The Robot Expert System is the governing and authorization layer within the Laws Of Robots architecture.
✓ EARLY PRODUCT · OWNER USE
RES is at early product maturity and in Owner use. This page does not claim a shipped third-party robot deployment.
Current scope
This section will explain how RES applies the Laws and Owner rules, records decisions, carries obligations, and remains available when outside services are unavailable.
Authority boundary
RES determines permission and obligation. It does not claim to supply perception, navigation, cognition, or low-level robot control.
What this section will add
A later development pass will add the decision path, audit relationship, failure posture, and links to the current Law and evidence records. Owner sign-in and registration remain in the separate RES Owner UI task.
The Robot Expert System (RES) is the governing and authorization layer. It applies the Laws and Owner rules to the available decision context, determines what the robot may and must do, and records why.
✓ EARLY PRODUCT · OWNER USE
RES is at early product maturity and in Owner use. This page describes the public governance contract; it does not expose a live RES system or claim a shipped third-party robot deployment.
One governor, four invariants
Govern the decision without inheriting every other job
01
Safety floor fixed
The Laws are evaluated in their fixed order. Owner rules may make the robot's limits stricter, but they cannot lower the safety floor.
02
Authority kept singular
Evidence, candidates, adapters, and specialist modules may contribute to the path. Only RES may turn that context into permission or obligation.
03
Reasons kept attributable
The decision remains connected to the governing rules and admitted context instead of becoming an unexplained yes, no, or command.
04
Decision kept distinct
What RES authorized remains distinguishable from what the robot later attempted, completed, stopped, or failed to do.
The governing path
From decision context to an accountable outcome
1
Decision context arrives
A request, available evidence, known uncertainty, and any bounded candidate work arrive as attributable context—not as permission.
2
The Laws set the floor
The Laws are applied in their fixed order. An Owner rule may tighten the available path, but it cannot weaken that floor.
3
RES makes the result explicit
RES may authorize, constrain, reject, or request clarification, and it identifies any obligation that follows from the decision.
4
Only authorized work crosses
An authorized goal carries its scope, constraints, cancellation semantics, and decision identity into coordination and integration.
5
Outcome returns as evidence
Progress, stops, faults, and terminal state are correlated to the decision without rewriting what RES originally authorized.
Authority map
Inputs can inform the decision; none can inherit it
Each neighboring part contributes something different while RES remains the only governing and authorization layer.
The Laws
Contributes
The fixed safety floor and governing order.
Authority limit
They are not optional inputs that another component may average away or override.
Owner rules
Contributes
Additional limits and requirements chosen by the Owner.
Authority limit
They may tighten the governed path; they cannot weaken the Laws.
RQR
Contributes
Attributable evidence, explicit uncertainty, and bounded candidate work.
Authority limit
A belief or candidate can inform RES, but it cannot authorize itself.
Robot Integration
Contributes
Capability status, observations, execution feedback, and bounded actuation.
Authority limit
It may execute or stop authorized work; it cannot decide the governing purpose or widen its scope.
Record and review
The record makes a decision checkable, not infallible
Surface
What it carries
What it does not claim
Decision record
The machine's decision and stated reason in a tamper-evident, append-only record that travels with the agent.
The record supports review of the recorded decision; it does not certify that the decision was wise.
Offline verification
A bounded check that the record is intact and that a decision can be re-derived and checked.
The check is designed not to depend on a server, dashboard, or a party's assurance.
Execution evidence
Returned observations, progress, stops, faults, terminal state, and unresolved uncertainty.
An outcome can be correlated to a decision, but it cannot silently revise the original authorization.
Independent verification
Outside review of the narrower claims the public record and certificate can actually support.
GRRC remains outside the product chain; its certificate is not certification of a robot or general safety.
Failure posture
A missing dependency cannot manufacture permission
If the network is unavailable
The judgment runs inside the machine and requires no network to decide. Its record is designed to remain verifiable offline.
If context is missing or contradictory
The gap remains visible decision context. Missing, stale, or conflicting evidence cannot silently become permission or proof of success.
If execution is unavailable
The affected work cannot be reported as complete. Unavailable capability, refusal, stop, or fault returns as evidence against the unchanged decision.
Deliberate limits
What RES Governance does not do
RES does not supply perception, navigation, motion planning, robot drivers, or low-level control.
RES does not make a robot more capable and does not certify that supplied evidence is true.
RES does not permit RQR, Integration, an extension, or an outside service to become a second governance engine.
RES does not turn a valid record into a claim that a decision was wise or that a robot is generally safe.
This public page does not expose a live decision endpoint, robot-control surface, account flow, or compatibility promise.