Telecommunication

Enhancing the Accuracy of Position Information through Superposition of Location Server DataS.D.Hermann,A.Wolisz Telecommunication Networks Group Technische Universit¨a t Berlin Einsteinufer25,10587Berlin,Germany Email:[hermann,wolisz]@tkn.tu-berlin.deM.SortaisMathematisches Institut,Fak.II MA7-4Technische Universit¨a t Berlin Strasse des17.Juni136,10623Berlin,Germany Email:sortais@math.tu-berlin.deAbstract—The localization of a mobile device is of essential need in order to provide location based services to the users, e.g.in an emergency situation.Several mechanisms have been proposed so far to determine the positions,requiring rather expensive enhancements of existing access network equipment or mobile devices.In this paper,we present a new approach which uses relatively coarse,cell ID based location determination mechanisms.If the location information precision provided by a carrier is not sufficient for a service,it can be refined with data from other location servers.In contrast to other approaches for the enhancement of location information precision,our approach relies on existing location determination mechanisms inside the networks and requires only updates for the servers providing the information.I.I NTRODUCTIONDuring the recent years,a wide variety of location based services(LBS)[1]has emerged,being designed for the provi-sioning of useful services and information.For instance,these mechanisms provide nearby points of interest,track persons or asset,send warnings to traffic participants who approach an accident or guide rescue teams to users of mobile wireless devices being in distress.Obviously,possessing a sufficiently precise information about the geographical positions of the user devices is one of the fundamental issues for the provisioning of LBS.While most of the wireless devices like cell phones or personal digital assistants are not capable of determining their position by themselves,e.g.by means of GPS[2],the LBS providers rely on location information emanating from the carriers of the wireless access networks.In order to make location information from mobile devices available to LBS providers,two different components are being required by a carrier.First,the position of a mobile device has to be determined by the carriers.Afterwards,the information about the current whereabouts has to be provided to the LBS providers.The provisioning of the location information is performed via so called location servers.These are often directly main-tained by the carriers and possess the required interfaces to access the device positions.Depending on the standard,these location servers are also known as Gateway Mobile Location Centers in GSM/UMTS or Mobile Positioning Centers in ANSI.For the determination of the positions,several proposals have been made based on mechanisms like triangulation, trilateration or traversing.A survey of such techniques can be found e.g.in[3].They require a rather intensive technological andfinancial enhancement of the existing network equipment and/or the handsets.Thus,the widespread deployment of high accuracy positioning systems is not a challenge that could be achieved in short term.As long as no additional mechanisms are being deployed, the determination of a wireless device position can be achieved by using the ID of the radio cell it is currently associated with[4],which can usually be provided via the mobility management of the carriers.The drawback of this approach is the relatively low precision of the location information:one only knows that a mobile resides somewhere in the area being covered by a certain cell.In order to overcome this drawback,we propose a loca-tion refining(LORE)mechanism that can be used when a mobile device can be localized by several carriers,either by the same or by heterogeneous radio technologies(e.g.for mobile devices with WLAN and GSM/UMTS transceivers). Whenever the precision of the reported mobile device position is insufficient for the intended service,e.g.to guide rescue workers to an accident,the client can contact another carrier and request the position of the device from its location server. If the other carrier can also determine the cell of its network where the device currently resides,the client has the possibility to refine the position information by a superposition of the different data and a computation of the intersecting area of the overlapping cell coverage areas.The advantages of the approach are that it relies on the ex-isting cell ID based positioning mechanisms,without requiring further technical enhancements of the devices or base stations, but it can nevertheless lead to significant improvements of the resulting information.The rest of the paper is organized as follows:in the next section,related work is presented.Then in Section III,the model for the considered system is being presented,together with the metrics for the rating of the location information accuracy.Afterwards,in Section IV,we present our approachfor the location server based information refinement and the required enhancements for the interchange of location infor-mation,considering the Mobile Location Protocol(MLP)from the Open Mobile Alliance as an example.Some exemplary numerical results are presented in Section VI,and the last section concludes the paper by giving an outlook into the next steps of our ongoing work.II.R ELATED W ORKIn the context of cell ID based localization,statements about the current position of a mobile device are derived through its cell ID and a lookup in the equipment database of the carrier,where the positions of the deployed base stations are associated with the IDs.Taking into account the transmission range of the base stations,the possible positions of a mobile device can be limited to the resulting coverage area.In[5], experimental studies are presented for the assessment of cell ID based positioning accuracy in different scenarios.If a mobile device can receive radio signals from several base stations simultaneously,it has to be located