RTS Smoother for GLMB filter

Abstract

In this paper, we implement a low-cost but effective smoothing strategy to smooth estimated tracks returned by the GLMB filter. While the forward filtering step is carried out via the GLMB filtering procedure, the backward smoothing step is recursively implemented from the final time step to the first time step via a smoothing algorithm. In particular, the smoothing algorithm is based on the Rauch-Tung-Striebel (RTS) of fixed-interval smoother. We demonstrate our smoothing strategy on a linear Gaussian model and the experimental results show consistent improved tracking performance over 100 Monte Carlo runs.