Robot swarm inspired by cells can keep moving even if its parts fail

A swarm of robots
A robot swarm

Columbia University

A swarm of robots inspired by living cells can squeeze through gaps and keep moving even if many of its parts fail.

Living cells gather together and collectively migrate under certain conditions, such as when inflammatory cells travel through the bloodstream to a wound site to help the healing process. To mimic this, Hod Lipson at Columbia University in New York and his colleagues created 25 disc-shaped robots that can join together.

Each is equipped with cogs that cause the robot’s outer shell to expand and contract and magnets around its perimeter that let it stick to neighbouring bots. Individually, the bots can’t move, but once stuck together, the swarm can slither across a surface by making individual bots expand and contract at different times.

It can use this process to travel or to surround an object and transport it. “This entire chunk can move around and slither through gaps,” says Lipson. If it needs to move through a tunnel, for example, the swarm can narrow itself down, traverse the gap, and reassemble as a blob on the other side.

Many biologically inspired robots look like animals, but are constructed from complex components that make it difficult for them to heal or adapt their shape, says Lipson. Robotic swarms, on the other hand, have the potential to self-assemble, and can change in scale.

Increasing the size of the robot swarm is as simple as adding more individual bots.

The swarm was still able to move around even when 20 per cent of individual bots weren’t working – an advantage over conventional robots that may completely malfunction when a single component fails.

Individual bots were programmed to vary the timing of their expansions and contractions according to the intensity of the light they sensed, enabling the swarm to move towards an object.

Although the swarm consists of only two-dozen robots in the real world, the team also tested computer simulations of up to 100,000 bots.

Journal reference: Nature, DOI: 10.1038/s41586-019-1022-9

Add Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.