BLADE-AV Governance Node: Authority-Governed Drive-by-Wire Safety Architecture for Autonomous Vehicles

Project Overview BLADE-AV is a hardware-enforced safety architecture designed to control when an autonomous vehicle is allowed to act. Instead of relying solely on perception and planning systems, BLADE-AV introduces an independent governance layer that evaluates sensor trust in real time and regulates drive-by-wire authority before commands reach vehicle actuators. At its core, the system answers a critical safety question: “Should the vehicle be allowed to execute this command given current uncertainty and potential threats?” To achieve this, BLADE-AV implements a ten-stage governance pipeline that processes sensor inputs through trust evaluation, identity verification, authority computation, consensus validation, and recovery mechanisms before allowing any action. The architecture combines probabilistic reasoning (Dempster-Shafer trust fusion), graded authority control with hysteresis, Byzantine fault-tolerant consensus, and hardware-level fail-safe enforcement via a normally-open relay. Unlike traditional safety approaches that operate at the planning or software level, BLADE-AV enforces safety at the hardware boundary, ensuring that unsafe commands cannot physically reach the drive-by-wire system. This makes the architecture robust against sensor spoofing, adversarial inputs, and partial system failures. The project includes a fully reproducible simulation environment, complete hardware specification, and open engineering artifacts, enabling independent validation and extension. While current results are simulation-based, the design targets real-world deployment under ISO 26262 ASIL-D safety requirements and aligns with SAE J3016 Level 4 autonomous driving constraints. BLADE-AV also demonstrates cross-domain applicability, showing that authority-governed autonomy is a general safety principle applicable to both civilian autonomous vehicles and defense systems. Technical Description BLADE-AV introduces a hardware-enforced authority gating architecture for autonomous vehicles using Dempster-Shafer trust fusion and a ten-stage governance pipeline: Sensors → ADARA → SATA → IFF → HMAA → MAIVA → FLAME → CARA → BDA → EFFECTOR Five core governance modules — SATA (trust fusion), HMAA (authority with hysteresis), MAIVA (2-of-3 Byzantine consensus), FLAME (deliberation windows), and CARA (GREP deterministic recovery) — execute on a Zynq UltraScale+ FPGA to regulate drive-by-wire authority based on real-time sensor trust. The platform demonstrates zero unsafe actions across 1,200 simulation runs (100 per scenario × 12 attack vectors) under defined threat scenarios, including adversarial machine learning, sensor spoofing, Byzantine faults, PCIe bus faults, and system-level failures. A fully reproducible dual-compute architecture (NVIDIA Jetson AGX Orin + Zynq UltraScale+ ZU7EV, 62 components, ~$16,287) is provided with complete engineering artifacts, including simulation, bill of materials, electrical and mechanical specifications, configuration files, and schematic diagrams. All results are derived from simulation; hardware validation is ongoing work. The architecture targets ISO 26262 ASIL-D functional safety requirements and SAE J3016 Level 4 constraints, and demonstrates cross-domain applicability between civilian autonomous vehicles and defense systems (BLADE-EDGE, DOI: 10.5281/zenodo.19177472).