Public Product Guide

KSF AI Offshore Autopilot Suite

We built this to steer more like an experienced sailor: calm, predictive, and aware of the sea instead of constantly hunting the rudder. It did not take long before we realized we wanted more onboard help than just a better autopilot, and that is why it became a full suite.

Suite Overview Safety + AI

By: Knut Steinar Fremme / Kai Nils Fremme

Critical Principle

Skipper Authority First

No automation system, no matter how capable, ever removes command responsibility from the skipper on board. The KSF AI autopilot suite is designed to improve awareness, prediction and decision support, but the final decisions and command of the boat always stay with the skipper.

Just as important: you are always free to disengage the automation and hand-steer whenever you want, including the moments when you simply feel like sailing the boat yourself.

What AI Means Here

In this project, AI does not mean a chatbot or a remote cloud model. It means the autopilot continuously watches the live sensor stream, including fore and aft 9-axis attitude modules with yaw, roll, pitch, accelerometer, gyro and magnetometer data, plus heading, heel, rate of turn, acceleration/G-force cues, wind, rudder command and boat response, to learn how the boat is behaving in the current waves.

That live learning lets the system adapt steering to the sea state, anticipate the next correction, and reduce the constant rudder hunting often seen in fixed-tune autopilots.

Why This Project Started

The project started with a simple frustration: many commercial autopilots chase heading with aggressive rudder oscillation driven by basic PID rudder-control logic. Offshore, that can mean more wear, more noise and more fatigue.

We wanted something different: first understand how the boat is actually moving in the waves, then steer with anticipation so rudder use is calmer, cleaner and closer to the way an experienced helmsman drives.

If you have hand-steered in a seaway, you have probably noticed that the boat often comes back close to the same heading and track after a full wave cycle. This AI-powered autopilot is built around that same idea: work with the sea instead of fighting it unless a real correction is needed.

Less rudder hunting and fewer violent corrections
More predictable course-keeping through changing sea states
More confidence when trying to sleep on singlehanded passages

I have worked professionally with software for many years, and because I also love both microcontroller programming and sailing, this project became a very natural direction for me.

Network And Compute Architecture

All modules interoperate over standard NMEA2000 for marine compatibility.
Every hardware module has its own local touch display: a 4.0-inch TFT on each module, except the Main unit, which uses a 7-inch helm display.
Full multilingual support is built into every module, including UI/LCD and iPad workflows.
The system is designed to work with current sensors, wind/log instruments, rudder command units, and related components from Raymarine, Garmin, Simrad and other NMEA2000-capable ecosystems.
Inside the AI autopilot cluster, the core nodes communicate over a private high-speed NMEA channel for deterministic control-data flow.
Current high-speed core topology: 3x Teensy 4.1 + 1x ESP32-WROVER.
Gateway and UI layers bridge the private control network with standard boat-network context.

Current hardware deployment: 7x Teensy 4.0, 3x Teensy 4.1, and 2x ESP32-WROVER. The active module stack now exceeds 240,000 lines of code, excluding external libraries, and the iPad app alone adds more than 20,000 lines.

Operational Modes, Including WORK

STBY, HDG, WIND and CRS/TRK cover normal passage handling.
WORK mode prioritizes boat stability to make deck and sail handling safer during high-workload operations.
In WORK mode, stability can take priority over strict course-holding, so temporary off-course drift is acceptable when it improves safety on board.
All critical transitions stay acknowledged and visible.

Traffic And Hazard Awareness

AIS/RADAR warnings and alarm flows are integrated across the helm UI, iPad app and voice system.
An avoid-and-resume route workflow supports temporary tactical deviations.
Hazard cues from attitude and force sensing can trigger advisories or prevention logic.
The goal is to keep the skipper in command while reducing reaction delay.
Barometric pressure-trend plus wind-shift and gust fusion gives earlier squall warnings, which is especially useful on night watches.

Long Passage AI Support

Weather-corridor guidance helps with safer and faster ocean passages.
CTS guidance balances VMG, comfort and overall risk.
Hazard-aware suggestions help avoid overpowered sectors and unstable sea states.
When the internet is unavailable, the system falls back to local logic with no hidden assumptions.
Route, tack and gybe planning are supported with operator-confirmation gates.
The sail-recommendation engine suggests reducing or adding fore and main sail based on wind, sea state and boat-handling stability.
The system can automatically report to SailingKSF with current position, heading, weather and sea state at configured intervals.
Dedicated web pages let supporters and family follow each boat with live voyage context.

