## Robot Pose Estimation in Unknown Environments by Matching 2D Range Scans (1994)

### Cached

### Download Links

- [www.cs.dal.ca]
- [users.cs.dal.ca]
- [www.cim.mcgill.ca]
- [www.cim.mcgill.ca]
- [robotics.caltech.edu]
- [web.cs.dal.ca]
- [web.cs.dal.ca]
- [ftp.cs.yorku.ca]
- DBLP

### Other Repositories/Bibliography

Citations: | 230 - 8 self |

### BibTeX

@MISC{Lu94robotpose,

author = {Feng Lu and Evangelos Milios},

title = {Robot Pose Estimation in Unknown Environments by Matching 2D Range Scans},

year = {1994}

}

### Years of Citing Articles

### OpenURL

### Abstract

A mobile robot exploring an unknown environment has no absolute frame of reference for its position, other than features it detects through its sensors. Using distinguishable landmarks is one possible approach, but it requires solving the object recognition problem. In particular, when the robot uses two-dimensional laser range scans for localization, it is difficult to accurately detect and localize landmarks in the environment (such as corners and occlusions) from the range scans. In this paper, we develop two new iterative algorithms to register a range scan to a previous scan so as to compute relative robot positions in an unknown environment, that avoid the above problems. The first algorithm is based on matching data points with tangent directions in two scans and minimizing a distance function in order to solve the displacementbetween the scans. The second algorithm establishes correspondences between points in the two scans and then solves the point-to-point least-squares probl...

### Citations

2232 |
A method for registration of 3D shapes
- Besl, McKay
- 1992
(Show Context)
Citation Context ...on between the scans and using an embedded least-squares procedure to solve for the relative translation. Previous methods which iteratively match points to curves or surfaces have also been reported =-=[1,4]-=-. They typically solve both rotation and translation from a least-squares procedure and use fixed-point iterations to improve the solution. To ensure convergence, an initial small rotation and transla... |

1996 |
Robot Motion Planning
- Latombe
- 1991
(Show Context)
Citation Context ...y known environment and perfect sensing. The optimality criterion in this class of problems is the minimization of the cost for traversal between a start and an end position, while avoiding obstacles =-=[18]-=-. The second type of problem is exploration of an unknown world with perfect range sensing and odometry information [23]. Here the issues are primarily the complete coverage of the environment and the... |

585 |
Robust Statistics: The Approach Based on Influence Functions. Probability and Mathematical Statistics
- Hampel, Ronchetti, et al.
- 1986
(Show Context)
Citation Context ...ization of a distance measure between the two scans, as a function of the rotation and translation. To exclude outliers in defining the distance measure, we can apply the concept of robust statistics =-=[11,2]-=-. 3 Search/Least-Squares Matching Algorithm 3.1 Method Overview Our approach to the scan matching problem is to define a distance measure between the two scans and search for an appropriate rigid tran... |

572 |
Object modeling by registration of multiple range images
- Chen, Medioni
- 1992
(Show Context)
Citation Context ...on between the scans and using an embedded least-squares procedure to solve for the relative translation. Previous methods which iteratively match points to curves or surfaces have also been reported =-=[1,4]-=-. They typically solve both rotation and translation from a least-squares procedure and use fixed-point iterations to improve the solution. To ensure convergence, an initial small rotation and transla... |

407 | A robot exploration and mapping strategy based on a semantic hierarchy of spatial representations
- Kuipers, Byun
- 1991
(Show Context)
Citation Context ... estimation in a known (or partially known) twodimensional polygonal environment as solved. Finally, exploration of an unknown world with imperfect range sensing and odometry information is addressed =-=[17]-=-. Here the self-localization of the robot is still an important issue. Distinct locations or landmarks detected from sonar data are used in [19]. Matching local models of line segments obtained from s... |

343 |
Object Recognition by Computer: The Role of Geometric Constraints
- Grimson
- 1990
(Show Context)
Citation Context ...h execution and exploration of unknown world. The main issue in our problem is the consistent alignment of two data sets (range scans in particular). This is quite different from model-based matching =-=[10]-=-. Although a range scan represents a 2D shape (contour of the visible world), this shape is only represented by noisy discrete points instead of a high-quality model, which makes it very difficult to ... |

