AI for Drone Warfare: An Experimental Survey

Artificial Intelligence (AI) is expanding drone autonomy across sensing, tracking, navigation, coordination, planning, action, assessment, and human supervision, yet experimental evidence remains fragmented across individual capabilities. This survey introduces the Autonomous Drone Warfare Stack (ADWS), a ten-layer framework for distinguishing strong component-level capability from experimentally demonstrated cross-layer integration across AI-enabled drone systems. The synthesis identifies substantial progress within individual functions but considerably weaker evidence of integrated operation, particularly in mission-state preservation, uncertainty transfer, recovery, feedback closure, resilience, and human authority. ADWS provides an evidence-derived reference architecture for examining how validated outputs move between autonomous functions, where integration remains experimentally unsupported, and which interfaces require stronger validation. The framework separates local performance from system-level integration while highlighting persistent limitations in reliable composition across the mission chain. These findings identify priorities for developing more robust, recoverable, resilient, and appropriately governed autonomous drone systems and for strengthening experimental evaluation of cross-layer interfaces overall.