Sample-efficiency-optimized auxiliary particle filter

We present a sample-efficiency-optimized auxiliary particle filter. We show that by computing the weights of the auxiliary variable through a simple optimization procedure, the variance of the resulting importance weights can be considerably reduced, in fact, minimized. In turn, the effective sample size and therefore the efficiency of the overall sampling both increase. Experiments have been extensively conducted using the bearings-only model. The experimental results fully support the theoretical conclusions that the proposed sample-efficiency-optimized auxiliary particle filter outperforms the nonoptimized auxiliary particle filtering, producing smaller tracking errors and larger effective sample sizes