OpenAIS
Toggle Dark/Light/Auto mode Toggle Dark/Light/Auto mode Toggle Dark/Light/Auto mode Back to homepage

AIS Message Parsing

The goal of this container is to do the first stage parsing of raw AIS nmea strings and to put them onto the pub/sub message broker. This allows parsed data streams to be ingested from multiple sources (one AIS-i-MOV container per input stream), apply tags to the datastreams (one routing-key applied to each data stream) and allows multiple next stages to ingest data from multiple streams.

The parsed data message includes the raw AIS message as well as the metadata attached to the message from the data provider and the metadata created for each incoming data stream.

This is not an AIS decoder service. Decoding the NMEA strings is left to another container/service.

Input Data

The input data streams can be from a network port or file and may contain different metadata formats. The most typical form of AIS metadata is a timestamp attached to each message. The AIS protocol does not include a full timestamp so data providers typically provide a timestamp per message but there is no standard method of doing this. They may be an epoch timestamp that prepends a message, or an ISO timestamp as a footer for each message, or maybe some kind of JSON key/value pairing going on.

In general the AIS data will come in from a network port or from a pre saved data file. The data will be in a raw NMEA format and can be one row per message or multiple rows per message, depending on the message type:

!AIVDM,1,1,,A,13aEOK?P00PD2wVMdLDRhgvL289?,0*26
!AIVDM,1,1,,B,16S`2cPP00a3UF6EKT@2:?vOr0S2,0*00
!AIVDM,2,1,9,B,53nFBv01SJ<thHp6220H4heHTf2222222222221?50:454o<`9QSlUDp,0*09
!AIVDM,2,2,9,B,888888888888880,2*2E

Network Data Streams

A common source for AIS Data is AISHub. The data received from AISHub does not include any header or footer metadata but is instead streamed from the AIS coastal receiver network, through the AISHub servers, and finally to a server port where the data can be listened to. Since the data is near real time the time that the message was transmitted, the “event time”, can be assumed to be the same (or close enough) as the “server time” (when the message was received at the server).

Header/Footer Metadata

Header and footer metadata is stripped from the AIS message and included as “header” and “footer” items in the parsed dictionary object. The definition of “header” and “footer” is the string that is before and after the AIS message for each row. In some cases this string is a dictionary of objects (IMIS styled AIS data) and these are parsed as a dict.

File Data Streams

Some work still needs to be done to ingest AIS messages from a file. The goal being to be able to reprocess old AIS data using the same tools as live AIS data. This could be of use to get a new instance of OpenAIS up and running or to test new decoders/streaming processors.

Output Data

The output parsed data is held in a JSON dictionary and fed to RabbitMQ with a routing key. Here’s an example of a message:

{"ais": "!AIVDM,1,1,,B,H6<u`S19DEV0V1<QT4pN373<000,2*23\r", 
 "header": null, 
 "server_time": "2022-11-03T10:06:10.904079", 
 "event_time": "2022-11-03T10:06:10.904084", 
 "routing_key": "encoded_ais.aishub.all", 
 "multiline": false}

This is a single line message and does not include any header or footer information.

Multi-Line Messages

Part of the issue with AIS decoding is the existance of multi-line messages. While these almost always follow each other in the message stream there is no guarentee in the protocol that this will be the case. The AIS-i-Mov ingestor stores multiline messages and tries to gather the different parts together before publishing them. The below message contains 2 parts and has header information for each part.

{"ais": ["!AIVDM,2,1,5,A,53m66:800000h4l4000pu8LD000000000000000S2`t666JW0;hQDQiC,0*2B\r", "!AIVDM,2,2,5,A,P00000000000008,2*49\r"], 
"header": [null, null], 
"server_time": "2022-11-03T10:28:08.190386", 
"event_time": "2022-11-03T10:28:08.190391", 
"routing_key": "encoded_ais.aishub.all", 
"multiline": true, 
"msg_id": "5"}

Routing Keys

RabbitMQ uses routing keys to gather and distribute messages. The routing keys are strings seperated by periods. Each individual string can be replaced by wildcards so, for example, if you wanted a decoder that ingested all AIS data, regardless of source, the following routing key could be provided: “.encoded_ais.”. This would pull in all streams that had a “encoded_ais” term in them.

The general routing_key convention in the OpenAIS project is to include the provider of the data, go from most generic to most specific, and to label the data as either “encoded_ais” or “ais”. The routing keys can be configured to be almost anything in the config file so not too much emphasis is put on this. A good routing key would be:

<class of data>.<source of data>.<sub type of data>.<sub-sub type of date>

  • class of data: The generic class of data (encoded_AIS, AIS, VMS, SAR, etc)
  • source of data: The source or provider of the data (IMIS, Spire, AISHub, etc)
  • type of data: Some descriptor that groups the data source together (Coastal vs satellite, port AIS, fishing vessels only etc)
  • Some further subtype for the data: Not really used but can be.

So for example; if coastal AIS data was being received from a local port authority, but only type A receivers were being sent to the data stream the encoded data routing key could look like:

encoded_ais.port_authority.coastal.class_A

Config

The service is configured using an environment variable file. An example can be found in ./config/sample.env. This config file is loaded into the docker container on run (generally with docker-compose).

The config file is split up into several different catagories. Below is a short explanation of them.

#----------------------
#Project
#----------------------
PROJECT_NAME=ais_pipeline   - The name of the project, this is prepended to all the containers being run
UID=1000                    - UID and GID of the user that will end up owning files created by the containers
GID=1000
AIS_STYLE=NONE              - The header/footer/meta style expected from the AIS stream

#----------------------
# Source Server Socket
#----------------------
SOURCE_HOST=source.com      - The server to connect when receiving AIS data
SOURCE_PORT=8888            - Port to connect to on server
SOCKET_TIMEOUT=60           - Socket timeout, should kill and restart the connection if too quiet
CHUNK_BYTES=300             - The size of the chunk to take from the socket for batch processing

#----------------------
# Logging Info
#----------------------
LOG_NAME=Coastal_AIS.log    - The log file to write the encoded data to (to preserve the data)
LOG_DIR=/usr/local/ais_i_mov/logs/  - the DIR to write data log files to on container
MOUNT_LOG_DIR=./volumes/logs        - the DIR to write data log files to on host machine (docker volume)

#----------------------
# Sink_RabbitMQ 
#----------------------
SNK_RABBIT_HOST=rabbit.host.com     - The server details for publishing AMQP messages
SNK_RABBIT_MSG_PORT=5672
SNK_RABBITMQ_DEFAULT_USER=rory
SNK_RABBITMQ_DEFAULT_PASS=rory_pw
SNK_RABBIT_EXCHANGE=ais_decoder_test
PRODUCE_KEY=sink.key.test
ON_ERROR_DROP_MSGS=True

Deployment

The system is generally deployed using docker-compose. There is a compose file stored in the repo that shows how to start the container, mount volumes, use config files in the .env file. The steps to get this running on your system is:

  • Clone/Pull the repo to the target machine
  • Copy the ./config/sample.env to .env and edit it to reflect your environment
  • run “docker-compose up –build -d”

That will get the service running on your machine but will be pretty useless without any of the other services. There is work being done on a generic deployment that can be found in the deployment project in the OpenAIS namespace.