Particle filters and high-dimensional systems

Abstract:

Particle filters offer an elegant solution to the problem of state estimation. They make no assumptions about the form of the underlying probability distributions and, in principle, are applicable in the presence of strong nonlinearity and non-Gaussianity. Driven in part by geophysical applications, much recent work has focussed on particle-filter algorithms for high-dimensional systems. I will review reasons that high-dimensional system are especially challenging for particle filters, give some bounds on the performance of certain classes of particle filters and discuss some potential paths toward more effective high-dimensional particle filters.