Out of Academia
Anchor Drag Prediction
PopEye: Autonomous Anchoring for Large Commercial Vessels
Innovation is the lifeblood of any industry, and this month we present — in their own words — insights on a device which combines AI and video to help a ship's captain predict anchor drag before it begins. Created by a group of five engineering seniors from the University of Pennsylvania, the innovation has caught the eye of some powerful potential collaborators, including the U.S. Coast Guard.
By Orestis Skoutellas, Jack Bendell, Mario Ferre, Ben Abt and Nick Anderson
Coming Together: Team First
We knew that working on our senior capstone project was unlike any other university assignment we had ever attempted before. This was the culmination of four years of intense engineering work at an Ivy League university on a project we had to actively work on for a whole year.
On a sunny morning of August, two months before our team formation deadline, we sat together outside Panera Bread in West Philadelphia, five of the brightest engineering seniors across campus. The first thing we had to align on was motivations and goals. Nobody was sitting on that dusty table in search of an "easy A" or to judge each other's technical competence - those were a given - but what more? We soon arrived to our core five principles of what our project should be:
Engineering. Software is cool, hardware is cooler, a combo is the coolest.
YC S23. Not just a technical solution, but there is a willing-to-pay customer.
A project is a project. If we're going to solve a problem, it may as well be big and noble.
No-BS. So, no Software as a Service, no VR, no smart coffee machines.
Awkwardly specific. Not too many dependencies, so it's within our wheelhouse.
With mutual agreement on the project basis, the arduous journey began of brainwriting and subsequent judging of ideas, not from our own biases but by cold-calling hundreds of industry experts across domains.
During a phone interview with a Greek Captain, he brought up the current practice of anchor-watch: the deckhand, "let's call him Popeye", walks every hour all the way to the front of the vessel to report back to the bridge the direction and tension in the chain. Popeye does this rain or shine, even at a moonless night at 4am, on a frozen deck. And that happens every hour at this moment on every anchored vessel worldwide.
"That can't be right", we naively thought. The safety hazard for the deckhand, let alone the myriad risks from re-active anchor dragging, seemed quite profound to us, five passionate engineers. There must be a way that technology can help boost safety and productivity, while reducing risk and cost.
Robotics, AI Meets Maritime
We happen to be engineers in one of the world's best universities for Robotics and Artificial Intelligence. Taking advantage of our knowledge in our past four years of classwork and with guidance from professors, like Sid Deliwala, Nick McGill, Micheal Carchidi and Kostas Daniilidis, we were able to re-define the problem from first principles and come up with some unique insights leading up to our novel solution.
Using state of the art Computer Vision techniques, we have developed the world's best chain-recognition model. By pointing a camera placed at the bow of the ship directly on the anchor chain, we're able to extract the tension and direction of the chain. Using that, we model the shape of the catenary and back-out the static friction coefficient at the specific geolocation at the anchorage. That's all pretty cool AI and math, but so what? Well, by doing all that, we're able to predict before you start dragging anchor. Knowing this, the captain can save fuel by avoiding unnecessary re-anchorings as well as protect Popeye. Storing the footage provides also a system of record for insurance purposes.
Beyond PopEye: Pirates, Overboard and Perimeter Modes
AAfter being invited to present at the US Coast Guard's HQ in Washington DC, three new directions were proposed to us. The 20 senior officials pointed out that our 900 rpm self-cleaning lens technology together with our infrared camera capability has perhaps even more important use cases. What else can a smart camera inside a weather-proof casing do?
Pirate mode. While the ship remains at anchor, it is at its most vulnerable stage. We heard from several ship operators that it has happened that unauthorized individuals gain access to the vessel by climbing the chain. PopEye comes to the rescue here, by detecting the human figure climbing the chain and proceeds to alert local port authorities and the captain so as to protect the crew and cargo.
Overboard mode. At 3am at a frozen deck on a moonless night, it can happen that the deckhand falls in while attempting to point their flashlight at the anchor chain. In that case, PopEye is able to detect that, alert the Coast Guard and protect our seamen 24/7.
Perimeter mode. Perhaps the biggest need amongst ship operators was to have eyes where they now don't. At 1 mile out, you can see everything from the bridge, but at 50 meters away you have no idea what's happening along the 200 meter vessel perimeter. By placing multiple PopEye devices across the ship railing, the captain can now have 360° awareness and receive alerts when unauthorized personnel or suspicious activity is detected.
Our team has been invited to speak at the main stage of international conferences such as Ocean Business 2023 as well as at the national level at the annual conference of the American Society of Naval Engineers and at the SAGE conference winning "Best Paper" award. At our engineering school, we've won the Leadership award across all other senior capstone projects and on a university-level we've won the most innovative, most interdisciplinary and the best application of AI in Business. We're fortunate and humbled by these recognitions, which are a testament to how far phenomenal team collaboration and work ability can bring you.
Learn more: www.popeyelabs.co