AIS and position data Fleet operations ETA and port calls

Custom vessel tracking software development

A ship transmits its position every few seconds, but the question ashore is a different one: when will it be alongside the quay, and what does that mean for tomorrow's planning? Between that AIS message and a usable answer lie processing, storage, an arrival estimate that keeps shifting and coverage that has gaps on the open sea. Appfront builds software that turns vessel positions into decisions, for shipowners, charterers, agents, terminals and shippers who plan around an arrival.

What makes tracking a ship different

Tracking a delivery van is a solved problem: an on-board computer or phone with mobile coverage, a position every few seconds, a route over a road network known down to the metre, and an arrival within the same shift. Tracking a seagoing vessel looks superficially similar, but in practice it is a different discipline. The position comes from a broadcast over the VHF band, picked up by shore-based receivers or satellites, with all the interruptions that entails. The route runs over water, not over a network of junctions and bends. And the arrival lies so far ahead that the estimate shifts dozens of times along the way, while pilots, tugs, cranes, staff and trucks have already been committed to it ashore.

On top of that, the operation itself is spread across parties that cannot see into each other's systems. The shipping company knows what the vessel is doing, the agent arranges the port call, the terminal determines the berth and the window, customs wants the data in advance, and the shipper only wants to know whether their cargo is feasible. Vessel tracking software is therefore rarely just a map. It is the layer that turns a stream of raw positions into a shared picture on which multiple parties can make their own decisions.

Are you in the right place? This page is about sailing, not shipbuilding. If your question concerns the yard, sections and outfitting, engineering and the build file of a vessel that has not yet entered service, you belong with software for shipbuilding. There, the ship is the product being created; here, the ship is a moving point carrying cargo, paperwork and an arrival time.

The position is public

AIS is a broadcast, not a private channel. Anyone with a receiver can see your vessel. What you protect is not the position, but what you conclude from it.

The message is not the truth

Destination, draught and estimated arrival are entered by hand. They are an intention, not a measurement, and sometimes still carry over from the previous voyage.

One arrival affects ten parties

Pilot, tug service, berth, crane, crew, transport and customs declaration all hang on the same moment. If it shifts, the whole chain shifts with it.

Processing AIS: from message to usable position

AIS stands for Automatic Identification System and works as a broadcast over two fixed VHF channels. A transponder on board transmits its position, course and speed at short intervals, and every receiver within range picks it up. Under the international convention for the safety of life at sea, seagoing ships above a certain size and all passenger ships are required to carry such a transponder and keep it switched on. Smaller vessels use a lighter class that transmits less often and with less power. You need to know that difference, because it determines how often you can expect a new position, and therefore when silence means a fault.

The messages come in several types. A position report contains the essentials: coordinates, course over ground, speed over ground, heading and rate of turn, plus a timestamp and an indication of the accuracy of the position fix. A static message contains the identity and the hull: name, call sign, ship type and dimensions. A third category contains the voyage data: destination, draught, number of persons on board and estimated arrival. That last group is entered by hand. This makes the distinction clear straight away: the dynamic fields are measurements, the voyage fields are intentions, and a system that stores the two mixed together produces unreliable planning.

Identity: the number you attach the history to

A ship has several numbers, and they do not all do the same thing. The number the transponder identifies itself with belongs to the radio registration and to the flag state; it therefore changes when the ship changes flag or owner, and it is reissued after deregistration. The International Maritime Organization's ship identification number belongs to the hull and stays the same for its entire life, even after renaming and re-registration. If you attach your sailing history to the radio number, you lose a ship's past as soon as it reflags, and one day you inherit the history of another ship that was given the same number. The practical solution is your own ship entity with a validity window per identifier, so that a past position stays with the ship that transmitted it at the time.

Names are even less reliable. They are mistyped, appear in capitals with or without a prefix, and occur for multiple ships. So never match on name, not even in an integration with a terminal or a freight forwarder. Record an explicit mapping between your own ship entity and the number the counterparty uses, manage that mapping in the system rather than in a spreadsheet, and log which assumption was made when a message could match more than one ship.

