Large scale mapping using submaps

The main contribution of this article the implementation of a new method for the usage of submaping algorithms, which surpasses the limitations arising from the need of statistical independence between the submaps, when mapping large scale areas. This method relies on the commonly used structure of SLAM (Simultaneous Localization and Mapping) mandatory for building submaps which share information, but stay conditionally independent. This approach is able to recover the final map, without introducing other approximations ignoring the ones already introduced by the EKF (Extended Kalman Filter), all this in a linear time. The algorithm has been tested for the case of local submaps because the effects of linearization errors are smaller in this way. The obtained maps are more accurate and consistent compared with ones obtained by EKF or EIF (Extended Information Filter) methods, which are not using local coordinates.