Implicit Sampling for Path Integral Control, Monte Carlo Localization, and SLAM

Abstract

Implicit sampling is a recently developed variationally enhanced sampling method that guides its samples to regions of high probability, so that each sample carries information. Implicit sampling may thus improve the performance of algorithms that rely on Monte Carlo (MC) methods. Here the applicability and usefulness of implicit sampling for improving the performance of MC methods in estimation and control is explored, and implicit sampling based algorithms for stochastic optimal control, stochastic localization, and simultaneous localization and mapping (SLAM) are presented. The algorithms are tested in numerical experiments where it is found that fewer samples are required if implicit sampling is used, and that the overall runtimes of the algorithms are reduced.