## An Atlas Framework for Scalable Mapping (2003)

### Cached

### Download Links

- [graphics.lcs.mit.edu]
- [cogvis.nada.kth.se]
- [groups.csail.mit.edu]
- [graphics.lcs.mit.edu]
- [graphics.lcs.mit.edu]
- DBLP

### Other Repositories/Bibliography

Venue: | in IEEE International Conference on Robotics and Automation |

Citations: | 152 - 18 self |

### BibTeX

@INPROCEEDINGS{Bosse03anatlas,

author = {Michael Bosse and Paul Newman and John Leonard and Martin Soika and Wendelin Feiten and Seth Teller},

title = {An Atlas Framework for Scalable Mapping},

booktitle = {in IEEE International Conference on Robotics and Automation},

year = {2003},

pages = {1899--1906}

}

### Years of Citing Articles

### OpenURL

### Abstract

This paper describes Atlas, a hybrid metrical /topological approach to SLAM that achieves efficient mapping of large-scale environments. The representation is a graph of coordinate frames, with each vertex in the graph representing a local frame, and each edge representing the transformation between adjacent frames. In each frame, we build a map that captures the local environment and the current robot pose along with the uncertainties of each. Each map's uncertainties are modeled with respect to its own frame. Probabilities of entities with respect to arbitrary frames are generated by following a path formed by the edges between adjacent frames, computed via Dijkstra's shortest path algorithm. Loop closing is achieved via an efficient map matching algorithm. We demonstrate the technique running in real-time in a large indoor structured environment (2.2 km path length) with multiple nested loops using laser or ultrasonic ranging sensors.

### Citations

2670 |
Random sample consensus: A paradigm for model fitting with applications to image analysis and automated cartography
- Fischler, Bolles
- 1981
(Show Context)
Citation Context ...ing data from the same experiment but using the Polaroid ultrasonic rangers instead of the laser scanner. The local navigation method used is a combination of Delayed Decision Making [LRNB02], RANSAC =-=[FB80]-=- and wide beam sonar interpretation [LDW92]. The key idea of this approach is to incorporate temporal as well as spatial correlations in the stochastic mapping process. This enables map features to be... |

1593 | A note on two problems in connexion with graphs
- Dijkstra
- 1959
(Show Context)
Citation Context ... map-frame with respect to another arbitrary mapframe by following a path formed by the edges between adjacent map-frames. These paths are computed using either (1) Dijsktra’s shortest path algorith=-=m [3], -=-or (2) breadth-first search (BFS). When Dijkstra’s shortest path algorithm is used, the uncertainties of the transformations of the edges of the graph serve as a statistical distance metric, with a ... |

1330 | Three-Dimensional Computer Vision - A Geometric Viewpoint - Faugeras - 1999 |

771 | Tracking and data association - Bar-Shalom, Fortmann - 1988 |

667 | Monte carlo localization for mobile robots - Dellaert, Fox, et al. - 1999 |

475 | FastSLAM: A factored solution to simultaneous localization and mapping
- Montemerlo, Thrun, et al.
- 2002
(Show Context)
Citation Context ...tion and mapping (SLAM) problem. A variety of approaches have been proposed for representing the uncertainty inherent to sensor data and robot motion, including topological [9], particle filter [16], =-=[12]-=-, and feature-based [14] models. Several highly successful SLAM approaches have been developed based on the combination of laser scan matching with Bayesian state estimation [7], [16]. These methods, ... |

413 |
Estimating uncertain spatial relationships in robotics
- Smith, Self, et al.
- 1988
(Show Context)
Citation Context ...problem. A variety of approaches have been proposed for representing the uncertainty inherent to sensor data and robot motion, including topological [9], particle filter [16], [12], and feature-based =-=[14]-=- models. Several highly successful SLAM approaches have been developed based on the combination of laser scan matching with Bayesian state estimation [7], [16]. These methods, however, incur computati... |

385 | A Solution to the simultaneous localization and map building (SLAM) problem
- Dissanayake, Newman, et al.
- 2001
(Show Context)
Citation Context ...tion, linear motion and measurement models, and Guassian error models [14]. The convergence and scaling properties of the Kalman filter solution to the linear Gaussian SLAM problem are now well-known =-=[4]-=-. Considerable recent research effort has been extended toward mitigation of the O(n2 ) complexity (where n is the number of features) of the Kalman filter SLAM solution. Efficient strategies for SLAM... |

304 | Incremental mapping of large cyclic environments
- Gutmann, Konolige
- 2000
(Show Context)
Citation Context ...particle filter [16], [12], and feature-based [14] models. Several highly successful SLAM approaches have been developed based on the combination of laser scan matching with Bayesian state estimation =-=[7]-=-, [16]. These methods, however, incur computational difficulties that make real-time performance impossible in closing large loops. The Kalman filter provides the optimal linear recursive solution to ... |

