How to Prevent the Wrong Policy Version from Reaching a ROS 2 Robot
A concrete ROS 2 workflow for binding immutable learned-policy identity to approval and local dispatch.
Read the guideLearned-policy release control, one boundary at a time.
Practical guides for immutable release identity, approval, authorization, robot configuration, and evidence-first Shadow Mode.
Why Model Approval Is Not Execution Authorization
Separate durable review decisions from short-lived, context-bound permission to run a learned robot policy.
Read the guideShadow Mode for Learned Robot Policies
Build an evidence-only shadow path that exercises learned-policy release controls without hardware dispatch.
Read the guideHow to Bind Robot Configuration to a Policy Release
Include controller and device configuration identity in learned-policy release approval and execution checks.
Read the guide