Sequential Monte Carlo is a family of methods that approximate evolving probability distributions using a set of weighted samples updated recursively as new observations arrive.
Semantic Classification
Content
- Sequential Monte Carlo represents a distribution by a population of samples, or particles, carrying importance weights, then propagates and reweights them over time, resampling to combat weight degeneracy. It is well suited to non-linear, non-Gaussian state estimation.
- The particle filter is its best-known instance, used for tracking and localisation in robotics and XR. It builds on Monte Carlo integration and importance sampling within a Bayesian inference framework.