Frontier-Guide Planner for UAV Navigation in Challenging Long-Continuous Obstacle Environments
Yuhang Liu, Yan Yang, Yanmin Liu, Yi Cui, Haikuo Liu, Zhen Li · 2025
Recent advancements in unmanned aerial vehicles (UAVs) path planning have shown promise in indoor obstacle-filled and outdoor forest environments, where obstacles are typically of low continuity. In such environments, suboptimal solutions are often close to safe solutions, allowing optimizers to efficiently refine dangerous trajectories. However, in environments with long-continuous obstacles, such as indoor buildings and complex mazes, existing local planning methods face significant challenges. These methods often fail to escape unsafe suboptimal solutions promptly, which can lead to mission failure. In this paper, the Frontier-Guide Planner is proposed to expand the application scenarios of UAVs. First, a mixed sampling method combining uniform and Gaussian approaches is employed to extract high-quality viewpoints from the frontier. Subsequently, the viewpoints are evaluated using the existing environmental information, and the optimal viewpoint is selected as the guiding point. Finally, a safe trajectory based on the perceptual space is generated to guide the flight. The carefully designed simulation comparative experiments have demonstrated the applicability of the proposed method in both dense forest and long-continuous obstacle environments. Compared with existing frontier-based planning algorithms, the task completion time is reduced by 2% to 17%, and the success rate is increased by 14% to 32%.