|Documentation :||Hardware | Software | Tutorials||Wardrobe :||Jackets | Applications|
|Papers :||Notes | Proposals | Research||Administration :||Stock | Minutes | Progress|
This report describes research into dynamically determining the accuracy of GPS. It outlines tests carried out in the Millennium Square of the @Bristol site. The results show an accuracy of around 2m, including error components introduced by the test procedure. Recommendations for use of the system are made, and on further appropriate testing.
The need to determine position in the Millennium Square has arisen as part of the Bristol Wearables Project in order to provide precisely positioned soundscapes for artistic and tourist applications. In particular we are proposing to provide a spatialised sound experience along the route of the Zenith artwork - a 50 metre analemma. We propose to use the GPS system to track users as they progress around the analemma. Our recent tests have shown that an accuracy of around 2m can be achieved using differential GPS. These tests were carried out with static GPS receivers over 24hrs - in practice we need to know what accuracy can be reasonably expected with an instananeous (<2sec) reading while in motion. this document describes our approach and the results achieved.
Before carrying out the tests a precise mapping of the Millennium Square was made by placing a GPS receiver (Garmin eTrex Legend) connected to a radiobeacon dGPS receiver (CSI SBX-2) at each of the four corners of the Explore building. A Jornada 568 was used to log the output of the GPS receiver over 24 hours at each position. By averaging the lat/lon values each of the four corners of the building could be located on the WGS84 datum to within one metre. The resulting positions enabled gridlines suitable for GPS positioning to be placed on an architects plan - see below:-
(click on image for full download).
|Each test was carried out by walking once around the
analemma carrying a GPS receiver. The total time taken for each test was two minutes. The
tester stopped for five seconds at eight predetermined positions on the analemma.
The GPS receiver was worn on different parts of the body to determine suitable locations for it's position in the wearable computer configuration.
The results may be compared with those for our indoor positioning system as the procedures are very similar.
Tests were carried out using the eTrex Legend again connected to the CSI SBX-2, however instead of a direct connection between them, an 802.11 wireless connection was established using a laptop as a server and the logging Jornada 568 as the client. The error signals were packetised and transmitted using IP protocol and sockets - see here for background information on this technique (dGPS_IP). Test were also carried out using the Pharos iGPS-180 (non-differential).
The results are shown graphically in the graphs below. The pink trace is an accurately positioned representation of the analemma, the blue diamonds are the actual GPS readings, and the blue line is the assumed path.
|Legend - Handheld
The 'big' end of the analemma is well tracked with maximum errors of around 1.5m.
Errors around the'small' end are larger giving an exaggerrated effect.
|Legend - Shoulder mounted
Again, a good set of readings with errors in the order of 2 metres - this may be due to satellite signals being blocked by the head.
There are serious problems with the small end. As it is only 2m wide, the 2m resolution make it difficult to resolve which side of the annalemma the receiver is located.
|Legend - Head mounted.
Apart from a serious anomaly close to the centre this is the best trace - perhaps the receiver slipped. Differential correction was consistent so it is unlikely that the anomaly was caused by system failure.
|Legend - Waist mounted.
Though a GPS fix was maintained throughout this test, waist mounting does appear to introduce errors in the order of 10metres.
|iGPS-180 - Shoulder mounted
The lack of differential correction, combined with the limitations of the iGPS-180 (low size; low power; low processing; low accuracy), has resulted in a offset of around 30metres.
Conclusions and recommendations.
The tests have shown that it is possible to track a user around the Millennium Square analemma with better than 2 metre accuracy most of the time. This accuracy is sufficient to determine which side of the 'big' end of the analemma that the user is on, but is insufficient to do this at the 'small' end.
While head mounting gives the most accurate results, shoulder mounting is nearly as good, and is recommended for our applications.
The iGPS-180, while not suited for tracking around the analemma, performed well enough to recognise the quadrant of the Millennium Square in which it was located, and hence can be considered suitable for most 'tourist' type applications.
A more detailed analysis is required of the data to produce 50% CEP and 95% error measurements - additional data would add weight to the resulting figures. Further testing with the eTrex Legend is recommended to evaluate the gain in performance from using dGPS_IP, and also to test whether chest mounting is a viable option. As the reference data collection, and the testing, was all carried out over a period of five days it would also be advisable to repeat the tests at another time when the ionospheric conditions may be different.
Thanks to Karen Charlton and Mike 'Amec' of @Bristol for assistance in gathering the reference data; and to Mike McCarthy for hand signals and other help with the testing.
14th Feb 2002