Multi-level Monte Carlo Simulation on Particle Filters and Extended Kalman Filters.

Abstract

the automation of vehicles depends on different observations that may represent the states in an inaccurate way. The probabilistic approach in the prediction and enhancement of measurement are already used in automation to model the noise and perceptual limitations in vehicles sensors. In this paper, we work on the statistical solutions of Extended Kalman filter and Particle filters that include noise in the prediction of the new state. We use different variance reduction technique as control variates to increase the precision of the simulation and make it statistically efficient.