#### DMCA

## Robust Monte Carlo Localization for Mobile Robots (2001)

### Cached

### Download Links

- [www.informatik.uni-freiburg.de]
- [www4.cs.umanitoba.ca]
- [isl.ecst.csuchico.edu]
- [www.informatik.uni-freiburg.de]
- [www2.informatik.uni-freiburg.de]
- [www4.cs.umanitoba.ca]
- [www-2.cs.cmu.edu]
- [www.cs.cmu.edu]
- [www.cs.cmu.edu]
- [www.cs.cmu.edu]
- [msl.cs.uiuc.edu]
- [www.cs.cmu.edu]
- [www.cs.cmu.edu]
- [robots.stanford.edu]
- [kobus.ca]
- [www.cs.cmu.edu]
- [www.cc.gatech.edu]
- [www.cc.gatech.edu]
- [www.cs.cmu.edu]
- [www.informatik.uni-freiburg.de]
- [www.cs.cmu.edu]
- [www.ri.cmu.edu]
- [www.cc.gatech.edu]
- [www.cc.gatech.edu]
- [www.cc.gatech.edu]
- [www-2.cs.cmu.edu]
- [www.cs.washington.edu]
- [www.ri.cmu.edu]
- [www.cs.sfu.ca]
- [www.cs.sfu.ca]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [reports-archive.adm.cs.cmu.edu]
- [www.cs.uml.edu]
- [www.cim.mcgill.ca]
- [www.cs.uml.edu]
- DBLP

### Other Repositories/Bibliography

Citations: | 838 - 85 self |

### Citations

11970 | Maximum likelihood from incomplete data via the EM algorithm
- Dempster, Laird, et al.
- 1977
(Show Context)
Citation Context ... max-range measurements, which frequently occur when a range sensor fails to detect an object. The specific parameters of the density in Figure 3b have been estimated using an algorithm similar to EM =-=[15,52]-=-, which starts with a crude initial model and iteratively labels several million measurements collected in the Smithsonian museum, while refining the model. A smoothed version of these data is also sh... |

8904 |
Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference
- Pearl
- 1988
(Show Context)
Citation Context ...n the same basic idea, but asymptotically approximates the desired posterior. The idea of sampling from the sensor measurement (the “evidence”) has also been proposed in the context of Bayes networks =-=[57,29]-=-, in particular in the context of marginalization using Monte Carlo sampling. Under the name of “arc reversal,” Kanazawa and colleagues [38] have proposed an efficient sampling algorithm that jump-sta... |

3856 | A new approach to linear filtering and prediction problems,” Trans. - Kalman - 1960 |

1734 |
A novel approach to non-linear and non-gaussian bayesian state estimation
- Gordon, Salmon, et al.
- 1993
(Show Context)
Citation Context ...wn from it. To update this density representation over time, we make use of Monte Carlo methods that were invented in the seventies [6], and recently rediscovered independently in the target-tracking =-=[7]-=-, statistical [8] and computer vision literature [9, 10]. By using a sampling-based representation we obtain a localization method that has several key advantages with respect to earlier work: 1. In c... |

1503 | CONDENSATION -- conditional density propagation for visual tracking,
- Isard, Blake
- 1998
(Show Context)
Citation Context ...ecent. In the statistical literature, it is known as particle filters [17,18,46,58], and recently computer vision researchers have proposed the same algorithm under the name of condensation algorithm =-=[33]-=-. Within the context of localization, the particle representation has a range of characteristics that sets it aside from previous approaches: (1) Particle filters can accommodate (almost) arbitrary se... |

1468 | The EM Algorithm and Extensions - McLachlan, Krishnan - 1997 |

1069 | A tutorial on learning with Bayesian networks.
- Heckerman, Geiger, et al.
- 1995
(Show Context)
Citation Context ...n the same basic idea, but asymptotically approximates the desired posterior. The idea of sampling from the sensor measurement (the “evidence”) has also been proposed in the context of Bayes networks =-=[57,29]-=-, in particular in the context of marginalization using Monte Carlo sampling. Under the name of “arc reversal,” Kanazawa and colleagues [38] have proposed an efficient sampling algorithm that jump-sta... |

