A Box Regularized Particle Filter for terrain navigation with highly non-linear measurements * * The authors would like to thank the COGENT Computing lab (Coventry University) for their financial support.

摘要

This paper addresses the design of a new set-membership particle filter named Box Regularized Particle Filter (BRPF) applied to terrain navigation. This algorithm combines the set-membership particle estimation (known as Box Particle Filter) with the Kernel estimation method. This approach makes possible to enhance significantly the filter’s robustness while reducing the computation time (only 200 particles are needed instead of 5,000 with a conventional Sequential Importance Resampling (SIR) Particle Filter). Numerical results are presented from 10,000 Monte-Carlo runs.