An Efficient Two-Stage Sampling Method in Particle Filter

Qi Cheng , Pascal Bondon (1)

1 L2S - Laboratoire des signaux et systèmes

We present a modified bootstrap filter (MBF) to draw particles in the particle filter (PF). The proposal distribution for each particle involves sampling from the state-space model a number of times, and then selecting the sample with the highest measurement likelihood. Numerical examples show that this filter outperforms the bootstrap filter (BF) with the same computational complexity when the state noise has a large variance.

Qi Cheng, Pascal Bondon. An Efficient Two-Stage Sampling Method in Particle Filter. IEEE Transactions on Aerospace and Electronic Systems, 2012, 48 (3), pp.2666-2672. ⟨10.1109/TAES.2012.6237616⟩. ⟨hal-00819784⟩