Autonomous navigation in rough terrains is challenging because physical properties of the terrain, such as slope and surface roughness, have a direct impact on driving stability, making it difficult to ensure reliable traversal using path planning bas...
Autonomous navigation in rough terrains is challenging because physical properties of the terrain, such as slope and surface roughness, have a direct impact on driving stability, making it difficult to ensure reliable traversal using path planning based solely on obstacle avoidance. This study compares the performance of representative local path planning algorithms, including Dynamic Window Approach(DWA), Timed Elastic Band(TEB), Regulated Pure Pursuit(RPP), and Model Predictive Path Integral(MPPI), based on 2.5D costmap that incorporates traversability information, and experimentally demonstrates that existing algorithms suffer from instability, inefficient paths, and collisions in unstructured terrains. To overcome these limitations, this paper proposes an integrated planning framework called Hybrid A*-guided MPPI. In the proposed method, Hybrid A* generates a reference path that prioritizes regions with low slope and roughness based on traversability evaluation, and MPPI optimizes local control inputs along this path. This structure enables global path strategy and local control optimization to be organically combined within a single framework, maintaining consistency and stability of the overall navigation path while adapting to real-time environmental changes. Experimental results conducted in various rough terrain scenarios show that the proposed method consistently improves key performance indicators, including reduced roll and pitch oscillations, enhanced path smoothness, and increased navigation success rate compared to existing approaches.