Vandi Verma, SebastianThrun, Reid Simmons, "Variable Resolution Particle Filter", International Joint Conference on Artificial Intelligence,
     August 2003.


 Particle filters are used extensively for tracking the state of
  non-linear dynamic systems. This paper presents a new particle
  filter that maintains samples in the state space at dynamically
  varying resolution for computational efficiency. Resolution within
  statespace varies by region, depending on the belief that the true
  state lies within each region. Where belief is strong, resolution is
  fine. Where belief is low, resolution is coarse, abstracting
  multiple similar states together. The resolution of the statespace
  is dynamically updated as the belief changes. The proposed algorithm
  makes an explicit bias-variance tradeoff to select between
  maintaining samples in a biased generalization of a region of state
  space versus in a high variance specialization at fine resolution.
  Samples are maintained at a coarser resolution when the bias
  introduced by the generalization to a coarse resolution is
  outweighed by the gain in terms of reduction in variance, and at a
  finer resolution when it is not. Maintaining samples in abstraction
  prevents potential hypotheses from being eliminated prematurely for
  lack of a sufficient number of particles.  Empirical results show
  that our variable resolution particle filter requires significantly
  lower computation for performance comparable to a classical particle