Autonomous object mapping using UAV-based remote sensing

Abstract

Safety plays a huge role in our world. As technology advances, we can harness it to negate risks in our everyday life. This thesis is made for a project by Työtehoseura, who are responsible for Finland's yearly warehouse pallet rack inspections. Their upcoming innovation is to automate this inspection process by introducing commercial quadrocopters capable of autonomous defect detection of these warehouse shelf structures. With University of Turku as a contractor for the project, in this thesis we build a digital playground for the team to experiment with. The simulation allows a fast, risk-free testing of different sensors, algorithms, drone models, warehouses and pallet racks. In this thesis we evaluate different existing methods in robot simulation, pathfinding, collision avoidance and simultaneous localization and mapping. As no suitable existing solutions for the problem are found, in this thesis a solution built on top of Microsoft AirSim drone navigation framework is offered. The proposed solution, and its components, are presented, explained and illustrated. The end result is a simulated quadrocopter equipped with a 3D-lidar sensor capable of mapping shelf structures of various shapes and sizes. A demonstration video of this simulation will be used as a promotion material for the project.