Using particle filters for modeling landmarks´ uncertainties in Bearing-only SLAM

The recurrent algorithm for solving 2D Bearing-only SLAM problem is proposed. This algorithm is based on the Sequential Monte Carlo method and Rao-Blackwellisation technique, decomposing the state-vector into two parts which are the robot´s and the landmarks´ positions. The trajectories of the robot are modeled independently while the landmarks´ coordinates are modeled as conditional distributions. These conditional distributions are found using the individual particle filters corresponding to each generated trajectory of the robot. Our method has linear complexity growth with respect to number of the landmarks. The generalization of the proposed algorithm to 3D case is trivial.