in the area common to these stations.By a superposition of these areas and a computation of the intersecting area,a carrier can reduce the possible whereabouts of the device to the resulting area.In[6],the improvements of the location information are determined,if base stations are aligned in hexagonal or mesh structures.Additionally,the effects of transmission power tun-ing have been investigated.Earlier experiments were presented in[7],where the localization is performed by computing the centroid of the positions of the base stations that a device can detect.Besides an adaption of the transmission power,various other mechanisms have been proposed for an enhancement of accuracy,e.g.by computing suited sites for new base stations [8].All of the presented approaches for the improvement of the location accuracy require an enhancement of the handsets or of the access network equipment.III.S YSTEM M ODELIn the sequel,we consider a large geographical territory T with an area of size|T|.The territory is covered by a wireless access network A.The access network has a set of base stations S A and the members of the set are denoted by s A.Observing real networks,the possible locations of base stations are mostly very constrained.Instead of being aligned in regular structures,e.g.in hexagonal or rectangular patterns, the locations of the base stations show some spatial variability, but their characteristics may be describable via stochastic models with parameters like the density of the stations. Following[9],we model the positions X A i of the base stations as realizations of a stationary Poisson point process Φ.The positions are identically and independently distributed over the territory T.The density of the base stationsλA denotes the number of base stations per square meter.The(a)SNR(b)SINR(c)Closest basestationFig.1.Resulting coverage area approximations for the different models.In cases(a)and(b),the mobile device can associate with any of the base stations providing a sufficient signal,in case(c)it is associated with the closest base stationprobability distribution function for the number of base sta-tions in the territory to be|S A|=n is thusp(n)=e−λA|T|(λA|T|)nn!if n∈{0,1,...}0else(1)The territory under consideration is populated by a set of mobile devices M.If required,the network can determine the ID of the base station in whose coverage area a certain device m(i)currently resides.Next,we consider a second network B with a different set of base stations S B in the same territory.Similar assumptions which have been made for network A also hold for B,i.e. the positions of the base stations X B i∈Ψare realizations of a stationary Poisson point process with densityλB and the network can determine the ID of the base station in which device m(i)currentlyfinds itself.The wireless access technology is identical all over S A.The same holds for S B,whereas the technologies from the net-works A and B may differ.For the latter case we assume that the mobile devices support heterogeneous access technologies, i.e.they have several transmitters.Coverage Model–For the estimation of the base station coverage areas,we use two different,simplified models[10]. The pathloss,i.e.the weakening of the received signal power P RX between a base station and mobile device at position y during the transmission is computed viaP RX=P T X·K|y−X i|α(2) P T X denotes the power being transmitted by the base station s.The constant K and distance power gradientαapproximate the attenuation characteristics of the considered environment. As long as the strength of the received signal and the ratio between the signal and some received noise(SNR)stays above a certain threshold,a communication with the station is possible and the device is in its coverage area.Assuming that the signals of the different base stations do not interfere at the receiver,the SNR can be computed viaSNR=P RXP N(3)where P N denotes the equivalent power of the noise.Consid-ering real systems,the mobile device does not only receive the signals of the base station it is currently associated with, but also the signals from neighboring stations.These signals interfere with the transmission,and this is reflected by an additional term in the denominator of equation3,which then becomesSINR=P RXP N+j=iωj·P T X·K|y−X j|α(4)Theωj factors account for the interference with signals of adjacent base stations.If a mobile device is in the coverage area of several base stations of network A or B,we distinguish two behaviors:in thefirst case,the device is associated with the station providing the best signal strength(respectively with the closest base station,if the transmission power is equal for all stations), whereas in the second case the device is only associated with a certain probability p a to it.In Figure1,typical shapes of the resulting coverage areas are displayed.In the sequel, we assume omni-directional antennas or directional antennas being aligned in such a way that a site has omni-directional characteristics.If a mobile device is associated with several base stations,e.g.during a soft handover,this mechanism yields the ID of one of the cells.Metrics–Cell ID based positioning methods can assess that a mobile devicefinds itself somewhere in the coverage area of a certain base station C A.The possible positions are assumed to be uniformly distributed in the area,which leads to a natural understanding of the size of the coverage area|C A| as a measure for the precision of the position determination mechanism.If any gain is to be achieved in the precision of the location information of a device,the area of possible positions in the new setting should be significantly smaller than in the earlier setting.Therefore,a natural measure for the information gain is provided byInformation Gain=1−|C A∩C B||C B|if C A∩C B=∅0else(5)when the information from network B is combined with that from A.IV.L OCATION I NFORMATION I NTERCHANGE ANDP ROCESSINGIn this section,the provisioning of location information to the location services(LCS)clients