Not Applicable
This invention relates to techniques for determining location and, specifically, to techniques that utilize the NAVSTAR Global Positioning System (GPS).
The NAVSTAR Global Positioning System (GPS) developed by the United States Department of Defense uses a constellation of between 24 and 32 Medium Earth Orbit satellites that transmit precise microwave signals, which allows devices embodying GPS sensors to determine their current location. Initial applications were predominantly military; the first widespread consumer application was navigational assistance.
With the explosive growth in mobile communications devices, a new wave of location-based applications is emerging. These applications are characterized by a requirement for device-centered maps. One example is a form of “Yellow Pages” in which a map centered at the location of the mobile device presents user-selected establishments (e.g., barber shops) in situ. Another example would be an application enabling one to locate, on a device-centered map, members of his or her social network. In these applications, location involves but two dimensions (North/South and East/West). Thus, while GPS is generally capable of providing three dimensional locations (or 3D GPS fixes), the altitude, or z dimension is typically inhibited. In the interest of simplicity, the narrative which follows ignores the three dimensional potential of GPS-based location.
Outdoors, GPS (using four or more satellites) is a reliable and accurate source of the location information essential to enable a device-centered map to be served across a network to a mobile device. Indoors, however, today's GPS receivers, even when operating in 2D mode (using just three satellites and, most commonly, a pseudo satellite located at the center of the earth), do not have sufficient sensitivity to provide reliable and accurate location information. As a result, location-based applications typically utilize GPS outdoors (where at least three satellites are generally available), and the services of any of several location service providers, where GPS fails to yield location information. An example of such a service is that provided by Google to cellular network subscribers, wherein the location of the requesting subscriber is estimated using the relative strength of signals from base stations in the vicinity of the subscriber.
Location services are built around proprietary databases, compiled by location service providers. These databases are essentially compilations of the locations of terrestrial transmitting towers (beacons), or compilations of the signal-strength contours surrounding these beacons, or compilations of the locations together with the associated signal-strength contours. The most commonly available beacons are cellular towers and wireless (e.g., Wi-Fi) access points. A subscriber (to a location service) provides the ID's and (optionally) the associated signal strengths of any beacons detectable by his or her mobile device, and the location service provider responds with its best estimate of the location of the device. For the purpose of this discussion, the system employed by a location service provider is referred to as a Mobile Device Location System (MDLS).
Whether the beacons are cellular towers or wireless access points—the measurement processes are largely identical. Measurements are taken using instruments similar to (and in some cases, identical to) the mobile communications devices the MDLS has been designed to locate. The standard mode of operation is to traverse the geography of interest with a measuring instrument, pausing periodically to record its location (using GPS) together with the ID's and signal-strengths of any beacons detectable by the device. As these measurements accumulate, beacon locations and/or beacon-associated signal-strength contours are generated and thereafter continually updated by the BLEN, to enable the generation of accurate and reliable estimates of location in response to subscriber requests.
To appreciate the limitations of an MDLS as described in
While the signal-strength contour approach is fundamentally sound and works indoors and out, commercial cellular systems do not support the accuracy expected by subscribers, who have become accustomed to the accuracy provided (outdoors) by GPS.
Given the ubiquity of GPS, viability compels location service providers to consider the alternative of an access-point-based service. Providing access-point-based services requires a different approach, inasmuch as the task of developing a database of wireless signal-strength contours is virtually impossible owing to the inability of GPS to provide thorough coverage in the vicinity of typically-indoor access points. Inasmuch as signal strength measurements are generally discontinuous across exterior walls, precluding the interpolation of signal strength contours across the length and breadth of buildings (from measurements taken on the periphery), there are really no practical alternatives that leverage the ubiquity of GPS.
Rao and Siccardo (U.S. Pat. No. 6,269,246) describe a method for estimating the location of a mobile device by comparing a fingerprint of the RF spectrum captured by the mobile device with an ensemble of fingerprints, each with an associated location, resolving the closest matching fingerprint, and selecting its associated location as the estimate of the location of the mobile device. However, without cost-effective means for developing and maintaining a database of tens of indoor fingerprints, for perhaps 500M indoor access points, the practicality of the method is limited.
