Auxiliary particle filter robot localization from high-dimensional sensor observations

Abstract

We apply the auxiliary particle filter algorithm of Pitt and Shephard (1999) to the problem of robot localization. To deal with the high-dimensional sensor observations (images) and an unknown observation model. we propose the use of an inverted nonparametric observation model computed by nearest neighbor conditional density estimation. We show that the proposed model can lead to a fully adapted optimal filter, and is able to successfully handle image occlusion and robot kidnap. The proposed algorithm is very simple to implement and exhibits a high degree of robustness in practice. We report experiments involving robot localization from omnidirectional vision in an indoor environment.

Probabilistic cooperative localization and mapping in practice

“…One approach that has gained popularity of late falls under the category of Monte Carlo Simulation (see [4] for an overview) and is known under different names in different fields. The technique we use was introduced as particle filtering by Gordon et al [5]; in mobile robotics, particle filtering has been applied successfully by different groups [3,7,15]. In vision this technique was introduced under the name of condensation [6] and particle filtering [1].…”

Spatial-temporal Collaborative Sequential Monte Carlo for Mobile Robot Localization in Distributed Intelligent Environments

“…Following this framework, an implementation algorithm using Augmented Particle according to the standard prediction and update processes in the PF scheme. Since the sampling/importance Resampling (SIR) [17] method is known not so robust to outliers for two different reasons: sampling efficiency and the unreliability of the empirical prediction density in the tails of the distribution, we apply Auxiliary Particle Filters (APFs) [18] [19] in the augmented sampling step, which can be coped with by a sensor node with reasonable processing power and memory. Since there are no data dependencies during the particle generation and evaluation, these steps can be easily parallelized and pipelined, maintaining N particles in K PUs, each carries k N particles.…”

Local optimization of cooperative robot movements for guiding and regrouping people in a guiding mission

“…The DTM predicts people and robot movements and gives the motion instructions to robots. DTM uses a Particle Filter formulation [1], [18], [25], [27], with the particularity that it incorporates realistic human motion models. The model assumes that obstacles, people and robots are modeled by potential functions.…”