This graph shows the heading/time relation as derived from the Canadian Bureau of the Census data.
In increasing numbers, visually impaired persons (VIP's) make use of the Global Positioning System (GPS) to improve their orientation. Next to the white cane and the guide dog, the GPS makes it possible to guide the way in known and unknown areas and informs the user of his or her position on the planet. However, many commercial available GPS units are not suited for low speed pedestrian use. A small number of units now have been made available addressing the need of VIP's, among which the Trekker, the Wayfinder and Loadstone GPS. Ultra portable, running on PDA or mobile telephone, these units provide spoken directions and information about the environment.
Until now, the users focus has been on the software packages that provide the spoken information and little attention has been given to the GPS receivers that translate the satellites signals into usable data for this software. Still, a navigational unit is as good as the attached receiver.
The Loadstone team,VIP's themselves and makers of the Loadstone GPS navigational software that runs on the Symbian operating system as found on Series60 mobile phones made by Nokia has followed the development of GPS receivers closely. With the introduction in 2006 of the SiRF3 chipset a vast improvement of the accuracy and reliability of GPS receivers was established and many Loadstone GPS users make use of receivers equipped with this chipset. Feedback from the active Loadstone community shows, that SiRF3 equipped receivers made by Royaltek, Holux and Globalsat function well combined with the Loadstone GPS package. In fact, the Royaltek RBT2010 and RBT2110 have become standards for reliability and accuracy when navigating at low speed.
Recently, new receivers appeared on the market, equipped with the MTK chipset. The makers of this chipset claim an even better performance at better battery consumption and lower prices. However, in spite of more channels and higher sensitivity, the results of these new receivers were disappointing in a number of cases, especially while used at low speed. Therefore, the Loadstone team decided to put three GPS receivers to the test, the Royaltek RBT 2010 SiRF3 equipped receiver and two receivers that make use of the MTK chipset, the Holux M1000 and the QStarz BT Q818. Below you find the results of this test as well as the Loadstone team's solution to some of the encountered problems.
The test was performed on 29June, 2007 in Burnaby, close to Vancouver Canada (latitude: 49.260490, longitude: -123.113940). Other major cities around this latitude are Dublin, Oslo, Stockholm, London, Copenhagen, Amsterdam, Brussels, Berlin, Warsaw, Prague and Moscow. The results are based on the synchronous logged NMEA data of the three receivers during a 10 minute walk around a more or less square city block. With the receivers side by side mounted on a tray with a two inch spacing the walk starts in a Southern direction for three minutes, then turns left to the east for little over one minute. The next turn is to the North for three minutes followed by a left turn to the North West for one minute again. Finally, the last curve returns the heading to the South again for 1 minute to end at the starting point.
Weather conditions were fine that day and the logging was started shortly after the receivers had obtained a fix. The buildings on the block are mainly apartment buildings of two to four stories high. The testers stayed on the sidewalk, about six to ten feet away from the walls of the buildings. There was nothing overhead to block the view of the sky except for a couple of small trees along the way. A normal walking speed was maintained, only stopping at the corners to test how well the receivers could get a new heading from a dead stop. The test conditions can be classified as 'normal' for an urban area.
The data used in this comparison was obtained as follows:
To compare the individual GPS receivers we calculated the 'real' route. Bearing and length of legs were calculated by analysis of street-lines and angles from the Roadmaps released by the Canadian Bureau of the Census. This resulted in the following time/course:
|
|
This graph shows the heading/time relation as derived from the Canadian Bureau of the Census data. |
Of course no street runs, and no VIP walks, in a perfect straight line, but this data serves well to calculate the average difference in GPS heading reported per receiver. Also the difference to the heading as calculated from the position data coming from the receiver by the Loadstone software can be compared this way. Since heading is a direct result of two positions, heading can be used to determine accuracy.
