FGIP: A Frontier-Guided Informative Planner for UAV Exploration and Reconstruction

This article proposes a frontier-guided informative planner for unmanned aerial vehicle volumetric exploration and 3-D reconstruction, which can explore a complex unknown environment and provide the accurate truncated signed distance function reconstruction simultaneously. Different from the existing methods, the key insight of the proposed method is that the hybrid surface frontier is proposed to guide both the tree expansion of dynamic rapidly exploring random tree star and the informative trajectory generation. As a result, the proposed planner can achieve more efficient volumetric exploration with higher reconstruction quality. Specifically, hybrid global–local surface frontiers are designed to guide the potential viewpoints sampling and tree expansion, which results in directional exploration. Then, the hybrid surface frontiers are further leveraged to guide the candidate paths generation. On the basis, the path maximizing the new comprehensive gain is selected for the following B-spline trajectory optimization, which can further improve the reconstruction quality. Comparative simulation and real-world experiments are conducted to demonstrate the superior performance of the proposed method including the exploration efficiency and reconstruction quality.