of service providers by means of location servers is presented.After a brief description of the existing mechanism,we present our enhancement for the superposition of information from several servers.In the last part,an additional enhancement to increase the efficiency of the mechanism ispresented.Fig.2.The considered scenario.a)The LCS Client sends a location request message to the location server of carrier A.b)A location response message containing the area covered by the base station of network A as possible whereabouts is returned.c)An enhanced location request message is sent, this time to the location server of network B in order to refine the position information,followed by its response d)A.Current Location Server Based MechanismLocation servers are established by the carriers for the provisioning of suited access to the positions of the mobile devices in their networks.If a service requiring device location is to be provided to a device user,the LCS client of the service provider contacts the carrier and sends a location request to the location server(see Figure2).This request contains the mobile device ID of the user and some additional information, e.g.the coordinate reference system that should be used for the position being returned by the server and a time frame within which a reply should be sent.After receiving the request,the location server checks if it possesses the current position of the device due to a request that has been answered shortly before,else it triggers the location determination process inside the carrier network which determines the ID of the cell the device is associated with.The concrete realization of the location determination inside the network depends on the access technology and the state of the rmation about the associated cell can usually be deduced directly from the mobility management of the network,which makes the procedure fairly attractive for the carriers.Before a reply can be generated for the LCS client,the carrier has to translate the cell ID information into the possible positions of the device,i.e.the area being covered by the cell. This is achieved by mapping the ID to a geographical position and by estimating the coverage area of the cell.Due to the fact that radio propagation is influenced by many, even time dependent factors,some heuristic assumptions have to be made for their estimation.In some cases,the carriers have a more detailed description of the deployed network equipment being used for coverage area approximations,e.g. for network planning in UMTS[11].These descriptions mayalso be used for the present localization purpose.An approach for the acquisition and provisioning of hotspot coverage infor-mation is presented in[12].Approximations of coverage areas can certainly not be exact.In[13],several methods taking the corresponding uncertainty into account are provided.When the coverage area has been determined,the location server sends a location response message to the LCS client which issued the request.The message contains the ID of the mobile device and the area denoting its possible whereabouts. The area can be approximated by appropriate geometrical shapes.Definitions of various shapes are usually provided by the protocol for the location information interchange,and these shapes may be circular,elliptical or polygonal.B.Superposition of Location Server Information Depending on the service to be provided,the LCS client has to decide if the precision of the returned information is sufficient or not.For instance,for the determination of the scene of an accident in an urban area,any improvement may be desirable,whereas for services like home zone billing,smaller improvements may be sufficient.If the information meets the requirements,the service will be executed,otherwise the LCS client has to try and refine it.Such refinement can be performed if other carriers are able to locate the mobile device in their networks,as well as the corresponding base station.If the device possesses more than one transceiver,enabling an association with different carrier networks,the LCS client can request the position from the responsible location servers.Since another carrier will normally use its own location server,a location request for the mobile device being sent to that carrier will result in a different location report than thefirst one.This second request necessitates the ID of the device,respectively the ID pertaining to the second transceiver,and has to be transmitted to the LBS provider.If the mobile device has one transceiver,the possibility for a second carrier to locate it depends on several factors,e.g. on the radio access technology,usage of the same frequency band or on the state of the mobile,and on the use of some further information by the carrier.For instance,received probe requests from a WLAN device can be used to determine and provide its location.In order to supply the LBS provider with an additional refinement,a list of alternative carriers is transmitted to it.The interchange of messages between the LCS client and the second location server is almost identical to thefirst inter-change being described in the previous section.The LCS client sends a location request message with the ID of the mobile device to the location server,and it receives an appropriate response with the area denoting the possible whereabouts of the device.From the second location server’s point of view, the request message can not be distinguished from a usual one,because no information is exchanged between different location servers.After the LCS client received the response,it possesses two different areas which have to be considered for the positions.Since the mobile device is located in the area part being covered by the two different cells together,the LCS client determines the positions of the mobile device by a combination of the information from the different carriers,i.e. by a computation of the intersection area of the shapes that have been reported as possible positions of the device.In case