Unmanned Aerial Vehicle (UAV) performing autonomous navigation exhibit high sensitivity to scene perturbations, even minor changes can drive mission outcomes to diverge sharply between success and failure. Therefore, constructing a mapping between scene parameters and the success rate of autonomous navigation missions can provide quantitative evidence for mission evaluation. However, the parameter space of autonomous navigation is effectively unbounded, and physical or simulation tests yield only discrete, single-trial observations linking parameters to success rates. Consequently, commonly used dense-sampling methods struggle to deliver stable and interpretable estimates with continuous coverage of the parameter space. To address this, the paper proposes a multi-scale entropy fusion–based method for constructing an uncertainty field that maps discrete parameters and execution outcomes to a full-domain, continuous parameter–success-rate estimation field. Simultaneously, the fused entropy is used as a spatial weight to suppress over-extrapolation in sparse regions, and Gaussian processes with kernel interpolation are employed for continuous reconstruction, thereby quantifying the relationship between parameters and success rate at arbitrary locations. Finally, experimental results on simulation-generated mission data demonstrate that the proposed method can stably construct a quantitative mapping between parameters and mission success under limited trials, and outperforms conventional baselines.

success-rate prediction,uncertainty quantification,multi-scale entropy fusion,deep learning enhancement