949 |
Tracking and Data Association,
- Bar-Shalom, Fortmann
- 1988
(Show Context)
Citation Context ...augmenting the sample set through uniformly 3distributed samples [21], generating samples consistent with the most recent sensor reading [43] (an idea familiar from multi-hypothesis Kalman filtering =-=[1,34,61]-=-), or assuming a higher level of sensor noise than actually is the case. While these extensions yield improved performance, they are mathematically questionable. In particular, these extensions do not... |

778 | A new extension of the kalman filter to nonlinear systems, - Julier, Uhlmann, et al. - 1997 |

776 | Filtering via simulation : auxiliary particle filters.
- Pitt, Shephard
- 1999
(Show Context)
Citation Context ...distribution [62]. The idea of estimating state recursively using particles is not new, although most work on this topic is very recent. In the statistical literature, it is known as particle filters =-=[17,18,46,58]-=-, and recently computer vision researchers have proposed the same algorithm under the name of condensation algorithm [33]. Within the context of localization, the particle representation has a range o... |

664 | Sequential Monte Carlo methods for dynamical system,
- Liu, Chen
- 1998
(Show Context)
Citation Context ...distribution [62]. The idea of estimating state recursively using particles is not new, although most work on this topic is very recent. In the statistical literature, it is known as particle filters =-=[17,18,46,58]-=-, and recently computer vision researchers have proposed the same algorithm under the name of condensation algorithm [33]. Within the context of localization, the particle representation has a range o... |

661 | Contour tracking by stochastic propagation of conditional density
- Isard, Blake
- 1996
(Show Context)
Citation Context ...g problems of practical importance. In computer vision, particle filters are commonly known as condensation algorithm, where they have been applied with remarkable success to visual tracking problems =-=[32,33,48]-=-. Their application to mobile robot localization has been proposed in [13,21] and since been adopted (and extended) by several other researchers [16,43]. In our own work, we recently extended the basi... |

591 |
Monte Carlo filter and smoother for non-Gaussian nonlinear state space models,
- Kitagawa
- 1996
(Show Context)
Citation Context ...particle filters. The poor performance of particle filtering in cases where the pro41posal distribution differs significantly from the target distribution has been observed by several authors, e.g., =-=[17,39,46,58]-=-. Typical “fixes” involve the design of a different proposal distribution that places more weight on the tails of a distribution. In this light, Mixture-MCL can be views as one way to deal with this m... |

576 |
An Analysis of TimeDependent Planning.
- Dean, Boddy
- 1988
(Show Context)
Citation Context ...ber of samples). The advantage of such an implementation is its adaptivity to the available computational resources. Such resource-adaptive algorithms are sometimes referred to as any-time algorithms =-=[11,76]-=-. They have the advantage that when ported to a different computer (e.g., a new, faster PC), they can exploit the additional computational power without any modification to the program code. An argume... |

531 | Globally consistent range scan alignment for environment mapping
- Lu, Milios
- 1997
(Show Context)
Citation Context ...[5,71]. The vast majority of existing algorithms address only the position tracking problem (see e.g., the review [4]). The nature of small, incremental errors makes algorithms such as Kalman filters =-=[28,37,47,68]-=- applicable, which have been successfully applied in a range of fielded systems (e.g., [27,44,42,63]). Kalman filters estimate posterior distributions of robot poses conditioned on sensor data. Exploi... |

520 |
Estimating uncertain spatial relationships in robotics,” in Autonomous Robot Vehicles,
- Smith, Self, et al.
- 1988
(Show Context)
Citation Context ...[5,71]. The vast majority of existing algorithms address only the position tracking problem (see e.g., the review [4]). The nature of small, incremental errors makes algorithms such as Kalman filters =-=[28,37,47,68]-=- applicable, which have been successfully applied in a range of fielded systems (e.g., [27,44,42,63]). Kalman filters estimate posterior distributions of robot poses conditioned on sensor data. Exploi... |

