## Robust range-only beacon localization (2004)

### Cached

### Download Links

Venue: | In Proceedings of Autonomous Underwater Vehicles |

Citations: | 43 - 8 self |

### BibTeX

@INPROCEEDINGS{Olson04robustrange-only,

author = {Edwin Olson and John Leonard and Seth Teller},

title = {Robust range-only beacon localization},

booktitle = {In Proceedings of Autonomous Underwater Vehicles},

year = {2004},

pages = {66--75}

}

### Years of Citing Articles

### OpenURL

### Abstract

Abstract — Most Autonomous Underwater Vehicle (AUV) systems rely on prior knowledge of beacon locations for localization. We present a system capable of navigating without prior beacon locations. Noise and outliers are major issues; we present a powerful outlier rejection method that imposes geometric constraints on measurements. We have successfully applied our algorithm to real-world data and have demonstrated navigation performance comparable to that of systems that assume known beacon locations. I.

### Citations

2756 | Normalized cuts and image segmentation
- Shi, Malik
- 2000
(Show Context)
Citation Context ...thus presumably not caused by noise), and using these intervals to help interpret noisier intervals [1]. Our outlier rejection method uses a form of spectral graph partitioning. Shi’s Normalized Cuts =-=[3]-=- and Ding’s MinMaxCuts [4] are typical approaches. These approaches attempt to find balanced clusters, but the problem of 2soutlier rejection is more akin to finding one cluster amidst noise. In this ... |

489 | A new extension of the kalman filter to nonlinear systems
- Julier, Uhlmann
- 1997
(Show Context)
Citation Context ... resolve ambiguous data. In the future, we plan to deploy our exploration ideas on a real vehicle. We are also interested in replacing the Extended Kalman Filter (EKF) with an Unscented Kalman filter =-=[10]-=- or particle filter which should be less vulnerable to the effects of linearization. ACKNOWLEDGEMENTS We would like to thank the MIT GOATS 2002 team, particularly Paul Newman, for providing us with th... |

434 |
Algebraic connectivity of graphs
- Fiedler
- 1973
(Show Context)
Citation Context ...spectral graph partitioning, which has been used in a number classification systems ([3], [4]). Much of the mathematical foundation for spectral analysis of adjacency matrices was established in [5], =-=[6]-=-, [7]. Most systems incorporate some type of outlier rejection strategy. When priors on beacon locations are available, extremely unlikely measurements can be discarded. Other methods include searchin... |

210 |
Linear Algebra
- Strang
- 1988
(Show Context)
Citation Context ... eigenvector. For the toy problem in Fig. 5, the eigenvector u is plotted in Fig. 6. The indicator values for the inlier measurements (1-5) are significantly larger than those for the outliers (6-8). =-=[8]-=-. The metric r(u) is also the Rayleigh quotient of matrix A, whose extrema values have previously been known Smaller eigenvalues of A correspond to alternative cuts with lower scores. Since A is symme... |

145 | Robust mapping and localization in indoor environments using sonar data
- Tardós, Neira, et al.
(Show Context)
Citation Context ...here [5]. Our method for initializing feature locations is reminiscent of Hough transforms. Hough transforms have previously been used to classify sonar returns from point and line features by Tardós =-=[6]-=-. Hough-style voting was also used by Wijk [7]. III. NOISE IN LBL DATA Most navigation systems use a Kalman filter for state estimation. Kalman filters produce optimal estimates when measurement noise... |

144 |
Lower bounds for the partitioning of graphs
- Donath, Hoffman
- 1973
(Show Context)
Citation Context ...m of spectral graph partitioning, which has been used in a number classification systems ([3], [4]). Much of the mathematical foundation for spectral analysis of adjacency matrices was established in =-=[5]-=-, [6], [7]. Most systems incorporate some type of outlier rejection strategy. When priors on beacon locations are available, extremely unlikely measurements can be discarded. Other methods include sea... |

100 |
Machine analysis of bubble chamber pictures
- Hough
- 1959
(Show Context)
Citation Context ... is not traveling in a straight line, some solutions will occur more often than others. We use a voting scheme implemented with a two-dimensional accumulator similar to that used in a Hough transform =-=[9]-=-. The physical world is discretized into a two-dimensional grid, with each grid cell corresponding to a rectangular area in the world. Each consistent measurement pair “votes” for its two solutions. I... |

45 | Preliminary Results in Range-only Localisation and Mapping
- Kantor, Singh
(Show Context)
Citation Context ...he algorithm is prone to divergence due to the search’s random initialization. Consequently, the approach did not robustly deal with ambiguities, such as those arising from baseline crossings. Kantor =-=[2]-=- addresses the problem in a terrestrial system, briefly discussing the case of unknown beacon locations. However, he assumes that the beacon locations are approximately known, and does not discuss the... |

43 | Pure range-only sub-sea SLAM
- Newman, Leonard
(Show Context)
Citation Context ...nd Leonard approached range-only SLAM by casting the problem as a nonlinear optimization over a search space including not just the beacon locations, but also the AUV’s position at each point in time =-=[1]-=-. As the authors acknowledge, the algorithm is prone to divergence due to the search’s random initialization. Consequently, the approach did not robustly deal with ambiguities, such as those arising f... |

36 |
I.Christensen, “Triangulation based fusion of sonar data with application in robot pose tracking
- Wijk, Henrik
(Show Context)
Citation Context ...locations is reminiscent of Hough transforms. Hough transforms have previously been used to classify sonar returns from point and line features by Tardós [6]. Hough-style voting was also used by Wijk =-=[7]-=-. III. NOISE IN LBL DATA Most navigation systems use a Kalman filter for state estimation. Kalman filters produce optimal estimates when measurement noise is Gaussian and stationary, but their perform... |

17 |
Single Cluster Graph Partitioning for Robotics Applications
- Olson, Walter, et al.
- 2005
(Show Context)
Citation Context ... a consistency-based clustering technique that is well-suited to outlier rejection. We developed SCGP to support a number of other clustering applications; the method is described in detail elsewhere =-=[5]-=-. Our method for initializing feature locations is reminiscent of Hough transforms. Hough transforms have previously been used to classify sonar returns from point and line features by Tardós [6]. Hou... |

13 |
A MinMaxCut spectral method for data clustering and graph partitioning
- Ding, He, et al.
- 2001
(Show Context)
Citation Context ... by noise), and using these intervals to help interpret noisier intervals [1]. Our outlier rejection method uses a form of spectral graph partitioning. Shi’s Normalized Cuts [3] and Ding’s MinMaxCuts =-=[4]-=- are typical approaches. These approaches attempt to find balanced clusters, but the problem of 2soutlier rejection is more akin to finding one cluster amidst noise. In this paper, we make use of Sing... |

6 |
property of eigenvectors of nonnegative symmetric matrices and its application to graph theory,” Czechoslovak
- “A
- 1975
(Show Context)
Citation Context ...ral graph partitioning, which has been used in a number classification systems ([3], [4]). Much of the mathematical foundation for spectral analysis of adjacency matrices was established in [5], [6], =-=[7]-=-. Most systems incorporate some type of outlier rejection strategy. When priors on beacon locations are available, extremely unlikely measurements can be discarded. Other methods include searching for... |