Skip to main content

AIS message decoding

Project description


PyPI license codecov downloads CI Documentation Status

AIS message encoding and decoding. 100% pure Python. Supports AIVDM/AIVDO messages. Supports single messages, files and TCP/UDP sockets. This library has been used and tested extensively in representative real-world scenarios. This includes tests with live feeds from Spire, the Norwegian Coastal Administration and others. I test each major release against a selection of public and non-public data sources to ensure the broadest possible compatibility.

You can find the full documentation on readthedocs.

I also wrote a blog post about AIS decoding and this lib.

Binary releases (Debian packages) are provided by the pyais-debian starting with version v2.5.6. They are downloadable under the Releases page.


Jetbrains Logo

This project is a grateful recipient of the free Jetbrains Open Source sponsorship. Thank you. 🙇


AIS (Automatic Identification System) is a communication system that allows ships to automatically exchange information such as vessel identification, position, course, and speed. This information is transmitted via VHF radio and can be received by other ships and coastal stations, allowing them to accurately determine the location and movement of nearby vessels. AIS is often used for collision avoidance, traffic management, and search and rescue operations. AIS messages are often transmitted via NMEA 0183.

NMEA (National Marine Electronics Association) is an organization that develops and maintains standards for the interface of marine electronic equipment. NMEA 0183 is a standard for communicating marine instrument data between equipment on a boat. It defines the electrical interface and data protocol for sending data between marine instruments such as GPS, sonar, and autopilot.

Here is an example of an AIS sentence:


This AIS sentence is known as a "Position Report" message and is used to transmit information about a vessel's position, course, and speed. AIS messages are transmitted in digital format and consist of a series of comma-separated fields that contain different types of data. Here is a breakdown of each field in this particular sentence:

  • !AIVDM: This field indicates that the sentence is an AIS message in the "VDM" (VDO Message) format.
  • 1,1: These fields indicate the total number of sentences in the message and the current sentence number, respectively. In this case, the message consists of a single sentence.
  • : This field is left blank. This field can contain the sequence number.
  • B: This field indicates the communication channel being used to transmit the message. In this case, the channel is "B".
  • 15MwkT1P37G?fl0EJbR0OwT0@MS: This field contains the payload of the message, which is encoded using a variant of ASCII known as "Six-bit ASCII". The payload contains information such as the vessel's identification, position, course, and speed. 0*4E: This field is a checksum that is used to verify the integrity of the sentence.

pyais is a Python modul to encode and decode AIS messages.


The project is available at Pypi:

$ pip install pyais


There are many examples in the examples directory.

Decode a single part AIS message using decode()::

from pyais import decode

decoded = decode(b"!AIVDM,1,1,,B,15NG6V0P01G?cFhE`R2IU?wn28R>,0*05")

The decode() functions accepts a list of arguments: One argument for every part of a multipart message::

from pyais import decode

parts = [

# Decode a multipart message using decode
decoded = decode(*parts)

Also the decode() function accepts either strings or bytes::

from pyais import decode

decoded_b = decode(b"!AIVDM,1,1,,B,15NG6V0P01G?cFhE`R2IU?wn28R>,0*05")
decoded_s = decode("!AIVDM,1,1,,B,15NG6V0P01G?cFhE`R2IU?wn28R>,0*05")
assert decoded_b == decoded_s

Decode the message into a dictionary::

from pyais import decode

decoded = decode(b"!AIVDM,1,1,,B,15NG6V0P01G?cFhE`R2IU?wn28R>,0*05")
as_dict = decoded.asdict()

Read a file::

from import FileReaderStream

filename = "sample.ais"

for msg in FileReaderStream(filename):
    decoded = msg.decode()

Decode a stream of messages (e.g. a list or generator)::

from pyais import IterMessages

fake_stream = [
for msg in IterMessages(fake_stream):

Live feed

The Norwegian Coastal Administration offers real-time AIS data. This live feed can be accessed via TCP/IP without prior registration. The AIS data is freely available under the norwegian license for public data:

Data can be read from a TCP/IP socket and is encoded according to IEC 62320-1:

  • IP:
  • Port: 5631

Refer to the examples/ for a practical example on how to read & decode AIS data from a TCP/IP socket. This is useful for debugging or for getting used to pyais.


It is also possible to encode messages.

:exclamation: Every message needs at least a single keyword argument: mmsi. All other fields have most likely default values.

