operators have been able to deviate from the rules of the road. However, computers operate in a robotic man- ner and are designed to strictly adhere to their pro- gramming.3 This significant difference in the capability to follow the COLREGs was evident in the MAS400 transit. Generally, when the MAS400 was in a crossing situ- ation with one other vessel, she was able to make the proper decision to avoid collision. The issue became more complex when there were several vessels in a crossing situation. Most notably, she lacked the ability to effectively communicate with another vessel. When there are inherent conflicts in the application of the rules for multiple ships, the complexity of the situation became problematic for MAS400’s computer as the COLREGs are a set of ship-to-ship rules. One of MAS400’s standing orders was to maintain a 10-mile Closest Point of Approach (CPA). On her first encounter out of the United Kingdom, MAS400 quickly had an encounter with two tankers ahead and one tanker at her starboard quarter. The larger ships were cruising at 15 knots while MAS400 was cruising at 6 knots. The three ships were to pass the MAS400, respectively, across her port bow, starboard bow, and overtake her on her starboard side. The MAS400 recommendation was to turn to star- board to avoid the two head-on oil tankers. However, this altering of course to avoid collision would have compromised the tanker on her starboard quarter. The MAS400’s presence caused the ships to act differ- ently, and this could have led to negative consequences. Because of the complexity of the situation, ProMare’s watchstander took over, maintaining speed and course, ensuring a safe encounter. Similarly, after leaving the Azores following an elec- multi-ship situations reliably and in a predictable man- ner. Decision-making is a rapidly changing process and autonomous ships must be able to anticipate the influ- ence of other vessels’ actions. When one vessel assumes the right of way, requiring another vessel to maneuver to avoid collision, it causes a domino effect on all other vessels in that vicinity. Imagine two autonomous vessels on track for a head- on collision. In consultation with a ProMare engineer, it was concluded that the vessels would have a challenge responding to such a constricted space where each ship is relying on hydrodynamic interaction with the other vessel to safely navigate. There is a tendency to move far away from the other vessel to avoid collision, but this may result in the MAS400 running aground or alliding with a sea wall. This is where human skill overtakes a computer’s ability to “judge” what is going on. What is Needed to Succeed Globally, regulations are still catching up to innovation, and we need to consider a change to the COLREGs to incorporate autonomous vessels. The Coast Guard needs to continue taking a proactive approach with evaluating risks and developing policies and regulations that will provide a safe, secure governance framework for autono- mous vessels. Specifically, there should be a standard of digital communication requiring autonomous vessels to com- municate with each other in route-planning to avoid col- lision. For now, we must make assumptions about the other ship’s intent based on our belief that the other ship also intends to follow the COLREGs. There is a need for voice-over-satellite and the ability to communicate intent, though it may not be the opti- mal solution due to latency. Natural language processing for discreet VHF communications at sea—intership—is very challenging and nowhere near ready for deploy- ment, so we must accept the latency of VHF to satellite communications while striving to predict and resolve potential issues. Additionally, there is a need for a stan- dard protocol for communicating intent. How can we measure this risk in such a complex situation? The MAS400 minimized this risk by using four redundant computers to run the necessary operat- ing programs in order to safely function. If a piece of software or hardware failed during the transit, another computer could run the operating program instantly and take control. Unfortunately, one cannot plan for every situation, but ProMare found ways to reduce risk with MAS400 by incorporating watertight compartments, adding “dead- man” switches for external safety shutdowns, and ensur- ing a neutral/midship reset during a power outage. Right now, we must continue to verify that the trical repair, MAS400 encountered another crossing situation that was made more complex due to sea state and maneuverability issues. MAS400 was on a course of 270 degrees with a container ship 20 miles to the south- west. MAS400 was content to maintain course and speed as the stand-on vessel with an expected 10-mile CPA. However, due to sea state and maneuverability concerns, the ProMare watchstander was concerned the other ship may not have considered MAS400 and overrode the decision, slowed down, and turned to port. At the same moment, the other ship, having seen MAS400, had changed course 10 degrees to starboard to open the CPA. The watchstander changed back to 270 degrees, and the vessels never came within 10 miles of each other. Lessons from Complex Encounters Route planning algorithms must be sophisticated enough to anticipate the assumptions of other vessels. Autonomous ships must be able to navigate complex 52 Proceedings Summer 2023