Agrawala and Youssef (U.S. Pat. No. 7,406,116) describe a method for estimating the location of a mobile device utilizing multiple radio maps (one for each set of access points detectable by the mobile device within the service area), which define the distribution of signal strength (for each access point) at surveyed locations throughout the service area. Their radio maps are compiled through the systematic calibration of signal strength(s) at surveyed locations within the service areas, a measurement process similar to that contemplated in Rao and Siccardo. The estimate of the location of the mobile device is then determined as the point of measurement at which the conditional probability of the reported vector of signal strengths is maximized. The practicality of this solution is likewise limited owing to the prohibitive cost of creating maps for all of the nearly 500M indoor access points operational today.
Morgan, et al (U.S. Pat. No. 7,403,762) describe a method of building a reference database of access point locations which involves traversing the target area in a programmatic route to avoid arterial bias. Measurements obtained for a given access point are then used to reverse triangulate the position of the detected access point. The location of a mobile device, reporting its detection of specific access points together with the associated signal strengths, is then accomplished using one of several location-determination algorithms. With the exception of access points deep in the interior of large buildings or high above the ground floor, which may go unobserved, this method leverages the ubiquity of GPS, and thus qualifies as being more practical than either of the two previous methods. Still, the task of systematically mapping 500M access points is daunting, if not prohibitively expensive.
In addition, U.S. Pat. No. 7,403,762 further describes a reference database of access point locations each with its associated signal fingerprint information, consisting of the signal strengths of the messages received from the access point as well as the locations recorded at the several points of measurement used to determine the location of the access point itself. With sufficient measurements per access point, this database could be used to construct signal strength contours, with the location of the mobile device determined by “intersecting” the contours associated with the access points/signal strengths reported by the mobile device, but its utility as such would fall considerably short, as the associated fingerprint information would largely be limited to signal strengths measured outdoors, and thus provide little insight as to signal strengths measured indoors. Useful or not, the task of developing and maintaining such a database remains daunting.
Externally, the growth in applications built on the assumption of broadband access suggests that as mobile broadband traffic searches for relief from the cost and congestion of the cellular networks, location-based applications will increasingly favor access-point-based location services. The demand for access-point-based services exposes a need in the art for an MDLS framework to enable the rapid, inexpensive compilation and maintenance of a comprehensive AP Survey Database, enabling the location of subscribers with the accuracy and efficiency required by current and contemplated location-based applications.
In general, the object of the present invention is to provide an MDLS framework to enable the rapid and inexpensive compilation and maintenance of a comprehensive AP Survey Database, to support the location of subscribers with the accuracy and efficiency required by current and contemplated location-based applications. As the repository for field measurements, applied ultimately to estimate the locations of subscribers' mobile devices, the AP Survey Database influences virtually every metric of importance to the operators as well as the subscribers of location services. Coverage must be complete; measurements must be easily and efficiently obtained, using accurate but inexpensive instruments. And the real-time computation associated with estimating subscriber locations with acceptable accuracy must be manageable.
The frameworks of mobile device location systems in use today are typically built on Beacon Survey Databases constructed with the aid of 2D GPS fixes. Estimates of subscriber location are then derived using cues (beacon ID's, signal strengths) supplied by the subscriber's mobile device. While cell-tower-based systems provide excellent coverage, the accuracy indoors is generally unacceptable. Access-point-based systems, with the potential for significantly improved accuracy, have yet to demonstrate their viability, owing to the cost of developing and maintaining the AP Survey Database.
To address the challenges facing today's access-point-based systems, a new MDLS framework (
A convenient corollary of this thesis is that it enables the enlistment of subscribers in the measurement process. Under the existing framework(s), subscribers do not participate in the measurement process, requesting location services when GPS has failed to yield a 2D fix. Under the proposed framework, a request for location service could include GPS data (even though GPS has failed to produce a 2D fix), along with the ID's and signal strengths of any AP's detectable by the subscriber's mobile device. (To the extent a subscriber supplies any information for the purpose of obtaining his or her location, he or she has voluntarily disclosed his or her location; hence the proposed enlistment introduces no issues of privacy.)
