Enhancing Unmanned Systems Autonomy Evaluation: A Scalable Framework for Adaptive Scenario-Task Integration

This paper tackles persistent limitations in evaluating autonomy levels of unmanned systems, such as unsystematic subject generation, poor complexity scaling, and rigid adaptability to emerging technologies. We introduce an innovative framework that dynamically integrates environmental and task dimensions to generate tailored assessment scenarios. By defining a three-tier autonomy standard (machine-assisted A1, human–machine collaborative A2, machine-autonomous A3) and leveraging the Analytic Hierarchy Process for quantifiable grading, our method establishes a robust indicator system covering geography, meteorology, hydrology, electromagnetics, task composition, requirements, conditions, and risks. Unlike traditional expert-driven methods, the proposed model features complexity quantification, intelligent enabling-point analysis, and dynamic adaptability mechanisms, enabling a closed-loop process of level definition, scenario-task design, and capability verification. Experiments across diverse domains demonstrate significant improvements in subject generation efficiency, contextual alignment, and scalability. The framework provides a standardized, engineering-oriented approach for cross-domain autonomy evaluation.