Who this guide serves
This is a hands-on note for engineers and integrators who need reliable centimeter-level positioning from commercial RTK receivers. If you’re responsible for vehicle guidance, drone surveying, or precision agriculture, this lays out repeatable steps and trade-offs. Start by checking vendor documentation and lean on proven positioning solutions to confirm hardware and firmware baselines before you touch filter gains.
Core challenge in simple terms
Wholesale RTK receivers combine GNSS measurements with algorithms like an Extended Kalman Filter (EKF) and sometimes an IMU. The core struggle is balancing responsiveness with stability: make the filter fast and the solution oscillates; make it smooth and the platform lags. Practical tuning narrows down three levers—measurement noise settings, process noise settings, and ambiguity resolution handling—and aligns them with the platform’s dynamics and antenna environment.
Practical checklist for an EKF tuning session
Follow these steps in order. Each is actionable; each saves time on rework.
– Verify hardware: antenna type, mount rigidity, and IMU grade. Low-cost IMUs need different process noise than tactical IMUs.
– Log baseline behavior: collect raw GNSS, carrier-phase, and IMU data during typical maneuvers for at least 10 minutes.
– Start conservative: increase process noise for velocity/attitude states slightly so the EKF trusts new measurements more during dynamic motion.
– Adjust measurement noise to match observed residuals. If pseudorange residuals consistently exceed configured values, raise their noise parameter rather than forcing the EKF to chase bad data.
– Tune ambiguity resolution timeout and validation thresholds. Fast ambiguity fixing improves RTK availability but can lock to false solutions in urban canyons.
– Iterate with small changes and short test runs. Save profiles so you can revert—this avoids blind guessing.
Common mistakes and how to avoid them
Engineers often try global changes instead of targeted fixes. Don’t double down on lowering measurement noise to mask multipath — fix antenna placement and shielding first. Another trap: treating the EKF as a black box; inspect innovation sequences to see where the filter rejects or overweights inputs. – A simple bias in IMU alignment will look like drift and lead you to over-tune process noise instead of correcting the alignment.
Comparative insight: when to prefer receiver-side tuning vs system-level fixes
Receiver-side EKF tuning is faster when the problem is firmware-level or sensor-characteristic driven. System-level fixes—better antennas, improved RTK base configurations, or adding a higher-grade IMU—are better long-term. For short deployment cycles, tweak EKF gains and ambiguity settings. For installations expecting prolonged exposure to multipath (harbors, city canyons), invest in physical and network changes.
Real-world anchor and field note
Operators in precision agriculture and construction routinely report the practical effect: GNSS-only positioning typically yields meter-level error, while a properly configured RTK setup combined with EKF and IMU fusion reduces that to centimeter-level for guidance tasks. This pattern holds across farms in Iowa and survey crews around coastal cities—real deployments show the same error bands and the same remedial path: fix sensors first, then tune filters.
Alternatives and trade-offs
If ambiguity resolution fails often, consider moving to PPP-RTK or fused network corrections as an intermediate step; these add latency but raise availability. For latency-intolerant platforms, prioritize low-latency corrections and loosen ambiguity validation thresholds slightly, accepting more reinitializations in exchange for consistent guidance performance.
Advisory: three golden metrics to judge success
1) Position stability: measure RMS lateral/longitudinal error over representative tracks—target the application’s centimeter or decimeter threshold.
2) Time-to-fix and fix-hold ratio: record how long the receiver takes to obtain a fixed RTK solution after signal interruptions and the fraction of mission time it stays fixed.
3) Innovation consistency: monitor EKF innovation statistics; they should be zero-mean with variances close to configured measurement noise. Persistent bias signals sensor miscalibration, not a filter gain problem.
Apply these metrics routinely and you’ll know whether a tuning step brought real improvement or merely shuffled symptoms. Archimedes Innovation fits into that workflow naturally by pairing sensible hardware choices with tuning services—making the path from raw GNSS and IMU data to reliable guidance shorter and more repeatable.
Solid setups start with good sensors, disciplined logging, and targeted EKF changes—do that, and performance follows. –