LawsOf Robots

The System

Cognition & Coordination

Cognition & Coordination maintains evidence-based belief and prepares work for RES judgment, then coordinates only what RES has authorized.

IN DEVELOPMENT

The architecture is in development. This page describes its intended contracts and authority boundaries; it does not claim production availability, robot compatibility, or a cleared product mark.

One job, four responsibilities

Maintain justified belief and carry authorized work without becoming the judge

Belief custody

Keep observations and assertions with their source, time, confidence, and history so an old claim is never presented as a fresh fact.

Uncertainty made explicit

Expose stale evidence, blind spots, known unknowns, and contradictions instead of silently filling gaps or choosing a convenient claim.

Candidate preparation

Prepare perception requests, candidate plans, and bounded goals for RES review. A candidate is a proposal, never permission to act.

Authorized coordination

After RES authorizes work, preserve its exact scope while correlating the tracked goal, feedback, result, and returned evidence.

Responsibility map

Each boundary keeps a different kind of control

The same event can involve all three layers without giving all three the same authority.

Evidence enters

Cognition
Attributes inputs and maintains the belief record without certifying them as true.
RES
Receives available evidence as context for an independent governing decision.
Robot Integration
Carries observations and health signals from sensors or middleware without deciding permission.

Uncertainty changes

Cognition
Ages confidence, records missing capability, and relates incompatible claims.
RES
Treats uncertainty and contradiction as evidence—never as an automatic authorization.
Robot Integration
Reports availability, diagnostics, covariance, or loss without inventing certainty or success.

Work is proposed

Cognition
Frames a bounded candidate inquiry, plan, or goal and preserves its assumptions.
RES
Alone may authorize, constrain, reject, or request clarification under the Laws.
Robot Integration
Remains idle; a candidate is not an executable command.

Authorized work runs

Cognition
Binds the unchanged RES decision to a tracked goal and correlates returned evidence.
RES
Retains the governing decision, conditions, and audit identity.
Robot Integration
Owns sensing, navigation, control, drivers, and actuation mechanics, then returns feedback and results.

From evidence to execution evidence

Authorization remains a visible crossing, not an implied side effect

  1. 1

    An observation or assertion arrives with attributable provenance.

  2. 2

    Cognition updates belief while preserving age, uncertainty, and conflict.

  3. 3

    A bounded candidate is prepared with its assumptions and required evidence.

  4. 4

    RES independently authorizes, constrains, rejects, or requests clarification.

  5. 5

    Only the authorized result crosses into tracked coordination and integration.

  6. 6

    Execution evidence returns and is correlated without rewriting the original RES decision.

Failure and degradation

Missing components reduce capability; they do not move authority

If Cognition is unavailable

The architecture requires RES to remain able to govern without this layer. Context and convenience may be reduced, but Cognition must never become a gate that can disable RES authority.

If RES is unavailable

Cognition may retain evidence and candidate work, but no consequential candidate becomes authorized work. The boundary fails closed rather than manufacturing permission.

If sensing or integration is unavailable

The loss becomes evidence: beliefs age, unknowns become explicit, and unfinished work cannot be reported as successful.

Deliberate limits

What Cognition & Coordination does not do

  • It does not certify truth. Confidence always remains attributable and inspectable.
  • It does not decide permission or obligation; those judgments remain exclusively with RES.
  • It does not own SLAM, motion planning, control loops, robot drivers, or body-level actuation.
  • It does not expose a live development system, unrestricted planning service, or public robot-control endpoint.

Back to System overview