## Towards constant-time SLAM on an autonomous underwater vehicle using synthetic aperture sonar (2003)

### Cached

### Download Links

- [cogvis.nada.kth.se]
- [oe.mit.edu]
- [cml.mit.edu]
- [albacore.mit.edu]
- [groups.csail.mit.edu]
- DBLP

### Other Repositories/Bibliography

Venue: | In Eleventh International Symposium of Robotics Research |

Citations: | 25 - 4 self |

### BibTeX

@INPROCEEDINGS{Newman03towardsconstant-time,

author = {Paul M. Newman and John J. Leonard and Richard J. Rikoski},

title = {Towards constant-time SLAM on an autonomous underwater vehicle using synthetic aperture sonar},

booktitle = {In Eleventh International Symposium of Robotics Research},

year = {2003},

pages = {409--420}

}

### Years of Citing Articles

### OpenURL

### Abstract

This paper applies a new constant-time, consistent and convergent Simultaneous Localization and Mapping (SLAM) algorithm to an autonomous underwater vehicle (AUV). A constant-time SLAM algorithm offers computation independent of workspace size and is one key component in the development of truly autonomous agents. The real-time deployment of such a system would be a landmark achievement for the mobile robotics community. This paper describes progress towards this goal focusing on the sub-sea domain — an area set to benefit massively from the autonomy afforded by SLAM. The primary sensor used in this work is a sixteen element synthetic aperture sonar (SAS) carried on the nose of the AUV “Caribou”. Using a novel target detection strategy, data gathered from a 40 minute survey is processed by the new SLAM algorithm and the results compared to both a ground truth and the quadratic time “gold standard ” full covariance SLAM algorithm. 1

### Citations

3158 |
A.: “Multiple View Geometry in Computer Vision
- Hartley, Zisserman
- 2004
(Show Context)
Citation Context ...of the range-only data) but also the problem is inherently ill conditioned over small vehicle path lengths. The range only SLAM problem has much in common with the structure from motion (SFM) problem =-=[1, 13, 9, 22, 21]-=- which in turn has a strong duality with bearing only and range only SLAM [4, 5]. The motivation for this approach is clear. Submerged, on-the-fly calibration of transponders would enable an AUV to la... |

448 | FastSLAM: A Factored Solution to the Simultaneous Localization
- Montemerlo, Thrun, et al.
- 2002
(Show Context)
Citation Context ...o prior information and using only onboard sensors, to move through its environment and build a consistent map of its surroundings as well as an estimate of its own trajectory. SLAM is a hard problem =-=[7, 23, 9, 22]-=- and may not always be the best technique to employ to achieve mobile autonomy. For example, in many land based applications navigation and mapping can be aided by in-place infrastructure - GPS, radio... |

380 |
Estimating Uncertain Spatial Relationships in Robotics," Appeared
- Smith, Self, et al.
- 1986
(Show Context)
Citation Context ...on 5 discusses areas of future and intended research that lie on the path to developing a truly autonomous, SLAM-enabled AUV. 2 Prior work and background The seminal work of Smith, Self and Cheeseman =-=[27]-=- proposed an algorithm (commonly known as the “Full Covariance solution”) with complexity O(n 2 ) where n is the number of features mapped. The Full Covariance solution is accepted as the “Gold Standa... |

347 | A solution to the simultaneous localization and map building (slam) problem
- Gamini, Newman, et al.
- 2001
(Show Context)
Citation Context ... filter provides the optimal estimate of this pdf, which is described by its mean [ˆxv(k) T ˆx1(k) T ... ˆxn(k) T ] T and covariance P(k). The properties of single-map LG SLAM solution are well-known =-=[6]-=-. Recent related work in SLAM includes submap decomposition methods [19, 10, 14, 32, 28], FastSLAM [23], sparse extended information filters (SEIF’s) [30], scan-matching [11, 29, 12] and topological a... |

281 | Incremental mapping of large cyclic environments
- Gutmann, Konolige
- 1999
(Show Context)
Citation Context ...AM solution are well-known [6]. Recent related work in SLAM includes submap decomposition methods [19, 10, 14, 32, 28], FastSLAM [23], sparse extended information filters (SEIF’s) [30], scan-matching =-=[11, 29, 12]-=- and topological approaches [16, 2]. To date, all published SLAM algorithms are subject to the scaling problem — the fact that the SLAM task gets harder as more and more features are mapped. The unbou... |

