Invalid MMSI numbers, a real problem
The above (bigger here) is clipped from an interesting report on MMSI “anomalies” that dropped into my e-mail box. It makes a good case for why the FCC decided that Class B AIS devices should have their MMSI numbers input in a controlled way, and why, as noted in the new Report and Order (page 20), it may extent those controls to Class A. As I understand it, invalid MMSIs don’t affect AIS’s primary collision avoidance function—unless there are duplicates in the same target area—but they can mess up advanced functions like DSC calling (and security monitoring). Here’s more detail on the report:
* Twice in the last year a U.S. Coast Guard system called NAIS (National Automatic Identification System) was used to “collect 30 to 40 million position reports from approximately 5000 AIS-equipped vessels off the US coast and in US waters during a 24 hour period.” (NAIS must have a pretty good receiver network!)
* The MMSI numbers were analyzed and, as you can see above, 3.7% were found to be invalid on one of the test days, 4.6% on the other. Specifically 4.2% were found to have an illegal MID, which is the three digit country ID that begins each ship’s MMSI, and which are only supposed to run from 200 to 800. (Here’s a USCG MMSI explanation, and here’s an ITU list of assigned MIDs.)
* The most common invalid MMSI was 1193046, which is the default of older Nauticast A units, and a problem noted here by commenter “deckofficer” on the 22nd. I know ACR/Nauticast is working on this, but it’s hard to fix electronics that are usually roaming the oceans.
* Interestingly, the primary purpose of this testing was not to find invalid MMSIs but to see if NAIS could use the GPS “Sync State” included in AIS messages to monitor for areas of GPS system interference or anomalies (kind of like GPS cell phones monitoring traffic). I don’t know the results, but if it works it would be yet another good use for the AIS system.
PS A study by Abbas Harati-Mokhtari—AIS: A Human Factors Approach (PDF here)—documents more static data errors, MMSIs included, and illustrate the “Swiss Cheese” model of potential accidents: