A SLAM system based on improved distributed UPF

SLAM based on distributed particle filter diminishes the calculation quantity with an acceptable result. However, the accuracy of the distributed particle filter is not ideal and the distribution of the particles is not adjusted with the observations. To tackle with this problem, this paper used distributed Unscented Particle Filter (UPF) arithmetic to estimate the states of SLAM. The distributed UPF was improved in two aspects to solve the problem when it is used in SLAM. First, the improved UPF arithmetic utilized the result of UPF as input information to adjust affect of the observation. Second, the particle distribution in each step of UPF is optimized by generate the initial particles in the local filter access to the land marks which have not been matched before (we call it the new adding local filters). The results of simulation proved the stability and the accuracy of the improved method.