MIT’s Autonomous ‘Roboats’ Learn to Shapeshift

Self-driving cars may be the future for most cities, but Amsterdam is about one-quarter water thanks to its extensive canal network. So, maybe autonomous boats are worth exploring? MIT has been working with the Amsterdam Institute for Advanced Metropolitan Solutions (AMS Institute) to create just that. When we last checked in on the so-called “Roboat” vessels, they had just learned to reliably link up while floating on water. Now, MIT reports Roboats can shapeshift into different conformations in just a few minutes.

Each individual Roboat is a fully fledged vessel, but the idea is not to have you hop on a single boat and use it as a taxi. From the start, MIT and AMS Institute have worked toward a multi-use role for the robotic vessels. Rather than designing different boats for different tasks, the Roboats can link up to become whatever people need. They could form bridges, stages, cargo transports, and floating busses.

Progress has been swift for the Roboat project. In 2016, MIT tested a prototype boat that could move along pre-programmed routes. In 2018, it developed a method to 3D print the boats and tested advanced location tracking algorithms. Earlier this year, MIT and AMS Institute demoed the latching system that lets the robots link together on the water.

The latest advance adds a layer of complexity to the Roboat docking system, marking a major step toward the goals of the project. MIT says it has developed new algorithms that allow the boats to smoothly reshape themselves in a few minutes. So, controllers can ask for a confirmation of Roboats like lines, squares, and L-shapes. The boats talk amongst themselves and determine the best way to morph into the desired shape.

The shapes demonstrated in a pool at MIT are admittedly simple, consisting of just a handful of boats. However, the programming that went into it is still incredibly complex. Engineers had to make sure each boat was aware of its location relative to others, as well as how the group could move while shapeshifting to avoid collisions. To make that happen, the team created a division of labor. Both classes of Roboat have four propellers, wireless communication gear, and multiple docking hardpoints. The coordinators also have GPS and inertial measurement units (IMUs) that allow them to form the “core” of a structure. One or more worker Roboats connect to the coordinator and use actuators to help steer it.

The current Roboats are one-quarter scale versions of the planned units. They’re about one meter long and half a meter wide. The team believes the trajectory-planning algorithms developed for the smaller boats will scale up to the full-sized ones when they exist in a few years.