Novel particle filtering algorithms for fixed parameter estimation in dynamic systems

Standard particle filters cannot handle dynamic systems with unknown fixed parameters. In this paper, we extend the recently proposed cost-reference particle filtering methodology (CRPF) to jointly estimate the time-varying state and the static parameters of a dynamic system. In particular, we introduce three strategies that allow assigning costs to the random samples in the state-space independently of the fixed parameters. Asymptotic results that illuminate the relationships among the methods are derived, and computer simulation results are presented to illustrate their practical implementation in a vehicle navigation problem.