The Fans Go Wild: How a Self-Driving Boat Took Victory Laps in Stevens' Annual Competition

The Gallois Autonomous Robot Competition has become a Stevens tradition, happening each year during the university’s annual Innovation Expo.

This spring it happened as usual, but with a twist.

A nautical one.

Connecting with history

To keep students engaged and broaden their hands-on exposure, the competition’s organizers — under the leadership of SES Dean Jean Zu and Associate Dean Frank Fisher — developed new contest rules and specs.

Students would program fan-powered boats, rather than wheeled motorized vehicles, around an oval watercourse with a few obstacles sprinkled in en route. Each boat would be required to deploy two electric motors, powering two propulsive fans.

“The basic approach was similar to the previous ground-vehicle project,” explained mechanical engineering lecturer Chang Beom Joo, who supervised the competition again this year. “The core skill taught in our class is the implementation of a feedback-control system that controls the moving direction of a vehicle, based on a robot's position information as acquired through a sensor.”

Students would need to learn to master those feedback control systems by understanding and experiencing how mathematical concepts can be applied to real-world engineering problems, Joo continues — in this case, by teaching a boat to move on its own through water and around an obstacle course, while also racing a fellow boat. An emphasis would be placed on hull design, stability and hydrodynamics, using the popular 3D CAD platform SolidWorks.

That new challenge also connected deeply with Stevens’ long nautical history.

The university’s founding family designed the first U.S. steam ferry and created the America’s Cup yachting competition, after all. And Stevens today provides life-saving maritime and river data, modeling and forecasting to local agencies. (A floating ship even once served as student residential campus housing for a time.)

With a shiny new theme and brand new specs in place, here’s how the competitors made it to Expo day.

Same great contest, more great duels

During the year-long runup to the contest, eight top-performing teams were selected from an original 160 (yes, 160) squads that formed and competed during events held in engineering class sections.

“We did this to make the event more dynamic and exciting,” says Joo.

The competition format was also switched up, pitting dual teams head-to-head against each other in the final rounds rather than entrants simply making solo runs around the course and being timed.

On Expo day, the final eight squads assembled and matched up in a tournament-style knockout competition. As competition progressed, a knot of onlookers grew around the rink-like venue and excitement built to a fever pitch. Robotics professor Long Wang mounted a stepstool and gave Kentucky Derby-style commentary while the miniature self-driving crafts circulated in pairs around the oval course.

Competitors and the healthy crowd cheered or groaned as the bot-boats attacked the oval for three laps each — occasionally doing 180’s or crashing each other out. Once the ripples cleared, the team of David Storch, Megan Koek and Ashley Tiglao had taken the victory.

Their flawlessly programmed boat made a perfect, Daniel Ricciardo-like run around the circuit to secure the $1,000 top prize. Catherine Bondi’s team finished runners-up in the final matchup for a $500 prize, while Ayaan Patel’s squad won the consolation race and split $250 third prize.

Keeping in the loops

How did the winning team do it?

The victorious craft’s onboard processing utilized not one but two asynchronous task loops. One near-continuously checked the boat’s position, systems and propulsion while the other checked sensing and telemetry data at half-second intervals.

Many of the techniques those winning algorithms utilized, such as low-pass filtering, a PID (proportional–integral–derivative) control algorithm and various approximations, came from the team’s coursework — which then crystalized around an idea Storch developed last fall.

“One night I decided I would begin making an Arduino-based drone, controlled by Bluetooth,” he recounted. “This task ended up feeding into the Gallois competition entry.”

Those two asynchronous task loops proved highly successful at staying the course and driving the boat home to victory, one looping every 5 milliseconds reading the boat’s onboard accelerometer — and directly controlling the fans and keeping the system stable — the other looping every 500 milliseconds or so, handling telemetry data and communications and deciding how the boat should navigate with respect to walls and obstacles.

“We had to spend countless hours in the lab getting this system to behave properly, but we learned so much along the way,” added Storch.

With boat races now under the contest’s belt, more changes are in store as the Gallois competition strives to keep things fresh.

The next one will utilize a third autonomous platform, known as Dune Runner, inspired by Mars-exploration vehicles. That system will feature a transformable chassis, custom wheel designs and ArUco marker-based vision tracking and sensing.

In other words, a fun new race to teach, build and plan for… and spectate.

“Together, these three autonomous challenges — land robot, boat and Dune Runner — will provide a rich and exciting educational ecosystem,” summed School of Engineering and Science (SES) lab manager Louis Oh, who helped prepare and direct the competition and its year-long runup, “empowering Stevens students to continue exploring some of the most advanced and engaging topics in autonomous systems design.”

The Gallois Autonomous Robot competition is made possible each year thanks to the generosity of Bernard Gallois, a former Dean of SES. During his tenure as Dean, Gallois and his academic team conceived of the Design Spine that remains a hallmark of SES’ curriculum.