Unmanned Aerial Vehicle Path Planning for Exploration Mapping

Abstract

This paper presents a new path planner for the exploration of previously unknown environments. The proposed algorithm uses a Next-Best-View (NBV) fashion to decide the movements ahead. A Rapidly-Exploring Random Tree (RRT) is built and the node with the highest gain is the next configuration to execute, based on the amount of unmapped volume in the field of view (FoV) of the sensor employed. The proposed planner is able to send online a 3D occupancy map of the volume and work with a variety of different depth sensors. The performance is analyzed in various challenging simulations, such as exploration of a city street or a narrow tunnel. Moreover, real experiments indoor and outdoor were conducted to further validate the proposed approach results. In cases with large scale volumes, it can cover a wide range in affordable flight time.