274 | The Spatial Semantic Hierarchy
- Kuipers
- 2000
(Show Context)
Citation Context ...scale simultaneous localization and mapping (SLAM) problem. A variety of approaches have been proposed for representing the uncertainty inherent to sensor data and robot motion, including topological =-=[9]-=-, particle filter [16], [12], and feature-based [14] models. Several highly successful SLAM approaches have been developed based on the combination of laser scan matching with Bayesian state estimatio... |

258 | Directed Sonar Sensing for Mobile Robot Navigation
- Leonard, Durrant-Whyte
- 1992
(Show Context)
Citation Context ...g the Polaroid ultrasonic rangers instead of the laser scanner. The local navigation method used is a combination of Delayed Decision Making [LRNB02], RANSAC [FB80] and wide beam sonar interpretation =-=[LDW92]-=-. The key idea of this approach is to incorporate temporal as well as spatial correlations in the stochastic mapping process. This enables map features to be consistently initialized using data from m... |

206 | Real-time obstacle avoidance for fast mobile robots in cluttered environments, in - Borenstein, Koren - 1990 |

202 | A probabilistic online mapping algorithm for teams of mobile robots
- Thrun
- 2001
(Show Context)
Citation Context ...calization and mapping (SLAM) problem. A variety of approaches have been proposed for representing the uncertainty inherent to sensor data and robot motion, including topological [9], particle filter =-=[16]-=-, [12], and feature-based [14] models. Several highly successful SLAM approaches have been developed based on the combination of laser scan matching with Bayesian state estimation [7], [16]. These met... |

195 | Data association in stochastic mapping using the joint compatibility test,” Robotics and Automation
- Neira, Tardos
- 2001
(Show Context)
Citation Context ...uces the total computational burden to O(n 2 log n). The approach we have adopted for Map-Matching is not unique. For example the Joint Compatibility test with branch and bound technique suggested by =-=[13]-=- could also be utilized. The freedom to choose a Map-Matching strategy highlights the modularity of the Atlas framework. E. Traversal Once the robot has mapped an area, it can reuse previously built m... |

192 | Optimization of the simultaneous localization and map-building algorithm for real-time implementation
- Guivant, Nebot
- 2001
(Show Context)
Citation Context ...SLAM solution. Efficient strategies for SLAM with feature-based representations and Gaussian representation of error include postponement [2], decoupled stochastic mapping [10], the compressed filter =-=[6]-=-, sequential map joining [15], the constrained local submap filter [18], and sparse extended information filters [17]. Each of these methods employs a single, globally referenced coordinate frame for ... |

145 | Robust mapping and localization in indoor environments using sonar data
- Tardós, Neira, et al.
(Show Context)
Citation Context ...ategies for SLAM with feature-based representations and Gaussian representation of error include postponement [2], decoupled stochastic mapping [10], the compressed filter [6], sequential map joining =-=[15]-=-, the constrained local submap filter [18], and sparse extended information filters [17]. Each of these methods employs a single, globally referenced coordinate frame for state estimation. The Kalman ... |

82 | Automatic reconstruction of piecewise planar models from multiple views - BAILLARD, ZISSERMAN - 1999 |

73 | A Non-divergent estimation algorithm in the presence of unknown correlations
- Julier, Uhlmann
- 1997
(Show Context)
Citation Context ...m the refined estimate i T j+ + i , and its uncertainty Σ ij . Since we do not maintain the cross-covariances between robot estimates in different maps, we advocate the use of Covariance Intersection=-= [8] to perform the update. Where � Σ + ij = ω (�-=-�ij) −1 +(1−ω) � Σ −� � −1 −1 ij T j � + + i = Σ ij ωΣ −1 xi xi +(1−ω)Σ −1 xj xj � (13) (14) � � ω = argmin�+ Σ � ω ij . (15) If the uncertainty in local ma... |

65 | Mobile Robot Navigation Using Active Vision
- Davison
- 1999
(Show Context)
Citation Context ...xity (where n is the number of features) of the Kalman filter SLAM solution. Efficient strategies for SLAM with feature-based representations and Gaussian representation of error include postponement =-=[2]-=-, decoupled stochastic mapping [10], the compressed filter [6], sequential map joining [15], the constrained local submap filter [18], and sparse extended information filters [17]. Each of these metho... |

62 | Gool, "Automated reconstruction of 3D scenes from sequences of images - Pollefeys, Koch, et al. - 2000 |