Finally, this thesis and the resulting framework favor a thin-client implementation, which, in turn, enables the application of sophisticated signal processing techniques to reduce significantly the minimum signal strength required to acquire GPS satellites indoors.
a illustrates the thin-client approach to the extraction of GPS satellite signals from the composite GPS signal. The GPS measurement data input to the MDLS consists of the composite GPS signal for processing offline, rather than a 2D fix as is the case in today's MDLS. Moreover, instead of compiling a database of AP locations from 2D fixes, the proposed framework envisions a database of wireless signatures for use in locating subscribers directly. The GPS Signal Processor (GPSP) extracts surfaces of location (e.g., spheres and hyperplanes) which are catalogued by signature (a signature being the set of AP's detected at the point of measurement together with their associated signal strengths) in the Surfaces of Location Database (SLDB). Periodically the SLDB is referenced by the Signature Mapping Engine (SMEN) for the purpose of updating the Signature Database—a database used to estimate the location of subscribers. To the extent that “spheres and hyperplanes” (requiring one and two satellites, respectively) are more readily acquired indoors than 2D fixes (requiring three satellites), this framework enables the faster, cheaper compilation of a market-ready Signature Database, improving coverage as well as accuracy. Furthermore, the enlistment of subscribers in the compilation of the database ensures that once up and running, the database will be maintainable at modest cost.
Instead of performing the compute-intensive (and therefore battery depleting) signal processing task at the point of measurement, it is performed offline, where powerful servers are available to implement sophisticated (and generally more compute-intensive) signal processing techniques to maximize the processing gain, reducing the minimum signal strength required to acquire a GPS satellite, and leveraging further the thesis articulated above. One such technique looks to large datagrams (recordings of the composite GPS signal) as, the means to improve signal processing gain. This technique has recently become practical (cf. pending U.S. patent application Ser. No. 12/587,096 and Ser. No. 12/587,099) with the application of a perfect reference (see below).
In the MDLS of
As indicated, the Signature Mapping Engine (SMEN) constructs/maintains the Signature Database. For each Signature, a proxy location is constructed utilizing the associated surfaces of location. The Signature Database thus becomes a directory indexed on wireless signatures, with a wireless signature consisting of a list of access points together with their associated received signal strengths, and the directory associating with each signature a location, which is derived from surfaces of location measured at different times using different measuring instruments. The task of the Subscriber Device Location Engine (SLEN), which estimates the location of subscribers, is thus reduced to a search of the Signature Database for the wireless signature or signatures closest to the subscriber-supplied signature, and interpolating between the associated locations.
There are many possible metrics available to measure closeness, when the subscriber-supplied signature cannot be found in the Signature Database. A “distance metric” could be constructed to measure the closeness of two signatures. One or more “acceptance criteria” (e.g., the number of wireless access points in common) could be applied to insure the integrity of the “distance metric”.
An alternative approach to the construction of the Signature Database is to begin with a comprehensive catalogue of wireless signatures. Working through this catalogue, signature by signature, the SMEN first assembles a relevant set of unique hyperplanes for use in estimating the location corresponding to the signature in question. In the ideal case, the Surfaces of Location Database contains several unique hyperplanes whose signatures match identically the signature in question, in which case, the choice is simple: the relevant set is comprised of those unique hyperplanes whose signatures match that provided by the subscriber.
Having constructed a relevant set of hyperplanes, SMEN proceeds to determine the point closest to those hyperplanes comprising the relevant set. A common technique is to determine the point for which the weighted sum of the squares of the distances from said point to the hyperplanes of the relevant set is minimized.
As an aside, it is important to note that the minimization of the sum of the squares of the distances to hyperplanes, for example, is but one of many objective functions, contemplated in the present invention, for minimizing subscriber location estimation error. Another well-known objective function is min/max, wherein the maximum distance from the estimated location to the surfaces of hyperplanes is, minimized. Simple modifications to the foregoing would be to weight the measurements within the minimization. Numerous weightings (e.g., relative signal strength, age of measurement, etc.) are available to enable high quality estimates.
Absent the ideal case, the choice of relevant hyperplanes involves the use of a “distance metric” with an “acceptance criteria”. The metric would allow signature comparison, and the criteria would determine the permitted variance from the ideal. It should be noted that the minimum number of hyperplanes required to estimate a subscriber's location is two, as SMEN applies a pseudo satellite at the earth's center to assure that reasonable estimates are obtained even when measurements are sparse.