181 |
Dynamic map building for an autonomous mobile robot
- Leonard, Durrant-Whyte, et al.
- 1992
(Show Context)
Citation Context ...ect range sensing and odometry information is addressed [17]. Here the self-localization of the robot is still an important issue. Distinct locations or landmarks detected from sonar data are used in =-=[19]-=-. Matching local models of line segments obtained from sonar data with a cumulative global model is adopted in [8,24]. In the context of the above literature, our work addresses the robot self-localiz... |

180 |
Blancheâ€”an experiment in guidance and navigation of an autonomous robot vehicle
- Cox
- 1991
(Show Context)
Citation Context ...bot "Blanche" assumes that a metric map of the environment consisting of polygonal obstacles is available, and it matches noisy range scans against the map frequently to keep the position er=-=ror small [6]-=-. At the heart of the method is an iterative least-squares algorithm that finds the congruence between a range scan and the map provided that the initial displacement is small. We can thus consider th... |

151 | World modeling and position estimation for a mobile robot using ultrasonic ranging
- Crowley
- 1989
(Show Context)
Citation Context ...mportant issue. Distinct locations or landmarks detected from sonar data are used in [19]. Matching local models of line segments obtained from sonar data with a cumulative global model is adopted in =-=[8,24]-=-. In the context of the above literature, our work addresses the robot self-localization problem by using laser range scans, similar to the work in [5]. However, we consider more general cases as (1) ... |

131 |
Qualitative Navigation for Mobile Robots
- Levitt, Lawton
- 1992
(Show Context)
Citation Context ...ll understood that odometry is not sufficient because it leads to unbounded position error [13]. A qualitative solution by tracking which side of landmark-defined lines the robot is on is proposed in =-=[20]-=-. If the robot is equipped with vision, then matching 3D models with 2D scene images is possible [9,16]. Because of the enormous computational requirements of using image data, using twodimensional la... |

106 |
Fast vision-guided mobile robot navigation using modelbased reasoning and prediction of uncertainties, CVGIP: Image Understanding 56 (3
- Kosaka, Kak
- 1992
(Show Context)
Citation Context ...alitative solution by tracking which side of landmark-defined lines the robot is on is proposed in [20]. If the robot is equipped with vision, then matching 3D models with 2D scene images is possible =-=[9,16]. Because -=-of the enormous computational requirements of using image data, using twodimensional laser range scans has also been proposed and demonstrated successfully [5]. The robot "Blanche" assumes t... |

79 |
Robust incremental optical flow
- Black
- 1992
(Show Context)
Citation Context ...ization of a distance measure between the two scans, as a function of the rotation and translation. To exclude outliers in defining the distance measure, we can apply the concept of robust statistics =-=[11,2]-=-. 3 Search/Least-Squares Matching Algorithm 3.1 Method Overview Our approach to the scan matching problem is to define a distance measure between the two scans and search for an appropriate rigid tran... |

60 |
Building 3-d models from unregistered range images, Graphical Models and Image Processing 57
- Higuchi, Hebert, et al.
- 1995
(Show Context)
Citation Context ...tial correlation methods which use arclength to reference shape contours (such as [15]) can not be easily applied to scan-matching. The registration method based on fitting Spherical Attribute Images =-=[12]-=- also has difficulty in matching partial views. Our approach is to start with an approximate alignment of the two scans (obtained from odometry) , and then iteratively improve the alignment by definin... |

55 |
Numerical Methods and Software
- Kahaner, Moler, et al.
- 1989
(Show Context)
Citation Context ...as simulated annealing) may also be applied to minimize the non-smooth distance function (such as [3]). These methods are usually computationally expensive. We use the search by golden section method =-=[14]-=- to find the minimum along the rotation dimension in the distance function. We choose this method because of its efficiency in terms of required function evaluations. The search by the golden section ... |

48 |
TwoDimensional, Model-Based, Boundary Matching Using Footprints
- Kalvin, Schonberg, et al.
- 1986
(Show Context)
Citation Context ...occlusion, since the range scans from two different robot positions will only partially overlap. This implies that spatial correlation methods which use arclength to reference shape contours (such as =-=[15]-=-) can not be easily applied to scan-matching. The registration method based on fitting Spherical Attribute Images [12] also has difficulty in matching partial views. Our approach is to start with an a... |

44 | Precise positioning using model-based maps
- MacKenzie, Dudek
- 1994
(Show Context)
Citation Context ...mportant issue. Distinct locations or landmarks detected from sonar data are used in [19]. Matching local models of line segments obtained from sonar data with a cumulative global model is adopted in =-=[8,24]-=-. In the context of the above literature, our work addresses the robot self-localization problem by using laser range scans, similar to the work in [5]. However, we consider more general cases as (1) ... |