190 | An probabilistic online mapping algorithm for teams of mobile robots,” Int
- Thrun
- 2001
(Show Context)
Citation Context ...AM solution are well-known [6]. Recent related work in SLAM includes submap decomposition methods [19, 10, 14, 32, 28], FastSLAM [23], sparse extended information filters (SEIF’s) [30], scan-matching =-=[11, 29, 12]-=- and topological approaches [16, 2]. To date, all published SLAM algorithms are subject to the scaling problem — the fact that the SLAM task gets harder as more and more features are mapped. The unbou... |

185 | Topological simultaneous localization and mapping (SLAM): Toward exact localization without explicit localization
- Choset, Nagatani
- 2001
(Show Context)
Citation Context ...elated work in SLAM includes submap decomposition methods [19, 10, 14, 32, 28], FastSLAM [23], sparse extended information filters (SEIF’s) [30], scan-matching [11, 29, 12] and topological approaches =-=[16, 2]-=-. To date, all published SLAM algorithms are subject to the scaling problem — the fact that the SLAM task gets harder as more and more features are mapped. The unbounded growth of computation with map... |

183 | Optimization of the simultaneous localization and map building algorithm for real time implementation, in
- Guivant, Nebot
- 2001
(Show Context)
Citation Context ...y its mean [ˆxv(k) T ˆx1(k) T ... ˆxn(k) T ] T and covariance P(k). The properties of single-map LG SLAM solution are well-known [6]. Recent related work in SLAM includes submap decomposition methods =-=[19, 10, 14, 32, 28]-=-, FastSLAM [23], sparse extended information filters (SEIF’s) [30], scan-matching [11, 29, 12] and topological approaches [16, 2]. To date, all published SLAM algorithms are subject to the scaling pro... |

144 | A computationally efficient method for large-scale concurrent mapping and localization
- Leonard, Feder
- 1999
(Show Context)
Citation Context ...y its mean [ˆxv(k) T ˆx1(k) T ... ˆxn(k) T ] T and covariance P(k). The properties of single-map LG SLAM solution are well-known [6]. Recent related work in SLAM includes submap decomposition methods =-=[19, 10, 14, 32, 28]-=-, FastSLAM [23], sparse extended information filters (SEIF’s) [30], scan-matching [11, 29, 12] and topological approaches [16, 2]. To date, all published SLAM algorithms are subject to the scaling pro... |

138 | Robust mapping and localization in indoor environments using sonar data, in: Int
- Tardos, Neira, et al.
(Show Context)
Citation Context ...y its mean [ˆxv(k) T ˆx1(k) T ... ˆxn(k) T ] T and covariance P(k). The properties of single-map LG SLAM solution are well-known [6]. Recent related work in SLAM includes submap decomposition methods =-=[19, 10, 14, 32, 28]-=-, FastSLAM [23], sparse extended information filters (SEIF’s) [30], scan-matching [11, 29, 12] and topological approaches [16, 2]. To date, all published SLAM algorithms are subject to the scaling pro... |

87 | Bootstrap learning for place recognition
- Kuipers, Beeson
- 2002
(Show Context)
Citation Context ...elated work in SLAM includes submap decomposition methods [19, 10, 14, 32, 28], FastSLAM [23], sparse extended information filters (SEIF’s) [30], scan-matching [11, 29, 12] and topological approaches =-=[16, 2]-=-. To date, all published SLAM algorithms are subject to the scaling problem — the fact that the SLAM task gets harder as more and more features are mapped. The unbounded growth of computation with map... |

75 | Experimental comparison of techniques for localization and mapping using a bearing-only sensor
- Deans, Hebert
- 2000
(Show Context)
Citation Context ...e path lengths. The range only SLAM problem has much in common with the structure from motion (SFM) problem [1, 13, 9, 22, 21] which in turn has a strong duality with bearing only and range only SLAM =-=[4, 5]-=-. The motivation for this approach is clear. Submerged, on-the-fly calibration of transponders would enable an AUV to lay and extend its own beacon network to fit adaptive mission and navigation crite... |

73 | A unifying framework for structure and motion recovery from image sequences
- McLauchlan, Murray
- 1995
(Show Context)
Citation Context ...of the range-only data) but also the problem is inherently ill conditioned over small vehicle path lengths. The range only SLAM problem has much in common with the structure from motion (SFM) problem =-=[1, 13, 9, 22, 21]-=- which in turn has a strong duality with bearing only and range only SLAM [4, 5]. The motivation for this approach is clear. Submerged, on-the-fly calibration of transponders would enable an AUV to la... |

