Most people think of the waterfront as the edge of the city. The MIT research team sees it as a energetic construction site, reminiscent of Lego bricks.
Their recent system, called “Floating form” is a swarm of small, square, robotic boats that come together into larger structures on the water, break apart, and reassemble into something new, all with minimal human guidance.
About the size of a dinner plate and 21 square centimeters in size, each robot is a self-contained ship with its own thrusters, sensors and magnetic latches. Together they point to a future in which floating infrastructure could become more adaptive: a temporary platform in the event of an emergency, a market on a canal, or a stage that appears at a festival and disappears when the crowd returns home.
“Our FloatForm designs envision a future in which the waterfront becomes a programmable extension of the city, where autonomous boats can self-organize into bridges, platforms, and other useful structures on demand,” says Daniela Rus, professor of electrical engineering and computer science at MIT and director of MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL). “This type of distributed robotics opens up new possibilities in mobility, emergency response, public spaces and on-water infrastructure.”
“With FloatForm, we essentially transform static water surfaces into dynamic, programmable spaces,” says Wei Wang, lead author of the new article about the project and a former MIT scientist who now directs the Marine Robotics Laboratory at the University of Wisconsin, Madison. “Imagine an urban environment in which public space is not fixed but can autonomously expand, contract, or reconfigure on demand.”
“We see it as creating infrastructure on the water using a modular system to create one larger system,” says Alejandro Gonzalez-Garcia, a former researcher at MIT CSAIL and Senseable City Lab. “In an emergency, you can create a recent bridge to ease traffic in the city. You can also create floating markets and floating stages. If you want to make the city more livable, you also want to operate the water.”
public works, published today in comes from the labs of Rusa and Carlo Ratti, professors of the practice of urban technology and planning at MIT and director of the Senseable City Lab, and comes from Roboat, their joint project with the Amsterdam Institute for Advanced Metropolitan Solutions that aims to place full-scale autonomous ships on Amsterdam’s canals. These canals were once used to transport city goods; today they mainly transport tourists.
“We investigated whether canals could be used for waste collection or transportation to transfer some of the stresses that occur on roads back to the water,” says Niklas Hagemann, an MIT architecture graduate, CSAIL division and former Senseable City Lab researcher who worked on the project from its early stages. “Urban areas are becoming denser, so can public space be expanded to include water that is currently underutilized?”
FloatForm reduces this vision to the tabletop scale to answer a more difficult question: How did dozens, and eventually thousands, of floating robots organize themselves?
Lessons from a raft of ants
The team found the answer in biology. Fire ants are famous for surviving floods by combining their bodies into living rafts, and no leader supervised the choreography of the gathering. Each ant follows simple local rules and a resistant structure is created.
“Each ant is an independent agent,” says Gonzalez-Garcia. “We wanted each robot to have its own capabilities, just like an ant colony forms a raft.”
Most existing self-assembling robotic systems on water and elsewhere rely on a central computer dictating every movement. This approach is susceptible to single points of failure and scales poorly: the scheduling math balloons as robots are added, and the swarm must assemble sequentially, with most robots idling, waiting for their turn. FloatForm reverses the balance. A lightweight central planner steps in sparingly, assigning each robot a final position for mesh refinement, a level of geometric precision that distributed methods alone can hardly guarantee. Everything else, including navigating towards a target shape, avoiding collisions and adapting to disturbances, works on the robots themselves, which coordinate actions by exchanging positions with their immediate neighbors. The entire swarm moves at once.
This parallelism is what makes this work stand out. The scheduling complexity of the FloatForms approach depends only on the robot’s local neighbors, not on the overall size of the swarm. “We try to keep central intervention to a minimum and get them all moving together at the same time,” says Gonzalez-Garcia.
In experiments at MIT, a fleet of eight robots repeatedly moved into random positions, assumed a target shape, snapped into a rigid structure, broke apart on command, reassembled into a new configuration, and then traversed the pool as a single craft, with each run lasting four to eight minutes. In this latter mode, called mass transit, the planner determines the trajectory of the entire structure and each robot calculates its own contribution. “Every robot becomes an actuator,” explains Gonzalez-Garcia. Simulations showed a smooth scaling of the structure to swarms after 64.
