High-altitude dynamic emergency routing optimization via a hybrid A*-CNN algorithm with AHP-enhanced modeling

In this study, an intelligent path planning framework integrating A*-CNN hybrid algorithm and analytical hierarchy method (AHP) augmented modeling is proposed for the path optimization challenges brought by three-dimensional terrain complexity, extreme climate conditions and dynamic disaster risks in emergency rescue scenarios in high-altitude disaster areas. The HSV color space transformation and median filtering are employed to construct a high-precision elevation model, which is then combined with an edge-protected gradient algorithm to extract terrain mutation features and achieve the joint quantitative characterization of slope, elevation, and risk. Innovative design of a convolutional neural network (CNN) cost enhancement model, using double convolutional layers to extract terrain edge features, retaining terrain depression information through Tanh activation function, and introducing a high-elevation mask and dynamic penalty mechanism to significantly strengthen the continuity of safe passage. In the field test of 15 settlements in Tingri County, Tibet, the hybrid algorithm shows significant advantages: the hybrid algorithm avoids the number of areas with >45° steep slopes by 68%, and the path elevation standard deviation σ is reduced by 42%; the convergence performance is optimized: the adaptation curve is continually optimized to, compared with the A* algorithm, converge to achieve a substantial increase in the fluctuation standard deviation is reduced by 86.9%. Experiments show that in the high-altitude region with risk coefficient > 0.6, the A*-CNN hybrid algorithm reduces the path risk exposure rate by 57.4%, and improves the comprehensive safety performance by 68% although the path length increases by 158,370.1 unit lengths. This study provides a theoretically reliable and practically feasible path planning paradigm for high-altitude emergency rescue, and will focus on dynamic disaster response and multi-intelligence body collaborative optimization in the future.