Storing data without the stream overwhelming you

Tracking a fleet means storing a continuous stream of small messages, and that comes with three pitfalls that only bite further down the line. The first is duplication: the same broadcast arrives via several receivers and sometimes via several providers. Without deduplication on vessel identifier and measurement timestamp, your archive grows faster than it needs to and every count is off. The second is ordering. Messages do not arrive in the order in which they were sent: a satellite capture can arrive later than a position that was picked up from shore afterwards. If you treat the last line to arrive as the current position, vessels visibly sail backwards. The latest position is the latest by measurement timestamp, not by processing timestamp, and that is a choice in the data model, not a detail in the display.

The third pitfall is wanting to keep everything for ever in the same form. That doesn't work and it isn't necessary either. Raw messages are valuable for a limited period, for troubleshooting and for recalculating events. For history, a simplified track is enough: line simplification preserves the shape of the voyage with a fraction of the points. Store the derived events separately as well, because they are the answer to the questions that will be asked later. A common set-up is a time-series store with a spatial index, with the raw stream in a separate layer and an increasing retention period per layer: fine-grained and short for raw data, coarse and long for history.

On the read side, the same problem applies in reverse. A map with an entire fleet on it is tempting to fill straight from the positions table, and that works exactly as long as the pilot set-up is small. With a growing fleet and a longer history, the map should be served in aggregated form: summarised points per zoom level, clustering where vessels are close together, and a track that is delivered in stages rather than all at once. The question itself also changes per user. A planner wants the latest status of ten vessels, an analyst wants three voyages side by side, and a customer portal wants one vessel without the rest. Those are three different read paths, and it makes sense to build them as three paths, with a separate view table for the current status alongside the archive. Finally, expect that the browser is not the place where you filter: what the user is not allowed to see should not be sent along in the first place.

At sea, there are gaps in coverage

This is the point where demonstrations look better than reality. Reception from shore is limited to the line of sight of the VHF link, so to a few dozen miles of coast. Beyond that you depend on satellite reception, which has two characteristics you need to build in: there is time between two passes, and in busy areas the broadcasts of many vessels overlap to such an extent that some cannot be read. In addition, a transponder can be switched off, jammed or report a position derived from a disrupted satellite navigation signal. The result is a track with gaps, with jumps and sometimes with a position in the middle of the land.

What a responsible system does about this is not make the data look better but make it honest. That means recording for every position which source it came from and how old it is, and keeping that age visible wherever someone makes a decision. It means plausibility checks: a jump that would require an impossible speed, a position on land outside a waterway, a heading that does not match the track. Such observations are flagged and kept separate, not silently discarded, because a series of anomalies is a signal in itself. And it means calculating between two measurements using the last known course and speed, provided those interim values are clearly shown as estimates and never end up in the archive as measurements. Finally: set an expected interval per source and raise an alert when a vessel stays silent for longer than that interval allows, because that is exactly what counts as news to a planner.

What a vessel tracking system needs to be able to do

These components depend on one another. An arrival estimate without source recording cannot be defended, geofences without a voyage definition produce isolated events, and an integration without unambiguous vessel identification leaves the error with the recipient.

Position processing with provenance

Deduplicating, sorting by measurement time and recording for every position which source it came from, how old it is and whether it was measured or estimated.

Identity with a timeline

A single vessel entity with a validity window per identifier, so that reflagging and renaming do not break up the voyage history.

Arrival estimate with a margin

An estimate based on the route over water, with an uncertainty band, a source and a timestamp, instead of one hard number.

Geofences that create events

Port limits, anchorage areas and berths as polygons that report entry, dwell time and departure, with thresholds to prevent flip-flopping.

Voyage file per voyage

Track, events, documents and consumption under a single voyage number, with a recorded definition of where a voyage begins and ends.

Integrations that can tolerate failure

Exchange with planning, terminal and declaration systems using repeatable messages, correlation identifiers and a queue for whatever did not arrive.

