A Specialized Kalman Filter Framework for IMU Aided Stereo SLAM
This thesis aims at the design, implementation and analyzation of a method for Simultaneous Localization And Mapping. For that purpose a hard- and software system is developed and analyzed, that makes use of a pair of stereo cameras and an inertial measurement unit. This enables the system to perceive it’s environment and ego-motion respectively. This information is used to build up a sparse environment map (Mapping) and estimate the traveled trajectory (Localization). Moreover, the relative pose between the camera and the IMU (camera to IMU calibration) is estimated. The main focus is, to use a novel state parameterization improving the estimations consistency. That is, the appraisal of estimation errors is more reliable compared to other representations. At first the task of SLAM, previous work and the thesis’ structure are discussed. Following that chapter 2 describes the used notation and theoretic fundamentals and chapter 3 the theory of linear probabilistic estimators. Chapter 4 introduces the developed framework and it’s models, which are analyzed in part 5. After a detailed description of the systems implementation an experimental evaluation is performed in chapter 7. The thesis closes with a conclusion and an outlook to future work.
Use and reproduction:
No license. The provisions of the German Copyright Act (UrhG) apply.
Please note that individual components of the publication may be subject to other licensing or copyright conditions.