July 27, 2021

Part 1 of 2: Interview with Professor Shimizu from Tokyo University of Marine Science and Technology

Groke Technologies cuts no corners when it comes to research and development. This time we would like to introduce you to one of the researchers we are collaborating with to achieve our goal of ensuring our systems utilize the safest and most accurate technologies available.

Professor Etsuro Shimizu from Tokyo University of Marine Science and Technology

Professor Etsuro Shimizu earned his doctorate in Control Engineering from the Tokyo Institute of Technology in 1999, and his research work now revolves around control theory, underwater robotics, and autonomous ships. Being an academic, he often finds himself somewhat tied up with theoretical concepts, but since he is also an engineer, he knows that his theories must be tested for practicality if he wants his research to have real meaning. That is one benefit of the relationship he is building with Groke.

Professor Shimizu explains how Groke’s situational awareness system has a large role to play in marine navigation.

“Automatic control technologies can already be found in ships – with autopilots and course tracking – but these techs only work because there are humans making judgements in advance, telling ships to go in a certain direction, follow a certain course on a chart, etc. These systems simply follow those instructions,” he says.

“But if we want to have true automated control, we need systems that will make the judgements that humans currently make. The research I’m doing with Groke is looking into how to automate that stage of the process.”

Professor Shimizu notes that this sort of automation won’t be appearing on ships soon. Step one will be to create a system that serves as a support tool for human operators – because humans make mistakes, too.

“The first stage will have human operators using the system as they work. It will warn the bridge crew when to take action. There an obstacle over here…a ship moving over there…a course correction is necessary.”

Professor Shimizu noted that as accuracy improves, the crew will eventually be able to leave ship operation up to the system itself, and truly automated systems will become possible. However, there is currently no situational awareness system that can recognize obstacles, know what they are, and then inform the crew. That is what led him to start working with Groke.

“A situational awareness system has to be able to distinguish between the different types of objects it will encounter. Right now, no systems are able to tell if an obstacle needs to be avoided or not, or if more attention needs to be paid to an object or not. My research lab was working on exactly that sort of machine-based evaluation technology when we first encountered Groke and they learned we shared this goal. So, we started working together to find a solution,” he explains.

Professor Shimizu is hoping to work with Groke to resolve another obstacle to building a safe and accurate situational awareness system: the insufficient level of data provided by ship-mounted AIS.

“If AIS provided all the information we needed it would be one thing, but a lot of information is missing. First, the rules only require AIS to be installed on large ships at the moment, so you won’t find them on fishing boats, pleasure craft, or other private vessels. There is talk overseas of changing the rules to require all vessels to have AIS installed, but even if we put AIS on every ship, information would still be lost. There just isn’t enough communication capacity.”

Professor Shimizu is working on machine learning using a process called annotation – which involves tagging data with descriptors to help the AI learn how to distinguish objects – but before annotation can be harnessed to teach the AI, he needs to understand how to account for irregularities between AIS data and the locations shown in video and photographic data. That inaccuracy is why he believes we can’t rely too much on AIS data.

“We also can’t see fishing nets or buoys used by fishing boats because there’s no AIS on them either. We need to look for them using cameras and other means because such small objects don’t show up on radar.”

For Professor Shimizu, before we can start talking about truly automated ship operations, we must first build a system that can make for the shortfalls of AIS, radar, and other current tools used by ship crews.

Fortunately, that is exactly what Groke plans to do.

Article published originally in Groke’s exclusive customer newsletter in March 2021

Part 2 of the interview is published here.