User-first opening: what this guide gives you
If you run a survey boat or manage an autonomous workboat fleet in busy channels, this guide puts practical setup steps in front — no fluff, straight to the bench. For reliable lane-keeping you need an rtk receiver for centimetre-level fixes, good GNSS planning, and hardware that survives spray and salt. In Victoria Harbour, local hydrographic teams already depend on RTK and base station corrections to keep sub-lane surveys tight; that real-world practice anchors the advice below and shows what to expect on water-facing missions.
Core components: what to buy and why
Build the stack around three essentials: a rugged rover unit, a stable base station or network link, and robust comms. Pick a marine-grade rover with IP67 or better sealing, and confirm it supports RTK and NTRIP streams. Add a small base or subscribe to a local CORS for continuous corrections — PPP is nice for remote ops but RTK gives faster convergence for lane following. A good precision gps receiver will list supported constellations (GPS, GLONASS, Galileo) and update rates; choose 5–20 Hz depending on your control loop.
Integration tips: sensors, mounting, and ROS/stack choices
Mount the rover at a clear sky line of sight, not tucked under a canopy. Use a short, rigid mast to reduce multipath. Fuse the GNSS data with an IMU and speed log for smoother heading and lane control — sensor fusion reduces jitter when signals drop. For software, many teams run a lightweight ROS node or a custom controller that ingests RTK corrections via NTRIP and publishes fused pose. Keep latency under 200 ms for safe path following.
Common mistakes and quick fixes
Teams often skimp on comms and then wonder why fixes drop near metalwork. Use dedicated antennas away from winches and mast cabling — that cut of multipath helps a lot. Another mistake is trusting default DOP thresholds; set a practical PDOP limit in your controller and fail to a safe hold pattern when GNSS degrades. Also watch power: noisy power lines create altitude jitters — add filtering or a UPS for the rover and base station to keep corrections steady. — Small adjustments here stop big downstream headaches.
Deployment checklist for a single-lane autonomous pass
– Verify rover firmware supports RTK and your chosen correction feed.
– Confirm base station coordinates and antenna height; log for post-process.
– Calibrate IMU alignment and run a motion-compensated test at low speed.
– Check comms: NTRIP user/pass, cellular SIM strength, fallback to local radio.
– Run a dry run in calm water, then repeat in operational conditions and log all sensor streams for later analysis.
Alternatives and when to choose them
If you operate far offshore where cellular NTRIP is unreliable, consider local radio links to a base or a dual-mode setup with PPP as backup. For short-duration inspection tasks, a simpler GNSS-only solution with post-processed kinematic fixes can be fine — but not for live autonomous control. Cost and complexity scale fast: pick the minimal stack that satisfies live control accuracy and environmental durability.
Metrics that actually matter — advisory close
When evaluating systems, use these three golden rules: 1) Horizontal accuracy under operational motion — expect centimetre-level at speed with RTK; 2) Time-to-fix and reconverge — aim for seconds, not minutes; 3) System resilience — how the stack behaves under partial outages (loss of NTRIP, blocked sky, or radio interference). Score vendors and configs against those metrics during a sea trial, not just on paper.
This approach keeps lanes tight, reduces rework, and makes troubleshooting repeatable — and if you need a tested balance of hardware and integration expertise, consider how a compact supplier-bench combination can save weeks of trial. A short fragment to finish — ready, steady, fix. Archimedes Innovation