Swarm Robotics

From the AI Research Compass wiki

Swarm Robotics: Many simple, cheap robots following local rules can achieve robust collective behavior that no single complex robot could.

Beliefs

Swarm Robotics takes its inspiration from ant colonies, bird flocks and fish schools. Each unit is deliberately simple: limited sensing, short-range communication, a handful of local rules. No unit holds a global map or a master plan. Yet together the swarm can aggregate, form shapes, explore terrain, carry objects or cover an area, and it degrades gracefully when individual units fail. Intelligence, in this view, is a property of the collective and its interactions, not of any single brain.

It sits near the far World edge, since everything happens through bodies in physical space, and at the very bottom, the Harness end, since behavior comes from a system of many interacting parts rather than one end-to-end model. Its paradigm level is Ultra-Revolutionary: an established field, but one that rejects the one-big-robot mainstream outright.

Arguments

  • Experiments with around a thousand small robots self-assembling into programmed shapes showed that purely local rules can produce reliable global structure.
  • Classic flocking models demonstrate that three local rules (separation, alignment, cohesion) suffice for realistic collective motion.
  • Redundancy gives robustness: losing a few units barely matters, unlike losing the one expensive humanoid.

Critiques

  • Designing local rules that reliably produce a desired global behavior is hard, and emergent behavior can be equally hard to predict or certify.
  • Swarm demos are often slow and confined to lab arenas, with limited tasks of practical value so far.
  • Foundation-model advocates argue that one capable general-purpose robot is more useful than many that can only do collective tricks.

Representative works

Relationships

How to get it

Land on this cell in the full or pro test: x ≈ +0.82, y ≈ −0.82, paradigm ≈ +0.84.

A first draft. Corrections and better arguments are welcome through a pull request. Edit this entry on GitHub Edit its data