Abstract
In this thesis, we address the problem of autonomous exploration and online 3D reconstruction of complexe environments using a quadrotor equipped with a localization sensor and a depth camera. To solve these problems, we present SplatPlanner, a novel autonomous exploration method that allows to simultaneously perform a collision-free exploration and a voxel-based reconstruction within a given environment. In order to perform this two task efficiently, we propose a new exploration algorithm based on a novel Permutohedral Frontier Filtering (PFF) which relies on a combination of highly efficient operations stemming from bilateral filtering. Specifically, our PFF aggregates spatial information about unobserved voxels which lie along the boundaries of unknown regions into density scores to guide the entire exploration. We propose a real flight experiment with a custom-built quadrotor which support the feasibility of our method in real conditions. In order to fairly and accurately assess the performance of autonomous exploration methods, we propose FLYBO, a simulation environment tailored for the task of autonomous exploration planning and online surface reconstruction using a quadrotor. FLYBO includes (i) 11 challenging realistic indoor- and outdoor scenes of increasing complexity and size, (ii) a comprehensive benchmark of 7 of the top-performing autonomous exploration algorithms, including methods without publicly available code, (iii) a unified experimental system factorizes the routines shared by autonomous planners in order to fairly and accurately assess their exploration performance in a controlled environment. In particular, SplatPlanner emerges as the best compromise regarding key performance metrics between exploration and surface reconstruction amongst the leading state-of-the-art systems.