swarm robotics

The engineering of large robot groups in which no unit is in charge and none sees the whole picture: each robot follows simple rules using only what it senses nearby and what its neighbors signal, and useful group behavior arises from those interactions, as in insect colonies, fish schools and bird flocks (swarm intelligenceswarm intelligence, n.: The teamwork that appears when crowds of simple creatures or machines, such as ants, bees or small robots, each respond only to what is nearby, with nobody in charge, yet together solve problems.). Its goals are scalability (performance holds as robots are added), robustness (losing individuals does not cripple the group) and flexibility across tasks. Typical behaviors include aggregation, dispersion, pattern and shape formation, collective transport, foraging, area coverage and consensus decision-making. Designers either hand-craft local rules and check the emergent result in simulation or search for rules automatically with evolutionary methodsevolutionary computation, n.: Problem-solving by simulated evolution, much as breeders improve crops: keep many candidate answers, let the better ones produce slightly altered offspring, drop the worse, and repeat for many rounds. or multi-agent reinforcement learningmulti-agent reinforcement learning, n.: Reinforcement learning with several AI agents learning at once in a shared world, where each must adapt to the others — as teammates, rivals or both.. Demonstrations range from Harvard's 1,024 Kilobots self-assembling into shapes (2014) to drone light shows, which are usually centrally choreographed rather than true swarms. Applications under development include environmental monitoring, agriculture, search and rescue, logistics and military drone swarms, the last raising arms-control concerns.

The swarm kept covering the field even after a third of the robots ran out of battery.