RADIO POSITIONING USING OBSERVED TIME DIFFERENCES
PRIORITY This application claims priority from United States provisional application 60/519,593, filed on November 13, 2003, and is hereby incorporated by reference.
BACKGROUND OF THE INVENTION
Field of The Invention
The present invention is related to wireless communications networks and, more particularly, to obtaining the geographic position of a wireless mobile unit.
Background
A number of positioning methods have been defined, and some deployed, to support location services in wireless networks. Using these methods, it is possible to obtain the geographic position, for example the latitude and longitude coordinates, of a wireless mobile unit, for example a mobile cellular telephone, that is being served within a wireless network. Two closely related positioning methods are Enhanced Observed Time Difference (E-OTD), which is applicable to Global System for Mobile Communication (GSM) and General Packet Radio Service (GPRS) wireless networks, and Observed Time Difference Of Arrival (OTDOA), which is applicable to Universal Mobile Telecommunications System (UMTS) wireless networks. With either method, a wireless mobile unit or, as it will sometimes be referred to here, a mobile unit is required to measure the transmission timing reference (e.g. frame number, timeslot and bit number for GSM) from three or more nearby base stations and obtain the transmission timing difference, sometimes called the Observed Time Difference (OTD), between each pair of base stations. Normally, one base station - the "reference base station" - will be common to all pairs of base stations for which OTDs are obtained and, in
most cases, this will be the base station serving the mobile unit. Thus, the mobile unit would normally obtain an OTD value between its serving base station and each of two or more other nearby base stations. The OTD values are then returned to the network, normally to a central network entity, such as a Serving Mobile Location Center (SMLC) in a GSM or GPRS network, where the mobile unit's position is calculated from the OTD values and other information known to the network such as the location coordinates of the base stations.
OTD values are based on measuring timing differences between base stations. Each base station in a wireless network typically maintains a source of time that can be observed by any mobile unit sufficiently close. The time "displayed" by any base station need not be in normal units of time such as minutes and seconds nor in units explicitly related to these such as specific fractions or multiples of a minute or second. More typically, the timing maintained and displayed by a base station is related to the wireless signaling and transmission used by the base station in units fundamental to this. A typical GSM implementation is described as follows. Transmission occurs within separate 200 kilohertz (KHz) frequency bands at a fixed rate of approximately 270.833 Kbits/second. The sequence of bits transmitted at this rate is organized hierarchically into timeslots, each normally containing 156.25 bits, frames, each containing 8 timeslots, and various assemblages of frames the longest of which, the hyperframe, contains 2,715,648 frames. These units of transmission are explicitly numbered. Thus, frames within each hyperframe are numbered consecutively from 0 up to 2,715,647 according to their order of transmission. Timeslots within each frame are numbered consecutively from 0 to 7 and bits within each timeslot are consecutively numbered from 0 up to 156, where bit numbers between 0 and 155 represent whole bits and bit number 156 represents the final 0.25 bit time in a frame. Quarter bit periods are also numbered in each time slot from 0 through 624. Frame numbers are explicitly conveyed by well defined bit fields within the transmission sequence, while timeslot and bit numbers can be inferred (e.g. counted) within each frame. Thus, a mobile unit is able to effectively observe the transmission
timing from any nearby GSM base station at any point in time in terms of the frame number, timeslot number, bit number and fractional bit that is currently being received. Moreover, each of these GSM transmission units has a fixed relationship with normal time - for example, each GSM bit has a precise duration of 48/13 microseconds. An OTD value between two base stations can then be obtained in terms of the difference in frame number, timeslot number and bit number currently observed from each base station. As is known in the art, the difference can also be obtained in terms of just bits and fractions of a bit relative to a single timeslot (where the difference in timeslots and frames is omitted). Alternatively, an OTD may be obtained as the difference in time between the arrival of some known signal from one base station (e.g. the start of a new timeslot or new frame) and the arrival of the same signal from another base station. For other wireless technologies, transmission times and OTD values could be expressed in other units - for example chips for a Code Division Multiple Access (CDMA) network.
To obtain a mobile unit's position from OTD values, the network (e.g. SMLC) normally needs to know the "Absolute Time Differences" (ATDs) between the pairs of base stations for which OTD values are provided as well as the coordinates of these base stations. For any pair of base stations, the ATD is the OTD that would be observed if either the two base stations were precisely co-located or the mobile unit was equidistant from them. The same timing units used to express OTDs (e.g. frames, timeslots and bits for GSM) can thus be used to express ATDs.
