Propagation-Aware Stabilization for Mission Assurance: Failure-Diverse Supervisory Control for Resilient Autonomous Aerial Teams
Autonomous aerial teams can remain internally coherent while collectively acting on a corrupted mission reference. This technical report introduces Propagation-Aware Stabilization for Mission Assurance (PAS-MA), a supervisory architecture that separates collective coherence from mission-reference correctness and governs when shared belief is permitted to acquire mission authority. PAS-MA preserves quorum-dominant autonomy while incorporating failure-diverse minority assurance mechanisms designed to challenge exceptional states without becoming an alternate command system. In matched synthetic C-sUAS experiments, PAS-MA achieved 50/50 mission successes under the tested overt and stealth spoofing conditions, compared with 22/50 for W-MSR and 0/50 for fixed consensus in each condition. Under deliberately harsher shared-failure conditions, assured mission success fell to 0/50 despite preserved local task opportunities, exposing a coherent-but-wrong failure boundary central to the proposed architecture. Additional computational experiments characterize evidence-dependence and availability limits, protected-reference failure, conservative minority governance, and capacity/tempo saturation. The results support PAS-MA as a falsifiable mission-assurance research framework rather than an operationally validated system. This work is an unclassified, platform-agnostic computational proof of concept using synthetic data. It does not claim operational effectiveness, technology readiness, validation for a particular aircraft or mission, or immunity to arbitrary compromise. Implementation-specific tuning rules, deployable software logic, detailed sensor-selection trade studies, operational tactics, and other engineering particulars are intentionally outside the scope of this public preprint.
Authors
- Matthew Dominik (ORCID: https://orcid.org/0009-0007-9449-2630)
Institutions
- Dominion (United States) (US)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-08-24
- DOI
- https://doi.org/10.5281/zenodo.22086803
- Primary Topic
- Distributed Control Multi-Agent Systems
- Type
- preprint