72 | Map building with mobile robots in populated environments
- Hähnel, Schulz, et al.
(Show Context)
Citation Context ...AM solution are well-known [6]. Recent related work in SLAM includes submap decomposition methods [19, 10, 14, 32, 28], FastSLAM [23], sparse extended information filters (SEIF’s) [30], scan-matching =-=[11, 29, 12]-=- and topological approaches [16, 2]. To date, all published SLAM algorithms are subject to the scaling problem — the fact that the SLAM task gets harder as more and more features are mapped. The unbou... |

61 | Mobile Robot Navigation Using Active Vision
- Davison
- 1999
(Show Context)
Citation Context ...use of the difficulties encountered by SLAM algorithms when applied to larger environments, the “map scaling” problem has been identified as one of the key issues for research in this area. Davison’s =-=[3, 15]-=- “postponement method” and later Guivant and Nebot’s [10] Compressed Filter allow computational resources to be focused on maintaining a representation of the local area while postponing the computati... |

44 |
Motion and structure from motion from point and line matching
- Faugeras, Lustman, et al.
- 1987
(Show Context)
Citation Context ...o prior information and using only onboard sensors, to move through its environment and build a consistent map of its surroundings as well as an estimate of its own trajectory. SLAM is a hard problem =-=[7, 23, 9, 22]-=- and may not always be the best technique to employ to achieve mobile autonomy. For example, in many land based applications navigation and mapping can be aided by in-place infrastructure - GPS, radio... |

44 | An efficient approach to the simultaneous localisation and mapping problem
- Williams, Dissanayake, et al.
- 2002
(Show Context)
Citation Context |

42 |
Navigation using affine structure from motion
- Beardsley, Zisserman, et al.
- 1994
(Show Context)
Citation Context ...of the range-only data) but also the problem is inherently ill conditioned over small vehicle path lengths. The range only SLAM problem has much in common with the structure from motion (SFM) problem =-=[1, 13, 9, 22, 21]-=- which in turn has a strong duality with bearing only and range only SLAM [4, 5]. The motivation for this approach is clear. Submerged, on-the-fly calibration of transponders would enable an AUV to la... |

41 | Pure Range-Only Sub-Sea SLAM
- Newman, Leonard
- 2002
(Show Context)
Citation Context ... localization - even though the built map of the artifacts used to achieve this may not be of interest per-se. One approach to the sub-sea navigation problem is ‘on-thefly’ acoustic feature deployment=-=[25]-=-. In this scenario instead of detecting naturally occurring landmarks low cost or recoverable transponders are deployed in unknown locations which enable range only measurements between vehicle and be... |

39 | A batch/recursive algorithm for 3D scene reconstruction
- McLauchlan
- 2000
(Show Context)
Citation Context ...o prior information and using only onboard sensors, to move through its environment and build a consistent map of its surroundings as well as an estimate of its own trajectory. SLAM is a hard problem =-=[7, 23, 9, 22]-=- and may not always be the best technique to employ to achieve mobile autonomy. For example, in many land based applications navigation and mapping can be aided by in-place infrastructure - GPS, radio... |

29 | Combined Doppler/LBL based navigation of underwater vehicles
- Whitcomb, Yoerger, et al.
- 1999
(Show Context)
Citation Context ... domain is a different matter however. No wide-coverage underwater GPS equivalent exists (although some small oil and mineral rich areas are well populated with acoustic beacons at surveyed locations =-=[31]-=-). At present only a few percent of the earth’s seabed has been explored. What publicly available bathymetry maps [20] do exist of explored areas are coarse and unsuitable for precision navigation (CE... |

25 | Invariant filtering for simultaneous localization and mapping
- Deans, Hebert
(Show Context)
Citation Context ...e path lengths. The range only SLAM problem has much in common with the structure from motion (SFM) problem [1, 13, 9, 22, 21] which in turn has a strong duality with bearing only and range only SLAM =-=[4, 5]-=-. The motivation for this approach is clear. Submerged, on-the-fly calibration of transponders would enable an AUV to lay and extend its own beacon network to fit adaptive mission and navigation crite... |

15 |
Building a million beacon map
- Julier, Uhlmann
- 2001
(Show Context)
Citation Context |