To obtain a location from OTD and ATD values, the following derivation, known in the art, can be employed with reference to Figure 1. It should be understood that this derivation is common to a whole class of position methods for different types of wireless networks for which E-OTD (for GSM and GPRS) and OTDOA (for UMTS) are just specific examples. Any positioning method within this class would require any mobile unit being positioned to measure and report OTD values between pairs of nearby base stations as previously described. A positioning method that employs this general principle will be termed an OTD method herein. Any positioning
method that does not rely on measurement or knowledge of observed timing differences at a mobile unit between base stations will be referred to herein as a non-OTD method.
Figure 1 shows a mobile unit 20 served by some wireless network that is able to make OTD measurements between n (n > 3) nearby base stations. In the illustration, consecutively numbered 0, 1 , 2, 3, 4 . . . i base stations are shown. A central network entity 30 (referred to herein simply as a central entity) is also shown which receives the OTD measurements from the mobile unit 20 and may obtain ATD values and position estimates according to the method disclosed herein. In one preferred embodiment of this invention, the central entity 30 is an SMLC.
Let the base stations be numbered 0, 1 , 2, ...n with 0 representing the reference base station
Let P| = Propagation delay for wireless signals from any base station i (0 < i < n) to the mobile unit
Let OTDj = Observed Time Difference between any base station i (1 < i < n) and the reference base station 0 as observed by the mobile unit
= transmission timing of base station 0 - transmission timing of base station i (as observed by the mobile unit)
Let ATDj = Absolute Time Difference between any base station i (1 < i < n) and base station 0
Let GTDj = Geometric Time Difference relative to the mobile unit between any base station i (1 < i < n) and base station 0. This is defined as follows:
GTDj = OTDj between base station i and base station 0 when both base stations are synchronized to a common timing
(i.e. when ATDj = 0) (1 < i < n). For unsynchronized base stations, GTDj can be derived as follows:
(1)
(2)
Let (Xj, yi, Zj) = Cartesian coordinates of base station i (0 < i < n)
And (x, y, z) = Cartesian coordinates of the mobile unit
Equation (2) enables the Geometric Time Difference, GTDj, between the reference base station 0 and any other base station i to be obtained from the measured OTD, OTDj, and the ATD, ATDj, between these two base stations. Assuming the ATDs are already known, the GTD values can then be obtained. Equation (1) then enables the location coordinates of the mobile unit to be obtained from the GTD values using the following known equation which is derived from it as shown.
Pi = [(x - Xi)2 + (y - y 2 + (z - z,)2]172 / c (0 < i < n)
GTDi = Pj - Po (1 < i < n) (from equation
(1 ))
[ [(x - O + (y -
- [(x - XoT + (V - VoT + (z
(1 < i ≤ n) (3)
where c = velocity of light
Solutions for the x, y and z coordinates of the mobile unit can normally be obtained from equation (3) if GTD values are obtained for three different pairs of base stations (thus requiring measurements for a minimum of 4 base stations) - although values for more pairs are preferred to increase accuracy.
Since differences between the vertical coordinates, z and
for the mobile unit 20 and each base station i are commonly insignificant compared to differences in the horizontal x and y coordinates, it is common in the art to exclude these in equation (3). When this is done, solutions for the horizontal x and y coordinates of the mobile unit can be obtained from OTD measurements for a minimum of only 3 base stations.
The calculation of the x and y (and z) coordinates depends on knowing the ATD values between pairs of base stations. If a network is synchronized such that timing at all base stations is identical, the ATD values will all be zero, enabling calculation of the mobile unit coordinates from the OTD values (and base station coordinates) alone. But normally, GSM, GPRS and UMTS networks are not synchronized. Values for the absolute time differences (ATDs) between base station pairs are then normally obtained by fixed measurement units, known as Location Measurement Units (LMUs) for GSM, GPRS and UMTS. These are entities at fixed known locations in the wireless network, that may be part of a base station or may be physically separate, that periodically measure OTD values between pairs of nearby base stations. For LMUs, equations (2) and (3) can be used to obtain the ATD values from both the measured OTDs and the known coordinates of each fixed LMU and each fixed base station.
LMUs, since they are specially designed to make accurate timing measurements, can normally obtain accurate ATD values containing smaller errors than those in the OTD values obtained by mobile units. Thus errors in the position coordinates obtained for any mobile unit will normally be due mainly to errors in the OTD values obtained by the mobile unit rather than due to errors in ATD values.