Calculating the arrival estimate and keeping it up to date

The first mistake is the simplest: taking the straight-line distance and dividing it by the speed. That routes a vessel straight across land, ignores canals, locks, straits and traffic separation schemes, and gives a result that is too early in almost every basin. A usable calculation works out the remaining distance over a navigable route, with the known passages as nodes. That immediately sharpens the question: which route will this vessel take, and what happens to the estimate if it chooses a different one along the way?

The second mistake lies in the speed. The last measured speed over ground jumps with every wave and every course correction, and on a river or in a current it says something different from the speed through the water. Anyone who projects the last measurement forward to an arrival produces an estimate that jumps with every message and so loses its credibility. A smoothed speed over a window is more useful, distinguishing between the sea passage and manoeuvring, and with a correction for the fixed delays that every approach involves.

The third and most important mistake is treating arrival as a single moment. In reality there are at least three: reaching the port limit, arriving at the berth, and the moment the vessel is made fast and handling can begin. In between come waiting at an anchorage, a tidal window for an access channel with draught restrictions, pilot and tug availability, and a berth that is still occupied. The planner ashore benefits most from that last moment, and that is precisely the one that cannot be derived from AIS. It follows from your own model plus what the port and the terminal tell you.

Just as useful is asking why a vessel deviates from expectation. Wind, wave height and current separate speed over ground from speed through water, and against a current or in heavy weather the same engine revolutions produce different progress. But not every delay is weather: a vessel that knows its berth is still occupied can deliberately slow down rather than anchor and wait, and that is a decision, not a delay. For those ashore the difference matters, because in one case the arrival slips, while in the other the arrival is deliberately aimed at an agreed window. A system that records the reason alongside the expectation changes the conversation: not that it will be later, but what there is still to choose.

Multiple sources, one expectation

In practice, several expectations for the same vessel circulate side by side: the manually entered field in the AIS message, the agent's notification, the captain's message, your own model's calculation and the window the terminal has scheduled. They contradict one another, and that is normal. The work lies in bringing them together: record for each expectation who issued it, when, and on the basis of which source, choose a precedence rule that can be explained, and show which source is currently leading. A system that shows a single figure with no provenance gets ignored at the first conflict.

Also keep every expectation you have ever issued. Only then can you measure the error afterwards: how far off were you at a horizon of one day, and how far off at a horizon of half a day, per route and per vessel type. That error curve is the only honest proof that your model beats an experienced estimate. It is also the basis for a margin instead of a single figure, and a planner who knows the margin plans differently. See also custom planning and scheduling software if the emphasis is on the scheduling itself.

Finally: a shift is only news if there is something to be done about it. So don't report every adjustment, only the adjustment that crosses a threshold that means something to the recipient: a shift change, a window at the terminal, a booked loading time, the moment at which transport can still be rebooked. Without those thresholds the result is predictable: everyone switches the notifications off and nobody looks any more.

  • Distance along a navigable route, not as the crow flies
  • Speed smoothed over a window
  • Port limit, berth and being moored named separately
  • Tide, pilot, tug and berth occupancy factored in
  • A source, a time and a sender for each expectation
  • Explainable precedence when expectations conflict
  • Every expectation issued is kept so the error can be measured
  • Notify only at a planning-relevant threshold
Not sure about a large project yet?

Test your idea first: a working prototype in 1 day

With OneDayBuild, we make your idea tangible in a single day for €1,150, so you know whether further development is worth the investment. Decide to go ahead with the full build? Then we deduct the cost in full.

View OneDayBuild →

Geofences, port approach and the voyage file

A geofence is nothing more than an area on the map that the system reacts to: a port area, an anchorage, a berth, a canal section, an emission zone, an area you want to watch for insurance or sanctions reasons. As soon as a track enters or leaves that area, an event is created. That is the idea, and in practice it is harder than it sounds, for three reasons.

The first is a vessel that lingers on a boundary. A tug working along the edge of a port area, or a ship anchoring just outside the line, then produces a series of entries and exits that mean nothing. The fix is not a tighter boundary but a threshold: a minimum dwell time before an entry counts, a margin between the entry boundary and the exit boundary, and a maximum number of transitions per period. The second is that zones are nested. A berth sits inside a terminal, which sits inside a port area. Events should follow that nesting, so that an arrival at the berth is not detached from the arrival in the port, and a report does not double-count. The third is the quality of the zones themselves. Official port limits and berth boundaries are a different thing from a hand-drawn box, and sooner or later the difference comes back as an argument about who was late.