464 | Stochastic Models, Estimation, and Control. - Maybeck - 1982 |

405 |
Sensor fusion in certainty grids for mobile robots”,
- Moravec
- 1988
(Show Context)
Citation Context ...nt, since it only involves a small number of random coin flips. Figure 14 illustrates our sampling algorithm in practice. Shown in Figure 14a is an example range scan along with an occupancy grid map =-=[19,54]-=- as described in Section 2.2. From this scan, our approach extracts the three features described 30(a) laser scan and map (b) tree for this scan (c) samples of poses generated from tree Fig. 14. Samp... |

361 | Markov localization for mobile robots in dynamic environments.
- Fox, Burgard, et al.
- 1999
(Show Context)
Citation Context ...ionally, it has been referred to as “ the most fundamental problem to providing a mobile robot with autonomous capabilities” [8]. The mobile robot localization problem comes in many different flavors =-=[4,24]-=-. The most simple localization problem—which has received by far the most attention in the literature—is position tracking [4,75,64,74]. Here the initial robot pose is known, and the problem is to com... |

344 |
Stochastic Models, Estimation, and Control, volume 141-1
- Maybeck
- 1982
(Show Context)
Citation Context ...present the robot’s belief at time before incorporating the sensor measurement . Belief distributions of this type are often referred to as predictive distributions in the Kalman filtering literature =-=[51]-=-, since they represent the prediction of state based on the ! " # action and previous data, but before incorporating the sensor measurement at time . To use this distribution for calculating importanc... |

343 | Monte carlo localization: Efficient position estimation for mobile robots
- Fox, Burgard, et al.
- 1999
(Show Context)
Citation Context ...ain independence assumptions—which will also be the case for the approach presented in this article. This article presents a probabilistic localization algorithm called Monte Carlo localization (MCL) =-=[13,21]-=-. MCL solves the global localization and kidnapped robot problem in a highly robust and efficient way. It can accommodate arbitrary noise distributions (and non-linearities). Thus, MCL avoids a need t... |

343 | Inadmissibility of the usual estimator for the mean of a multivariate distribution. - Stein - 1956 |

342 | Tracking and Data - Bar-Shalom, Fortmann - 1988 |

329 | Experiences with an interactive museum tour-guide robot
- Burgard, Cremers, et al.
- 1999
(Show Context)
Citation Context ...rophic localization failures. Finally, all these problems are particularly hard in dynamic environments, e.g., if robots operate in the proximity of people who corrupt the robot’s sensor measurements =-=[5,71]-=-. The vast majority of existing algorithms address only the position tracking problem (see e.g., the review [4]). The nature of small, incremental errors makes algorithms such as Kalman filters [28,37... |

308 | Directed Sonar Sensing for Mobile Robot Navigation
- Leonard, Durrant-Whyte
- 1992
(Show Context)
Citation Context ...the density p(xkjZ k ) will remain Gaussian at all times. In this case, equations (1) and (2) can be evaluated in closed form, yielding the classical Kalman filter [12]. Kalmanfilter based techniques =-=[13, 14, 15]-=- have proven to be robust and accurate for keeping track of the robot’s position. Because of its concise representation (the mean and covariance matrix suffice to describe the entire density) it is al... |

293 | Probabilistic robot navigation in partially observable environments.
- Simmons, Koenig
- 1995
(Show Context)
Citation Context ...much of the information acquired by a robot’s sensors. Markov localization algorithms, in contrast, represent beliefs by piecewise constant functions (histograms) over the space of all possible poses =-=[7,24,30,36,40,50,55,56,66,70]-=-. Just like Gaussian mixtures, piecewise constant functions are capable of representing complex, multi-modal representations. Some of these algorithms also rely on features [36,40,50,55,66,70], hence ... |