In accordance with this invention, a method for estimating the location of a mobile Wi-Fi signal receiver from a database of independently obtained survey data; i.e., survey data measured at different times, in different places, and with different instruments, is disclosed. Each survey datum includes a surface of location derived from a composite GPS signal, together with a Wi-Fi signature measured concurrently with the GPS signal measurement, at some point on said surface of location. The method comprises receiving a Wi-Fi signature, measured and recorded by said mobile Wi-Fi signal receiver, at the location to be estimated; extracting from the database, an algorithmically-determined subset of surfaces of location, utilizing the Wi-Fi signature recorded by said mobile Wi-Fi signal receiver, and estimating the location of said mobile Wi-Fi signal receiver from said algorithmically-determined subset of surfaces of location.
In one embodiment, the algorithmically-determined subset consists of those surfaces of location with Wi-Fi signatures identical to the mobile Wi-Fi signature; and the estimate of the location of said mobile Wi-Fi signal receiver is determined as the point for which the sum of the squares of the distances to each of the surfaces of location included in said algorithmically-determined subset is minimized.
Also disclosed is a system for building a database of Wi-Fi signatures, each with an associated location within a geographical area of interest, said system comprising:
one or more mobile Wi-Fi database-building devices, each having a GPS receiver and a Wi-Fi signal transceiver;
a survey process, whereby each of said one or more mobile Wi-Fi database-building devices, operating within said geographical area of interest, periodically measures and records a time-stamped composite GPS signal, hereinafter referred to as the GPS datagram, together with the identifiers and signal strengths of any Wi-Fi access points detectable by said mobile Wi-Fi database-building device, the set of access point identifier/signal strength pairs hereinafter referred to as a Wi-Fi signature;
GPS signal processing means for deriving, from each GPS datagram, the start-of-transmission and the time-of-arrival (TOA) from, and implicitly the distance to, each acquirable GPS satellite:
spherical surface of location derivation means for deriving a spherical surface, centered on an acquired satellite, at a prescribed distance from the point of measurement;
spherical surface of location database means for storing spherical surfaces of location, together with their associated Wi-Fi signatures;
Wi-Fi signature location estimation means for estimating, from spherical surfaces of location sharing a common Wi-Fi signature, the approximate location of the one or more mobile Wi-Fi database-building devices at such times as said common Wi-Fi signature was measured and recorded; and
Wi-Fi signature database means for storing Wi-Fi signatures, together with their associated estimates of location.
In one embodiment of the foregoing system, the GPS signal processing means derives the distances to acquired satellites from GPS datagrams spanning multiple cycles of GPS' 50 Hz data overlay, extending processing gain to its practical limit.
In another embodiment, the GPS signal processing means uses a perfect reference, to enable efficient processing of large GPS datagrams.
Those skilled in the art will understand that the methods and apparatus of the present invention may be applied to satellite positioning systems evolved from the GPS satellite positioning system, including but not limited to the Galileo and Glonass systems.
Various aspects and features of the present invention may be understood by examining the drawings here listed.
a shows a system diagram of an MDLS employing the present invention
b shows a system diagram of an MDLS employing the present invention
a) describes a robust GPS assistance system
b) describes a robust GPS assistance system
In general, the object of the present invention is to provide an MDLS framework to enable the rapid and inexpensive compilation and maintenance of a comprehensive AP Survey Database, to support the location of subscribers with the accuracy and efficiency required by current and contemplated location-based applications. As the repository for field measurements, applied ultimately to estimate the locations of subscribers' mobile devices, the AP Survey Database influences virtually every metric of importance to the operators as well as the subscribers of location services. Coverage must be complete; measurements must be easily and efficiently obtained, using accurate but inexpensive instruments. And the real-time computation associated with estimating subscriber locations with acceptable accuracy must be manageable.
The framework of mobile device location systems in use today are built on Beacon Survey Databases constructed with the aid of 2D GPS fixes. Estimates of subscriber location are then derived using cues (beacon ID's, signal strengths) supplied by the subscriber's mobile device. While cell-tower-based systems provide excellent coverage, the accuracy indoors is generally unacceptable. Access-point-based systems, with the potential for significantly improved accuracy, have yet to demonstrate their viability, owing to the cost of developing and maintaining the AP Survey Database.