A disadvantage of obtaining ATD values using LMUs is the cost of developing, manufacturing and deploying LMUs and the additional costs of supporting them by the rest of the network and managing and maintaining them after deployment. Furthermore, for the most accurate positioning, one physical LMU can be needed for every base station in the network. It would
thus be an advantage to develop a method for obtaining accurate ATD values without the need to deploy LMUs, or the need to deploy so many LMUs. It would also be an advantage to develop a method of obtaining an accurate positioning estimate for a mobile unit from OTD measurements made by the mobile unit but without the need to know ATD values in advance.
An OTD method, such as E-OTD or OTDOA, may be supported without the need to deploy LMUs. For example, additional OTD measurements from a plurality of mobile units in the same local area may be used. In one such method, OTD values are needed at nearly the same time from two mobile units located at different positions for the same 5 base stations, which enable the positions of both mobile units to be derived without knowing the ATD values between the measured base stations. In addition, by combining OTD values from many mobile units for pairs of base stations throughout a network, it becomes possible to obtain ATD values between pairs of base stations without LMUs. Once ATD values have been obtained, it becomes possible to position any mobile unit (e.g. using equations (2) and (3)) even when only two OTD values are provided. Thus, while additional OTD values (beyond the minimum of two) are needed from a significant number of mobile units to obtain ATD values, it is not necessary that all mobile units contribute more than two OTD values. However, because OTD values contain higher errors than those in ATD values obtained by LMUs, the accuracy of the ATD values thereby obtained and thus the accuracy of the ensuing position estimates could be worse than those obtained using LMUs. In addition, it is necessary that a significant number of mobile units contribute more than the minimum number of two OTD values. In certain areas of a wireless network where transmission from only a few base stations can be received, for example in a rural area or in a dense urban area, this may not be possible, meaning that some other method would be needed to obtain a position without LMUs.
It would thus be an advantage to develop a method for obtaining a position estimate for a mobile unit using OTD measurements and without
LMUs but without requiring additional OTD measurements from some mobile units and without lower accuracy than is possible using LMUs.
Another role also exists for OTD positioning methods in relation to another positioning method known as Assisted GPS (A-GPS). A-GPS is based on the Global Positioning System (GPS), offers an accuracy of around 5 to 25 meters in many environments and does not require the deployment of additional measurement units like LMUs. As is known in the art, A-GPS has already been deployed for CDMA systems and is being developed for GSM, GPRS and UMTS systems. However, A-GPS can be unreliable and inaccurate in indoor and dense urban environments. Therefore, there is a preference to support a backup positioning method to enhance or replace A- GPS in these environments. For example, for CDMA, a positioning method known as AFLT (Advanced Forward Link Trilateration), that is less accurate than A-GPS, has already been deployed as a backup in certain CDMA networks. AFLT is similar to E-OTD (i.e. provides measurements similar to OTDs) and can be used in either of two ways. If no GPS measurements can be obtained by a mobile unit (e.g. signals from all satellites are too heavily attenuated), then positioning would be performed using AFLT alone. Alternatively, if measurements of some GPS satellites can be obtained by a mobile unit that are insufficient by themselves to obtain an accurate location estimate for the mobile unit, then the GPS measurements can be combined with AFLT measurements to yield an overall more accurate location estimate than either method alone is capable of providing.
For GSM, GPRS and UMTS networks, E-OTD or OTDOA could be used in the same way as a backup for A-GPS. Because E-OTD and OTDOA are normally less accurate than A-GPS, they would not be preferred as a primary method, but could provide measurements that would be combined with GPS measurements to improve accuracy and reliability in a similar manner to that provided by AFLT in CDMA networks. To avoid any expensive upgrade to a network in supporting both A-GPS and E-OTD or OTDOA, it
would be an advantage to develop a method for supporting E-OTD or OTDOA without the need to deploy LMUs.
More generally, a network may support any accurate non-OTD positioning method, including but not limited to A-GPS, and a second OTD based method, including but not limited to E-OTD or OTDOA. The non-OTD positioning method might normally be used alone to obtain an accurate position estimate without the need to invoke the OTD method. But if the non- OTD method is not always reliable - for example, cannot always deliver an accurate position estimate in certain environments - then the network could use the OTD method when needed to obtain a location estimate either alone or in combination with the non-OTD method. Any method to support the OTD method without LMUs would then be of benefit to providing a more reliable and accurate positioning capability.
SUMMARY OF THE INVENTION
It is an object of the present invention to provide a system and method for obtaining accurate ATD values without the need to deploy LMUs, or the need to deploy so many LMUs.