of a failed determination of the position by the second network,the LCS client may try to contact another carrier or try to process the service using the prior information from thefirst one,if no more carriers are available.In the present approach,messages are only exchanged between the LCS client and the location servers.It should be noted that a direct exchange between the location servers is also technically feasible.For example,whenever the server of the carrier cannot attain a prescribed precision,it might automatically send an enhanced request to another carrier location server.Although such an approach might save some bandwidth,it has some drawbacks.For instance,the LCS client might have only limited influence on the forwarding of its request.Another important point is the scalability of the approach.Since the refinement of location information is performed by geometrical computations that require a certain amount of processing power,the execution of this task by the LCS clients might significantly reduce the workload of the limited number of location servers.C.Protocol Enhancements for Efficiency Improvement Location determination by a network can be a relatively expensive task;in order to ensure that the additional location determination leads to an expedient refinement,information about the locations and parameters of the base stations of the second carrier are required.But these locations are often considered as sensitive information by the carriers and are rarely offered to the LCS clients or anyone else,mostly due to reasons of business competition.Thus,we propose an additional mechanism enabling a carrier to obtain predications about the information gain that could be achieved if it were to determine the coverage area of the cell where the mobile device currently resides.After the evaluation of the information precision from thefirst carrier,the LCS client can choose a minimum level of desired refinement depending on the service.The desired information gain is added together with the area that has been returned by thefirst location server to the second request.After receiving the request,the location server of the second carrier performs a lookup in the equipment database and determines the base stations being responsible for mobile devices in that region.If the requested information gain can be achieved(which may be impossible,e.g.if this area lies entirely within the coverage areas of the base stations),the position of the node is determined and reported to the LCS client,else an error message is sent back.By using the area,the direct exchange of base station positions between the involved entities is not required.Ad-ditionally,it is difficult to determine precise positions out ofan approximated coverage area,unless the carrier does not use circular areas with the base stations as their centers.<hdr ver="3.0.0"><client><id>carrier_b</id><pwd>carrier_b_pwd</pwd><serviceid>0005</serviceid><requestmode type="PASSIVE"/></client><requestor><id>req_lcs_client</id><serviceid>0003</serviceid></requestor></hdr><slir ver="3.0.0"res_type="SYNC"><msids><msid type="ASID">75462543748</msid></msids><area_extension><shape><CircularArea srsName="#4326"> <coord><X>523056.5N</X><Y>131931.8E</Y></coord><radius>67</radius></CircularArea></shape></area_extension><infogain_extension>0.6</infogain_extension><eqop><resp_req type="LOW_DELAY"/><hor_acc>1000</hor_acc></eqop><geo_info><CoordinateReferenceSystem><Identifier><code>4326</code><codeSpace>EPSG</codeSpace><edition>6.1</edition></Identifier></CoordinateReferenceSystem></geo_info><loc_type type="CURRENT"/><prio type="HIGH"/></slir>Fig.3.MLP message header and body of the second requestD.Required Protocol EnhancementsFor the refinement of the location information,additional data has to be added to the requests and responses being exchanged between the LCS client and location servers of the different carriers.In this section,we illustrate the integration of the required extensions,using the Mobile Location Protocol (MLP)[14]from the Open Mobile Alliance as an example. This is an application level protocol using XML encoding for the data.Extensions can be defined via separate document type definitions and applied in the messages.Figure3shows an example of the request being sent to the second carrier, with two extensions for the possible whereabouts of the device(area extension)and for the required refinement of the information(infogain extension).cr(a)Circlearrc12(b)Ellipsevvvvvv123456(c)PolygonFig.4.Different shapes for geographical areas representing the possible whereabouts of a mobile deviceV.C OMBINING D IFFERENT L OCATION S ERVERI NFORMATIONAs described in the previous section,a LCS client has to combine the information of the different location servers in order to refine the location information.This information consists of different areas denoting the possible whereabouts of a mobile device,and the refinement is performed by a geometrical computation of the intersection of these areas.In this section,we present the required mechanisms for different kinds of geometrical area descriptions.A.Area DescriptionsFor the description of the area which represents the possible whereabouts of a mobile device,several kinds of shapes are defined in the location protocol specification.Point–A point is described by its coordinates and denotes the exact position of a mobile device.Circle–A circle is characterized by the center coordinates c and a radius r(see Figure4(a)).The position of the mobile device is at a distance to the center being less than or equal to the radius.Further circle based shapes(e.g.arcs)can be described by introducing additional parameters.Ellipse–An ellipse is described by four parameters,the center coordinates c,an angle of orientation a,a semi-minor and a semi-major axis r1and r2,respectively(Figure4(b)).The orientation of the ellipse is given by the clockwise measure of the angle between the northern direction and the semi-major axis.The coordinates c describe the center of the ellipse.The