1. 0 Flying™ RTK Solution as Effective Enhancement of Conventional Float RTK Dmitry Kozlov, Gleb Zyryanov Magellan, Russia ION GNSS 2007 Session D1: Algorithms and Methods 1 September 26,2007

2. 1 Scope (18 slides) Summary Float and Fixed RTK Flying RTK Magellan products with Flying RTK Convergence performance Concluding remarks Acknowledgment

3. 2 Flying RTK algorithm: Summary We present new RTK algorithm which: Can be positioned between standard Float and Fixed RTK Has all the external attributes of Float RTK Uses internally some ideas of Fixed RTK Insures convergence performance better than Float RTK Integrated into 2 latest Magellan products OEM DG14 RTK board Handheld ProMark3 Surveyor

4. 3 Float and Fixed RTK: difference Float RTKFixed RTK Receiver typeUsually L1 onlyTypically L1&L2 Accuracy levelMeter to decimeterCentimeter Dependence on baseline length Not so dramaticalNoticeable (working distance is usually less than 50 km) Initialization timeDecimeter after 3-10 minutes Centimeter after some convergence time depending on baseline Treating DD ambiguity As float unknown value As known integer value

5. 4 Float and Fixed RTK: commonality Float RTKFixed RTK Core processing engine Kalman Filter or similar recurrent estimator Input dataReceiver raw observations and external reference RTK data Treating ambiguityAmbiguity is processed as float all the time Ambiguity is processed as float initially Behavior at star upSlow convergenceSame convergence until ambiguity fix Fixing ambiguity to integer Not appledApplied, but by quite a separated search and validation procedure

6. 5 Float and Fixed RTK: tentative diagram Float RTK Fixed RTK

7. 6 Flying RTK: backlog Actual DD is unknown integer (finite number of alternatives) There is theoretical foundation how to treat it Optimal multi-channel algorithm is too complicated Let us use Float RTK scheme Let us process DD ambiguity as unknown float, but consider as unknown integer Let us use DD ambiguity search results to generate Flying RTK correction to Float RTK solution

8. 7 Fixed and Flying RTK: tentative diagram Fixed RTK Flying RTK

9. 8 RTK slogans Float RTK: always process DD ambiguity as unknown float variable Fixed RTK: first process DD ambiguity as unknown float variable, then (after fix) as known integer value Flying RTK: always process DD ambigity as unknown float variable, but always consider as unknown integer

10. 9 Different RTK: typical convergence tubes Flying RTK convergence is as flat as Float RTK convergence Flying RTK converges faster than Float RTK Flying RTK never fixes wrong ambiguity as it can be with Fixed RTK Flying RTK steady state accuracy can be as good as cm

11. 10 Flying RTK implementation: DG14 RTK L1 GPS+SBAS RTK OEM board (base and rover) Fixed and Flying RTK RTCM-2.3 (base and rover) RTCM-3.0 and Magellan proprietary (rover) 20 Hz Raw data 10 Hz RTK (with extrapolated base) position 5 Hz synchronized (matched tags) RTK position OTF Fixed RTK initialization RTK with moving base Heading function

12. 11 Flying RTK implementation: ProMark3 RTK L1 GPS+SBAS RTK handheld (base and rover) Real time and post-processed Surveying and Mobile Mapping functions Fixed and Flying RTK RTCM-3.0 (base and rover) RTCM-2.3,3.0 and Magellan proprietary (rover) Compatibility with VRS, FKP and MAC Networks NTRIP&DIP rover (with external GPRS module) External license free radio (base and rover) OTF Fixed RTK initialization Initialization of known point Initialization on kinematics bar

13. 12 Flying RTK performance: evaluation principles Apple to apple comparison PC version of RTK Static data but kinematics processing RTK auto-reset each 10 minutes Statistically sufficient estimates CEP convergence pattern Cases: DG14 data (base and rover) Zmax data (base and rover, L1 portion only) ProMark3 data (against Ntrip Network)

14. 13 Flying RTK: Convergence with DG RTK data DG14 base and rover 24 full days data at different times and locations Baselines from few meters to 10 km Open sky to partly shaded sky environment Kinematics processing

15. 14 Flying RTK: convergence with Zmax data (L1 only) Zmax data: base and rover L1 CA portion only Open sky baselines Baselines from 7 to 52 km >48 hours for each baseline Accuracy after 3 minutes Kinematics processing

16. 15 Flying RTK: convergence with ProMark3 data ProMark3 data (rover) Orpheon NTRIP Network, France 10 km baseline Open sky 48 hours Kinematics processing

17. 16 Conclusions Many GNSS users want decimeter accuracy Standard DGPS can deliver only sub-meter One of the choices is RTK L1/L2 RTK with fixed ambiguity can be too expensive Alternative is much cheaper Float L1 RTK L1 RTK Float convergence is not always fast Flying RTK algorithm shows improvement compared to classic Float RTK ‘in the same box’ Flying L1 RTK is good compromise for these users

18. 17 Acknowledgments Magellan System Test group: for their careful testing and validation efforts with release DG14 RTK and ProMark3 RTK Yves Le Pallec, Eugeny Sunitsky (all Magellan), and Bill Cottrell (Cottrell Navigation Services): for help with data collection

19. 18 Final slide Flying™ RTK Solution as Effective Enhancement of Conventional Float RTK THAHK YOU FOR YOU ATTENTION QUESTIONS?