It is another object of the present invention to provide a method of obtaining an accurate positioning estimate for a mobile unit from OTD measurements made by the mobile unit without the need to obtain ATD values using LMUs.
It is yet another object of the present invention to provide a system and method for obtaining a position estimate for a mobile unit using OTD measurements and without LMUs and without requiring additional OTD measurements from some mobile units and without lower accuracy than is possible using LMUs.
It is yet another object of the present invention to reduce the expense associated with upgrading an existing network to support E-OTD and/or OTDOA when the network already supports GPS or A-GPS.
The present invention obtains a position estimate of a wireless mobile unit in a wireless communications network by measuring observed time differences (OTDs) between pairs of base stations and without the need to measure absolute time differences (ATDs) between base stations using additional equipment (e.g., Location Measurement Units). Mobile units are positioned using a second position method, for example GPS, that does not require the use of OTDs. Such positioned mobile units also provide OTD measurements. From the OTD measurements and the position estimates obtained using the second position method, ATDs are obtained between pairs of base stations. Other mobile units for which the second position method is either not applicable or not accurate in their current environment may be positioned using measured OTDs and the obtained ATDs
The invention is applicable to wireless networks such as GSM, GPRS and UMTS networks. It may be used where an accurate non-OTD positioning method such as A-GPS is deployed and where a second OTD positioning method, such as E-OTD or OTDOA, is deployed. In a preferred embodiment there are a significant number of mobile units in the network supporting both position methods. In an alternative preferred embodiment, the non-OTD method is not dependent on any support by mobile units. The invention supports OTD methods like E-OTD and OTDOA without the need for LMUs. The invention enables position estimates to be obtained for mobile units using the OTD method alone and without LMUs. The invention also enables the OTD method to be combined without LMUs with the non-OTD method to provide more accurate and reliable positioning whenever the non-OTD method is not able by itself to provide an accurate position estimate. The invention does not require provision of extra redundant OTD measurements from mobile units.
In a preferred embodiment of this invention, mobile units themselves are used as LMUs when an accurate position estimate can be obtained using the non-OTD method. For a network supporting positioning of a large number of mobile units, a significant fraction are likely to be in "good" environments
where accurate positions can be obtained using the non-OTD method - for example outdoors in the case of A-GPS. Whenever this occurs and the mobile unit is able to support the OTD method, the network may request OTD measurements from the mobile unit. Knowing the position of the mobile unit from prior application of the non-OTD method, the network would be able to treat the mobile unit like an LMU and obtain or update values for ATDs being base stations.
For any mobile unit being positioned, the network has the following options to obtain a position estimate.
(a) Obtain a position estimate using both methods. (b) Obtain a position estimate using the OTD method only. (c) Obtain a position estimate using the non-OTD method only.
In the event that it is possible to obtain an accurate location estimate using only the non-OTD (option (c) above), then the network may invoke the OTD method if supported by the mobile unit and use the OTD measurements provided by the mobile unit to obtain or update ATD values between pairs of ( base stations. The ATD values thus obtained or updated can then be used to help determine location estimates for other mobile units using the OTD method, either alone or in combination with the non-OTD method (option (a) or (b) above). In this way, use of the non-OTD method (option (c)) for some fraction of mobile units helps support the OTD method for other mobile units.
BRIEF DESCRIPTION OF THE DRAWINGS
Fig. 1 is an illustrative example of OTD measurements for a mobile unit with respect to surrounding base stations.
Fig. 2 is a flow diagram showing an example of a method for obtaining ATD values without LMUs in accordance with a preferred embodiment of the invention.
Fig. 3 is a flow diagram showing an example of an OTD method for obtaining the position of a mobile unit using derived ATDs in accordance with a preferred embodiment of the invention.
Fig. 4 illustrates the theoretical maximum error in GTD due to an error in estimating the position of a mobile unit.
Fig. 5 illustrates errors in propagation delay for a mobile unit relative to a base station and associated with errors in the position estimate for the mobile unit.
DETAILED DESCRIPTION OF THE INVENTION
Figure 2 shows an example of a flow diagram 100 for obtaining or updating ATD values between pairs of base stations.
