A 3D reconstruction pipeline enables robots to leverage sensors of opportunity鈥攃ollocated, heterogeneous, consumer-grade sensors embedded in the operational environment鈥攖o augment onboard robot sensing capabilities. A modular infrastructure integrates popular visual odometry and point cloud generation algorithms, allowing customization based on environmental conditions. In addition, a novel iterative closest point (ICP) algorithm supports fusing point clouds into a single data product without relying on a priori scale relationships, allowing diverse sensors to contribute to map construction.
Associated Researchers
Dan Szafir
Assistant Professor of Computer Science, UNC Chapel Hill • Former ATLAS faculty member