Stanford AI Grads Launch Low(ish)-Cost Underwater Robot

Photo: Tekla Perry
Eduardo Moreno (right) and Shuyun Chung started O-Robotix to build low-cost underwater robots and other automated gear for fish farms.

SeaDrone, the underwater robot coming out of a new company founded by two Stanford AI lab veterans, is aiming to make fish farming a lot easier—particularly for smaller aquaculture operations—by making underwater inspection cheaper and easier.

The ocean ROV’s story is not an unusual one for Silicon Valley: two Stanford students meet over a lab bench, get an idea that something they’d been tinkering around with for themselves could be turned into a product and the basis of a company. It’s a story Silicon Valley loves.

Eduardo Moreno met Shuyun Chung in the Stanford AI lab in 2013. Moreno, in the thick of his studies for a master’s degree in mechanical engineering, was working on underwater robot hardware in collaboration with King Abdullah University of Science and Technology in Saudi Arabia. Chung, a postdoctoral scholar, was working on the software for SupraPed robots, which are designed to walk over rough terrain using trekking poles. The two were assigned seats next to each other in the labs, even though their projects were vastly different.

One day back in 2014, another researcher in the lab brought in a DJI quadcopter that he had just purchased, and showed off the drone’s capabilities.

“I was amazed that this $600 robot had much better technology than underwater robots that cost $10,000 or $15,000,” Moreno said. “That’s when I realized how far behind underwater robots were.”

He sketched out a design for a small, low-cost, underwater observational robot that used many of the same parts—like brushless DC motors, cameras, inertial sensors, and batteries—commonly used in consumer aerial drones. Moreno showed his sketch to workplace-neighbor Chung; he was immediately intrigued, and asked how he could help turn the sketch into a useful product.

Moreno was thrilled with the offer. For one, he says, “Once you get someone else to sign on to an idea, then you know you’re not crazy.” For another, he knew that to design really good robotics hardware you need to be creating the software at the same time, because it informs the design. Chung had the software chops that Moreno didn’t. And, finally, Moreno says, he knew Chung was good at helping people, because he’d already been helping Moreno with his homework assignments for his graduate classes.

Before they began working on the detailed specs for the robot, Moreno took some Stanford entrepreneurship classes to try to figure out how to make an underwater drone into a commercial business. Figuring out the market, says Moreno, “took longer than expected. It’s the hardest thing for someone with an engineering background to do, I think, putting on the marketing hat.”

They were thinking of marketing the gadget to city governments, for use in inspecting bridges and other infrastructure and in police work (looking for weapons or even bodies in lakes and rivers), until a Stanford professor told him that was a tough path, given government entities have a long and complicated purchasing process that is a red flag for potential investors. They explored the idea of inspection in general, but couldn’t find an appealing niche that could serve as an entry point.

Finally, they came to aquaculture. “We’d looked at the markets for aerial drones, and agriculture is a big one. So we searched online to see if there was anything like that for aquaculture, and, it turns out, there is some underwater observational technology. But it’s a really new market,” Moreno said. In aquaculture, nets, lines, and anchors must be inspected regularly; most fish farmers today send divers out to perform this task.

Market identified, they started focusing seriously on designing their underwater drone, and had their first prototype built in six months. They tossed it into a swimming pool at a local apartment complex.

It worked better than they had hoped. They even tested it up against a commercial underwater observational robot, borrowed from a local dive shop, and their home-built drone was more stable and easier to control, even for novices. (The commercial systems are typically used to inspect ocean-based oil drilling platforms and pipelines, dams, pilings, and boats.)

It was also one-quarter its commercial cousin’s size, measuring just 0.3 by 0.25 meters. Enabling the compact size, Moreno said, is the choice of brushless motors. They also redesigned the underwater thruster into something that costs tens of dollars to build instead of thousands. They followed that up by  and using off-the-shelf pressure sensors, humidity sensors, inertial navigation units, and camera modules—all of which have come down dramatically in price in the past few years thanks to demand from the mobile phone market. They put it all together, and waterproofed the entire system.

The toughest thing for Chung was coming up with an easy-to-use control system that could handle different numbers of thrusters. That was a necessity because they decided early on that they’d have to offer versions at different price points. Generally, says Moreno, the system is image-based and uses what the robot “sees” in the camera as a navigation guide.

Moreno’s biggest challenge was creating the propellers. “I had never done any propeller design before, or created molds for injection molding—any of that. And I built something with 50 separate plastic parts that are mass producible.” Designing an expensive robot, as he had done in the past, he says, “is significantly easier.”

Chung and Moreno both worked on the electronics design. The drone sends an HD video stream to a tablet or smartphone; the mobile device also acts as the controller.

At that point, they realized that they had a gadget that did pretty much everything commercial underwater observational robots do, but in a much smaller and vastly cheaper package. The only similar gadget out there was the OpenROV, an open source effort that offers kits for hobbyists. That system, Moreno said, is inexpensive to build and can move quickly, but smaller movements lack precision and it struggles to capture stable video.

Now, O-Robotix is testing beta versions of its underwater robot, called the SeaDrone, at an offshore fish farm near the coast of Baja, Mexico. The company will have production versions of its hardware ready to ship by the end of the summer, ranging in price from $2700 to $3300 depending on the number of thrusters. More thrusters give the drone operator finer control of the robot. “We’re starting with inspection drones, but, long term, our goal is to automate multiple parts of the process of aquaculture, including feeding, maintenance, and cleaning,” Moreno says.

Though large, comprehensive, automated systems are on the market, Moreno says O-Robotix is aimed at small, independent fish farmers who don’t have a million dollars to invest in a massive fish-factory, and instead want to gradually add low-cost technology.

While aquaculture will be the main market for SeaDrone, Moreno and Chung think there’s another, smaller market out there—one that is particularly meaningful to them: the education and research market. “Working on robotics research,” Moreno says, “we would spend a significant amount of time either building hardware, or hacking something we bought to make it work for our purposes, instead of focusing on the research we really wanted to do.” So, he says, they are selling a reconfigurable developer’s kit designed for researchers. So far, researchers at MIT, Stanford, and several universities in China have purchased early versions, Moreno says.

Moreno and Chung have been getting mentoring and other help this spring from the Start-X accelerator. The company was to officially unveil its technology at the TechCrunch Startup Battlefield today.


View From the Valley

IEEE Spectrum’s blog featuring the people, places, and passions of the world of technologists in Silicon Valley and its environs.
Contact us:

Tekla Perry
Palo Alto, Calif.