Flow chart 100 may be applied in any network where both position methods, OTD and non-OTD, are applicable to at least some proportion of mobile units, such as mobile unit 20 shown in Figure 1. In a first step 102, a request is received within the network to position a mobile unit. The request could be generated by some client user outside the network, for example by a roadside assistance service, or by a client user within the network, for example an application performing billing, or by the mobile unit itself, for example instigated by the user of the mobile unit. The request could also be generated within the network for the specific purpose of obtaining or updating ATD values rather than obtaining the position of the mobile unit. Next, in step 104, the network, for example central entity 30, verifies whether both the non- OTD and the OTD position method can be used to obtain a position estimate for this mobile unit. This would include verifying whether the mobile unit supports the OTD method and, if applicable, the non-OTD method. If at least one method cannot be used (e.g. is not supported by the mobile unit), the mobile unit cannot be used to obtain or update ATD values and the procedure ends in step 106. Otherwise, in step 108, the network, for example central entity 30, instigates positioning of the mobile unit using the non-OTD method
to obtain, if possible, an accurate position estimate. In the case of A-GPS for example, this would include requesting GPS measurements or a GPS computed position estimate from the mobile unit. Then in step 110, the network, for example central entity 30, verifies whether a position estimate was obtained using the non-OTD method and whether this is accurate. In the case of A-GPS, for example, as is known in the art, the accuracy of a position estimate can be estimated with some reliability. In the case of no position estimate or an inaccurate position estimate, the procedure terminates in step 106. Otherwise, in step 112, the network, for example central entity 30, instigates OTD measurements from the mobile unit using the OTD method. Step 112 may be performed before, after or at the same time as step 108. For example, if steps 108 and 112 are performed at the same time and the non- OTD method requires support by the mobile unit, the network, for example central entity 30, might send positioning instructions to the mobile unit to perform both methods simultaneously. This is allowed for example in GSM, GPRS and UMTS networks when the non-OTD method is A-GPS and the OTD method is E-OTD or OTDOA. Finally in step 114, the network, for example central entity 30, uses the OTD measurements obtained in step 112 and the position estimate obtained from the non-OTD method in step 108 to obtain or update ATD values between pairs of base stations. If ATD values had earlier been obtained for some pairs of the base stations, the newly derived ATD values would be used to update them, for example using a running weighted average. The position estimate obtained as a result of step 108 would also be returned to any requesting user or application.
The derivation of the ATD values in step 114 could make use of the equations (2) and (3) above. First, equation (3) could be used to obtain GTD values relative to the mobile unit between pairs of base stations based on the known positions of the base stations and the accurate position estimate for the mobile unit obtained in step 108. Then, equation (2) could be used to derive an ATD value between each pair of base stations for which the mobile unit had provide an OTD measurement in step 112. In using equation (2),
the already obtained GTD values for pairs of base stations relative to the mobile unit would be applied.
It should be clear that the example flow diagram 100 is applicable for any non-OTD method including but not limited to A-GPS. Examples of other non-OTD methods include, but are not limited to, conventional GPS, pattern matching techniques and GALILEO, a satellite based positioning system being developed in Europe.
Figure 3 shows an example of a flow diagram 200 for making use of ATD values (obtained, for example, according to the flow diagram 100) for obtaining the location of a mobile unit using an OTD method such as E-OTD or OTDOA. The flow diagram 200 may be used together with the flow diagram 100 within a network as explained herein below.
In a first step 202, a request is received within the network to position a mobile unit. The request could be generated due to any of the conditions described in association with step 102 in flow diagram 100. Next, in step 204, the network, for example central entity 30, verifies whether the non-OTD method can be used for positioning. If not, step 212 described further down is executed. Otherwise, the network, for example central entity 30, instigates non-OTD positioning of the mobile unit in step 206 to obtain, if possible, a position estimate with the needed accuracy. Then in a step 208, the network, for example central entity 30, verifies whether a position estimate was obtained and whether it is accurate enough for the requesting user. If so, the procedure can terminate in step 210. If not, in step 212, the network, for example central entity 30, verifies whether the mobile unit can support the OTD method and whether ATD values are available to support the OTD method should the mobile unit support it. If not, the procedure terminates again in step 210 without a sufficiently accurate position estimate. Otherwise, in step 214, the network, for example central entity 30, instigates OTD measurements by the mobile unit for the OTD method. Next in step 216, the returned OTD measurements are combined with ATD values, obtained for example according to the flow diagram 100, to obtain a position estimate for
the mobile unit. If measurements or a position estimate were also obtained in step 206 for the non-OTD method but were not accurate enough for the requesting user, then a position estimate may be obtained in step 216 by combining the OTD measurements obtained in step 214 with both the ATD values, obtained for example using the flow diagram 100, and the measurements or position estimate obtained for the non-OTD method in step 206.
