From individual robots to robot societies.

Dario Floreano Hod Lipson Most of today’s robots operate in isolation. The coordinated motion of tens of robotic arms in manufacturing plants, hundreds of wheeled robots on warehouse floors, or thousands of drones in night skies is no different: Each of those robots is unaware of its conspecifics and obeys orders issued by a central computer that leaves no room for unexpected interactions or unsolicited initiatives of the individuals, not to speak of emerging collective behaviors. Robotic and biological individuals, however, have limited energetic autonomy, strength, perception, and decision-making abilities when taken on their own. The transition from solitary individuals to societies has been described as one of the eight major transitions in the evolution toward higher levels of biological complexity ( 1 ). There is ample evidence from biology that self-organized groups of individuals with limited capabilities can act as super-organisms that are more robust to individual failures and more resilient to environmental change and that can carry out more complex tasks and build more complex structures. Computer scientists have taken inspiration from principles of biological self-organization among interacting agents with limited capabilities, loosely labeled as swarms, to devise distributed and adaptive algorithms capable of solving complex, noisy, and changing computational problems ( 2 ). The more recent field of swarm robotics shares similar scientific roots and ambitions ( 3 ), as described by …