A Performance Assessment of Low-Cost RTK GNSS Receivers
John M. C. Jackson, Brian Davis, Demoz Gebre-Egziabher, University of Minnesota, Twin Cities
Alternate Number 3
The proliferation of low-cost, real-time kinematic (RTK) GNSS receivers in the commercial market has the potential to enable novel applications requiring sub-centimeter accuracy. Up until now, there has been no comprehensive literature available that studies the performance and reliability of these receivers. Original equipment manufacturers (OEMs) typically offer an accuracy specification (e.g. 2cm +/- 1ppm) without context on application or environment. This study compared the performance of several low-cost, RTK GNSS receivers against established positioning products to quantify their relative performance differences.
There were three criteria for choosing low-cost RTK receivers to test. These were that the receiver should cost around $500 or less, the receiver can compute an RTK solution with correction data via a continuously operating reference service (CORS) network, and the receiver is commercially available with an acceptable lead time. Based on these criteria, the receivers that were tested are
1) ublox NEO-M8P,
2) Skytraq S2525F8-RTK,
3) Emlid Reach (using the u-blox NEO-M8T module),
4) NVS Technologies NV08C-RTK, and
5) Swift Piksi Multi GNSS
The receivers used as a reference for performance were the Navcom SF-3050 and Trimble BX982 receivers.
The experimental conditions were split into three categories – antenna quality, test environments and application type. Two antenna types were used for testing; a high-quality, multi-frequency GNSS antenna and an L1-only patch antenna. Test environments were chosen to reflect the qualities of urban, suburban and rural skyscapes. Two application types were considered for this study. The first type of tests were static data collections conducted over National Geodetic Survey (NGS) markers. The second type of tests were on-vehicle dynamic tests. Each of these test types were performed for the different test environments and antenna qualities.
To make objective claims on receiver performance, metrics were defined for position and continuity quality. The primary metric for position quality is the error of a receiver’s calculated position with respect to a truth position. The truth positions that are used for calculating the errors are NGS markers of sub-centimeter accuracy for the static tests or the high-end GNSS receivers mentioned above for the dynamic tests. These position errors were used to determine a receiver’s 2? accuracy.
The continuity metrics are indicative of a receiver’s ability to calculate and maintain an RTK solution. The statistics for continuity include:
1) Time to first RTK floating-point solution
2) Time to first RTK fixed-integer solution
3) Total time spent in RTK fixed-integer or floating-point mode
4) Average number of RTK loss of solution occurrences
5) Average time to reacquire the RTK solution after a loss
The experimental setup consisted of a single antenna whose signal was split 7-ways using a low-attenuation splitter. Each receiver was connected to the splitter with a low-loss coaxial cable. A USB splitter and RS232 hub were used to connect each receiver to a laptop computer. The computer was connected to a cellular modem to access observational correction data from the Minnesota Continuously Operating Reference Station (MnCORS) network. Software written by the authors was used for data collection and logging, running experiments, and interfacing with the MnCORS network.
The static test results revealed interesting accuracy characteristics of the low-cost receivers. In general, there is a correlation between how obstructed the test environment is and the accuracy of a receiver. While the mean positional error of low-cost receivers was near manufacturer claims (e.g. 2cm +/- 1ppm), the 2? accuracy of the position for each low-cost receiver was higher, much higher in some test regimes. While the reference receivers had an East-direction 2? accuracy of 3cm or better in all test conditions, the low-cost receivers had 50cm+ 2? accuracy values in certain cases. The worst accuracy values occurred when the low-cost receivers were in the suburban or urban testing environments using the low-quality patch antenna.
The full static and dynamic test results of this study will shed light on the operating characteristics of low-cost RTK GNSS receivers. The results will quantify all the metrics outlined above to give an objective method of comparing the receivers, which can guide potential applications of the low-cost receivers and the focus of future studies.