36 |
Model-directed mobile robot navigation
- Fennema, Hanson, et al.
- 1990
(Show Context)
Citation Context ...alitative solution by tracking which side of landmark-defined lines the robot is on is proposed in [20]. If the robot is equipped with vision, then matching 3D models with 2D scene images is possible =-=[9,16]. Because -=-of the enormous computational requirements of using image data, using twodimensional laser range scans has also been proposed and demonstrated successfully [5]. The robot "Blanche" assumes t... |

29 |
Blanche: Position estimation for an autonomous robot vehicle
- Cox
- 1990
(Show Context)
Citation Context ...2D scene images is possible [9,16]. Because of the enormous computational requirements of using image data, using twodimensional laser range scans has also been proposed and demonstrated successfully =-=[5]. The robo-=-t "Blanche" assumes that a metric map of the environment consisting of polygonal obstacles is available, and it matches noisy range scans against the map frequently to keep the position erro... |

21 |
Optimal global pose estimation for consistent sensor data registration
- Lu, Milios
- 1995
(Show Context)
Citation Context ...gorithm to each pair. Our matching algorithm successfully aligned the scans in all 84 cases. All of these scan matching results were used by a global scan registration algorithm which is discussed in =-=[22,21]-=-. Two examples of the matching results are shown in Figures 15 and 16. In both examples, we exaggerated the pose errors for testing the algorithm. The odometry error from the actual robot is much smal... |

18 |
Shape registration using optimization for mobile robot navigation
- Lu
- 1995
(Show Context)
Citation Context ...gorithm to each pair. Our matching algorithm successfully aligned the scans in all 84 cases. All of these scan matching results were used by a global scan registration algorithm which is discussed in =-=[22,21]-=-. Two examples of the matching results are shown in Figures 15 and 16. In both examples, we exaggerated the pose errors for testing the algorithm. The odometry error from the actual robot is much smal... |

17 |
A Comparative Study on the Path Length Performance of Maze-Searching and Robot Motion Planning Algorithms
- Lumelsky
- 1991
(Show Context)
Citation Context ...ost for traversal between a start and an end position, while avoiding obstacles [18]. The second type of problem is exploration of an unknown world with perfect range sensing and odometry information =-=[23]-=-. Here the issues are primarily the complete coverage of the environment and the complexity of the algorithm as a function of the complexity of the environment (number of vertices and edges of objects... |

16 | Robust Statistics: The Approach Based on - Hampel, Rousseeuw, et al. - 1986 |

2 |
Registering Multiple Range Data to Create 3D Computer Objects
- Blais, Levine
- 1995
(Show Context)
Citation Context .... One possible way of searching for a minimum is to sample the search space. Stochastic methods (such as simulated annealing) may also be applied to minimize the non-smooth distance function (such as =-=[3]-=-). These methods are usually computationally expensive. We use the search by golden section method [14] to find the minimum along the rotation dimension in the distance function. We choose this method... |

2 |
On the congruence of noisy images to line segments models
- Cox, Kruskal
- 1988
(Show Context)
Citation Context ...primary issue in solving this problem is how to match sensed data (vision, sonar, laser, infrared etc. ) against map information. A comprehensive collection of literature in this area can be found in =-=[7,13]-=-. The robot selflocalization problem can be addressed either by metric means, by ensuring that the difference between the robot's actual position and the position where the robot thinks it is remains ... |

2 |
Autonomous Mobile Robots, Vols. 1 and 2
- Iyengar, Elfes
- 1991
(Show Context)
Citation Context ...primary issue in solving this problem is how to match sensed data (vision, sonar, laser, infrared etc. ) against map information. A comprehensive collection of literature in this area can be found in =-=[7,13]-=-. The robot selflocalization problem can be addressed either by metric means, by ensuring that the difference between the robot's actual position and the position where the robot thinks it is remains ... |

2 |
Design of ARK, a sensor-based mobile robot for industrial environments. Intelligent Vehicles
- Nickerson, Jenkin, et al.
- 1993
(Show Context)
Citation Context ...equentially combining the two algorithms. We apply this algorithm in the following experiments. 5.2 Sensing Strategy If we have control over the sensing directions (as is the case with the ARK sensor =-=[25]-=-), we can choose to take a range scan in such a way that the sampling points are evenly distributed in space. In other words, we want the distances between adjacent points to be approximately equal. T... |