Key points are not available for this paper at this time.
Abstract. This study presents a voxel-based path planning framework for unmanned aerial vehicles (UAVs) operating in complex indoor dynamic environments. To overcome the limitations of traditional 2D grid maps and static planning methods, the proposed system integrates real-time voxel modeling, Kalman filter-based dynamic obstacle prediction, and an improved A* algorithm with kinematic constraints. The environment is reconstructed from LiDAR-acquired point clouds and discretized into uniform voxel grids to support efficient 3D spatial queries. Predicted obstacle trajectories are incorporated into a risk assessment mechanism that triggers path replanning when safety thresholds are violated. The enhanced A* algorithm introduces directional continuity constraints and Z-axis motion suppression to reduce path oscillations and improve trajectory feasibility. Experimental results in simulated warehouse-like environments demonstrate improved path smoothness, fewer vertical oscillations, and higher success rates in avoiding dynamic obstacles compared to conventional approaches. The framework offers a practical solution for real-time UAV navigation in cluttered indoor spaces such as logistics facilities and rescue scenarios.
Liao et al. (Wed,) studied this question.