The pursuit of control has led to unprecedented experiments with not only the physical design but also advanced modeling software and control systems. One such experiment included automated systems commonly found in drone technology. In early 2016, through a series of serendipitous connections and recommendations, automated flight expert Ricardo Bencatel was recruited to do something that had never really been done before. At the time, Bencatel was finishing his postdoctoral research at U-M’s Aerospace Engineering Department and had been collaborating on an experimental project with the Italian team, Luna Rossa. However, when the Italians withdrew from the 2017 Cup in response to rule changes, Ricardo and his advisors at Michigan were approached by ORACLE TEAM USA.
Although automated controls are strictly forbidden in actual races, they can be used in interesting ways for experimentation and training purposes. The goal was to see if an automated system could fly the boat better than a manned crew. If so, what could the crew learn from the computer to better prepare for the America’s Cup?
“In the last cup, foiling was new. Just doing it was a big advantage,” says Bencatel. “Now, it’s about refining a way of handling something that’s already known. We’re trying to get some disruptive ideas to give us the advantage.”
This cup, we’re trying to get some disruptive ideas to give us the advantage.Ricardo Bencatel, ORACLE TEAM USA
The catamarans are covered in sensors that measure a variety of data, like the pitch, flight height and accelerations. That was all fed into Bencatel’s computer model that attempted to optimize the flight by leveling it off to keep it going steady – basically like an autopilot for an airplane, but in a completely different application. In sailing, the wind forces that provide thrust can also tip the boat over. There is a constant balance between what is known as the righting moment and the heeling moment. The side forces against the sail produce the heeling moment. In traditional sailboats, weighted hulls helped counteract these forces and keep the boat upright. With flying catamarans, the heeling moment is counteracted by the rudder and the hydrofoil on the leeward side (the side the wind is pushing to) and the rudder on the windward side (the same side the wind is pushing against). The rudder on the windward side actually creates inverted lift forces – pulling the hull back down into the water and keeping the platform level.
By nature, the AC45 operates under conditions that make it prone to rolling and tipping. An automated control system would have to sense and respond to these forces. The two main components of such a system were already on the boat. The sensors were already there to capture performance analytics. A joystick in the helmsman’s hand controlled the actuators. Bencatel’s mission was to connect the two. Doing so required a tedious programming loop.
He spent most of his time writing code for the controller and its analysis tools. “You go to the water and do tests and tuning,” says Bencatel. “Then you come off the water and you do analysis. And because this is an evolving problem and system, your analytical tools from the previous day are not enough for the new things you want to analyze. So you need to code the analysis tools. If everything goes well, you can identify the issue or bug in the controller code, go back, and correct it. Then you run simulations. And then back to the water.”
“We stuck with reasonably traditional controls in part because of the risk,” says Anouck Girard, Associate Professor of Aerospace Engineering at the University of Michigan and Bencatel’s advisor at the time. “We can try a lot riskier things on a quadrotor in my lab than on a boat that has six humans on board and costs $50 million. We haven’t yet pushed the boundaries of what we know how to do.”
“It’s very similar to aerospace – dealing with air flow, the technology,” says Bencatel. “In practice, the system will do unexpected things. Is that good or bad? Sometimes it’s a better way of controlling the machine than what humans might assume to do.”
One of the most exciting possibilities of the automation work was the possibility to figure out how to foil through turns and other maneuvers.
“I’m frankly amazed with what he’s able to do,” says Ferguson. “It’s not quite like taking your hands off the wheel of a car, but almost.” After months of experimentation, the autopilot was not able to outperform the sailors, and the team was forced to move on due to time constraints. “We learn from things and modify our path,” says Ferguson. “The positive influence he’s had doesn’t change. There are great takeaways from every experiment.” Bencatel stayed on with the team as a control systems analyst.