The port call as a sequence of events

A port call is not a state but a chain: arrival at the port limit, anchoring and anchor up, pilot on board, first line, all fast, start and end of cargo handling, last line, departure. Some of these moments can be derived from position data with reasonable certainty, for example by combining speed below a threshold inside a berth zone with a dwell time. Others cannot, and only come from a message from the agent, the terminal or the port authority. For each event, record how it came about: derived, reported or set manually. That distinction is exactly what you need when an invoice or a claim depends on it later.

International agreements are pushing this exchange towards a single digital window per port, where the vessel's notification and the cargo and crew data are submitted electronically instead of on forms. In the Netherlands this runs through the national single window and through the port community systems that pass notifications on to authorities and terminals. For your own system that rarely means you file the declaration yourself, but it does mean you must have the right data ready at the right time and be able to show when it was ready. You can read more about the systems around the port call in port software development.

Voyage file and voyage history that survive a question after the fact

A voyage is not a technical fact but an agreement. Does a voyage run from departure berth to arrival berth, per charter, per cargo, or per set of ports that together form one round trip? Each choice is defensible, but one choice has to be fixed and applied consistently everywhere, because otherwise figures in two reports will never match again. Around that voyage you attach the track, the events, the revised estimates, the documents, the port costs and the consumption. That is how a voyage file becomes something other than an archive: it is the place where a later question gets answered.

Those questions will come. Why was this vessel later than reported. Was the waiting time charged correctly. Where was the vessel at a given moment. Which route passed through which area, and does that match what the transport documents say. Reporting on fuel consumption and emissions under European rules also works per voyage and distinguishes between the ports involved in a trip, so how you delimit voyages is no longer a cosmetic choice. If you only set that up afterwards, you get to recalculate the history using zones and rules that have since changed.

Three design rules follow from this. Record events immutably, with source and measurement time, and correct by adding a new event rather than overwriting the old one. For every derived event, keep the version of the zone and of the rule that produced it, so that a redrawn geofence does not rewrite history. And separate measurement from interpretation in the model, so you can improve your rules without losing your measurements.

AIS decoding De-duplication on measurement time Vessel identity with timeline Route over water Geofence events Voyage segmentation Track simplification Time-series storage Spatial index Webhooks and API

Integrating with planning, terminal and customs

The value of vessel tracking lies not in the map but in the handover. A position that only appears on a screen changes nothing; an arrival forecast that lands automatically in your transport planning changes a planner's day. On the landside, there are usually three directions. The first is your own planning, whether that is a transport management system, an order package or a custom board. See building a TMS if that side is still missing, and a track and trace system if the customer needs to be able to follow along themselves.

The second direction is the terminal and the hinterland. This is about berth planning, about windows for delivering and collecting cargo, and about booking time slots for road transport. A shifting arrival is only truly processed once the corresponding slot shifts with it, and that requires an integration that not only reads but also changes, and sends a confirmation back. What such a booking flow looks like is described under the slot booking portal.

The third direction is the declaration. Cargo data must be known before arrival, and the pre-arrival notification and the manifest are tied to the same arrival moment you are tracking. Your system is rarely the place where the declaration is made, but it is the place that knows when each piece of data became available and which change came afterwards. For that side, automating customs handling is the adjacent building block.

Technically, these integrations run through two worlds. In one, structured messages are exchanged according to agreements that have lasted for decades; in the other, it is all about web interfaces with tokens and event streams. Both require the same foundation: messages that survive being sent twice without a double booking, a correlation identifier that keeps request and response together, a queue with retries, and a place where failed messages remain visible rather than disappearing. And the hardest part is not technology but master data: your vessel list, the terminal's and the declaration's must be explicitly mapped to one another, managed in the system and not in one employee's head.

