Fusion of Visual and Inertial Measurement Information for Unmanned Aerial Vehicles Autonomous Navigation in Unknown Environment

An autonomous navigation scheme for unmanned aerial vehicles is presented based on visual and inertial measurement information fusion without the known ground cooperative target. The UAV relative translation and rotation motion parameters are estimated by inter-frame image feature detection and tracking. Then the relative motion parameters are considered to be the relative pose measurements of two consecutive images and the observation models explicitly related to the state are provided. In particular, the reduced-order state augmentation algorithm is used to account for the dependence of the measurements on the state at two different measurement time instants. The optimal fusion of visual and inertial measurement information is performed using the framework of the extended Kalman filter. Finally the feasibility and validity of the proposed navigation algorithm are demonstrated by mathematical simulation.