Position estimation with Bayesian filters by using 3-dimensional environment models

Abstract

This paper presents a method for including spatial environment information into a Particle Filter for position estimation. The proposed method is targeted for indoor and outdoor scenarios where distance measurements to static nodes of known position are basis for the localization. Such scenarios are likely in industrial and logistic applications where maps or 3-dimensional model data of the relevant playground are available. Those environmental information supplement the noisy measurements of positioning systems and could be directly evaluated by the position estimator. Two approaches, Axis Aligned Bounding Boxes (AABB) and point cloud have been evaluated in combination with a Particle Filter estimator in this work on the base of a high level simulation environment. Different use cases with varying motion trails have been simulated. The results show that including spatial environment data reduces the position error and thus positively influences the estimation quality. With this knowledge a previously published hardware implementation of a Particle Filter has been enhanced by spatial information analysis on register transfer level using an efficient pipeline structure. The resulting implementation maps on a ZYNQ 7000 SoC hardware/software platform that provides an accelerated low power solution for the position estimation.