Localization in unknown environments is one of the most important challenges in mobile robotics, especially when GPS signals are weak or inaccessible. In these circumstances, robots can use Simultaneous Localization and Mapping (SLAM) techniques for localization. However, its implementation is often impractical in environments with huge number of landmarks. This problem is caused by high computational complexity of SLAM. In addition, the non-linearity of motion and observation models in common SLAM methods cause that the SLAM algorithms diverge. The current proposed approaches introduced so far only postpone the divergence of SLAM at the expense of increasing the computational costs. In this thesis, we first propose novel motion and observation models, which are linear with respect to state vector. The calculation volume of the proposed method is equal to the conventional SLAM based on Extended Kalman Filter. However, in this approach the accuracy of the landmarks and robot positioning is increased and also the convergence of SLAM is ensured. Next in the thesis, we have focused on relative map filter (RMF). The RMF approach considers the relative distances between the landmarks which do not depend on the robot positions. Also its observation model is linear, so it is a convergent method. However, RMF methods have two main challenges, data association and disambiguation of robot and landmarks positions. Here, novel approaches are presented for RMF which overcome these problems. In this method with proposing a data association algorithm based on successive switching between absolute and relative space, not only the problem of relative data association is resolved, but relative and absolute maps are created concurrently as well. In addition, to solve disambiguation of robot and landmarks positions, two different methods are presented. Lowest Position Error (LPE) and Minimum Variance Position Estimator (MVPE) are disambiguation procedures which have very low computational complexity. Their amounts of calculations are related to average number of landmarks observed in each robot scan. In this thesis, equations and algorithms to achieve the position of the robot and landmarks are presented and the empirical studies on the proposed methods show a better accuracy and more stability in comparison with the extended Kalman filter SLAM and one of the newest SLAM method.