To address the challenges facing today's access-point-based systems, a new MDLS framework (
A convenient corollary of this thesis is that it enables the enlistment of subscribers in the measurement process. Under the existing framework(s), subscribers do not participate in the measurement process, requesting location services when GPS has failed to yield a 2D fix. Under the proposed framework, a request for location service could include GPS data (even though GPS has failed to produce a 2D fix), along with the ID's and signal strengths of any AP's detectable by the subscriber's mobile device. (To the extent a subscriber supplies any information for the purpose of obtaining his or her location, he or she has voluntarily disclosed his or her location; hence the proposed enlistment introduces no issues of privacy.)
Finally, this thesis and the resulting framework favor a thin-client implementation, which, in turn, enables the application of sophisticated signal processing techniques to reduce significantly the minimum signal strength required to acquire GPS satellites indoors.
a illustrates the thin-client approach to the extraction of GPS satellite signals from the composite GPS signal. The GPS measurement data input to the MDLS consists of the composite GPS signal for processing offline, rather than a 2D fix as is the case in today's MDLS. Moreover, instead of compiling a database of AP locations from 2D fixes, the proposed framework envisions a database of wireless signatures for use in locating subscribers directly. The GPS Signal Processor (GPSP) extracts surfaces of location (e.g., spheres and hyperplanes) which are catalogued by signature (a signature being the set of AP's detected at the point of measurement together with their associated signal strengths) in the Surfaces of Location Database (SLDB). Periodically the SLDB is referenced by the Signature Mapping. Engine (SMEN) for the purpose of updating the Signature Database—a database used to estimate the location of subscribers. To the extent that “spheres and hyperplanes” (requiring one and two satellites, respectively) are more readily acquired indoors than 2D fixes (requiring a minimum of three satellites), this framework enables the faster, cheaper compilation of a market-ready Signature Database, improving coverage as well as accuracy. Furthermore, the enlistment of subscribers in the compilation of the database ensures that once up and running, the database will be maintainable at modest cost.
Instead of performing the compute-intensive (and therefore battery depleting) signal processing task at the point of measurement, it is performed offline, where powerful servers are available to implement sophisticated (and generally more compute-intensive) signal processing techniques to maximize the processing gain, reducing the minimum signal strength required to acquire a GPS satellite, and leveraging further the thesis articulated above. One such technique looks to large datagrams (recordings of the composite GPS signal) as the means to improve signal processing gain. This technique has recently become practical (cf. pending U.S. patent application Ser. No. 12/587,096 and Ser. No. 12/587,099) with the application of a perfect reference (see below).
In the MDLS of
As indicated, the Signature Mapping Engine (SMEN) constructs/maintains the Signature Database. For each Signature, a proxy location is constructed utilizing the associated surfaces of location. The Signature Database thus becomes a directory indexed on wireless signatures, with a wireless signature consisting of a list of access points together with their associated received signal strengths, and the directory associating with each signature, a location, which is derived from surfaces of location measured at different times using different measuring instruments. The task of the Subscriber Device Location Engine (SLEN), which estimates the location of subscribers, is thus reduced to a search of the Signature Database for the wireless signature or signatures closest to the subscriber-supplied signature, and interpolating between the associated locations.
There are many possible metrics available to measure closeness, when the subscriber-supplied signature cannot be found in the Signature Database. A “distance metric” could be constructed to measure the closeness of two signatures. One or more “acceptance criteria” (e.g., the number of wireless access points in common) could be applied to insure the integrity of the “distance metric”.
An alternative approach to the construction of the Signature Database is to begin with a comprehensive catalogue of wireless signatures. Working through this catalogue, signature by signature, the SMEN first assembles a relevant set of unique hyperplanes for use in estimating the location corresponding to the signature in question. In the ideal case, the Surfaces of Location Database contains several unique hyperplanes whose signatures match identically the signature in question, in which case, the choice is simple: the relevant set is comprised of those unique hyperplanes whose signatures match that provided by the subscriber.
Having constructed a relevant set of hyperplanes, SMEN proceeds to determine the point closest to those hyperplanes comprising the relevant set. A common technique is to determine the point for which the weighted sum of the squares of the distances from said point to the hyperplanes of the relevant set is minimized.
