Represents object (0) and the other base station (BS1, BS2, BS3) of the pair.

5. Method according to one of claims 2 to 4, in which the time-based variables in the entries of the feature vectors (C _{MEAS /} C _RM ) of the radio network are channel impulse responses, where ^¬ a channel impulse response in a respective entry of a feature vector (C _{MEAS /} C _RM) is in particular the Kanalimpulsant ^¬ word for the propagation time difference for each pair of base stations (BS1, BS2, BS3).

6. The method of claim 5, wherein at the beginning of the method, the reference card is initialized such that for a respective reference position (RP), the channel impulse response for the propagation time difference is represented on a respective pair of base stations (BS1, BS2, BS3) by a normal distribution whose average ence positions, a time difference (LD) ba ^¬ sierend on estimated distances between each refer- (RP) and the base stations (BS1, BS2, BS3) of the respective pair.

7. The method according to any one of the preceding claims, wherein in the pattern matching in step a) a number of feature vectors (C _RM ) at reference positions (RP) with the greatest agreement with the measured feature vector (C _meas ) at a respective object position (OP ) is determined and based on the reference position or positions (RP) for the number of feature vectors (C _RM ), in particular by averaging the reference positions (RP), the respective object ^¬ position (OP) is estimated.

8. The method according to any one of the preceding claims, are updated in the n step a) in parallel, the feature vectors (in the context of pattern matching C _RM) of the reference map, by the feature vectors (C _RM) of the reference map for Refe ^¬ ence positions (RP) in a predetermined environment are corrected by an estimated using the pattern matching object position (OP), each having a correction term, the correction term of the difference between the feature vector (C _RM) at the estimated via the pattern matching object Posi ^¬ tion (OP) and the (respectively to be corrected feature vector C _RM ) depends on the reference card.

9. The method of claim 8, wherein the size of the correction turterms with increasing distance of the respective reference Posi tion ^¬ (RP) in said predetermined around the estimated via the pattern matching object position (OP) decreases.

10. The method of claim 9, wherein the correction term depends on a center Gaussian function on the object position (OP) estimated over the pattern matching.

11. Method according to one of claims 8 to 10, in which the method is carried out iteratively, wherein after a number of updates of the feature vectors (C _RM ) of the reference card in step a) the steps b) and c) are carried out and subsequently is returned to step a).

12. The method according to any one of the preceding claims, wherein the predetermined movement model used in step b) sets a maximum speed for the movement of the mobile object (0) as a restriction.

13. The method as claimed in one of the preceding claims, in which the cost function used in step b) is a sum of Gaussian functions, each of which is based on the absolute difference between the respective object position (OP) estimated via the pattern matching and the optimized estimated object position ( OP).

14. The method according to any one of the preceding claims, wherein in step c) the updated feature vectors (C _RM ) of the radio network at the reference positions (RP) of the reference card are determined such that the optimized ge ^¬ estimated object position (OP), which has the lowest Distance to a respective reference position (RP) of the reference card has, is determined and the feature vector (C _RM ) at the respective reference position (RP) by the measured feature vector (C _MEAS ) at the optimized estimated Objektposi ^¬ tion (OP) tion with the smallest distance is replaced to the respective reference position (RP) of the reference card.

15. The method according to any one of the preceding claims, wherein in step c) the updated feature vectors (C _RM ) of the radio network at the reference positions (RP) of the reference card are determined such that for a respective reference position (RP) a weighted sum of Watch measured times ^¬ vectors (C _MEAS) is determined on optimized estimated Objektpositio ^¬ NEN (OP), wherein the weighted sum of the refreshes ^¬ alisierten feature vector (C _RM) at the respective Referenzpo- sition (RP) and in which the respective weights of the summands in the sum are smaller, the further an estimated optimized object position (OP) is away from the respective reference position (RP).

16. Method according to claim 15, in which the weights of the weighted sum are modeled by Gaussian functions, wherein a respective Gaussian function depends on the absolute distance between a respective reference position (RP) and that of the respective optimized estimated object position (OP).

17. The method according to any one of the preceding claims, wherein in the implementation of step a) and / or

Step b) fixed points with fixed spatial positions and feature vectors are considered, where one or more object positions (OP) are associated with pre-fixed points in the series of measurements and / or measured feature vectors (C _meas) of the measurement series are assigned fixed points, wherein the räumli ^¬ chen positions of Fixed points in step a) are treated as estimated object positions (OP) and / or treated in step b) as optimized estimated object positions (OP).

18. Method according to claim 17, in which measured feature vectors (C _meas ) of the measurement series are assigned fixed points via a heuristic.

19. The method of claim 17 or 18, wherein the spatial positions of the fixed points correspond to the spatial positions of the base stations (BS1, BS2, BS3).

20. A method for the computer-assisted localization of a mobile object (0) based on a reference map, which is learned with a method according to one of the preceding claims, wherein a feature vector (C _meas ) of the radio network is measured at the object position (OP) to be located and based on a pattern matching, which the measured feature vector (C _MEAS ) with feature vectors (C _RM ) at the reference position (RP) of the learned reference map, the Ob ektposition (OP) is determined.

21. A device for computer-aided learning of a reference map based on measurements of features of a radio network, wherein the measurements are made such that for a mobile object (0), which via the radio network with a plurality of base stations (BS1, BS2, BS3) of the radio network communicates, respective feature vectors (C _meas ) of the radio network are measured at a plurality of unknown object positions (OP) of the mobile object (0) and thereby a measurement series of a multiplicity of temporally successive feature vectors (C _MEAS ) for respective object positions (OP) jewei ^¬ time measurement points is obtained, wherein the reference card has a plurality of spatial reference positions (RP), wherein the device comprises a processing unit with which based on the series of measurements, the respective feature vectors in the operation of the apparatus (C _RM) of the radio network at the respective reference positions (RP) of the reference ^card with a method g be harvested, in which:

c) by means of the optimized estimated object positions (OP) updated feature vectors (C _RM ) of the radio network the reference positions (RP) of the reference card.

22. The apparatus of claim 21, which is configured such that with the device, a method according to any one of claims 2 to 19 is feasible.

23. Device for computer-assisted localization of a mobile object (0) based on a reference map, which is learned by a method according to one of claims 1 to 19, wherein the device comprises a measuring and processing device, with a feature vector (C _meas ) of the radio ^¬ network is on the measured locating Whether ektposition (OP), and compares with based on a pattern matching which is the measured feature vector (C _MEAS) to the feature vectors (C _RM) at the reference positions (RP) of the learned reference ^¬ card, the object position (OP) is determined.

24. Computer program ^product with a program code stored on a machine-readable carrier for carrying out a method according to one of claims 1 to 20, when the program runs on a computer.