13 |
High performance doppler-inertial navigation experimental results
- Larsen
- 2000
(Show Context)
Citation Context ...ant, Mike Bosse, Henrik Schmidt and the MIT Acoustics Group. April 2003 1 tion. The state-of-the-art in AUV navigation technology uses a fusion of Doppler Velocity Log (DVL) and inertial measurements =-=[17]-=-. Although impressive performance has been reported by such systems (0.1% of distance travelled) they are essentially odometric in nature and prone to the accumulating drift and error growth inherent ... |

8 |
Consistent convergent constant time slam
- Newman, Leonard
- 2003
(Show Context)
Citation Context ...ations that require empirical testing to verify state estimation consistency. 3 A Constant-Time SLAM Algorithm This section summarizes a new Constant-Time SLAM (CTS) algorithm, described in detail in =-=[18]-=-. Figure 2 illustrates the broad structure of the algorithm. The CTS algorithm is characterized by the following elements and criteria: 1. SLAM processing occurs within local maps. 2. Map management —... |

7 |
efficent solution to the slam problem using geometric projection
- Newman, Durrant-Whyte
- 2001
(Show Context)
Citation Context ...be completed - once again placing a limit on the size of environment in which the algorithm can be deployed. In a similar vein, the Constrained Sub-map Filter [32] and the Geometric Projection Filter =-=[24]-=- seek to delay the full O(n2 ) computation. Other techniques such as decoupled stochastic mapping [19] and SEIFs [30] achieve O(1) performance, but make approximations that require empirical testing t... |

5 |
Segmentation of bathymetric profiles and terrain matching for underwater vehicle navigation
- Lucido, Popescu, et al.
- 1996
(Show Context)
Citation Context ...mineral rich areas are well populated with acoustic beacons at surveyed locations [31]). At present only a few percent of the earth’s seabed has been explored. What publicly available bathymetry maps =-=[20]-=- do exist of explored areas are coarse and unsuitable for precision navigation (CEP < 20 m). Underwater SLAM is a key technology in development of Autonomous Underwater Vehicles (AUVs). A SLAM enabled... |

5 | Trajectory sonar perception
- Rikoski, Leonard
(Show Context)
Citation Context ...he complete set of all measurements. This sequence of estimates is used as the ground truth for the following results. 4.2 SAS processing using Trajectory Sonar Perception Trajectory Sonar Perception =-=[26]-=- is a method of tracking locally curved objects using sonar. Over small regimes, it is assumed that surfaces can be described by a radius of curvature ρ and a center of curvature [xc yc zc] T . A poin... |

5 |
Simultaneous mapping and localization with sparse extended information filters
- Y
- 2002
(Show Context)
Citation Context ... of single-map LG SLAM solution are well-known [6]. Recent related work in SLAM includes submap decomposition methods [19, 10, 14, 32, 28], FastSLAM [23], sparse extended information filters (SEIF’s) =-=[30]-=-, scan-matching [11, 29, 12] and topological approaches [16, 2]. To date, all published SLAM algorithms are subject to the scaling problem — the fact that the SLAM task gets harder as more and more fe... |

4 |
Bistatic synthetic aperture target detection and imaging with an auv
- Edwards, Schmidt, et al.
(Show Context)
Citation Context ...esults using the CTS algorithm with real data in a truly challenging environment. 4 Experimental Results — Application to an AUV This section presents CTS SLAM results using data from the 2002 “GOATS”=-=[8]-=- experiment near the coast of Italy. The experiment was one component of an ongoing program of research into the to the search, detection and identification of subsea mines (the mine counter measure (... |

2 |
Towards Fully Autonomous Mobile Robot Navigation
- Knight
- 2003
(Show Context)
Citation Context ...use of the difficulties encountered by SLAM algorithms when applied to larger environments, the “map scaling” problem has been identified as one of the key issues for research in this area. Davison’s =-=[3, 15]-=- “postponement method” and later Guivant and Nebot’s [10] Compressed Filter allow computational resources to be focused on maintaining a representation of the local area while postponing the computati... |

1 |
Towards simultaneous localization and map building (SLAM) in large unstructured environments
- Durrant-Whyte, Dissanayake, et al.
- 2000
(Show Context)
Citation Context |

1 |
Anefficient solutionto the slam problem using geometric projections
- Durrant-Whyte
- 2001
(Show Context)
Citation Context ...be completed - once again placing a limit on the size of environment in which the algorithm can be deployed. In a similar vein, the Constrained Sub-map Filter [32] and the Geometric Projection Filter =-=[24]-=- seek to delay the full O(n 2 ) computation. Other techniques such as decoupled stochastic mapping [19] and SEIFs [30] achieve O(1) performance, but make approximations that require empirical testing ... |