Autonomous flight in unstructured and unknown indoor environments

Abstract

This thesis presents the design, implementation, and validation of a system that enables a micro air vehicle to autonomously explore and map unstructured and unknown indoor environments. Such a vehicle would be of considerable use in many real-world applications such as search and rescue, civil engineering inspection, and a host of military tasks where it is dangerous or difficult to send people. While mapping and exploration capabilities are common for ground vehicles today, air vehicles seeking to achieve these capabilities face unique challenges. While there has been recent progress toward sensing, control, and navigation suites for GPS-denied flight, there have been few demonstrations of stable, goal-directed flight in real environments. The main focus of this research is the development of real-time state estimation techniques that allow our quadrotor helicopter to fly autonomously in indoor, GPS-denied environments. Accomplishing this feat required the development of a large integrated system that brought together many components into a cohesive package. As such, the primary contribution is the development of the complete working system. I show experimental results that illustrate the MAV's ability to navigate accurately in unknown environments, and demonstrate that our algorithms enable the MAV to operate autonomously in a variety of indoor environments.