Yesterday's News

A short essay on something that happened the day before.

Can Many Hear the Shape of a Room?

A group of 25 small robots has learned to recognise the shape of the room around it without cameras, maps, or a central computer – using mathematics that reveals exactly as much as it conceals.

Yesterday, Tech Xplore reported that a swarm of 25 small robots had identified the shape of the room they were in. No cameras, GPS, maps, or central computer were involved – the robots just exchanged numbers with their neighbours. In 30 out of 30 trials, the swarm correctly told a circular room from a ring-shaped one.1

»No matter how carefully you listen, you can’t tell them apart«

The experiment draws on a question the mathematician Mark Kac posed in 1966: »Can one hear the shape of a drum?« Picture two drums, each stretched across a differently shaped frame. If you strike them and listen, could you derive the shape of the frame from the sound alone? Does a vibrating surface’s mathematical signature contain enough information to reconstruct its geometry?

The answer, it turns out, is both yes and no.

This is the territory of spectral geometry. A vibrating drum produces a spectrum, a collection of natural frequencies, or eigenvalues – the characteristic pitches at which it can vibrate. In 1911, mathematician Hermann Weyl showed that the spectrum reveals the drum’s area. Kac showed that it also reveals its perimeter, and conjectured that it reveals its topology: whether it has holes. The following year, mathematicians Henry McKean and Isadore Singer proved him right.2 This matters because a disk and a ring differ topologically, and in mathematical principle, we should hear that difference.

The robots never build a map. Each carries a simple numerical value and repeatedly exchanges it with its neighbours; the values gradually even out, like heat spreading through a material, at a rate that depends on the robots’ connections. The researchers measure that rate to estimate an eigenvalue of the swarm’s network, and from that single number tell one shape of room from another.3

Roboticists already know intelligent behaviour doesn’t require central intelligence. In 1987, computer scientist Craig Reynolds’s Boids project showed that coordinated behaviour could emerge from simple local rules.4 In 1989, electrical engineering professor Gerardo Beni and collaborator Jing Wang named the phenomenon »swarm intelligence«.5 And in 1991, roboticist Rodney Brooks argued that robots could act intelligently without building a comprehensive model of the world.6

But telling a circle from a ring isn’t the same as answering Kac’s question in general. The answer is a hard no: in 1992, mathematicians Carolyn Gordon, David Webb, and Scott Wolpert proved that two differently shaped drums can share exactly the same spectrum. They constructed two different drums from seven identical triangles: same area, perimeter, topology, and spectrum. No matter how carefully you listen, you can’t tell them apart.7 This isn’t a limitation of the instrument or the listener. The information just isn’t there, and a better microphone or longer listening won’t change that.

But what if the robots could do more than listen? What if they could move the walls, change the room and observe how its properties respond? Georgia Tech’s Robotarium, founded by Swedish roboticist Magnus Egerstedt, is a remotely accessible lab where researchers submit algorithms to run on real robot swarms, taking collective robotics beyond any one laboratory.8 A natural next step might be to introduce movable walls, allowing the robots to configure the space they inhabit.

One experiment could ask whether a room can be built by ear alone, without ever being drawn. Give the swarm a target pattern of diffusion rather than a floor plan, and let it move the walls until the space produces the desired response. A second could start with two isospectral but geometrically different rooms, and ask whether the swarm could distinguish them by making small changes to their boundaries and listening to how the spectrum responds. A third could set musical goals: reach a target frequency, build as full a chord as possible, or create dissonance. Could a workable layout emerge from local listening and local action alone?

This would move the research beyond recognising fixed geometries. The swarm would listen to the very room it was changing, using the effects of its own interventions to decide what to do next. It would become, in a limited but quite literal sense, both architect and inhabitant, building by ear.

Kac asked whether one could hear the shape of a drum. Gordon, Webb, and Wolpert showed that some shapes can’t be told apart by listening, however perfect the ear. The next question might be whether a swarm can do more than just listen: change the room until it becomes worth hearing.

1 Ingrid Fadelli (2026) »Robots recognize the shape of an arena by messaging each other«. Tech Xplore, 14 November 2026. https://techxplore.com/news/2026-11-robots-arena-messaging.html

2 Mark Kac (1966) »Can One Hear the Shape of a Drum?«. The American Mathematical Monthly, 73(4), pp. 1–23. https://www.jstor.org/stable/2313748

3 Leo Cazenille et al. (2026) »Hearing the Shape of an Arena with Spectral Swarm Robotics«. Science Robotics, 11(119). https://doi.org/10.1126/scirobotics.adr9578

4 Craig Reynolds (1987) »Flocks, Herds, and Schools: A Distributed Behavioral Model«. Computer Graphics, 21(4), pp. 25–34. https://doi.org/10.1145/37401.37406

5 Gerardo Beni & Jing Wang (1989) »Swarm Intelligence in Cellular Robotic Systems«. NATO Advanced Workshop on Robots and Biological Systems, Tuscany, Italy, 26–30 June 1989. https://doi.org/10.1007/978-3-642-58069-7_38

6 Rodney Brooks (1991) »Intelligence Without Representation«. Artificial Intelligence, 47, pp. 139–159. https://doi.org/10.1016/0004-3702(91)90053-M

7 Carolyn Gordon, David Webb & Scott Wolpert (1992) »One Cannot Hear the Shape of a Drum«. Bulletin of the American Mathematical Society, 27(1), pp. 134–138. https://doi.org/10.1090/S0273-0979-1992-00289-6

8 Georgia Tech Robotarium. https://www.robotarium.gatech.edu


Previous editions

See the full archive →

Subscribe

Get it daily, or just the weekly roundup.