fh
17 International Conference on Applied Electromagnetics and Communications
34 1 - 3 October 2003, Dubrovnik, Croatia
A GPS Based Navigation Aid for the Blind
Wamadeva Balachandran, Franjo Cecelja, Piotr Ptasinski
Systems Engineering Department, Brunel University, Uxhridge, Middlesex, U.K., UB8 3PH
Phone: +44 1895 274 000, ext. 2925, Fax: +44 1895 812 556
Email:f,.[Link]@[Link]. irk
Abstract
111 this paper we present develapmei~tof the GPS based navigation uid,/ur blind. That aid utilizes inverse DGPS
positioning module which, along wit11other hen& for the blind user, enables implementati~mo/mure cutnplu
navigation algorithms. In particular, we have implemented DGPS algorithm angmented by altiiude data avail-
able from digital map. This solution improved the availabilin/ ofpositioning by more that 23% and its accuracy
by more than IS%. AI1 the results were [Link] verified.
1 INTRODUCTION 2 DESCRIPTION OF THE SYSTEM
A navigation system for blind and visually im- The Brunel Navigation System for the Blind is
paired pedestrians is being developed in the Electronic based on inverse DGPS positioning with mobile unit
System and Information Technology Research Group, and stationary base station, namely navigation center,
at Brunel University [ I ] with the main aim to increase linked using cellular mobile network, as shown in Fig.
their mobility. The main positioning module utilizes a I . Pseudo-range measurements from the GPS receiver
Differential Global Positioning System (DGPS), placed within mobile unit is sent via general packet
which is the Global Positioning System (GPS) oper- radio service (GPRS) terminal of the mobile phone to
ated in a differential mode, in order to provide more the navigation center. The base station calculates posi-
acceptable positioning accuracy for the blind user tion offset using DGPS correction, which is either
PIPI. generated by the stationary GPS receiver with its an-
Both GPS and DGPS, however, share the same tenna position surveyed with an accuracy of better
problem of insufficient availability in signal-blocked than 0.5 in or received from a commercial provider
environments [4]. In urban canyon areas, the availabil- through UHF radio link. The base station also calcu-
ity and accuracy of GPS can be degraded due to lates the position using altitude information retrieved
blocking of the satellite signals. More precisely, dur- froiii digital datasets (digital map). The position of the
ing the normal GPS receiver operation, pseudo-ranges user is then graphically displayed on the digital map,
from at least four GPS satellites are needed to solve which is then used by the operator to provide naviga-
the equations for the receiver’s antenna position be- tional information in a form o f spoken messages trans-
cause of the four unknowns: three position unknown ferred to the user via GSM communication channel.
x , y and z and the receiver’s clock bias Ar [5]. If
less than four satellites are available, or the geometry
of satellites is poor, the GPS positioning error rapidly
increases.
a_C*II)
In this paper we present solution, which increases
both the GPS positioning accuracy and availability. &
-d, cm_+
MI
We have implemented inverse DGPS algorithm y_n
which, because of its centralized architecture, allows Mobile Unit
for more complex navigational algorithms to he used. n a y station
In particular, it is two-dimensional (2-D) DGPS algo- Fig. 1. Brunel inverse DGPS system
rithm [Link] altitude data available from
Along with better positioning accuracy and avail-
map datasets replacing pseudo-range from one GPS ability, such a configuration entails additional bcnefits
satellite. Consequently, DGPS navigation solution
to the users. Placing the entire responsibility of operat-
requires pseudoranges from only three satellites to he
ing and maintaining the system with a sighted opcrator
available, hence, increasing the positioning availability
in the navigation ccntre minimizes the level of mental
for more than 20%. Alternatively, when more than
demand imposed on the user during navigation proces.
four satellites are available, the positioning processor
Additionally, availability of the remote sighted guide
can sclect only those threc satellites that will yield the
reduces the travel related stress experienced by the
best possible navigation solution, hcnce increasing the
visually impaired pedestrians in the same way as the
position accuracy [SI. real sighted guide, still allowing for considerably
higher level ofjoumey independence.
