Self-Organised Saliency Detection and Representation in Robot Swarms

Self-assembly of shapes using robot swarms has applications spanning architecture, functional materials, and art. Typically, the amount of robotic material needed to represent the shape is relative to the shape's area, with robots filling in the whole shape. To save in robotic material, we explore the ability to automatically aggregate on the frontier of shapes, essentially forming outlines, using a self-organised decentralised mechanism. An image is projected on top of a swarm of robots, and the robots sense the light colour and communicate this information to their neighbours to detect salient features. Results in simulation show the swarm can detect salient features, draw with different line thicknesses, adapt to changes in the feature over time, and scale to 300 robots in reality and up to 1000 robots in simulation. We then show a dynamic performance where robots continuously reconfigure in simulation to form a “video”.