Multi-Level-Frontier Empowered Adaptive Path Planning for UAVs in Unknown Environments

In unknown and complex environments, UAV path planning has found extensive applications in areas such as terrain exploration, disaster rescue and power line inspection. Despite significant advancements, real-time and safe path planning remains a critical challenge. This paper introduces an approach of multi-level frontier empowered path planning for UAVs, which takes iterative environmental perception, modeling and local planning, enabling UAV to continuously advance and gradually reach its destination. The proposed approach incorporates a scheme of multi-level frontier-based environmental modeling, allowing dynamic adjustment of modeling levels. This adaptability reduces the computational overhead of constructing frontier voxels, thereby speeding up the process of path planning Utilizing environmental information, the planner adopts a frontier-based method to plan local paths. Moreover, a backtracking and replanning mechanism is designed to deal with the dead-end dilemma, which is often encountered during UAV navigation. It leverages Historical State Record Tree (HSRT) to replan feasible paths from the reverted least feasible viewpoints, helping UAV escape from dead ends in environments. Simulation results across three diverse scenarios demonstrate the effectiveness of the proposed method in complex environments. Compared to existing methods, the proposed approach reduces planning time by approximately 20%, highlighting its practicality and efficiency in navigating in complex environments.