Encode data using a dictionary

You can pass a dict that has a set of key-value pairs:

  • use from pyais.encode import encode_dict to import encode_dict method
  • it takes a dictionary of data and some NMEA specific kwargs and returns the NMEA 0183 encoded AIS sentence.
  • only keys known to each message are considered
    • other keys are simply omitted
    • you can get list of available keys by looking at pyais/
    • you can also call MessageType1.fields() to get a list of fields programmatically for each message
  • every message needs at least two keyword arguments:
    • mmsi the MMSI number to encode
    • type or msg_type the type of the message to encode (1-27)

NOTE: This method takes care of splitting large payloads (larger than 60 characters) into multiple sentences. With a total of 80 maximum chars excluding end of line per sentence, and 20 chars head + tail in the nmea 0183 carrier protocol, 60 chars remain for the actual payload. Therefore, it returns a list of messages.

from pyais.encode import encode_dict

data = {
    'course': 219.3,
    'lat': 37.802,
    'lon': -122.341,
    'mmsi': '366053209',
    'type': 1,
# This will create a type 1 message for the MMSI 366053209 with lat, lon and course values specified above
encoded = encode_dict(data, radio_channel="B", talker_id="AIVDM")[0]

Create a message directly

It is also possible to create messages directly and pass them to encode_payload.

from pyais.messages import MessageType5
from pyais.encode import encode_msg

payload = MessageType5.create(mmsi="123", shipname="Titanic", callsign="TITANIC", destination="New York")
encoded = encode_msg(payload)

Under the hood

graph LR
    raw -->|"!AIVDM,1,1,,B,6B?n;be,2*4A"| nmea
    nmea[NMEASentence] -->|parse NMEA sentence layer| ais[AISSentence]
    ais -->|decode| final[AISMessage]

Decoding each AIS message is a three step process.

At first, the NMEA 0183 physical protocol layer is parsed. The NMEA layer is the outer protocol layer that is used by many different sentences/protocols for data transmission. Just like Ethernet can be used as a data link protocol to transfer data between nodes, the NMEA protocol can be used to transmit data between maritime equipment.

After the raw message was parsed into a NMEASentence, the inner protocol layer is parsed. While there are tons of different inner protocols that build upon NMEA, pyais currently only supports AIS sentences. Every AISSentence holds basic information about the AIS message like:

  • the AIS message ID
  • the number of fill bits required for ASCII6 encoding
  • the fragment count and fragment number
  • the actual AIS payload
  • the sequence number

Finally, the AIS payload is decoded based on the AIS ID. There are 27 different types of top level messages that are identified by their AIS ID.

Tag block

Some messages may look strange at first. Typical AIS messages look roughly like this:


But sometimes such messages look something like this:


These three messages are the same messages as above - only with a prefix, the so called tag block. Tag blocks are essential key-value pairs that are wrapped between \s. Every valid NMEA sentence may have one of the these tag blocks. Tag blocks are used to hold extra information and somewhat similar to Gatehouse messages.

A tag block consists of any number of comma-separated key-value pairs, followed by a checksum:

  • s:2573135,c:1671620143*0B -> s:2573135 & c:1671620143 & 0*B

The checksum is the same as for all NMEA messages. Regarding the key value pairs:

  • each key is a single letter
  • each letter represents a field:
    • c: Receiver timestamp in Unix epoch (e.g. 1671620143)
    • d: Destination station (e.g. FooBar)
    • n: Line count (e.g. 123)
    • r: Relative time
    • s: Source station (e.g. APIDSSRC1)
    • t: Text (e.g.g Hello World!)

Some things to keep in mind when working with tag blocks and pyais:

  • tag blocks are optional (a message may or may not have a tag block)
  • tag blocks are lazily decoded by pyais to save resources (need to call tb.init())
  • only some fields are supported by pyais (c,d,n,r,s,t)
    • unknown fields are simply omitted

How to work with tag blocks

from import IterMessages

text = """

messages = [line.encode() for line in text.split() if line]

with IterMessages(messages) as s:
    for msg in s:
        if msg.tag_block is not None:
            # Not every message has a tag block
            # Therefore, check if the tag block is not None
            # Also, it is required to call `.init()`, because tag blocks are lazily parsed
            # Print the tag block data as a dictionary

Gatehouse wrappers

Some AIS messages have so-called Gatehouse wrappers. These encapsulating messages contain extra information, such as time and checksums. Some readers also process these. See some more documentation here.

As an example, see the following, which is followed by a regular !AIVDM message


Such messages are parsed by pyais only when using any of the classes from e.g. FileReaderStream or TCPStream.

Such additional information can then be accessed by the .wrapper_msg of every NMEASentence. This attribute is None by default.

Communication State

The ITU documentation provides details regarding the Time-division multiple access (TDMA) synchronization.

Such details include information used by the slot allocation algorithm (either SOTDMA or ITDMA) including their synchronization state.

