| Popis: |
This research concentrates on both weak GPS signal-processing and GPS time error correction in parallel with the development of an innovative GPS tag – “TrackTag”. The technology is aimed at animal tagging applications and therefore has particular challenges. The tag has to be small enough so as not to impact the behaviour of the animal under study as well as robust enough to operate wherever the animal goes. TrackTag‟s architecture differs form that of a conventional GPS receiver in that it uses very short snapshots and there is no processing performed on the tag itself. This enables the tag to operate for over a year on a small battery. Weak signal processing algorithms studied include FFT techniques and conventional correlators for demodulation of the GPS signals using both non-coherent and coherent integration. The techniques developed enabled the tracking of signals down to 24dB-Hz, which is comparable to today‟s best conventional GPS units. However, the algorithms developed during this research achieve such performance using only 76 milliseconds of data while conventional designs require many seconds of data. The performance of the navigation solutions was not compromised in terms of accuracy. A 2dRMS accuracy of 29 metres was observed under forest canopy and the research even demonstrated over 50% success rate in the Amazon Rainforest canopy where conventional GPS tags had never been shown to work. New time correction techniques were also required as it was estimated that a 100ms clock error would result in 100m positional error. Temperature calibration of the RTC crystal was developed as the first step. However, with that, the clock drift over a month could still be up to 5 seconds (2ppm) and so it became clear that for long-duration (>3 months) studies another method for maintaining accuracy was required. Over-determined navigation solutions, and their subsequent position residual error, were used to estimate the time offset with some success. Pattern-matching against WAAS signals offered exceptional performance and was demonstrated on real-world tracking studies that were running concurrently with this research. The accuracy was considered to be limited only by the front-end sample period, i.e. 200 nanoseconds for any given deployment duration. |