!c@@m %@ lfhlnternational Conference on Applied Electromagnetics and Communications
1 - 3 October 2003, Dubrovnik, Croatia 35
3 INVERSE DGPS ALGORITHM effectively a distance from the centre of the Earth to
The GPS receiver in the mobile unit uses pseudo- the user's antenna, the data that is available from the
range measurements PRM to each satellite, corrected map dataset. Then the fourth equation in the system of
by ionosphere and troposphere corrections ATM cal- equations (2) takes the form
culated by using standard single frequency models [6]. liI1CP.S- I
c, = IC, - u l>(i17 I (7)
This data is then sent to the base station where pseu- where i = ( G t , i v , i y ) is the estimated initial user po-
dorange measurements are correctcd using DGPS
pseudorange correction PRC for each satellite as: sition and c,?is the centre of the sphere representing
PRD, = P R M , PRC,+ (1) the altitude aiding equation. For simplicity, further in
this paper we will denote this method simply as 2-D
where PRDj is the corrected pseudorange measure-
solution. However, the shape of the Earth is not
ments for j-th satellite. spherical; instead elliptical WGS-84 model is used. To
To determine the user position in three dimen- correct for this difference, we placed the centre of the
sions u = ( x , y , z ) and the receiver clock offset t , sphere at the centre of the ellipsoid. The sphere radius
pseudorange measurements are made to four satellites, was the distance between the centre and the predicted
which results in four equations with four unknowns point location. We used computer simulation and
X , y , z and t of the following form; changed the position ofthe point on the sphere within
the distance of f 100m. Observing the maximum dis-
PRD, =I S , -~m;psI+cfnmy (2) tance from the sphere over the reference WGS-84
where j = 1-4 indicates the j-th satellite, s,! is the ellipsoid wc found that for the point with the latitude
position of the j-th satellite at the time the pseudo- 45'N, longitude O', which is the most inaccurate sce-
range measurement was taken, u ~ ~ ,is, a. DGPS
~ posi- nario, the change from [Link] distance was not
more than 0.4 m.
tion solution, c is the speed of light and is a
4 RESULTS
GPS receiver clock offset. The DGPS position is then
calculated using linear approximation [7], which re- For the experimental verification of the developed
sults in the following set oflinear equations: inverse DGPS algorithm and altitude augmentation we
used system shown in Fig. I and the Land-Form
AP = HAu i x x - m s (3)
PROFILE map in 1:10,000 scale from Ordinance Sur-
= H-'Ap
AU,,~;,,~.~;,,~ (4) vey. The high accuracy of this map dataset depends on
where Auocs,,s-~;i.r-
is a vector between the DGPS and the nature of the ground and is typically better than the
half of the vertical interval of the source contour data,
GPS position solution, H is a cosine direction matrix
and which is either 5 or IO m.
For the purpose of accuracy measurement the per-
son with mobile unit walked along the marked route.
Ap=l 1$ 1I=/
14%
ATM, - PRC,
A T M , -PRC,
LA TM.4 - PRC.4
iJ (5)
The route was carefully selected to simulate a 'typical'
urban area with parts where signals from the satellites
are very likely to be blocked. Some points along the
route were surveyed with accuracy better than I m.
If the PRC correction value for the j-tb satellite Those points were used for calculation of the accuracy
of the system. In order to account for different constel-
is not available, it is assigned to be ATM,, hence pro- lation of GPS satellites, the experiments were repeated
viding Ap, =ATM,-PRC,=O. four times at different time of the day.
Along with the accuracy measurements, we havc
In order to minimize the error contribution using also assessed the availability of the positioning. The
redundant measurements to more than four satellites, positioniug service was considered available whcn
which results in an overdetermined solution set of GPS receiver tracked at least 4 satellites for 3-D type
equations, the position calculation can be processed by of positioning solution or when using at least 3 satel-
least square estimation techniques, which leads to the lites for 2-D type of solution. The elevation mask was
solution [ 7 ] ; set up to 5' as a commonly used value for perform-
A " , m . m =(H")-"'AP (6) ance assessment [5].In all the experiments, the thresh-
However, if only three pseudorange measure- old for the dilution parameter, was set up to 5 for
ments to different satellites are available, then the sys- Horizontal Dilution of Precision (HDOP). Selecting
tem of equations (2) is under-determined, and only this value should maintain the horizontal accuracy of
solution to two dimensional positioning is available. If 30 m and 20 m for GPS and DGPS systems through
range measurement from additional sensor or source is 95% of the time.