When an existing platform is enough

Being honest about the limits of custom is part of this. There are mature providers of AIS data and tracking platforms with worldwide coverage, satellite reception, port call data and ready-made map layers. Building the reception yourself is a bad idea for almost everyone: you buy that data in. And if you mainly want to view, filter, track a fleet and receive a notification when something happens, a subscription to such a platform is the wiser route, and we would rather tell you that now than halfway through a project.

Custom starts to pay off once the decision logic is yours. That is the case when you need to combine multiple sources into one accountable expectation, when your voyage definition and reporting obligations do not fit a platform's model, when the history must be kept in your own environment because claims and accountability depend on it, or when the position only becomes valuable in combination with your orders, contracts and agreements. Often the best form is a combination: buy in data and coverage, and build the layer that turns it into decisions yourself. If your question goes beyond position and arrival, custom maritime software is the broader entry point, and for the chain beyond the port, supply chain management software.

One practical warning goes with that choice. Read the licence terms of your data source before you set up an archive. Storing, passing on and showing data to third parties are all covered there, and the outcome partly determines how you set up storage and permissions. If you build it yourself, keeping up with changing regulations and message formats is down to your own organisation.

  • Buy data and coverage, build the logic yourself
  • A platform when the question is viewing, filtering and alerting
  • Custom when you need your own forecast across multiple sources
  • Custom when the history must live in your own environment
  • Match on identifier, never on vessel name
  • Messages that can be repeated, with a queue and visible failures
  • Master data mapped and managed in the system
  • Data source licence terms read first

Frequently asked questions about vessel tracking software

Because that ETA is entered by hand, as are the destination and the draught. The field only changes when someone on board updates it, and it often contains an abbreviation, a port code or even the value from the previous voyage. Read it as an intention, not as a planning figure, and calculate your own arrival forecast from position, course, speed and the route over water.

It shows the difference between a measured and an estimated position. Shore-based reception reaches only as far as the VHF horizon, satellite reception has repeat intervals, and in busy areas messages overlap. A sound system therefore shows the age of the last measurement, extrapolates between two measurements using course and speed, marks those interim values as estimates, and raises an alert as soon as a vessel has been silent for longer than is normal for its source.

That depends on the licence, and the question needs to be settled up front. Commercial data sources set conditions on retention, onward delivery and making data visible to third parties, and those conditions partly determine how you set up your archive. Also pay attention to the positions of small vessels: as soon as a position can be linked to a person, data protection comes into play and a purpose and a retention period must be defined.

No. Shipbuilding software supports the building process: sections, engineering, outfitting and the build file of a vessel that is not yet sailing. This page is about operating a vessel that is already at sea: where it is, when it arrives and what that means for planning ashore. If you are looking for the shipyard side, the page on shipbuilding software is the right starting point.

For viewing, filtering and alerting, often yes. There are mature AIS data providers and tracking platforms with worldwide coverage, satellite reception and port call data, and rebuilding the reception yourself rarely brings any advantage. Custom software becomes interesting once the decision logic is yours: your own arrival forecast across multiple sources, your own voyage definition, or history that has to live in your own systems for claims and accountability.

The source code and documentation belong to the client and are kept in a repository you have access to yourself. Handover includes a description of the architecture, the data models, the integrations and the deployment, so that another party can take over without having to reverse-engineer anything first. We work with mainstream technology, because that determines how many parties can pick up the maintenance.

Related services

Shipbuilding software

If you are not looking for operations but for the shipyard, with sections, engineering, outfitting and the build file of a vessel that is not yet sailing, shipbuilding software is the right starting point. There, the vessel is the product; here, it is a moving point with cargo and an arrival time.

Custom maritime software

If your question goes beyond position and arrival, covering maintenance, certificates, crew, charters or invoicing per voyage, then custom maritime software is the broader starting point, with vessel tracking as one of its components.

Port, terminal and the chain beyond

For the approach and handling in the port itself: port software development. For scheduling transport against an arrival, building a TMS, and for letting customers follow along, a track and trace system.

From AIS message to quay planning

Tell us which sources you use now, how the arrival forecast reaches the planner and what happens when a vessel drops out of coverage. Those three answers usually show whether you need to buy in data, build logic, or both.

Edit content