For the past several years, the debate about whether autonomous vessels can comply with COLREGs, or the International Regulations for Preventing Collisions at Sea, has heated up. Those in the industry question whether autonomous or unmanned vessels can follow the “rules of the road” that were designed to be used by human sailors at sea. The next question is whether COLREGs, published 1972, require modification to accommodate modern-day vessels that may utilize reduced-crew or unmanned configurations. As our industry is revolutionized by advanced technologies, such as autonomy and advanced perception, these questions have never been more pertinent.
A closer look at COLREGs
As background, the International Maritime Organization (IMO) established COLREGs to define the navigation rules to be followed by ships and other vessels at sea, for the purpose of preventing collisions between them. There are dozens of rules in place, which provide guidance on everything from safe speeds and lookouts to overtaking other vessels and crossing situations. Complicating matters, despite the regulations’ practical and objective guidance, they sometimes necessitate subjective deviations in accordance with the vessel’s circumstances or a captain’s experience.
Despite the international adoption of COLREGs, accidents between vessels still occur, at rates only slightly lower than what was seen in when the regulations were first put into place. It’s widely accepted today that a root cause of most incidents on the water are highly attributable to human error. These missteps range from fatigue to distraction, even boredom and negligence – despite a myriad of passive technologies and sensors that provide reliable real-time navigational data to operators.
So, how can commercially available autonomous technology help to mitigate the risks of human error while also following guidelines that sometimes require subjective action? This has been a primary area of focus for industry leaders like Sea Machines for years, and the short answer is that it’s an evolving solution.
Autonomy: Factoring in COLREGs and mitigating human risk
Sea Machines’ autonomous systems are proven to increase the productivity and efficiency of workboats and ships, while reducing the risk of incident via tested collision avoidance technology. Across all industries, automated systems relieve fatigue and allow human operators to focus on higher-level jobs that can help keep them alert and on-task. Additionally, autonomy offers redundancy for crews, providing a second set of “eyes” on vessel navigation, preventing avoidable accidents related to human error.
Logging more than 2,500 hours of on-water testing, the company’s SM300 autonomous command and control system is a self-piloting system for both manned and unmanned vessels, that also enables remote command. The SM300 provides operational domain awareness, telemetry and attitude to mariners located on or off the vessel via a chart-based user interface, tracked targets, video and warnings. These are generated by cameras, sensors and other equipment located on or off the vessel.
According to Sea Machines’ Jocelyn Lorrey, software engineer, the SM300 was developed around COLREGs, meaning that the system is designed to make the same decisions an experienced mariner might to keep crew and assets safe while vessels are underway.
“Our technology continually processes data from trusted sources – including radar, AIS, GPS, IMU and video – to develop a real-time understanding of a vessel’s surrounding marine domain. Based on this data, the SM300 prioritizes potential strategies for accurate navigation and avoiding collisions.”
She went on to explain that the SM300 automatically ranks perceived risks to the vessel using criteria such as speed, distance and more. Based on the resulting risk hierarchy, the SM300 chooses optimal maneuvers, with safety always as the priority. But any good mariner knows that a marine domain isn’t static and risk factors can change quickly. For this reason, the algorithm powering these automated decisions is designed for agility. Thus, as situational factors change, the perceived risk hierarchy adapts in an instant, and the the vessel’s behavior adjusts on a dime.
“The great part of COLREGs is that they have been clearly defined by experts,” Lorrey continued. “Engineers don’t have to guess what the boat should do – the success criteria is clearly defined and the boat responds according to the rules.”
This means that an autonomous vessel is programmed to automatically navigate at safe speeds, use all available means to determine risk of a collision, and follow the prescribed guidelines for avoiding a collision while underway – exactly as a human should. But what happens when there are no good options for a vessel? How could an autonomous system possibly make decisions in edge-case scenarios, when all possible actions only yield negative outcomes?
“Managing the edge-case scenarios requires a certain level of intelligence, because the system needs to recognize when and how to depart from the explicitly prescribed rules. While the SM300 system is programmed to prioritize human safety above all else, it is important to know that autonomy helps to prevent in extremis situations in the first place. Because the autonomy is constantly monitoring the environment without the human risks of distraction or fatigue, we know that the system should execute a safe and predictable behavior that prevents risk of collision, in some cases earlier than a human can.”
Humans are still important, but their roles will likely change
Since reacting correctly to a potential collision requires timely and accurate situational awareness, the role of a lookout is essential. The rule for maintaining a proper lookout states:
“Every vessel shall at all times maintain a proper lookout by sight and hearing as well as by all available means appropriate in the prevailing circumstances so as to make a full appraisal of the situation and of the risk of collision.”
Understanding this, how can unmanned vessels be programmed to stand in as a human lookout? The short answer is, they probably can’t, but we shouldn’t be convinced that “seeing” and “hearing” should be solely human functions. Based on the industry’s largest library of marine data sets, Sea Machines’ next product will incorporate a combination of AI and Computer Vision (AICV).
“With AICV, our system will use cameras to ‘see’ and classify what types of vessels are in our path, so that we can better prioritize our reactions,” said Lorrey. “In the future, the system could also use microphones to ‘hear’ whistle blasts to recognize other vessels and their alerts. This type of advanced perception provides higher confidence and will contribute to significantly reduced risk of collisions at sea.”
Today, autonomous systems are most frequently used to support on-board crews. In these circumstances, a human lookout can continue to serve the traditional role, as aided by the autonomy system. In the future, humans may see and hear oncoming vessels not from the bridge, but from a shoreside control station or second vessel, via camera feeds and speakers. From a remote location, tomorrow’s mariners will be supported by an intelligent autonomy system that either recommends or autonomously executes proper vessel behaviors to avoid collisions and ensure safe navigation.
The driver: Emerging technology or COLREGs?
In August, Lorrey presented on the topic of safe integration of autonomous systems at AUVSI’s XPONENTIAL ’21 event. From the stage, she examined the assured autonomy framework through the lens of building an autonomous system that aligns with COLREGs.
“It’s clear that autonomy can exceed what is possible for a human to do,” said Lorrey. “A human lookout gets tired, hungry and distracted. If autonomy is involved, then humans can manage the tasks that we’re better suited for. Autonomy is always cool under pressure. It can always determine the best move, without human emotions or stress.”
While the question about whether COLREG rules should be revised to accommodate new technology remains outstanding, it’s clear that we’re experiencing a seismic shift in the industry.
“All of us at Sea Machines are excited for what’s to come. Few others are using marine autonomy in real-world applications, like we are. Most are still in the research phase. In the near future, marine operators won’t be able to tell from a distance whether a vessel is driven by a human or autonomy. When that happens, we’ll be in a good place.”