In a variation of the flow diagram 200, steps 212 and 214 might occur before steps 204 and 206 or at the same time. For example, both the OTD method and non-OTD method, if both supported by the mobile unit, might be instigated at the same time in the mobile unit. In another variation of the flow diagram 200, steps 204, 206 and 208 might be omitted - for example, if the non-OTD method was temporarily unavailable or if the needed location accuracy could be supported by the OTD method alone. Furthermore if the non-OTD method is found to deliver an accurate position estimate in step 208, then instead of terminating in step 210, the procedure could invoke the flow diagram 100 starting at step 104 and skipping steps 108 and 110 in order to obtain or update ATD values. Similarly, if step 106 is reached in the flow diagram 100 due to either both position methods not being applicable or an insufficiently accurate position estimate being obtained using the non-OTD method, flow diagram 200 could be invoked starting at step 204 to obtain a position estimate for the mobile unit using either the OTD method or, if not already tried, the non-OTD method.
The method disclosed herein, for example in one preferred embodiment the flow charts 100 and 200, can be implemented by software resident in one or more network entities, for example in the central entity 30.
The reliability and accuracy of the ATD values derived as described hereinabove, for example using the flow diagram 100, is now evaluated. As is known in the art, the reliability and accuracy of GPS measurements can be assessed both from information obtained by the mobile unit during measurements and the degree to which different GPS satellite measurements
are consistent with a single position estimate. Thus, the network can evaluate whether GPS measurements alone will produce an accurate location estimate, for example one with a 67% probability of an error of less than 25 meters.
Suppose that a GTD value for two base stations relative to a certain mobile unit is obtained using equation (3) from a location estimate for the mobile unit that contains an error of E meters. This error will give rise to some other error in the derived GTD value. The theoretical maximum error in the GTD will occur when and only when both the mobile unit and its location estimate lie on the straight line connecting the two base stations as shown in Figure 4. In the example of Figure 4, the correct value for the GTD is (Pj - P0) where Pi and P0 are the propagation delays from each of the two base stations to the mobile unit. If the location error vector from the mobile unit to its location estimate has length E, the correct GTD value will be reduced by an error factor of 2E/c due to an apparent increase in Po of E/c and an apparent reduction in Pi of E/c. For an error vector of length E in the opposite direction, the correct value for the GTD would be increased by this error factor.
If the mobile unit or its location estimate do not lie on the straight line connecting the two base stations, the error in the GTD will be reduced because either the apparent change in each propagation delay is reduced or, in the special case that the mobile unit and location estimate lie on an extension of the straight line joining the two base stations, the apparent changes cancel or partially cancel. In fact, there is a known quantity in location services known as the Geometric Dilution of Precision (GDOP) (so named because it depends only on geometry) which relates an error in a positioning measurement, when converted into units of length, to the resulting error in a location estimate according to:
GDOP= resulting location error/ error in measurement (4)
= resulting location error / (error in GTD * c) (5)
Equation (4) provides the GDOP for an OTD positioning method when there is some error in a GTD value. As is known in the art, theoretical GDOP values for OTD methods range from a minimum of 0.5 (illustrated hereinabove in Figure 4) to a typical maximum of around 5. Thus, when a GTD value is obtained from a location estimate, for example in step 114 in the flow diagram 100, the error in the GTD value will range from around 0.2 E/c to 2 E/c for a given error E in the location estimate. If the non-OTD method used to obtain the location estimate is A-GPS or GPS, a location estimate would normally be possible with an error of at most 25 meters in good conditions (e.g. where a mobile unit is outdoors). This means that the ensuing error in the GTD for any pair of base stations relative to the mobile unit would normally be in the range of 0.017 to 0.17 microseconds (μs) at most.
The error in any OTD measurement in the best case will be around 0.13 μs when the OTD method is either E-OTD or OTDOA. This is based on the minimum measurement resolution for each positioning method which is normally around 1/32 of a bit for GSM and 0.5 chips for OTDOA. However, higher errors may result - e.g. due to multipath effects, interference and noise.
Statistics for the magnitude of the errors in the ensuing ATD values are now derived for the case where many ATD values for a specific pair of base stations, consisting of the reference base station 0 and another base station i in Figure 1 , are obtained from position estimates and OTD measurements for many mobile units, and where each ATD value is derived from one OTD value for the pair of base stations and one GTD value, as described herein above.
