<p>Amphibious unmanned ground vehicles (AUGVs) require a robust and adaptive path planning framework to operate effectively in complex amphibious environments. However, conventional algorithms often struggle with inconsistent path quality, suboptimal cross-domain decisions, and limited adaptability to dynamic environments. To address these limitations, this paper proposes the Amphibious Adaptive Dynamic A* (AAD-A*) algorithm, which integrates two core mechanisms: the Adaptive Path Planning Strategy Selection Mechanism (APPSM) and the Adaptive Cross-Domain Point Selection Mechanism (ACPSM). A hierarchical map is constructed using a Coarse–Fine Dynamic Voxel Grid, integrating 2D occupancy grids and a 2.5D digital elevation model (DEM), enabling precise terrain analysis and modeling in amphibious environments. AAD-A* algorithm enhances heuristic evaluation by incorporating terrain-aware features, thereby improving path stability and reliability. APPSM dynamically selects between single-environment and cross-domain planning strategies based on the environmental relationship between the start and goal, avoiding unnecessary transitions and improving route feasibility. ACPSM further refines transition point selection by jointly considering geometric constraints, environmental cost, and domain crossing safety, ensuring reliable navigation across domains. Extensive experiments demonstrate that AAD-A* significantly outperforms traditional A* and representative baselines in terms of path stability, terrain adaptability, and transition safety. These results validate the effectiveness of AAD-A* for amphibious navigation and highlight its potential for real-world deployment in intelligent amphibious robotic systems.</p>