position of the mobile device is assumed to be within or on the border of the ellipse.Polygon–A polygon is described by a list of point coordinates V.The points v∈V represent the vertices of the area inside which the mobile devicefinds itself(Figure4(c)).These vertices are assumed to be connected by straight lines.If coordinates are to be exchanged,the World Geodetic System1984(WGS84)[15]is chosen by default as reference system in the location protocol.The selection of proper shapes is directly influenced by the outline of the cell coverage areas that should be approximated. Generally speaking,the carriers have to choose those shapes whichfit the coverage areas of their cells best.For instance, ellipses might be used for the approximation of directional antennas,and polygons may be deployed in urban areas fori1i2Fig.5.Intersection points i1and i2are computed,and segments of the original borders become intersection area bordersirregular coverage areas that emerge from the influence of the buildings on the radio propagation.B.Intersection ComputationAfter receiving the descriptions of the possible whereabouts, the LCS client has to compute the intersection area.The description of the area will be used as refined location in-formation of the mobile device and consists of the borders of the area.Depending on the kind of original area shapes, the description will again be a point,a circle,an ellipse or a polygon,and the latter may also have curved segments.If the description of an intersection area becomes too complex for an appropriate usage as location information,it might be useful to perform an approximation of the shape(e.g.by the smallest circle enclosing the area).As afirst step for the determination of the area,the intersec-tion points of the original area borders have to be computed. Afterwards,the segments of the original area descriptions being borders of the intersection area will be used for the description of the new area(see Figure5).If no intersection points between the original areas exist,either one of the areas lies completely inside the other one,or they do not intersect at all.In thefirst case,the intersection area is identical to the smaller area.In the second case,an appropriate handling needs to be determined,if required.For the computation of the intersection area,several algo-rithms have been proposed and refined in the recent decades. While the intersection of a point with another shape will always be the point,other shapes require more complex algorithms.Circle Intersection–In the case of two intersecting circle borders,the intersection area is given by the area between the arcs of the two circles enclosing their common area.The intersection points of the circles are computed via algorithms solving the system of quadratic polynomial equations for the circles.Ellipse Intersection–The computation of ellipse intersection points can be performed e.g.by the algorithm being proposed in[16].By using a suitable coordinate transformation mapping one of the ellipses onto a unit circle and the second one onto a new ellipse,the problem can be reduced to the computation of the intersection area of this circle with the remaining ellipse, which is similar to the circle intersection problem. Polygon Intersection–For the computation of the intersection area of two polygons,first the intersection points of the edges of the polygons have to be determined.The intersection points are considered as new vertices.Together with the edges and vertices of the two polygons,they lead to edges surrounding the intersection area as well as the original areas of the remaining polygons.Then,the vertices of the edges from the intersection area are extracted.For non-convex polygons,the intersection area may be disconnected.A fast algorithm for the computation of intersection points between polygons has been presented in[17].It is asymptotically optimal with a time complexity of O(n log(n)+k),where n is the total number of polygon segments and k the number of intersection points. The kinds of area shapes being returned by the location servers may differ from one response to the other.Thus, computing the intersection area between different kinds of shapes may also be required.Intersections of ellipses and circles can be handled as a special case of ellipse intersections. By splitting the descriptions of circular and elliptical areas into multiple segments,intersections of these shapes with polygons can also be computed by the polygon intersection algorithm.VI.E XEMPLARY N UMERICAL R ESULTSIn this section,we present some exemplary numerical results for the information gain that can be achieved through a superposition of location information from different networks. In order to generate these results,we implemented the mech-anism and integrated it in the discrete event simulation tool OMNeT++[18].TABLE IP ARAMETERS FOR THE DIFFERENT TECHNOLOGIESParameter WLAN GSM UMTSSNR min14dB9dB9.8dBP T X10mW300mW500mWK-40dB-37.5dB-46.8dBα 2.8ω-17dBP N-100dBmFor the investigation,we have chosen three different tech-nologies,WLAN,GSM and UMTS.The parameters for the simplified coverage models of the technologies are listed in table I.In order to analyze the effect of the coverage models, we conducted simulations for4different cases:(i)SNR model for the coverage,and the mobile device is associated only with a certain probability to the closest base station,i.e.the possible whereabouts of the device are in the whole coverage area.(ii)SNR model for the coverage,but the mobile device is associated to the closest base station.(iii)SINR model with the device being possibly in the whole coverage area.(iv) SINR model with the device being associated with the closest station.In the simulations,we investigated a section of a geograph-ical territory with a size of1km×1km for WLAN and a size of3km×3km if GSM or UMTS are involved.Intervals being displayed in the graphs denote the95%confidence intervals. For each base station density,500different realizations of station positions have been simulated.。