Smartphone Quadrotor Flight Controllers and Algorithms for Autonomous Flight

Some of the most interesting recent research in quadrotor flight is related to fully autonomous systems, i.e. only using onboard processing and sensing. Many researchers have demonstrated different aspects needed for autonomy and within the last couple of years some have even demonstrated fully autonomous flight. [1], [2] One of the biggest challenges is finding powerful lightweight mobile processors, along with the state estimation and control algorithms that are fast enough to be run on those processors. In this paper we present some of the results of our use of a smartphone as a quadrotor flight controller. Modern smartphones have all the necessary sensors needed for quadrotor flight and also provide a powerful mobile computing environment. Additionally, smartphone vendors are rapidly increasing the computation performance and, since our work uses the standard Android operating system, software is easily transferred to new phones as they are released. The contributions of this paper are twofold. First, we describe the algorithms used towards the goal of fully autonomous flight. While many of the algorithms are not new, we have demonstrated that they are suitable for use on our smartphone flight controller. Secondly, this paper presents the first online implementation of a velocity estimation algorithm we recently proposed, [3] which generates image-space feature location prior distributions and then uses Bayesian inference to create “soft” point correspondences and calculate the maximum a posteriori velocity and height.