Adapting Pathfinding Algorithms for 3D Environments: Performance Analysis and Real-World Applications

Pathfinding algorithms are essential for autonomous navigation in fields such as robotics, UAV systems, and game AI. While extensively studied in two-dimensional (2D) spaces, their performance in three-dimensional (3D) environments, characterized by higher complexity and search space volume, remains less explored. This thesis benchmarks two foundational algorithms, Dijkstra’s and A*, in voxel-based 3D grids under varying conditions of movement complexity, grid size, and obstacle density. Using a simulation-based experimental approach, it analyzes scalability, efficiency, and solution quality using metrics such as computation time, memory usage, node expansions, and path characteristics. The results demonstrate that heuristic-guided A* variants significantly outperform Dijkstra’s in computational efficiency, particularly in large or dense environments. A* Manhattan performs best in constrained, axis-aligned spaces, while A* Euclidean excels in diagonal-rich, flexible movement models. In contrast, Dijkstra’s, though optimal in path quality, scales poorly and becomes impractical in 3D scenarios with high branching factors. These findings extend prior work by offering a broader, rigorous evaluation and comparison of unmodified algorithms, highlighting both their strengths and limitations in real-world-inspired 3D navigation tasks.