Scale And Expansion

From a software perspective, UI/LCD modules on the bus are effectively unlimited; the real limit is practical bus and network capacity. UI/LCD links run over a private CAN bus between the Main unit and the UI/LCD units.
From a software perspective, iPad clients are also effectively unlimited; the practical limits are bandwidth and onboard operating policy. iPads can run either on a dedicated private WLAN or on the shared WLAN/LAN infrastructure on board.
Current simulation and validation cover Hallberg-Rassy 57, Hallberg-Rassy 69 and Bavaria 32 profiles.
The roadmap includes dual cameras, bow and aft, for incoming-wave interpretation and predictive sea-state assistance.
Last impolemted is a weather-safety feature: pressure sensors are fused with gust and shift analytics for earlier squall detection when visual cues are weak.

And yes, the Hallberg-Rassy 57 is still the dream boat.

System Modules At A Glance

AI Autopilot Core

Mode logic, tack/gybe orchestration, anti-hunt steering, active prevention, and timed recovery.

Attitude Module

Heading, heel, pitch, and motion-quality signals that keep steering calm as the sea state changes.

Weather + Corridor AI

Onboard weather logic, cloud tactical corridors, and local JSON API export for other onboard systems.

Timer + Watchkeeper

Watch reminders, sleep-safe alarm rhythm, acknowledgement flow, and clearly mirrored status.

Blackbox

Traceability for what happened, why it happened, and what changed before and after each event.

Logging + Polar

SD-card and server logging, same-condition comparison, and owner-performance polar review on iPad and server.

Helm UI + iPad App

Fast command surfaces, route and waypoint workflows, track export, and tactical maps.

Voice Alarm

Spoken alarms and context callouts that reduce silent failures and mode confusion.

Sensor Safety

Lightning, air-quality, gas, and fire monitoring integrated into marine safety workflows.

SailingKSF Weather

72-hour corridor products, INTELS layers, full weather/position/boat-status tracking, and local onboard JSON feeds.

Sea-Trial Photo Set

We trimmed the image count to keep the pages cleaner. Where possible, I now prioritize iPad screenshots because the image quality is better.

**Compass Main (iPad)**
Primary compass/helm surface with vectors and command bar.

**Main Menu (iPad)**
Main menu layout and content overview.

Public Pages

Use the section menu above to move between the overview and each module page.

More Features Across The Suite

Current page focus: Why the platform exists and how the modules work together offshore.

The suite is much broader than a traditional autopilot. It is a complete singlehanded offshore assistance stack covering steering, weather intelligence, safety sensing, logging and post-passage analysis.

Adaptive anti-hunt steering tuned to real sea motion

Standard NMEA2000 interoperability plus a private high-speed NMEA core exchange

Cross-vendor NMEA2000 compatibility with Raymarine, Garmin, Simrad and more

Maneuver-aware route behavior with avoid-and-resume workflow

AI prevention layer for broach, surf and pitchpole tendency

WORK mode for high-workload deck and sail operations

AI sail recommendations for fore/main trim and reefing stability

Hazard diagnostics with recovery counters and reason traces

Onboard and cloud weather fusion with deterministic offline fallback

72-hour weather-corridor planning and INTELS tactical context

Track and voyage logging for weather, position and boat-status history

Owner-performance polar review on iPad and server, with same-condition better/worse comparison

Local Weather Box JSON export for weather, INTELS, corridor and wide-area hazard feeds

Automatic periodic SailingKSF reporting with position, heading, weather and sea state

Dedicated supporter and family follow pages with live voyage context

Automatic local track logging every 10 minutes while underway, with export for analysis

Emergency 1/2/3 escalation with automatic shore updates and SMS/email contact alerts

AIS/RADAR warning and alarm workflow across helm, iPad and voice

Pressure-trend plus gust/shift fusion for earlier squall warnings, especially at night

Voice alarms and structured escalation for fatigue-resistant awareness

Sensor safety integration for lightning, gas, fire and air quality

Configurable 4x analog + 8x digital protected inputs with relay and voice actions

Native iPad workflow for routes, waypoints and tactical map use

Roadmap: dual bow and aft cameras for incoming-wave prediction support

Feedback And Comments

This project is still work in progress, and I would genuinely like feedback on what is missing, what feels unclear, and what would be most useful offshore. The first sea trial is planned for spring 2026, and every comment is reviewed by me before it appears publicly.

No public comments yet. You can be the first to add feedback.

Built For Singlehanded Offshore Reality

These public pages focus on what the suite actually delivers offshore: steadier steering, fail-safe automation, and practical AI support that reduces workload without taking control away from the skipper.