available that is exactly in the vertical direction rela- The accuracy was expressed using the most com-
tive to the earth surface, this is then equivalent to hav- monly used accuracy measures: . ~ ' '
ing a satellite at the centre of the Earth [SI hence still 0 RMS - the square root of the average of the
enabling 3-D positioning. This range measurement is squared errors;
I”
17 Internat onal Conference on Appliea [Link] and [Link]
36 1 . 3 October 2003. DLbrovn k, Cr0at.a
CEP - a circle’s radius centered at the true posi- having in mind higher number of satellites available
tion and containing 50% of the points .in the for 2-D solution.
horizontal scattered plot; Table 2 The comparison of the accuracy for 2-D
R95 - a circle’s radius centered at the true posi- and 3-D positioning
tion and containing 95% of the points in the
horizontal scattered plot.
For the reason of comparison, all the experiments
were carried out with the system operating in both 2-D GPS
GPS and DGPS mode. 2-D DGPS
Fig. 2 shows typical experimental results plotted
on the corresponding Ordinance Survey Land-Line.
The reference path obtained from surveying points
coincides with the true walking path on the map
within 0.5 m.
Table 1 summarizes results of the assessment of 80
the accuracy when both 2-D and 3-D positioning were
available. It becomes apparent that 2-D positioning
-
5
always gives better accuracy in both GPS and DGPS -.-aE 6o
modes for at least 15% This is primarily due to the -
fact that the 2-D algorithm has higher number of op-
‘5 40
4
tions in selecting the best positioning solution (see
equation (6)). 20
n I
0 2 4 6 8 10
HDOP
Fig. 3. Service availability
.5 CONCLUSION
Performance of the GPS and DGPS positioning
. -.can be significantly degraded due to the blockage of
-signals from the satellites. Inverse DGPS configura-
tion offers an efficient solution to improve both the
accuracy and availability of the positioning as it en-
ables implementation more complex navigational al-
gorithms. It is also apparent that the navigation aug-
. Fig. 2. The results plotted on the map mented with altitude data from map datasets increases
Accuracy of 2-D positioning when 3-D position- the positioning accuracy for I5%, and positioning
ing was not available are shown in Table 2. It is ap- availability for at least 23%.
parent that positioning accuracy maintained the level
of the accuracy of that when 3-D positioning was 6 REFERENCES
available.
[I] V. Garaj. R. Jirawimut, P. Ptasinski. F. Cccelja. W. Bala-
Table 1 The comparison of the accuracy for 2-D chandran: The System far Remote Sighted Guidance of Visually
Impaired Pedestrians, acccptcd for publication in The British Jour-
and 3-D positioning nal o f v i s u a l Impairment. publication expected in M a y 2003.
[2] Ptasinski, P. Cecelja F. Balachanfran W, A n Asscssmenl of
DCPS Performance in Personal Navigation and Location Systems.
i n I O N GPS 2000.2000. Salt Lake City, Utah, USA.
131 Ptasinski, P., et al., Btsnel Inverse DGPS Pmilioninbr S.v~rrni.
3-D GPS NAVIGATION, Journal of the Institute of Navigation, USA (ac-
2-D GPS cepted).
[4] R. Jirawimut, P. Ptasinski, M. A. Shah. V. Gamj. F. Cecelja.
3-D DGPS 12.7 W. Balaehandran: A Method for Dead Reckoning Parameter Cor-
2-D DGPS 5.1 3.2 9.7 rcclion in Pedestrian Navigation System, IEEE Trans. on Instrumcn-
tation and Measurement. voI I, January 2003
[SI [Link], F. Cecelja, [Link], Altitude Aiding for
Availability of the service plotted against the GPS Systems using Elevation Map Datasets, The Journal o f Naviga-
value of HDOP for both 2-D and 3-D positioning so- tion - The Royal Institute o f Navigation, England, Vol. 55. pp. 452-
462 A u y s t 2 0 0 2
lution are shown in Fig. 3. It becomes apparent the 161 U.S. Department o f Defense, Tcchnical charactcristics of the
difference in availability of 2-D and 3-D service is NAVSTAR GPS (public release version), June 1991
[7] Kaplan. E. D. ed., Understanding GPS, principles and applica-
significant: 2-D service is improves the availability for tions, Artech House, 1996
more than 23%, from 63% to 86%. This is expected [SI Parkinson B W,Spilker J J, Global Positioning System Theory
and Application, vol. 2 American Instirule o f Aeronautics and Ar-
tranautics, 1996
