ADD is not a rulebook. It is the engine that lets your team write, validate, and deploy your own driving rules — at any scale, for any region, without retraining a model.
Out-of-the-box demo library — what ships on day one
Real-world driving has infinitely many scenarios. Any fixed rulebook — no matter how large — will always fall short. ADD solves this at a different level.
ADD gives your engineering team the language, toolchain, and runtime to express any driving behavior as explicit, verifiable rules — and deploy it without retraining a model.
The 184 starter rules and 10 reference scenarios are a proof-of-concept demo. The product is the infrastructure beneath them.
Vehicle A-IV's trajectory compresses B-III, which in turn threatens Ego. ADD detects this indirect threat chain and suppresses the lane change — without waiting for direct contact risk.
Click any zone to inspect its live threat status and the rule that fired.
These 10 scenarios demonstrate ADD's range — from simple reflex rules to multi-vehicle game theory to semantic reasoning. They are a starting point, not a ceiling. Your team extends the library using the same toolchain.
Who goes first? Symmetry-breaking for simultaneous arrival, multi-vehicle priority arbitration.
Same intersection, different outcomes. Jurisdiction hot-swap — zero retraining required.
Full ISO 26262 causal chain output — 100ms to reproduce any past decision with complete rule_id sequence.
Rain × pedestrian × wheelchair — three conditions firing simultaneously. SWRL/OWL semantic reasoning layer.
Two AVs face-to-face, both yielding indefinitely. Hash-parity tie-breaking resolves the paradox.
Barrier state determines action. Safety-layer rule — cannot be overridden by any lower-priority rule.
Blocked front, rear, and left simultaneously. Two-hop indirect threat cascade inference.
Heavy truck rules, additional edge cases, and client-specific custom scenarios built on the same framework.
Rules are plain YAML — readable, editable, versionable by your own engineers. No supplier dependency for policy changes. Your ODD, your regional laws, your brand's driving character — expressed and enforced directly.
Every decision is the output of explicit, auditable rules — not opaque neural weights. Full causal chain on every output. ISO 26262 ASIL-B responsibility tracing at 0.18ms. Engineers can read and certify each rule path.
Rule modules are versioned and swappable without system restart. Switch jurisdictions in 0.08ms. Update driving policy at software pace — not hardware revalidation cycles. Roll back any change instantly.
ADD replaces only the Safety Check layer in a conventional E2E pipeline — adding white-box auditability while preserving the planning layer's implicit generalisation. Compatible with your existing architecture from day one.