Nonlinear Real-Time Pose Estimation of Quadrotor UAV

Abstract

Object pose (rotation and translation) estimation problem arises in several domains of application. This chapter presents simplified nonlinear equations of motion for a quadrotor derived based on Newtonian mechanics before discussing and comparing the performance of two real-time image-based quadrotor pose estimation methods: one based on a commercial motion capture and tracking system utilizing six infrared (IR) cameras mounted around the test area aliased OptiTrack; the other based on classicPOSIT, an iterative pose estimation algorithm using a single image of a target object taken by an onboard camera. The geometry of the target and intrinsic parameters of the camera are known a priori; the image coordinates of five noncoplanar feature points on the target are extracted through a real-time image processing algorithm for pose estimation. Test results prove that onboard camera pose estimation is an attractive solution for autonomous real-time control of a quadrotor unmanned aerial vehicle (UAV).