Refer to readthedocs for more information.

AIS tracker

pyais comes with the the ability to collect and maintain the state of individual vessels over time. This is necessary because several messages can give different information about a ship. In addition, the data changes constantly (e.g. position, speed and course).

Thus the information split across multiple different AIS messages needs to be collected, consolidated and aggregated as a single track. This functionality is handled by the AISTracker class.

NOTE: Each track (or vessel) is solely identified by its MMSI.

import pathlib

from pyais import AISTracker
from import FileReaderStream

filename = pathlib.Path(__file__).parent.joinpath('sample.ais')

with AISTracker() as tracker:
    for msg in FileReaderStream(str(filename)):
        latest_tracks = tracker.n_latest_tracks(10)

# Get the latest 10 tracks
print('latest 10 tracks', ','.join(str(t.mmsi) for t in latest_tracks))

# Get a specific track

Unlike most other trackers, AISTracker handles out of order reception of messages. This means that it is possible to pass messages to update() whose timestamp is older that of the message before. The latter is useful when working with multiple stations and/or different kinds of metadata.

But this comes with a performance penalty. In order to cleanup expired tracks and/or to get the latest N tracks the tracks need to be sorted after their timestamp. Thus, cleanup() and n_latest_tracks() have a complexity of O(N * log(N)). Depending on the number of messages in your stream this may or may not be good enough.

If you know that your messages in your stream are ordered after their timestamp and/or you never pass a custom timestamp to update(), you should set the stream_is_ordered=True flag when creating a new AISTracker instance. If this flag is set AISTracker internally stores the tracks in order. Thus, cleanup() and n_latest_tracks() have a complexity of O(k).


It is possible to register event listeners as callbacks, so that you are is instantly notified whenever a track is created, updated, or deleted.

import pyais
from pyais.tracker import AISTrackEvent

host = ''
port = 5631

def handle_create(track):
    # called every time an AISTrack is created
    print('create', track.mmsi)

def handle_update(track):
    # called every time an AISTrack is updated
    print('update', track.mmsi)

def handle_delete(track):
    # called every time an AISTrack is deleted (pruned)
    print('delete', track.mmsi)

with pyais.AISTracker() as tracker:
    tracker.register_callback(AISTrackEvent.CREATED, handle_create)
    tracker.register_callback(AISTrackEvent.UPDATED, handle_update)
    tracker.register_callback(AISTrackEvent.DELETED, handle_delete)

    for msg in pyais.TCPConnection(host, port=port):
        latest_tracks = tracker.n_latest_tracks(10)

Performance Considerations

You may refer to the Code Review Stack Exchange question . After a some research I decided to use the bitarray module as foundation. This module uses a C extension under the hood and has a nice user interface in Python. Performance is also great. Decoding this sample with roughly 85k messages takes less than 6 seconds on my machine. For comparison, the C++ based libais module parses the same file in ~ 2 seconds.


This module is a private project of mine and does not claim to be complete. I try to improve and extend it, but there may be bugs. If you find such a bug feel free to submit an issue or even better create a pull-request. :-)


Currently, this module is able to decode most message types. There are only a few exceptions. These are messages that only occur in very rare cases and that you will probably never observe. The module was able to completely decode a 4 hour stream with real-time data from San Francisco Bay Area without any errors or problems. If you find a bug or missing feature, please create an issue.

Known Issues

During installation, you may encounter problems due to missing header files. The error looks like this:


    bitarray/_bitarray.c:13:10: fatal error: Python.h: No such file or directory
       13 | #include "Python.h"
          |          ^~~~~~~~~~
    compilation terminated.
    error: command 'x86_64-linux-gnu-gcc' failed with exit status 1


In order to solve this issue, you need to install header files and static libraries for python dev:

$ sudo apt install python3-dev

For developers

After you cloned the repo head into the pyais base directory.

Then install all dependencies:

$ pip install .[test]

Make sure that all tests pass and that there aren't any issues:

$ make test

Now you are ready to start developing on the project! Don't forget to add tests for every new change or feature!


Use Docker to run your application inside a container. At first you need to build the image locally:

docker build . -t pyais

Afterwards, run the container (bash):

docker run -it --rm pyais /bin/bash

You can then run the examples inside the container:

python ./examples/


Python3.11 is faster

With Python3.11 significant improvements to the CPython Runtime were made:

Some results from the internal performance test:

3.10: Decoding 82758 messages took: 3.233757972717285

3.11: Decoding 82758 messages took: 2.5866270065307617

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pyais-2.5.9.tar.gz (80.4 kB view hashes)

Uploaded source

Built Distribution

pyais-2.5.9-py3-none-any.whl (44.0 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page