Over 90 percent of global trade travels by sea. What makes this overseas commerce possible is access to safe, reliable shipping lanes. But as ships grow in size and number, collisions are on the rise.
From 2011 to 2017, collisions doubled from 2,000 to nearly 4,000. Almost 75 percent of these incidents were due to human error, resulting in losses totaling well into the billions of dollars. Legacy navigation systems are also partly at fault, especially when traversing busy and narrow waterways. To adapt to more crowded oceans, crews need smarter tools. One potential solution is AI-powered navigation, which promises to make modern ships safer and more efficient. But convincing the cautious maritime industry to embrace AI will require carefully steering the conversation and process to overcome any obstacles in its way.
Exploring the Potential of Maritime AI
AI has already spurred innovation in the automotive, medical, and agriculture sectors. Semi-autonomous systems like Tesla’s Autopilot continue making cars safer. Google’s AI system for detecting lung cancer outperformed doctors in recent tests. And IBM’s Watson AI now helps farmers better predict weather and optimize yields. Seeing these breakthroughs, many forward-thinking members of the shipping industry now long to adopt similar innovations. And who can blame them? The potential applications of AI for the industry are staggering.
The most pressing need for AI is in the realm of maritime safety. AI can improve the detection of ships in crowded waterways, under poor visibility, and at a longer range. In high-risk scenarios, this will give crews the crucial time necessary to adjust course and avoid collisions. Another benefit of AI-based systems is that they can help in remediating human error. An AI-driven platform never tires or loses track of objects. In other areas, AI can learn vital skills from human crews. Working with humans, AI can observe the human actions taken to avoid a collision, then extrapolate lessons for reacting to similar difficult situations.
AI-powered navigational systems will also enhance reliability and efficiency. Current systems struggle to detect obstacles under less-than-ideal conditions. This often forces ships to either slow down or stop in poor visibility and at night. Improved detection capabilities will let ships travel at full speed for longer spans of their journeys.
Fewer delays in overseas commerce will mean smoother logistics and better business relationships. AI-powered navigation can also provide crews with more reliable information. By learning from real-time voyage data, AI systems can spot patterns and distill actionable insights. For crews drowning in alerts from manual sensors, having a way to filter important signals from amidst the noise will be a godsend.
Building AI systems for maritime navigation requires a few key considerations. Data is the lifeblood of any AI system, but collecting data at sea poses unique challenges. Ships spend a majority of their time in the open ocean, where scenarios from which AI systems need to learn congested waters and interactions with other vessels, for instance, are infrequent. Connectivity at sea is also limited, which makes storing and transmitting data difficult. One upside is that ships already have many sensors installed on board. Connecting these existing sensors to a system that can coordinate them and process their data is the best approach. Such hybrid systems can provide crews with better navigational data right away, while only requiring minimal training.
Overcoming the Challenges
Despite its interest in leveraging AI’s benefits, the maritime industry’s conservative approach poses a challenge when it comes to adoption. Strict regulations and high financial stakes leave decision-makers wary of taking very expensive risks on systems they are not sure they can rely on. However, it’s impossible to deny that with today’s systems and sensors falling short, the current safety situation is untenable. With millions of dollars at stake per voyage, just one collision can knock an entire company off course. If incorporating new sensors or enhancing existing ones can reduce this risk, early adopters might be open to it. This will mark a critical first step. These initial installations will, in turn, begin collecting vital training data to kickstart the AI learning process. Each mile logged through the system will strengthen the underlying AI and produce powerful insights to further improve navigation.
However, ensuring the adoption of maritime AI will require a few things. First, ships running AI systems must be equipped with sensors that identify and collect only the most crucial data and do so reliably. Second, these systems must be globally accessible with adjustments made for different languages and local protocols. And last but not least, key industry players must be open to piloting AI-backed navigation systems that gather data during voyages. Not only will this catalyze Deep Learning, but once these AI-backed systems start saving lives or preventing accidents, the results will speak for themselves. Assuming that maritime AI can weather the storm of early skepticism, it can help ensure smooth sailing for an industry in great need of innovation.