Indoor Path Planning and Navigation Of Quadrotors

Abstract

In this thesis we study the problem of indoor quadrotor navigation. Specifically we focus on the design of an optimal controller, and an optimal state estimator. We then provide a working example of an autonomous quadrotor navigating through a static environment. The implemented controller is a Linear Quadratic Regulator (LQR) augmented in such a way to solve the Servo- Problem. The state estimation is achieved by coalescing via a Kalman Filter, the measurements from an Inertial Measurement Unit (IMU), an Ultra-WideBand localization system, and a monocular camera-based object detection algorithm. To the best of the author’s knowledge, this is the first work that combines these three technologies for an onboard real-time implementation of a localization mechanism. The implementation is achieved with a Hardware-in-the-loop (HIL) simulation technique, where the dynamic model of the quadrotor is simulated in an open-source 3D robotics simulator, and the navigation system is implemented on an Artificial Intelligence embedded computer. Our results show that the camera-based solution is a viable option capable of further improving the performance of the purely UWB approach.