Swarm Rule¶
Method — instantiates Local Rule Design
Implements local rule design by giving many agents simple proximity, movement, following, separation, or alignment rules whose aggregate behavior forms coordinated motion or coverage.
A swarm rule loads all coordination into the behavior of a single, identical agent, then runs many copies of it so that coordinated motion or coverage emerges as a real-time physical side effect. Every agent executes the same short rule over only what its sensors can reach — who is nearby, how fast, which way — and acts by moving. No agent holds, plans, or represents the global pattern; the flock, the swept field, the merging lane is nowhere on record, only in the aggregate of thousands of local moves. What makes it this mechanism and not its modeling twin is that a swarm rule is deployed onto real agents in a real medium: the rule is a live controller whose output is motion, not a hypothesis you watch converge on a screen.
Example¶
A vineyard hires a fleet of about 40 palm-sized drones to photograph every vine row for early signs of mildew before harvest — with no per-drone flight plan and no central choreographer. Each drone runs three rules at once: stay at least ~4 m from any drone it can see (separation), roughly match the heading and speed of neighbors within ~10 m (alignment), and drift toward ground it senses has not yet been photographed (coverage bias). On launch they fan out, thread between rows without collisions, and sweep the block. When eight drones peel off to recharge, nobody reassigns their territory — the coverage bias in the survivors simply re-weights toward the freshly-uncovered ground, and the gaps close on their own. The outcome is complete photographic coverage that degrades gracefully rather than failing, produced entirely by identical local rules.
How it works¶
- One policy, many copies. The design object is a single agent's rule; scale comes from replication, not from a bigger plan.
- Neighbor-scoped sensing. Each agent reads only agents and cues within a sensing radius; "the swarm" is never an input.
- Act by moving. The rule's output is a velocity or heading — the pattern is kinetic, happening in the world in real time.
- Weighted drives. Typically a vector sum of a few pulls (separation, alignment, cohesion, goal-gradient); their relative weights set the character of the emergent motion.[1]
- Robustness from redundancy. No agent is special, so losses and additions are absorbed rather than repaired.
Tuning parameters¶
- Sensing radius — how far each agent looks. Wider gives tighter coordination but more crowding, noise, and compute; too narrow and the swarm fragments.
- Drive weights — the balance of cohesion versus separation. Tip toward cohesion and the swarm clumps; toward separation and it scatters and loses the pattern.
- Density — agents per unit space. The same rule that coordinates at one density stampedes or disperses at another, so density is a first-class dial.
- Response gain and damping — how sharply an agent reacts to a sensed change. High gain is snappy but breeds collective oscillation; damping trades responsiveness for stability.
- Goal-gradient strength — how hard the task pull (coverage, target) competes with the flocking drives.
When it helps, and when it misleads¶
Its strength is scale with resilience: coordination that costs nothing per added agent, survives individual failures, and needs no central route for anyone. It shines exactly where a central planner would be a bottleneck or a single point of failure.
Its honest failure mode is emergent pathology — a rule that is sensible for one agent can, in aggregate, produce collective oscillation, gridlock, or a stampede, and these appear only at certain densities and gains. The classic misuse is tuning the weights until the swarm "looks right" in one test arena and then deploying it at a different density, where the identical rule now clumps or shatters. The guarding discipline is to test the rule across the full range of densities and losses it will actually meet, add damping before gain, and treat the aggregate as something a human operator (or a modeling step) must watch — the rule itself runs open-loop and cannot see the pattern it is making.
How it implements the components¶
local_rule— the short per-agent policy (separation / alignment / coverage) that is the whole design.local_actor_model— the rule is written to a specific agent profile: a given sensing range, speed, and actuation, replicated identically.interaction_medium— the physical space and the neighbor topology ("within range") the rule plays out in; the same rule behaves differently in open air than in a corridor.feedback_signal— each agent's live sensing of neighbors and environment is the feedback that steers its very next move.
Swarm Rule does not implement emergent_pattern_monitor or rule_revision_loop — a deployed swarm runs open-loop at the agent level and never observes or searches over the aggregate it produces; that observe-the-pattern-and-iterate-the-rule work is Cellular Automata Rule. A swarm runs a rule on hardware; a cellular automaton is where you discover which rule to run.
Related¶
- Instantiates: Local Rule Design — a swarm rule is the mobile-agent instance of shaping a system by shaping one participant's behavior.
- Sibling mechanisms: Cellular Automata Rule · Market Rule · Protocol Rule · Team Working Agreement · Community Norm · Routing Rule · Decentralized Governance Norm
Editorial Notes¶
Form Classification¶
Form family: Control, Automation & Runtime
Rationale: Swarm Rule is defined in the frozen evidence as: Implements local rule design by giving many agents simple proximity, movement, following, separation, or alignment rules whose aggregate behavior forms coordinated motion or coverage. Its operative deployed or enacted form is therefore Control, Automation & Runtime.
Nearest alternative: Rule, Policy & Commitment — Rule, Policy & Commitment can support this mechanism, but the evidence centers the concrete operation described above rather than the alternative family's defining operation.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Robotics & Automation
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Local interaction rules producing collective motion are foundational swarm robotics.
Related originating lineages:
- Biology & Ecology — Flocking and eusocial behavior supply biological models.
- Computer Science & Software Engineering — Computer science and software-engineering practice supplies a parallel or contributing lineage for the mechanism's defining operation: implements local rule design by giving many agents simple proximity, movement, following, separation, or alignment rules whose aggregate behavior forms coordinated motion or coverage.
- Engineering & Design — Engineering design, reliability, and systems-safety practice supplies a parallel or contributing lineage for the mechanism's defining operation: implements local rule design by giving many agents simple proximity, movement, following, separation, or alignment rules whose aggregate behavior forms coordinated motion or coverage.
- Systems Thinking & Cybernetics — Emergent coordination follows decentralized feedback.
Review resolution: The blind reviewers agree that robotics_automation is the primary origin and differ only on alternate origin disagreement, origin mode disagreement, domain reach disagreement, encyclopedia synthesis disagreement. I preserve every independently explained alternate from both records rather than imposing a numeric cap. I retain cross_disciplinary_synthesis because the combined evidence shows material contributions from several lineages. The broader reach of multi_domain records portability separately from historical provenance; encyclopedia_synthesis=true preserves the affirmative synthesis judgment where either reviewer identified one.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
A swarm rule is worth pre-testing in a Cellular Automata Rule-style sandbox before it touches hardware, because the pathologies that matter (oscillation, stampede) are aggregate effects the agent can never sense — but the pre-test is a separate modeling step, not part of the swarm rule itself.
References¶
[1] Craig Reynolds' 1987 "boids" model produced lifelike flocking from three local rules — separation, alignment, and cohesion — and is the canonical demonstration that complex aggregate motion needs no leader, only a shared local rule. withdrawn registry ↩