Let X = true value of any quantity X which may be a GTD, OTD, ATD or propagation delay X* = approximate value of X containing measurement errors
e(X*) = error component in X* X* - X E[e(X*)] = mean error in X* over many values for X* obtained for different mobile units in different locations σ[e(X*)] = standard deviation of the error in X* over many values
Assuming that positive and negative OTD errors of any magnitude occur with approximately the same probability, then the mean OTD error will tend to zero. Thus, for an OTD, OTDj, between the base station i and reference base station 0 the following can be assumed:
E[e(OTDj*)] 0 (1 ≤ i ≤ n) (5)
If similarly the error in a location estimate for any mobile unit is such that for any magnitude of error, a particular direction for the error vector occurs with the same probability as for the opposite direction, then the error in the approximate propagation delay, Pj*, calculated between a distant base station i and any mobile unit can be shown to satisfy:
E[θ(P,*)] = 0 (O ≤ i ≤ n) (6)
Equation (6) can be demonstrated in association with Figure 5 which shows a base station A and a mobile unit O for which many location estimates are obtained with errors of the same magnitude e, and thereby lying on a circle with center O and radius e.
With respect to Figure 5, an arbitrary location estimate for the mobile unit, with magnitude error e, is shown at point B. Corresponding to B, there is another location estimate at point C directly opposite on the circle to B which theoretically must occur with equal probability due to the prior assumption. If the mean value of the distances AB and AC can be shown to equal the distance AO, then the mean value of the distance between the base station at A and all location estimates with magnitude error e will also be equal to AO (since B is arbitrary). Since e is arbitrary, the same would apply for any value
of e, implying that the mean distance "between the base station A and all location estimates for the mobile unit at O will equal the true distance AO between the mobile unit at O and the base station at A. As propagation delay is proportional to distance, this will demonstrate equation (6) with respect to any one position for a mobile unit and thence with respect to all positions, provided none are too close to the base station.
To show the mean of AB and AC is AO, proceed as follows referring to Figure 5 in which some additional lengths are shown and where it is assumed that the distance s from the base station to a point inside the circle of location errors is much greater than the error e (and thus much greater than the other distances h and y).
AB = (h2 + s2)1 2 = s (1 + Vz (h/s)2)
AC = (h2 + (s + 2y)2)1 2 = (s + 2y) (1 + Vz (h / (s + 2y))2) (s + 2y) (1 + Vz (h/s)2 (1 / (1 + 2y/s))2) = (s + 2y) (1 + Vz (h/s)2 (1 - 4y/s))
[AB + AC] / 2 = [ (2s + 2y) + h (h/s) + Vz (h/s)2 (2y - 8y(y/s) - 4y) ] / 2 = (s + y) + (h/2) (h/s) AO + (h/2) (h/s) (7)
Equation (7) shows that the error in the original assumption, that the mean of the distances AB and AC is equal to AO, is of maximum order (e/2) (els). For a small ratio (els) given when the location estimation error e is much smaller than the distance s to the base station, this error can be neglected. This leads to equation (6) as described above.
Equation (6) allows the expected error in the GTD, GTDj, for base stations i and 0 relative to location estimates for many mobile units to be obtained as:
E[e(GTDj*)] E[e(Pj* - Po*)J (1 < ≤n) E[(Pι* - Po*) - (Pi - Po)] (1 ≤ ≤n) E[(Pι* - Pi) - (Po* - Po)] (1 < ≤n) E[e(Pj*) - e(P0 *)] (1 < ≤n) E[e(Pi*)] - E[e(P0 *)] (1 < ≤n) 0 (1 < ≤n) (8)
Equations (5) and (8) imply the following for the ATD value derived from a measured OTD and a GTD value obtained from a measured location estimate.
E[e(ATDj*)] = E[e(OTDj* - GTDj*)] d ≤ ≤n) E[(OTDj* - GTDj*) -(OTDj • ■ GTDi)] ≤ <n) E[(OTDi* - OTDj) - (GTDi* ■ ■ GTDi)] ≤ <n) E[e(OTDj*) - e(GTDi*)] (1 ≤ <n) E[e(OTDi*)] - E[e(GTDi*)] O ≤ <n) 0 0 ≤ <n) (9)
Equation (9) shows that the expected error in an ATD value (i.e. considering both positive and negative errors) will be zero. This does not mean that errors are necessarily insignificant, only that positive and negative errors cancel out on average. To evaluate the error in this case (i.e. with an expected value of zero), it is normal to consider the absolute magnitude of the error ignoring any sign using its root mean square which will also equal its standard deviation, as is known, due to equation (9). This leads to the following.
