Authority-Governed UAV Autonomy for Contested Environments: Integrating Sensor Trust Fusion, Dynamic Authority Control, and Deterministic Recovery

This record presents the HMAA-UAV (Human-Machine Authority Architecture for UAVs), a governance-centered autonomy framework designed for operation in contested and adversarial environments.

The system integrates three core components into a unified flight-decision pipeline:

• Sensor-Anchored Trust Assessment (SATA): multi-sensor trust estimation using weighted Dempster-Shafer fusion

• Human-Machine Authority Architecture (HMAA): trust-conditioned authority computation across four graded levels

• Control Authority Regulation Architecture (CARA): deterministic recovery enforcement

The architecture is defined as a dual-compute UAV system design and evaluated in a browser-based simulation environment (100 Hz loop rate) across five adversarial scenarios, including GPS spoofing, RF communication loss, sensor disagreement, and compound failures.

Simulation results indicate (all timings are based on simulated clock execution, not physical hardware latency):

• Sub-200 ms trust-collapse detection under modeled conditions

• ~150 ms authority transitions

• Deterministic recovery activation under severe degradation

This record includes full reproducibility artifacts, including simulation configuration and per-run results data (n=250), as well as detailed system design specifications covering mechanical, electrical, and software architecture components.

This work contributes a governance-based autonomy paradigm in which trust assessment, authority regulation, and recovery enforcement are integral components of the control loop, rather than post-hoc safety mechanisms.

This research supports broader applications in resilient autonomous systems, safety-critical artificial intelligence, and national security-relevant UAV operations.