The following graphs and sonifications show the test results per receiver. We start with the Royaltek RBT 2010.
|
Royaltek RBT2010 the red line shows the heading as reported by the receiver, the green line shows the heading as calculated by the Loadstone software. Differences between reported and calculated values are minimal, the average difference to the 'real' route is 5.73 degrees. |
In the above graph and sonification heading data from the chipset (reported heading) and data obtained by calculating the heading from the position data (calculated heading) were compared. A GPS receiver outputs data that confirms to the NMEA protocol. The NMEA data contains a set of different 'sentences', each containing specific values that can be used directly by navigational software or used to calculate the desired values. In theory these results should be the same, and as the above example shows, this is true in the case of the Royaltek RBT 2010.
The next graph and sonification represent the results obtained from the Holux M1000.
|
Holux M1000 the red line shows the heading as reported by the receiver, the green line shows the heading as calculated by the Loadstone software. Differences between reported and calculated values are slightly bigger than in the case of the Royaltek RBT 2010, the average difference to the 'real' route is 7.16 degrees. |
A chipset is, as the name suggests, a set of chips, each chip performing one or more tasks, mostly communicating between processors and peripherals. A navigational chipset is a collection of receivers, transmitters, protocol translators and a processor, with as main task receiving satellite signals, processing this data, translating the results to the NMEA protocol and transmitting the resulting data to the navigational software. During the translation phase a certain amount of influence can be exercised on the data, like calculating thresholds and setting mask values. In the following graph you see and hear the result of this kind of preprocessing. The computed heading no longer has a relation to the real heading because of a high Static Navigation (SN) setting.
Data obtained from the QStarz BT Q818 receiver.
|
QStarz BT Q818 the red line shows the heading as reported by the receiver, the green line shows the heading as calculated by the Loadstone software. Differences between reported and calculated values are unacceptably big, the average difference to the 'real' route is 92.18 degrees. |
Static Navigation (SN) is a function in GPS receivers that 'freezes' the continuous reading of the satellite signals. A GPS unit needs to know the heading in which it is moving in order to make predictions. When the receiver is stationary, no heading can be determined and this results in a pseudo random behavior (drift)of the navigational unit. To avoid this behavior, some manufacturers of receivers configured them to stop updating when the speed dropped beyond a certain threshold. For use in cars, the threshold was set to about 4 to 5 kph. For pedestrian use this value is too high. when GPS navigation became popular with hikers, many manufacturers decided to switch static navigation off in their receivers. In the above example we see that this is not the case with the QStarz BT Q818, thus making this unit unusable for pedestrian use.
| unit | Ofs_from_calc | Ofs_from_actual | Ofs_calc_from_actual |
| RoyaltekRBT2010 | 7.60 | 5.73 | 7.94 |
| HoluxM1000 | 18.47 | 7.16 | 20.44 |
| QStarzBTQ818 | 92.18 | 90.45 | 21.85 |
Ofs_from_calculated is the degree offset from the calculated baring. It says more about the calculation method than the accuracy of the receiver.
Ofs_from_actual is the offset from the actual baring determined by calculating the angles between the street corners. This value determines the difference between 'real' heading and 'calculated' heading and can be used as a reliability indicator.
Ofs_calc_from_actual again says more about the calculation routine than gps accuracy.
A GPS receiver also provides speed measurement. While not as important as heading accurate speed measurements are a vital part of a navigational program. When speed of travel is known, predictions can be made about the moment when a certain point will be reached, thus providing the user warnings. A guide dog user needs to be able to give the right commands to the dog well in advance.
The below values were taken from second 30 to second 530 to eliminate the static part of the walk during startup and finish.
| Unit | Max_recv_spd | Av_recv_spd | Av_calc_spd | Av_dif_spd |
| RoyaltekRBT2010 | 5.64 | 3.67 | 3.94 | 0.58 |
| HoluxM1000 | 6.68 | 3.65 | 4.25 | 1.16 |
| QStarzBTQ818 | 6.12 | 3.58 | 4.13 | 1.11 |
Max_recv_speed is the maximum reported speed as derived from the receivers output, Av_recv_spd is the average speed as reported by the receiver, Av_calc_spd is the average speed as calculated by the Loadstone program and
Av_diff is the difference between reported and calculated average speed.
