| name | carve-q-quantum-driving-repair |
| category | quantum |
| description | Verifier-shielded quantum-AI search architecture for certified autonomous driving repair. Combines Grover/Dür-Høyer minimum finding with classical safety certification. |
| arxiv | 2606.06531 |
| published | 2026-06-03T00:00:00.000Z |
| categories | cs.AI, quant-ph |
| activation | carve-q, quantum-ai, verifier-shielded, autonomous-driving, quantum-minimum-finding, grover-search, certified-autonomy |
CARVE-Q: Quantum-Proposed, Classically Certified Interactive Driving Repair
Overview
CARVE-Q introduces a verifier-shielded quantum-AI search architecture for autonomous driving repair after a correct driving veto. It applies quantum minimum finding (Grover/Dür-Høyer algorithm) to multi-agent repair lattices while keeping all safety authority classical.
Core Methodology
Verifier-Shielded Architecture
- Quantum proposes: Quantum minimum finding searches the repair lattice
- CARVE certifies: Classical authority validates all proposed repairs
- Black-box lattice: Product lattice M = ∏|Aⱼ| for multi-owner repair
- Query separation: O(√M) quantum vs Θ(M) classical worst-case queries
Key Technical Components
- Repair lattice construction: Finite lattice of feasible joint repairs
- Quantum oracle design: Black-box access to repair lattice
- Classical verifier: Certificate preservation and safety validation
- Priority non-elicitation: Finite-precision reversible oracle constructibility
Performance
- Validated up to 65,536 assignments
- 100% right-of-way respect on Lanelet2-grounded INTERACTION replay
- 100% blame consistency, zero priority false positives
- Quadratic speedup in query complexity
Implementation Patterns
Trust-Bounded Quantum-AI
┌─────────────────┐ ┌──────────────────┐
│ Quantum Search │───▶│ Classical Verify │
│ (O(√M) queries)│ │ (Certificate) │
└─────────────────┘ └──────────────────┘
Black-box Safety Authority
Lattice Access Non-negotiable
Certificate Structure
- Binding rule identification
- Selected joint repair
- Right-of-way-scaled cooperation envelope
- Responsibility-weighted cost split
- Ego-only fallback specification
When to Use
- Multi-agent autonomous systems requiring certified safety
- Quantum-classical hybrid architectures with trust boundaries
- Safety-critical AI with verifiable decision-making
- Any system needing quantum speedup with classical certification
Pitfalls
- Quantum speedup only applies to black-box search, not the entire pipeline
- Classical verifier must remain fully authoritative
- Reversible oracle construction requires careful finite-precision handling
- Query complexity advantage assumes oracle access is the bottleneck
Verification Steps
- Validate oracle constructibility (reversible, finite-precision)
- Verify certificate soundness (safety constraints satisfied)
- Check priority non-elicitation (no false positives)
- Confirm query separation (O(√M) achieved in practice)
