A partly virtual robot swarm practises surrounding a moving target at ESA

A swarm can be larger than the robots in the room. At ESA’s Orbital Robotics Laboratory in the Netherlands, the University of Bologna’s STAR-BOTS team combined real platforms with virtual agents to test how autonomous robots approach, surround and follow a target. ESA reported the completed campaign on 22 October.[1]

Here, autonomy sits in the cooperative control software. The longer-term engineering problem is how several machines could work around an object that does not help them: a piece of debris or another non-cooperative target. The project’s earlier brief included formation control and tracking a tumbling target, with potential applications in orbital servicing and debris removal.[2]

Why the floor matters

ESA’s ORBIT facility provides a 43-square-metre floor whose highest and lowest points differ by less than a millimetre. Air bearings hold the platforms above it on a thin air gap, allowing extremely low-friction movement. The result reproduces free-floating behaviour in two dimensions. It offers a physical way to study drifting machines and their interactions while keeping the experiment accessible on the ground.[3]

A wheeled robot on an ordinary floor constantly encounters contact forces. A floating platform makes momentum harder to ignore: a movement persists until a counteracting influence changes it. In a swarm, that means the route chosen by one member changes the space available to its neighbours. Surrounding a target is therefore a coordination problem as much as a navigation problem.

The mixed setup is a useful middle stage in development. Pure simulation offers freedom to change the number of agents; physical testing exposes software to a particular machine and environment. Connecting the two lets an experiment retain both advantages. It also leaves a specific question for the eventual performance analysis: how does the behaviour change as the balance between physical and virtual agents changes?[2]

Getting the software onto the machines

ESA says the team first had to adapt software developed in simulation to the laboratory platforms. After troubleshooting, the students completed their planned experiments across multiple configurations. Detailed analysis remains the next step; the campaign account reports that the main objectives were achieved but supplies no numerical success rate.[1]

The value of this step is concrete. A coordination rule has now encountered real equipment, a bounded floor and moving targets. The data can show where the swarm maintains its formation, where geometry constrains it, and which cases deserve the next experiment. That is how an abstract collection of agents begins to become an engineered robotic system.

Report an error or suggest a correction