PR-SLAM in particle filter framework

Simultaneous localization and mapping is an important task for autonomous mobile robot. To let the robot explore a new environment without any prior map, real-time estimation of the geometrical relation between the robot and the environment is necessary. Extended Kalman filter (EKF)-based approaches are the most common. However, they always have the risk of collapse where the assumption of Gaussian distribution is not applicable. It is well known that state estimation with a particle filter is very robust against clutter in dynamic and noisy environments because of its ability to represent non-Gaussian distributions. Unfortunately, particle-based posterior representation in high dimensions is extremely expensive. We propose an approach, named partitioned recursive SLAM, that overcomes the complexity problem arising in adopting a particle filter in SLAM. By partitioning the state and alternating the turns for the state update, the computational capacity required to process SLAM is reduced to scale linearly with the number of landmarks in the map.