Research

Mixed-species bird communities can produce more useful predator warnings when different species contribute complementary information. Researchers studied babblers, tits and their associated bird communities across ten nature reserves spanning 1,650 kilometres in China.

Basking sharks are usually solitary, yet sometimes gather in enormous groups, and a new model suggests that neither food nor social attraction alone adequately explains why.

The distinctive patches and bands of vegetation found in drylands may help ecosystems withstand drought, but their relationship with plant evolution is more complicated than expected.

A fish school may achieve coordinated movement without each fish tracking a crowd of neighbours. Researchers at CNRS and collaborating institutions placed a real rummy-nose tetra in a closed-loop virtual-reality environment alongside four simulated fish whose interaction rules could be controlled.

A simulated lunar robot swarm has independently developed two cooperative survival behaviours resembling those of emperor penguins. Researchers trained a shared reinforcement-learning policy for twenty robots using only the temperature of the five coldest members as the final reward.

A distributed drone-search framework addresses the difficulty of maintaining coordinated flight when satellite navigation is unavailable. The researchers use distance measurements between neighbouring drones to estimate positions, while explicitly modelling how localisation uncertainty accumulates through multi-hop connections.

Researchers argue that decentralised AI societies face a coordination problem fundamentally different from ordinary message delivery. Autonomous agents have limited capacity to read and retain incoming information.

Researchers have demonstrated a decentralised security mechanism that allows cooperating robots to reject malicious information while continuing to trust legitimate neighbours.

A new mathematical analysis establishes a surprising equivalence between collective decision-making in bee swarms and reinforcement learning. Researchers Karthik Soma, Yann Bouteiller, Heiko Hamann and Giovanni Beltrame demonstrate that a population following simple imitation rules can behave mathematically like a single learning agent evaluating multiple alternatives simultaneously.

Researchers have developed a method for robotic swarms to construct collective shapes through local interactions rather than a predefined deployment-wide attachment list.

Networks of autonomous language-model agents can develop polarised social structures through repeated local interactions, without being instructed to form opposing groups.

Marine animals repeatedly evolve collective behaviours because their shared environment rewards coordinated responses to food scarcity, predation and physical conditions.