Monocular vSLAM using a novel Rao-Blackwellized particle filter

This paper presents the theoretical framework and experimental results of a generalized formulation of a Rao-Blackwellized particle filter (RBPF) in which higher order state variables and a modified undelayed initialization scheme are incorporated to solve the 3D monocular visual SLAM problem (vSLAM). As an example of the proposed formulation, velocity has been included in the state variables so that filtering progresses based on sampling from velocity distribution, not the displacement. The new sampling posterior has been obtained with respect to observations, control inputs and the robot path. The proper importance weight for resampling has been derived in this paper. To solve the bearing-only problem, the proposed approach features a modified initialization scheme that uses an inverse depth of the landmarks. The results of the offline experiment indicate the feasibility of the proposed approach.