56 | Calibrated, registered images of an extended urban area - Teller, Antone, et al. - 2001 |

54 | Decoupled stochastic mapping
- Leonard, Feder
- 2001
(Show Context)
Citation Context ...tures) of the Kalman filter SLAM solution. Efficient strategies for SLAM with feature-based representations and Gaussian representation of error include postponement [2], decoupled stochastic mapping =-=[10]-=-, the compressed filter [6], sequential map joining [15], the constrained local submap filter [18], and sparse extended information filters [17]. Each of these methods employs a single, globally refer... |

46 | An efficient approach to the simultaneous localisation and mapping problem
- Williams, Dissanayake, et al.
- 2002
(Show Context)
Citation Context ...entations and Gaussian representation of error include postponement [2], decoupled stochastic mapping [10], the compressed filter [6], sequential map joining [15], the constrained local submap filter =-=[18]-=-, and sparse extended information filters [17]. Each of these methods employs a single, globally referenced coordinate frame for state estimation. The Kalman filter can fail badly, however, in situati... |

41 | Mapping partially observable features from multiple uncertain vantage points
- Leonard, Rikoski, et al.
- 2002
(Show Context)
Citation Context ...ent began. Figure 7(b) shows results using data from the same experiment but using the Polaroid ultrasonic rangers instead of the laser scanner. The local navigation method used is described fully in =-=[11]-=-. Additional results, including concurrent processing of both laser 140 120 12 45 11 13 14 20 46 44 15 21 10 43 16 22 19 1723 31 30 29 28 24 18 2532 26 27 33 100 9 842 47 48 34 80 60 41 7 95 6 540 49 ... |

36 | I.Christensen, “Triangulation based fusion of sonar data with application in robot pose tracking - Wijk, Henrik |

36 |
A Bayesian algorithm for simultaneous localization and map building. Unpublished manuscript
- Durrant-Whyte, Majumder, et al.
- 2001
(Show Context)
Citation Context ...vide global results. An alternative to the use of local linearization would be to adopt a fully nonlinear formulation of the SLAM problem, such as FastSLAM [12] or SLAM using a sum of Gaussians model =-=[5]-=-. The computational requirements of these methods, however, remain poorly understood in large cyclic environments. In future research, it may be possible to implement Atlas using one of these techniqu... |

23 | Approches to mobile robot localization in indoor environment - Jensfelt - 2001 |

20 |
Large scale sonarray mapping using multiple connected local maps
- Chong, Kleeman
- 1997
(Show Context)
Citation Context ... shown below in Figure 6(b) in Section V provides a dramatic illustration of this type of situation. One of the appealing aspects of a hybrid metrical/topological approach to mapping and localization =-=[1]-=-, [9] is that uncertain state estimates do not need to be referenced to a single global reference frame. This is the strategy advocated in this paper. With Atlas, we obtain the best of both global and... |

20 | Localisation in large-scale environments - Bailey, Nebot |

16 | A Bayesian method for certainty grids - Moravec, Cho - 1989 |

5 |
Simultaneous mapping and localization with sparse extended information filters
- Y
- 2002
(Show Context)
Citation Context ... include postponement [2], decoupled stochastic mapping [10], the compressed filter [6], sequential map joining [15], the constrained local submap filter [18], and sparse extended information filters =-=[17]-=-. Each of these methods employs a single, globally referenced coordinate frame for state estimation. The Kalman filter can fail badly, however, in situations with large angular errors and significant ... |

1 |
A Bayesian algorithm for simultaneous TABLE
- Durrant-Whyte, Majumder, et al.
- 2001
(Show Context)
Citation Context ...vide global results. An alternative to the use of local linearization would be to adopt a fully nonlinear formulation of the SLAM problem, such as FastSLAM [12] or SLAM using a sum of Gaussians model =-=[5]-=-. The computational requirements of these methods, however, remain poorly understood in large cyclic environments. In future research, it may be possible to implement Atlas using one of these techniqu... |

1 | T j i � = ⎡ ⎣ = ⊖T j ⎤ ⎦ −ci −si sixi − ciyi si −ci cixi + siyi 0 - J⊖ |

1 | Feature based exploration. Submitted for publication for ICRA - Newman, Bosse, et al. - 2003 |

1 |
Subsea range only SLAM. Submitted for publication for ICRA
- Newman
- 2003
(Show Context)
Citation Context ... enough to prevent the recognition of an already mapped region — i.e., loop closing is hard [Thr01]. • We expect to encounter featureless regions in which navigation must rely on dead reckoning al=-=one [New02]-=-. Building a single monolithic map results in an un-mangaged growth of complexity and computational burden. This in combination with the fact that spatially distant features can be decoupled has led s... |