Autonomous ships could improve navigation security

Researchers at Viterbi School of Engineering (University of South California) are developing an automated system that relies on both data analytics and artificial intelligence, as an attempt to remove or minimize the need for human decision in ship navigation

In 2017, collisions made up nearly 40% of all marine accidents, and over half of the total casualties. These incidents were primarily caused by human error.

“The main purpose of autonomous ships is safety,” said Yan Jin, member of the Department of Aerospace and Mechanical Engineering and project lead. “We’re human and sometimes we make mistakes, but if we had an autonomous kind of decision-making computer program, it would be different.”

Knowing the locations of other ships and objects, the system can predict  movements and determine the best possible course of action to minimize the chance of collision.

Mechanical engineering Ph.D. student Xiongqing “Vincent” Liu was responsible for developing the AI portion of the new system. Initially, he planned on using data on how ship captains avoid collisions, to train his system to replicate this behavior. Unable to get this data, he turned to a machine learning method called reinforcement learning, a method that uses simulations of different scenarios to teach the computer how to achieve its goal of not hitting another object.

“At first the computer agent doesn’t know anything. It has to explore the simulated environment by itself,” Liu said. “If the agent collides with the obstacles, then it will receive a negative penalty. But if it reaches the goal, then it receives a positive reward.”

After running the simulation thousands of times, the agent learns from its past experiences what trajectory to take to avoid a collision.

But the AI system alone is not fully error-proof. Major variations on the scenarios can cause confusion and lead to a dangerous trajectory.

The model developed by aerospace engineering Ph.D. student Edwin Williams helps fill some of these gaps, adding historical data on trajectories and outcomes, to predict what other vessels are going to do.

In simulations, Williams’ system had a 100 percent success rate for avoiding marine collisions. But, just like the AI, it is limited by the scenarios provided by the data. By using the two systems together, they have an additional layer of security in case an unexpected situation occurs.

In addition to marine vessels, this new method could be applied to air traffic control and space traffic management.

The research team is now working with a three-year grant funded by the Maritime Technology Division of the Monohakobi Technology Institute in Japan. In the summer, they will begin another three-year grant to continue their work and develop the system even further. By the end of that time, they plan on performing a full-scale test using simulators.