Particle Filtering
Overview
Particle filtering uses weighted samples to approximate a changing probability distribution over hidden states, supporting Bayesian estimation when Gaussian assumptions are inadequate. Methods improve the proposal distributions that place particles, manage weight collapse through resampling, and extend filtering with smoothing, learned proposals, or differentiable updates. Applications include robot and maritime tracking, tactile pose estimation, and data assimilation, with evaluation on nonlinear state-space systems and real sensor data. A central challenge is balancing posterior accuracy and coverage against computational cost, especially in high-dimensional settings where particle degeneracy remains difficult to avoid.
Latest papers
28 papers4 papers in the last four weeks, level with the four weeks before. 0.0% of all new papers.