“The beauty of this largely decentralized approach is that the computations don’t get bogged down as the swarm grows,” Wang says. “Whether you’re working with eight or 80 boats, the entire fleet coordinates and moves simultaneously. Because overall assembly time is generally not significantly increased, the system remains highly scalable.”
Sticking together also has physical benefits. “Our boats become more stable, coming together like a raft of ants if there are waves or currents,” Hagemann says.
Origami handshake
The robots connect via a latch mechanism hidden entirely inside each hull. A single servo motor positioned in the center drives an origami-inspired auxetic structure whose geometry contracts evenly in all directions simultaneously, pulling permanent magnets on all four sides inward for release or pushing them outward to grab a neighbor in 10- to 15-centimeter increments. The magnets are arranged with variable polarity, allowing the boats to reliably connect to clean square grids.
The elegant part is what the mechanism doesn’t do: use (a lot of) power. The 3D printed gear keeps the latch in any state with the engine turned off. “It uses energy to latch and unlock, but it doesn’t use any energy in between,” says Hagemann. This is important for infrastructure that can store configuration for many hours. “Because the robots are so small, you can only have a battery that big,” adds Gonzalez-Garcia. “If they use less energy snapping, they can use more energy calculating or actually moving.”
It took humility and engineering to get there. Four miniature thrusters arranged in an “X” provide each robot with omnidirectional motion, including rotation in place, but they exert large forces compared to the robot’s low inertia, which made early prototypes twitchy and prone to aggressive spins at low speeds. The team added stabilizing ribs to increase hydrodynamic resistance and tuned the controllers to maintain the strength of the robots, which at this scale are never completely identical. The magnets presented their own problem: They held on so well that detaching sometimes required the robots to wriggle away.
From the reservoir to the canal
In 10 trials, the system performed its tasks without human intervention 90 percent with four robots and 70 percent with eight. When something went wrong, the architecture showed its resilience: a robot that became momentarily disoriented could rejoin the structure on its own without stopping the entire swarm, and robots stuck in formation stalemates learned to shake themselves off and try again.
Moving from a controlled internal reservoir to a real canal or port will require more than just confidence. “There is always a relationship between the size of a boat and the amount of disturbance it can withstand,” says Gonzalez-Garcia. “These boats are very small, so they can’t work in very rough water.” Scaling up will mean strengthening latches, potentially with mechanical locks like the full-size Roboat used, and replacing indoor ultrasonic lab positioning with GPS or vision sensors. Interestingly, the coordination algorithm is designed to be sensor-agnostic: swap sensors, keep logic.
The team envisions applications beyond urban channels, from creating temporary platforms for inspection and maintenance at sea, to adaptive sensor networks for studying migratory species, to reconfigurable docking stations for emergency response in hard-to-reach areas. There is also the potential to conduct activities at sea and at a distance, from temporary construction platforms to environmental monitoring and scientific expeditions.
And the geography is wide open. “Venice, the Netherlands, Belgium, the fjords and lakes of Norway, basically any city with a river can benefit from this,” says Gonzalez-Garcia. “The design takes advantage of spaces where water is already important, but it also raises the question: Where else can water be used for something more?”
“This is an exciting step forward in the study of distributed collective behavior on water,” says University of Michigan assistant professor Steven Ceron, who was not involved in the research. “Assembly, self-reconfiguration, and collective movement are tough enough in a arid environment, but achieving these behaviors in a predominantly distributed manner over water poses significant additional challenges, and this team has plausibly managed to overcome them. By offloading the computational burden to the robots themselves, they have built a more resilient system that could, in the near future, enable the deployment of such robot collectives in open water environments for search operations, environmental monitoring, and reconfigurable marine infrastructure.”
Gonzalez-Garcia, Hagemann and Wang wrote the paper with senior authors Ratti, who is also a professor at the Politecnico di Milano and Rus. Gonzalez-Garcia is additionally associated with the MECO Research Team at KU Leuven. The research was supported by a grant from the Amsterdam Institute for Advanced Metropolitan Solutions and additional support from the University of Wisconsin at Madison. The team thanks MIT Sea Grant and Professor Michael Triantafyllou for providing the test tank.