ABSTRACT
Collision avoidance and path planning technologies for unmanned aerial vehicles (UAVs) have become essential for improving operational safety and efficiency in applications such as delivery drones and future aircraft. Furthermore, research is...
ABSTRACT
Collision avoidance and path planning technologies for unmanned aerial vehicles (UAVs) have become essential for improving operational safety and efficiency in applications such as delivery drones and future aircraft. Furthermore, research is rapidly progressing on integrating high-precision sensor technologies like LiDAR and vision sensors to enable high-difficulty missions in complex environments, such as indoor exploration, building entry, and disaster relief. This study proposes an algorithm for solving 3D path planning problems in complex environments with high obstacle density, such as urban areas or industrial facilities. Various algorithms exist, and grid-based algorithms have the advantage of generating the shortest path in complex environments and avoiding local minima. Furthermore, considering dynamic obstacles and real-time path regeneration, the grid-based algorithm D* Lite was selected for this study. To generate a UAV-flyable path, there are two methods: expanding the dimensionality by adding state information to search nodes and post-processing the generated path. Adding state information to search nodes can lead to the algorithm suffering from the curse of dimensionality, exponentially increasing computation time. Furthermore, post-processing and interpolating the generated paths carries the risk of generating paths that collide with obstacles. Therefore, this study proposes an approach that prioritizes rapid maneuvers by reflecting the dynamic constraints of UAVs. The proposed approach separates the global computation and path extraction processes of the D* Lite algorithm, utilizing the map information obtained during the global computation as a heuristic function. Furthermore, concepts for parent nodes, current nodes, and next nodes are added, adding angles and weights for the direction of travel. Based on this, the algorithm was designed to effectively generate a path that reflects the dynamic constraints of the UAV. The weights were expressed as parameters reflecting heuristics and rotations. Furthermore, to verify the algorithm, the heuristic and search range were expanded to three dimensions, and an appropriate cost function was selected for heuristic calculation. We also compared and analyzed the algorithms through the initial path generation and path regeneration processes. Initial path generation involves performing a global search to calculate distance cost values for each node and then performing initial path generation. At this time, we compared the performance of existing algorithms and the algorithm proposed in this study by varying the rotation weights. For path regeneration, we measured the time required for regeneration by adding information about arbitrary obstacles to the algorithm, and confirmed that the proposed algorithm generates efficient paths while maintaining dynamic feasibility.