As you have seen and heard, in all three cases the Loadstone software was able to produce heading data that had a more or less strong relation to the data as provided by the Canadian Bureau of the Census. In the case of the Royaltek RBT2010 and the Holux M1000 receivers, the heading and speed data as obtained from the NMEA sentences (the receivers output) also had this relation. In the case of the QStarz BT Q818 receiver the data obtained from the NMEA sentences proved not usable for pedestrian use and showed the highest diversion of the 'real' course when calculated by the Loadstone software. Still, the calculated data was usable for determining a position with acceptable accuracy.
Since most manufacturers have switched off static navigation in their chipsets, we believe that the makers of the QStarz BT Q818 will follow soon. But no control over the 'drift' that appears without the appliance of static navigation at all is not desirable. Therefore the Loadstone GPS software offers the option to set static navigation to an acceptable threshold for pedestrian use. A setting of one to 2 KpH has proven to be a fine compromise between a loss of heading announcement and the effects of drift.
From this test we have learned, that the position data as reported by all three receivers is adequate for accurate location announcements. The tests also show, that the algorithms used by the Loadstone software suffice for accurate heading and speed calculation. Our next step will be the implementation of the option to choose from announcements directly provided by the receiver or provided by the Loadstone software. With this option in place, the Loadstone GPS program will be able to work together with any receiver on the market today.
|
The three lines show the latitudinal relations between the tested receivers. The longitudinal relation is equivalent. |
| Units | Max_diff_lat | Av_diff_lat | Max_diff_lon | Av_diff_lon |
| RoyaltekRBT2010->HoluxM1000 | 19.34 | 4.66 | 14.66 | 4.59 |
| RoyaltekRBT2010->QStarzbt318 | 19 | 4.99 | 11.16 | 3.15 |
| HoluxM1000->QStarzbt318 | 21.66 | 4.02 | 12.5 | 4.18 |
The above table shows the maximum and average differences in latitude and longitude between the test receivers in meters. All average differences lie below 5 meters, confirming the possibility to use the position data obtained from the RMC and GGA sentences in the NMEA code as a source for calculation of heading and speed by the Loadstone GPS program.
We like to stress, that the use of the word 'accurate' in this article refers to the commonly accepted accuracy of 10 to 15 meters as obtained by the Global Satellite System. We also like to point out, that developers are constantly improving GPS receivers and that the results of this test are but an impression of the facts on 29 June 2007.
At this moment we see no confirmation of the claimed superiority of the MTK chipset over the SiRF3 chipset. In fact, the Royaltek RBT2010 outperforms the Holux M1000 by a small margin and the QStarz BT Q818 by a larger margin.
We are not surprised by this conclusion. The limits of usable sensitivity in reception of satellite signals has been reached and increasing the number of satellites in the solution hardly offers a means for improvement of the performance of today's receivers in general. Any further improvement to be expected from positioning systems will have to come from augmentation systems such as WAAS and EGNOS, provided that sufficient groundstations will be built and enough geo-stationary satellites will be put into orbit.
Geographical information about Canada:
http://www12.statcan.ca/english/census06/geo/index.cfm
Mathtracks, a program to sonify graphs:
http://prime.jsc.nasa.gov/mathtrax/bvihomepage.htm
GPS receivers, how they work:
http://www.loadstone-gps.com/articles/gps_rec.html
The Loadstone GPS home page:
http://www.loadstone-gps.com/
Although Loadstone-GPS is free to use, it has used up thousands of development and testing hours since February 2004. In order for us to continue funding this project, adding features, improving functionality, and strengthening the support and documentation we need your help.
Please consider making a donation (Large or small - any amount is gratefully received) and will ensure development of this project continues well into the future.
Select this link to donate to the Loadstone-GPS project