The New Mate on the Bridge: Artificial Intelligence
March 1, 2020
It seems every week we read of new developments in artificial intelligence (AI) and autonomous vessels, aircraft and vehicles. It is likely that in the time between writing this article and its publication there will have been announcements of new AI vessel applications – be it navigation, traffic management, machinery controls or communication. This may give the impression that a switch will soon be thrown and artificial intelligence will control the majority of vessels plying our harbors and oceans.
The truth is there has been an incoming tide of technology that has been flooding into vessel operations for hundreds of years and artificial intelligence is the most current component of that rising tide. Even though the pace of AI innovation and implementation has accelerated, the autonomous systems being proposed and implemented will require a “human-in-the-loop” for some time to come, whether that human is on the bridge or at some remote location. If the human/AI partnership leverages our collective strengths and manages our collective weaknesses it will result in improved vessel safety and efficiency. But in order to accomplish this we must identify the strengths and vulnerabilities of the mariner and AI and ensure the interface between the two is practical and functional.
The Single Person Bridge is a good example of how AI and Humans can be partners, not competitors on the bridge. The single person bridge is common on workboats, tugs and even some ships. The man or woman on the bridge, acting alone requires a nimble mind, one that must engage in a continuous decision-making process of acquiring, analyzing, and prioritizing data, taking appropriate action and evaluating the results. In addition he or she must have a “mind’s eye” that can rapidly shift perspectives – from looking out the pilothouse window to the 30,000-foot view of a chart with its implications of routing and navigation hazards.
The technological evolution that has been taking place on vessels has always used one benchmark for measuring success – can technology perform task(s) that equal or surpass the performance of an intelligent, skilled mariner. One of those benchmarks is the ability of a human to make effective, complex, risk-based decisions in a short time period. The gift of AI research and development is that it has required a deep exploration of the human brain and how our mind acquires, analyzes and filters information to create good decisions. As a result we have a clearer picture of the human mind and its capabilities.
When humans are on our game our brains are exquisite data processors and decision makers; we make not only data driven decisions, but decisions based on experience and intuition. A trained, competent and experienced mariner has the capability to draw conclusions from low amounts of data, create new, successful solutions that are not always supported by data, rapidly shift perspectives, filter out non-relevant data in the moment and apply general intelligence, while also mastering conversational skills. Think of the tug master going to a new port in Alaska that he or she has never been to before. There is low “data” but experience, intuition and the ability to transfer knowledge to a new situation (general intelligence) results in a successful transit.
We are however also vulnerable to having that process interrupted – the “human factor” in incidents. We are subject to fatigue, boredom, distractions and other factors that diminish our brains capability.
Artificial Intelligence has the capability to outperform humans in completing singular repetitive tasks, compiling and analyzing large amounts of data, and making rapid, complex mathematical computations. An ARPA radar has the capability to rapidly acquire vessel targets and constantly calculate courses, speeds, timing and distance of approach.
Like the human brain, Artificial Intelligence also has vulnerabilities. AI relies on massive amounts of data and will make poor decisions or provide false conclusions if there is a small amount of data, or the data is interrupted, corrupt or inaccurate. Additionally AI has no tolerance for loss of power or power variance. The electronic chart program that places the vessel “on land” due to a weak or broken GPS signal or voltage surge from the vessel’s generator is an example of AI vulnerability. Data is the lifeblood of AI and energy is its heart.
The key to leveraging the strengths of the human mind and artificial intelligence on the bridge is the interface between the two. If done well the use of Artificial Intelligence on the bridge will be like having a perfect trusted mate. AI will be capable of completing simple or complex assigned tasks, simultaneously integrate critical functions, problem solve, monitor non-critical conditions or tasks, communicate with a “like mind” to the human in control, all while remaining contextually relevant to the situation on the bridge.
For the Human/AI relationship to work effectively and seamlessly the User Interface (UI) must be intuitive, graphical, fast switching, allow both associative and sequential information retrieval, and be customizable. In other words it should be an extension of the mariner’s mind, the way he or she retrieves, filters, organizes and processes information and problem solves the situation at hand.
Take the example of our tug master towing a barge, navigating a narrow channel in the Inside Passage to Alaska, with opposing traffic, in fog, and experiencing a steering failure that turns the tug to towards a meeting vessel.
In this situation the tug master acting alone will apply different “thinking modes” in rapid succession. He will apply sequential or “checklist” troubleshooting of the cause. Is there a steering pump or power failure? Has the autopilot or jog lever failed? Is there debris jammed in the rudder? The operator will be fast switching between a horizontal view out the pilothouse window and the mind’s eye, birds eye view of the tug in relation to surrounding navigation hazards and vessel traffic– all while using general intelligence and associative thinking to create possible solutions. In other words gathering real time data and pulling information from mental files of experience that apply to the context of this new situation.
Today’s Human in the Loop/Artificial Intelligence systems have the capability of providing key information – detailed machinery conditions, vessel traffic, and own vessel position, course and speed within the geographic area. AI can function as that mate on the bridge but must be linked with the user so that critical information can be quickly retrieved and its accuracy trusted.
The interface must make “sense” to the user. It should empower the human to “think” outside his or her head with the ability to rapidly shift perspective from the global to detail, retrieve and filter data and hypothesize solutions as quickly as our brains can process. In an ideal world the mariner will be able to interact with AI to retrieve information with no more than one or two steps, switch rapidly between perspectives and not be overwhelmed by volumes of irrelevant data.
In this idealized scenario artificial intelligence would have rapidly diagnosed that steering pump No. 1 had failed, switched to steering pump No. 2, provided collision avoidance options and communicated this through an intuitive interface with the tug master. The tug master would still have the option to query AI to confirm that critical data was accurate and relevant within the context of the situation at hand. In other words confirming that the proposed solutions will result in a safe meeting with the oncoming vessel and not place either vessel in navigation danger.
Like an incoming tide, there’s no stopping the implementation and influence of Artificial Intelligence into vessel operations. However, we are at a critical juncture in defining the interface between AI and users. Mariners should not “pass the helm” to non-mariners in the development and implementation of artificial intelligence on vessels. Rather there should be collaboration between software developers, programmers, hardware engineers and mariners to ensure that the human/AI link leverages the strengths and recognizes and manages the vulnerabilities of both. This is not limited to a touch screen – it includes collaboration in developing safeguards, controls, training and security measures. Only then will a successful partnership between the mariner and artificial intelligence make vessel operations safer and more efficient.
Captain Slesinger’s maritime career has spanned over 40-years as a tug master, educator, author, auditor and surveyor. He brings a unique and broad perspective to issues facing the maritime industry and can be reached at email@example.com.