SDB-1: Safety-Binding for Estimator Disagreement in Autonomous Systems
Authors/Creators
Description
SDB-1 is a narrow safety-binding rule: when disagreement across independent estimators exceeds a declared threshold δ, the system must not increase irreversible or high-risk actuation and should prefer reversible degradation until coherence improves. This note defines the invariant, a minimal disagreement metric (placeholder), a hysteresis stage model, a reversible-action policy sketch, and a minimal logging contract. It is an insertion-point artifact, not a perception model, planner, or certification framework.
Notes
Reference repository:
https://github.com/putmanmodel/SDB-1
Part of the broader Spanda architectural framework repository:
https://github.com/putmanmodel/spanda-architectural-framework
Related bridge paper:
Weak-Signal Ladder for Early AV Edge-Case Triage
https://zenodo.org/records/19750669
This note focuses on disagreement-triggered reversible degradation under estimator conflict; the AV edge-case triage paper addresses an earlier weak-signal promotion layer.
Files
PaperBridge8_SDB-1_Safety-Binding_for_Estimator_Disagreement_in_Autonomous_Systems_v0.2.pdf
Files
(196.0 kB)
| Name | Size | Download all |
|---|---|---|
|
md5:e5e4015738b32b789361fc2f4aefba9e
|
196.0 kB | Preview Download |