As an aside, it is important to note that the minimization of the sum of the squares of the distances to hyperplanes, for example, is but one of many objective functions, contemplated in the present invention, for minimizing subscriber location estimation error. Another well-known objective function is min/max, wherein the maximum distance from the estimated location to the surfaces of hypeiplanes is minimized. Simple modifications to the foregoing would be to weight the measurements within the minimization. Numerous weightings (e.g., relative signal strength, age of measurement, etc.) are available to enable high quality estimates.
Absent the ideal case, the choice of relevant hyperplanes involves the use of a “distance metric” with an “acceptance criteria”. The metric would allow signature comparison, and the criteria would determine the permitted variance from the ideal. It should be noted that the minimum number of hyperplanes required to estimate a subscriber's location is two, as SMEN applies a pseudo satellite at the earth's center to assure that reasonable estimates are obtained even when measurements are sparse.
In addition to its primary role; namely, to refresh the Signature Database as new surfaces of location are acquired, SMEN is uniquely positioned to assume a quality assurance role. With access to the Signature Database, SMEN can spot anomalies in the Signature data and isolate “corrupted” hyperplanes in the SLDB.
While a database of wireless signatures is conceptually quite different than a database of AP locations, the proposed framework has a number of similarities with today's MDLS. Whether the beacons are cellular towers or wireless access points (or the beacon of the future)—the proposed measurement processes are largely identical. As with the existing framework, the measuring instruments employed are similar, if not identical, to the mobile communications devices which the MDLS has been designed to locate. The standard mode of operation is modified, but slightly; namely, to traverse the geography of interest with the instrument, pausing periodically to record GPS data (rather than 2D fixes) together with the ID's and signal-strengths of any beacons detectable by the instrument. It is at the point that the processing of the measurements begins that the frameworks diverge dramatically. In the proposed framework, the GPS data are processed offline, as it were, with a resulting improvement in sensitivity achieved through the use of advanced signal processing technology (see below). The output of the signal processor is a surface of location (rather than a fix), which forms the Surfaces of Location Database. In another departure from today's MDLS, the subscriber's mobile device (assuming on-board GPS) serves as an instrument in the ongoing measurement process.
While a database of wireless signatures provides a foundation for providing more accurate estimates of subscriber device location, the improvement in accuracy depends on the resolution of signal strength measurements. Where signal strength measurements are relatively coarse, and the incidence of overlapping wireless coverage is high, it may be appropriate to compile a Signal-Strength Contour Database from the Signature Database, enabling the use of signal-strength-contour intersection techniques in lieu of or in addition to the signature matching techniques, enumerated in a previous paragraph.
To exploit the availability of these signal-strength contours, the functionality of SLEN would necessarily be expanded to include 1) logic to choose between signature matching and signal-strength-contour intersection techniques, based on the attributes of the subscriber-supplied wireless signature, and 2) an algorithm to estimate the location of the subscriber device using signal-strength contours. In an example of the former, signal-strength contour intersection is chosen if the number of available contours “fit” to sets of 10 or more points is 3 or more; signature matching, if less than 2. In an example of the latter, the estimate of the subscriber device is determined as the point for which the sum of the squares of the distances to each of the signal-strength contours is minimized.
As indicated above, GPS data can be recorded in a variety of formats, varying from perhaps the most compact format, a 2D GPS fix, to perhaps the least compact format, a digitized recording of the composite GPS signal. In between are numerous possibilities, from the start-of-transmission and the time of arrival (TOA) of a transmission from an identified GPS satellite, to the start-of-transmission for each of two identified GPS satellites, together with the difference in their times of arrival (TDOA).
Mindful that the objective is to map wireless signatures, where the availability of a 2D GPS fix is a rarity, we are compelled to examine other possibilities. If, instead of 2D fixes, measurements are taken at different times and in different places, and with each measurement, the recorded GPS data includes the start-of-transmission and the TOA (measured from the start-of-transmission) from identified GPS satellites, the effect would be to generate spheres of location, centered on the identified satellites, each with radius equivalent to the (computed) distance to said satellite. While the actual 2D locations of these measurements would forever be unknown, the general locations of these measurements would be known by the spheres on whose surfaces they lie. Given an ensemble of spheres each with the same wireless signature, the mapping (or location) of said signature could be accomplished by determining the point for which the sum of the squares of the distances to the surfaces of the associated spheres is minimized. (This approach is equivalent to determining the point for which the sum of the squares of the errors in the associated TOA's is minimized.) In this case, all that is needed is one (not three) acquirable satellites per measurement—and an accurate clock, to measure TOA's. It should be noted that the accuracy required of the clock is severe, and in the general case may make single-satellite measurements impractical. In the special case, where four or more satellites (and, implicitly, an accurate clock) are available, one to three TOA's could be recorded, depending on the extent to which privacy considerations may impose a limit on the amount of third-party location information accessible to a location service provider.
