Maritime Applications for Machine Learning and Data Science

Network Communication Hardware Shares Data Between Instruments. J. Durham


As data collection and analysis seeps into the conservative maritime industry we are beginning to see cracks in this grand old system based on tradition and trial and error.

When I say old, I don’t mean the 1980’s or even the 1880’s. Opinions vary on the exact time shipping became modern as any sailor or longshoreman would recognize today. When the English and Dutch began standardizing shipping practices between their two countries to increase safety and profit the practices soon spread.

This was happening in the late 1600’s and if you wanted to be part of the shipping economy you looked to the English, Dutch, and to a lesser extent, the Spanish.

Today we can see another example of this technology clustering having a lasting effect on a growing industry. Starting in the 1960’s California became the place to be if you were a part of the new generation of electronics companies. Standards were set and the jargon and culture of Silicon Valley we have today is a direct result of this small but powerful geographic area. In addition to soft concepts like jargon, deep architectural standards like eight digit binary numbers were solidified. The same sorts of transactions and relationships were also true of shipping as it became a standardized business.

Global shipping today represents many cultures and values and it must be responsive in the era of pervasive media and digital content, or it will be demonized and lose the minimal goodwill available to a largely invisible industry.

Yet when they see a good idea, which is one that will save money, it is quickly adopted by the upper levels of management. Workers are sometimes resistant to change for fear of job loss. Both of these behaviors occurred when the intermodal shipping container was introduced in the 1950’s as a cost saving measure.

Automation of ships and ports will be a much more difficult journey than the one fought by the proponents of the modular container in the early days. Job loss among longshoremen was real and the sealed container ended the common practice of pilfering some of the cargo. This was common, and still happens occasionally today, with some Masters sanctioning the activity. The fact was it took much less labor to load a ship with large boxes than it did individual sacks or grain or crates of equipment that varied in size and weight.

Automated ships and ports will eliminate some jobs that are hazardous or dirty and most people will not miss this kind of work. Jobs that have a high value are a different story. A totally autonomous ship is in the future and that means less risk for deck hands while increasing profits substantially for ship owners. The savings are similar to autonomous car savings, less risk, less insurance costs, more efficient operation, better traffic management, and elimination of human error.  

The elimination of human error on the operational level is important since most accidents happen because of failure due to poor design or human error in some aspect of vessel operation.

Machine learning is giving us insights into the marine world we never had before, and some of the revelations are contrary to accepted beliefs. A good example of this is the Digital Deck product for commercial fishermen that was developed by the company Point 97. Digital tracking of fishery data collected by fishermen in their daily operation led to discoveries local regulators used to manage fish stocks and reduce the resources needed to search for illegal fishing activity. Automatic importing of data allows for near real-time insights not only for regulators, but also fishermen.

Now a new class of data is emerging with the announcement from MIT that they have developed an algorithm that monitors wave data in order to predict rouge wave formation. Rouge wave are giant and often deadly waves that form in the open sea where two wave fields combine.

Rouge waves are often in the form of a peak and not a long running wave like those produced by tsunami.

This is a new class of data because it needs quick action to work. Automatic avoidance systems are not generally accepted and permission to change course could take minutes. Rouge waves form and do their damage quickly so the best use of this data is in an automatic system that will change course or turn to face the wave bow-on. This will make mariners uncomfortable but the alternative is worse.

Classification societies, insurers, and regulators all stand in the way of more automation but like self-driving cars, they will be accepted because of greater convenience and cost savings.

We have already reached a point where there is too much data for one human to absorb. All that data on the helm displays can be better managed by computers which already run many parts of a modern ship. The few sailors that do remain on ships of the future will likely be technicians with few hands on duties unless automated maintenance and repair systems fail.

mla apa chicago
Your Citation
Bruno, Paul. "Maritime Applications for Machine Learning and Data Science." ThoughtCo, Mar. 1, 2016, Bruno, Paul. (2016, March 1). Maritime Applications for Machine Learning and Data Science. Retrieved from Bruno, Paul. "Maritime Applications for Machine Learning and Data Science." ThoughtCo. (accessed March 21, 2018).