Image-Aided Navigation Using Cooperative Binocular Stereopsis

摘要

This thesis proposes a novel method for cooperatively estimating the positions of two vehicles in a global reference frame based on synchronized image and inertial information. The proposed technique - cooperative binocular stereopsis - leverages the ability of one vehicle to reliably localize itself relative to the other vehicle using image data which enables motion estimation from tracking the three dimensional positions of common features. Unlike popular simultaneous localization and mapping (SLAM) techniques, the method proposed in this work does not require that the positions of features be carried forward in memory. Instead, the optimal vehicle motion over a single time interval is estimated from the positions of common features using a modified bundle adjustment algorithm and is used as a measurement in a delayed state extended Kalman filter (EKF). The developed system achieves improved motion estimation as compared to previous work and is a potential alternative to map-based SLAM algorithms.