particle filters
Particle filters, also known as sequential Monte Carlo methods, are a type of recursive Bayesian filter used for state estimation in dynamic systems. They are based on the idea of representing the state using a set of particles, where each particle represents a possible state hypothesis. These particles are propagated forward in time by sampling from a proposal distribution and then weighted according to how well they match the available measurements. The weights are then used to resample particles, favoring those with higher weights, in order to focus the estimation on regions of high likelihood. By iteratively repeating this process, particle filters provide an approximation of the posterior distribution of the system state given the measurements. They are particularly useful in scenarios where the underlying system and measurement models are nonlinear or non-Gaussian.
Requires login.