To get around the requirement for an accurate clock, another alternative is available, requiring two acquirable satellites. Consider the case in which measurements are taken at different times and in different places, and with each measurement, the recorded GPS data includes the start-of-transmission for each of two identified satellites, together with the TDOA, the effect would be to generate hyperplanes of location, each with its apex on, and its axis collinear with, the line joining the two identified satellites. If a third satellite were acquirable, and the difference in its time of arrival relative to that of either of the first two satellites were measured, the effect would be to generate two hyperplanes of location for that measurement. While the actual 2D locations of these measurements would forever be unknown, the general locations of these measurements would be known by the hyperplanes on whose surfaces they lie. Given an ensemble of hyperplanes each with the same wireless signature, the mapping (or location) of said signature could be accomplished by determining the point for which the sum of the squares of the distances to the surfaces of the hyperplanes is minimized. Alternatively, the location of the signature may be estimated by determining the point for which the sum of the squares of the errors in the associated TDOA's is minimized. However the estimate is derived, the measurements employed require a minimum of two acquirable satellites—and an ordinary clock.
The simplest approach is to record the composite GPS signal, together with the ID's and signal strengths of any AP's detectable by the instrument/device, and extract the available satellite signals offline. The disadvantage of this approach (occurring when a subscriber device operates as a measurement device) is the bandwidth utilized to transmit the GPS data (also called the datagram); the advantage is the potential to apply sophisticated signal processing techniques to decode long datagrams (hundreds of milliseconds in length), and enhance further the sensitivity of GPS receivers (see below). Using powerful servers, spheres (times of arrival) and hyperplanes (differences in times of arrival), invisible to commodity GPS receivers, are extracted from the datagrams, and posted to the Surfaces of Location Database.
The signals from all GPS satellites are broadcast synchronously, using the same carrier frequency, 1.57 GHz in the case of the NAVSTAR system. However, each satellite has a unique identifier, or pseudorandom noise (PRN) code having 1023 chips, thereby enabling a GPS receiver to distinguish the GPS signal from one GPS satellite from the GPS signal from another GPS satellite. In addition, each satellite transmits information allowing the GPS receiver to determine the exact location of the satellite at a given time. The GPS receiver determines the distance (pseudo range) from each GPS satellite by determining the time delay of the received signal. Given the exact locations and the pseudo ranges, the estimation of 2D location coordinates can be accomplished with as few as two satellite pseudo ranges, provided they have been measured using an accurate time reference. Since this is impractical with current GPS navigational platforms, the computation of 2D location coordinates is generally accomplished using three pseudo ranges. Once the pseudo ranges for at least three GPS satellites have been determined, it is a straightforward process to determine the location coordinates of the GPS receiver.
Indoors, satellite signals suffer severe path losses as they are forced to penetrate windows, walls, and ceilings enroute to the receiver. Commercial buildings, in particular, introduce severe path losses (
To obtain a first fix, GPS receiver 50 must (1) acquire a minimum of three GPS satellites by tuning the local frequency 53 and the code phase of the local PRN code replica 54 in the GPS receiver to match the carrier frequency and the PRN code phase of each of the electronically visible (i.e., decodable) satellites. The search for correlation peaks of sufficient strength to enable the extraction of reliable pseudo range information is a time-consuming process, and failure-prone in indoor and urban canyon environments.
To minimize the time to first fix (TTFF) of GPS receivers such as GPS receiver 50, the concept of a GPS assistance system has been introduced (see
Care must be taken in the generation of GPS assistance data, to insure the integrity of data for satellites at or near the horizon, as these may be the satellites most visible indoors. If GPS assistance data are generated from GPS signals taken in the clear, for example, strong overhead satellite signals could compromise the integrity of GPS assistance data generated for weaker satellites, for example, satellites at or near the horizon.
