An exploration vehicle made of cooperating spherical robots

Ants form rafts and bridges by linking their bodies, and in doing so they change how they interact with the environment. This project carries that idea over to an exploration vehicle: three commercial spherical robots couple to a passive frame and move it as a single unit, with the option of separating to explore on their own and assembling again. Joint movement was achieved. Autonomous assembly was not, and the reason lies in the mechanics of contact.

Context

Search and rescue robots have to be small and portable, which limits what each one can do. One alternative is to use several simple units that cooperate. The units in this project are Sphero robots: a sphere with an internal cart that shifts the center of mass and makes it roll. They are robust and sealed, but they can only push, and they start from rest with difficulty when something opposes the rotation.

Frame design

The first frame was a triangular plate with three spherical sockets that act as ball joints.

To let the robots enter and leave, the design evolved into a frame with access ramps and channels that guide each sphere to its socket. Ramp height, fit in the channel and the height difference between channel and socket were determined by trial on a test bench with one third of the frame.

Robot characterization

Actual speed as a function of commanded speed was measured on a 5 m track, along with the response of the internal sensors. The robots were programmed in JavaScript, with a routine that waits for all three to connect before sending the same command to all of them.

Results

Joint movement. The keyboard sets the heading of the three robots, and the 130 g frame moves at the speed of the individual units. The dominant friction is between the spheres and the floor, not between the frame and the spheres.

Starting is imprecise. The three robots do not begin moving at exactly the same time, which tilts the frame, and the factory command rolls each robot for a full second regardless of speed.

Locked position. When the sphere is in simultaneous contact with the floor and the frame, static friction at both points exceeds the torque the internal cart can generate, and the robot cannot start. It only leaves that condition if it arrives with speed.

Assembly. On the bench, with a single robot, entering by the ramp works over a range of speeds, distances and angles.

With all three robots and the full frame it did not succeed in any attempt. The best result was two out of three. At the speed needed to avoid locking, the impact is abrupt and the spheres slip and lose orientation. At low speed, they lock after turning toward the ramp.

Exit. A deformable exit mechanism was designed and tested separately. It also requires the robot to arrive with speed.

What is missing

The vehicle cooperates but does not self-assemble. The limitations come from the commercial unit: starting torque cannot be controlled and the onset of motion cannot be synchronized between robots. The author proposes smoother surfaces to reduce friction with the frame and, above all, designing custom units. The project did not implement autonomous navigation or environment sensing either.

How it fits in Robiolab

This is the group's first work on collective behavior, a topic that returns years later with the cooperative drone system. Its lesson is mechanical before it is algorithmic: cooperation between physical robots depends on contact conditions, and a design inspired by biological self-assembly needs units whose body is built to couple.