Title :
Single cluster PHD SLAM: Application to autonomous underwater vehicles using stereo vision
Author :
Nagappa, Sharad ; Palomeras, Narcis ; Chee Sing Lee ; Gracias, N. ; Clark, Daniel E. ; Salvi, Joaquim
Author_Institution :
Comput. Vision & Robot. Group, Univ. of Girona, Girona, Spain
Abstract :
This paper considers the application of feature-based simultaneous localisation and mapping (SLAM) using a random finite sets (RFS) framework for an autonomous underwater vehicle. SLAM allows for reduction in localisation error by tracking features which provide a fixed external reference. The SLAM problem is addressed here using a single-cluster probability hypothesis density (PHD) filter. The filter uses a particle approximation for the vehicle position with a conditional Gaussian mixture PHD for the feature map. Map features are selected as unique point features generated from a stereo camera on-board the vehicle. We demonstrate the improvement in localisation applying the algorithm to a dataset obtained in an indoor test tank.
Keywords :
Gaussian processes; SLAM (robots); approximation theory; autonomous underwater vehicles; cameras; feature extraction; mobile robots; position control; robot vision; set theory; statistical analysis; stereo image processing; PHD filter; RFS framework; autonomous underwater vehicles; conditional Gaussian mixture PHD; feature-based simultaneous localisation and mapping; localisation error reduction; map features; particle approximation; random finite sets framework; single cluster PHD SLAM; single-cluster probability hypothesis density filter; stereo camera; stereo vision; vehicle position; Cameras; Clutter; Equations; Feature extraction; Kalman filters; Simultaneous localization and mapping; Vehicles;
Conference_Titel :
OCEANS - Bergen, 2013 MTS/IEEE
Conference_Location :
Bergen
Print_ISBN :
978-1-4799-0000-8
DOI :
10.1109/OCEANS-Bergen.2013.6608107