The potential for strong satellite signals to interfere in the tracking of weak satellite signals is an artifact of the correlation process which serves as the foundation for GPS satellite signal acquisition and tracking techniques. This is illustrated in
To insure the integrity of GPS assistance data for weak or overpowered satellites, exemplified by those at or near the horizon, a global alternative to GPS assistance system 69 is proposed. One embodiment of this Global GPS Assistance System (GGAS) is described in
b describes a minimal implementation of the Global GPS Assistance System of
The utility of GGAS extends beyond conventional thick-client implementations of GPS for sensor location; indeed, GGAS provides an ideal foundation for sophisticated thin-client implementations of GPS for subscriber location, such as the MDLS of
In the compilation of AP Survey Databases utilizing the MDLS of
As described previously, each measurement contains access point information as well as location-related information. In the category of location-related information are the time cue and the GPS datagram. The GPSP of
As shown in
The unconventional element of the GPSP of
To understand the rationale for and operation of PRG 123, consider the impact of the size of a datagram. Correlator 121 and correlators, in general, are better able to detect signals buried in noise when they are able to examine longer datagrams. Signal processing is compute-intensive, and, as a consequence, the limits on the size of datagrams are generally imposed, by limits on available computing resources. Today's thick-client GPS systems are constrained to datagrams of the order of 10 milliseconds. Thin-client implementations, leveraging powerful servers capable of processing datagrams spanning hundreds of milliseconds, open the door to dramatic increases in processing gain, resulting in dramatic reductions in the minimum signal strength required to acquire and track GPS satellites. The key to the efficient realization of dramatic increases in processing gain is the perfect reference.
A Perfect Reference is a waveform representing the signal transmitted by a given satellite (inclusive of its 50 Hz data overlay) as said signal would be observed at a time prescribed by the time cue, and a location prescribed by the location cue. The perfect reference is used to correlate client-supplied GPS data in much the same way as the PRN code replica is used by prior art GPS receivers Using the Perfect Reference constructed fora particular satellite, correlator 121 processes the client-supplied datagram in order to acquire said satellite over periods which can cross multiple low frequency overlay bit boundaries. The resulting processing gain from correlating the client-supplied datagram across multiple bit boundaries has direct bearing on the increased receive sensitivity of GPSP.
The construction of perfect reference waveforms for potentially acquirable satellites is a straightforward process. The first step involves the accurate decoding of the 50 Hz data overlays, enabled by the deployment of GPS receivers so as to insure access to strong signals from each of the GPS satellites. Next, the 50 Hz data overlays are combined with the PRN code replicas for said satellites. Finally, using the satellite positions and trajectories embedded in the 50 Hz data overlays to determine the satellite-specific carrier frequencies, and their rates of change, as well as their relative code phase offsets, in conjunction with the time and optional location cues, the individual waveforms are scaled (in time) and biased to represent the satellite signals as they would be observed at the time and location implicit in the time and location cues.
Accordingly, given a time cue and an optional location cue, the streamed 50 Hz data overlay for each of the potentially acquirable satellites (from GGAS), and PRN code replicas from each of the potentially acquirable satellites, PRG123 generates the perfect references critical to the accurate and efficient extraction of surfaces of location from the received GPS data.
The time cue provides an estimated start time of the GPS data received by the GPS receiver or sensor and the location cue is an estimated location of the receiver. The optional location cue may take the form of coordinates, a wireless access point ID, a cellular base station ID, a ZIP code, or a metropolitan area. The main benefit to using accurate time and location cues is to reduce the amount of computing resources required to process the received GPS data.
This application claims the benefit of U.S. Provisional Application No. 61/458,371; entitled “System Framework for Mobile Device Location”, filed on Nov. 23, 2010. This application further claims priority under 35 U.S.C. §120 from pending U.S. patent application Ser. No. 12/924,618, filed on Oct. 1, 2010, and Ser. No. 12/807,463, filed on Sep. 7, 2010, entitled “System Framework for Mobile Device Location”, a Notice of Allowance for U.S. patent application Ser. No. 12/807,463 having been mailed on May 25, 2011.
Number | Date | Country | |
---|---|---|---|
61458371 | Nov 2010 | US |