A deliberately separated autonomy architecture

System Overview

Laws Of Robots brings governance, cognition, robot integration, and proving tools together without allowing any one layer to quietly inherit another layer's authority.

One authority, explicit responsibilities

RES alone determines what the robot may and must do. Cognition and integration components may inform, prepare, and carry out that work, but they cannot clear their own actions or override the Laws.

The operating relationship

One trustworthy loop

Information travels in both directions. Authority does not: consequential work must pass through RES before it reaches the robot's body.

Decision and authorization

RES — Governance

Applies the Laws and Owner rules, determines what the robot may and must do, and records why.

Beliefs and candidate work move toward RES; authorization and obligations return
Belief and preparation

Cognition & Coordination

Maintains what the robot has reason to believe, exposes uncertainty, prepares possibilities, and coordinates work RES has authorized.

Observations and evidence move upward; authorized goals and cancellations move downward
Sensing and actuation

Robot Integration

Connects sensors, navigation, drivers, and physical action through ROS 2, vendor middleware, or simulation adapters.

System parts

Different jobs, visible boundaries

Decision and authorization

EARLY PRODUCT · OWNER USE

RES — Governance

Applies the Laws and Owner rules, determines what the robot may and must do, and records why.

Boundary: RES governs and authorizes; it does not perform perception, navigation, or low-level control.

Understand RES governance

Belief and preparation

IN DEVELOPMENT

Cognition & Coordination

Maintains what the robot has reason to believe, exposes uncertainty, prepares possibilities, and coordinates work RES has authorized.

Boundary: It may propose and coordinate, but it cannot authorize work or override an RES decision.

Explore cognition and coordination

Sensing and actuation

IN DEVELOPMENT

Robot Integration

Connects sensors, navigation, drivers, and physical action through ROS 2, vendor middleware, or simulation adapters.

Boundary: Integration executes authorized goals; no particular middleware becomes the system's governing authority.

See the integration boundary

Planned specialist modules

PLANNED

Extensibility

Defines narrow, auditable ways for future specialist capabilities to contribute analysis or perform approved work.

Boundary: A specialist module cannot override the Laws, clear an RES decision, or become required for safe governance.

Review the planned extension model

Independent verification

GRRC remains outside the product chain

GRRC independently certifies and verifies claims. Keeping it outside the product hierarchy is part of the system's accountability model.

View the GRRC registry

Proving environment

Scenario Lab

Scenario Lab will use controlled TRES and RobotSim evidence to show how the same system contracts behave before physical hardware is involved.

See the Scenario Lab plan