Root mean square of error in ATDi = E [e(ATDf **2 )\ι] 1/2 (1
≤i≤n)
σ[e(OTDj*) - e(GTDi
*)] (1 < i < n) [σ [e(OTDj*)]
2 + σ [e(GTDj*)]
2]
2 (1 ≤i≤n)
(10)
Equation (10) follows from a well known result in statistics for the case where errors in OTD values and errors in GTD values are independent. This latter assumption can be justified because OTD values and GTD values are each obtained using different positioning methods (an OTD method for OTD values and a non-OTD method for GTD values). As discussed hereinabove, theoretically the worst case error in a GTD value will be around 0.17 μs, whereas the best case error in an OTD will be around 0.13 μs. This implies that the corresponding standard deviations will be less than 0.17 μs for the GTD and more than 0.13 μs for the OTD. The impact of this on the standard deviation (equals root mean square) for the ATD is as follows.
Let m = σ [e(OTDi*)] / σ [e(GTDj*)]
Then σ [e(ATDj
*)] = σ [θ(OTD|
*)] [1 + 1/m
2]
1/2 (1 ≤ i ≤ n) (11) = σ [e(OTDι
*)] [1 + 1/(2m
2)]
(12)
Equation (12) follows when OTD errors are significantly greater than GTD errors (i.e. when m is large). In this case, the error in a derived ATD will be approximately the same as the error in the OTD value from which it was derived. In the best case when OTD errors are close to their absolute minimum, OTD and GTD errors will be approximately the same, giving m roughly equal to 1 and, from equation (11), the error in an ATD would be roughly root 2 times (i.e. roughly 1.4 times) an OTD or GTD error.
Derived ATD values may, however, be averaged to yield a more accurate value. By averaging ATD values the error will trend to zero as demonstrated in equation (9). In particular, while the error in any individual ATD value will be approximately the same as an error in an OTD value, averaged ATD values will be much less. This means that when averaged ATD values are employed in equations (1 ) to (3) to obtain the position of a mobile
unit from OTD values measured by the mobile unit, the dominant error in the GTD value obtained from equation (2) will be due to the OTD error and not the ATD error. This implies that the accuracy of location estimates for any OTD method, for example E-OTD or OTDOA, obtained using this invention will be close to that obtainable when the OTD method is used together with LMUs to obtain ATD values.
To average individual ATD values, one of two approaches can be taken. The individual values may be accumulated and stored in the network and averaged using a simple arithmetic average either when a predetermined number of values have been collected or after a predetermined interval of time has elapsed. This method is applicable when ATD values are stable and change only slowly over time. When ATD values may occasionally change rapidly, for example due to re-initialization of timing at a base station, a single running weighted average is more appropriate obtained using the following equation.
Let ATDn = n.th individual ATD value for any pair of base stations (n > 1 ) ATDn = running weighed ATD average obtained from
ATDi ATDn
Then ATD. = ATD! (13)
And ATDn+1 = (1-w) ATDn + w ATDn+. (n > 1) (14) where (0 < w < 1)
The value of the weight w in equation (14) should normally be small (for example between 0.01 and 0.1) in order to reduce errors when the true ATD value is stable.
In the case that the error in each individual ATD value can be estimated - for example, by means of an estimation of the errors in the OTD value and position estimate for a mobile unit from which it was derived - then a variable weight w may be used in equation (14) dependent on the error. For
example ATD values with smaller estimated errors might be assigned higher weights than those with higher errors. If the standard deviation for the errors in the current average ATD value, ATDn, and each individual ATD value, ATDn+ , in equation (14) are known or can be estimated and individual ATD values are independent of one another, then the following weight assignment in equation (14) will minimize the standard deviation of the error in weighted averaged ATD value.
w = (1 / σ[e(ATDn+1 *)]2) / (1 / σ[e(ATDn+1 *)]2 + 1 / σ[e(AIDn *) ) σ[e(AJDn *)]2 / (σ[e(ATDn*)]2 + σ[e(ATDn+1*)]2) (15) giving: σ[e(ATDn+ι*)] = (1 / 0 / σ[e(ATDn+1 *)]2 + 1 / σ[e(AJDn *)]2)) % (16)
Equation (16) shows the resulting standard deviation for the error in the new weighted average ATD value, ATDn+ι, following application of equation (14). Equations (15) and (16) follow from known results in statistics for independent random variables and the fact that the variance (standard deviation squared) of any individual or weighted average ATD value is the same as the variance of the error in this value provided the true ATD value is constant.