Visual Odometry : Autonomous UAV Navigation using Optic Flow and Stereo

Abstract

Visual odometry is vital to the future of mobile robotics. In this paper, we demonstrate a method that combines information from optic flow and stereo to estimate and control the current position of a quadrotor along a pre-defined trajectory. The absolute translation in 3D is computed by combining the optic flow measurements between successive frames and stereo-based height over ground. The current 3D position, as estimated from path integration of the incremental translations, is controlled in closed loop to follow the prescribed trajectory. The performance of the system is evaluated by measuring the error between the initial and final positions in closed circuits. This error is approximately 1.7% of the total path length.