268 | An improved particle filter for non-linear problems,”
- Carpenter, Clifford, et al.
- 1999
(Show Context)
Citation Context ... associated with them, and in doing so a new set Sk is obtained that approximates a random sample from p(xkjZ k ). An algorithm to perform this resampling process efficiently in O(N) time is given in =-=[19]-=-. After the update phase, the steps (i) and (ii) are repeated recursively. To initialize the filter, we start at time k =0 with a random sample S0 = fsi 0g from the prior p(x0). 4.1 A Graphical Exampl... |

261 | A probabilistic exclusion principle for tracking multiple objects,”
- MacCormick, Blake
- 2000
(Show Context)
Citation Context ...g problems of practical importance. In computer vision, particle filters are commonly known as condensation algorithm, where they have been applied with remarkable success to visual tracking problems =-=[32,33,48]-=-. Their application to mobile robot localization has been proposed in [13,21] and since been adopted (and extended) by several other researchers [16,43]. In our own work, we recently extended the basi... |

254 | Stochastic Models, - Maybeck - 1982 |

251 | On Sequential Simulation-Based Methods for Bayesian Filtering,”
- Doucet
- 1998
(Show Context)
Citation Context ...distribution [62]. The idea of estimating state recursively using particles is not new, although most work on this topic is very recent. In the statistical literature, it is known as particle filters =-=[17,18,46,58]-=-, and recently computer vision researchers have proposed the same algorithm under the name of condensation algorithm [33]. Within the context of localization, the particle representation has a range o... |

223 |
An experiment in guidance and navigation of an autonomous robot vehicle”,
- Cox
- 1991
(Show Context)
Citation Context ...ms (see e.g., [10,25,31,45,59,65,74] and various chapters in [4,41]). Occasionally, it has been referred to as “ the most fundamental problem to providing a mobile robot with autonomous capabilities” =-=[8]-=-. The mobile robot localization problem comes in many different flavors [4,24]. The most simple localization problem—which has received by far the most attention in the literature—is position tracking... |

218 | Bayesian statistics without tears: A sampling-resampling perspective.
- Smith, Gelfand
- 1992
(Show Context)
Citation Context ... a set of N random samples or particles Sk = fsi k ; i =1::N g drawn from it. We are able to do this because of the essential duality between the samples and the density from which they are generated =-=[17]-=-. From the samples we can always approximately reconstruct the density, e.g. using a histogram or a kernel based density estimation technique. The goal is then to recursively compute at each timestep ... |

217 |
Tools for Statistical Inference,
- Tanner
- 1996
(Show Context)
Citation Context ...distribution " # 9 " #2 . Obviously, the pair q6 c Wed " 7J 7 Wed # r is distributed " G"H#J . " # . In accordance with the literature on the SIR algorithm (short for: Sampling importance resampling) =-=[67,62,69]-=-, we will refer to this distribution s as the proposal distribution. Its role is to “propose” samples of the desired posterior distribution, which is given in Equation (9); however, it is not equivale... |

206 | Using the sir algorithm to simulate posterior distributions. In - Rubin - 1988 |

200 | Estimating the absolute position of a mobile robot using position probability grids,
- Burgard, Fox, et al.
- 1996
(Show Context)
Citation Context ...much of the information acquired by a robot’s sensors. Markov localization algorithms, in contrast, represent beliefs by piecewise constant functions (histograms) over the space of all possible poses =-=[7,24,30,36,40,50,55,56,66,70]-=-. Just like Gaussian mixtures, piecewise constant functions are capable of representing complex, multi-modal representations. Some of these algorithms also rely on features [36,40,50,55,66,70], hence ... |

200 | Dynamic map building for an autonomous mobile robot
- Leonard, Durrant-Whyte, et al.
- 1990
(Show Context)
Citation Context ...s pose (location, orientation) relative to its environment. The localization problem is a key problem in mobile robotics. it plays a pivotal role in various successful mobile robot systems (see e.g., =-=[10,25,31,45,59,65,74]-=- and various chapters in [4,41]). Occasionally, it has been referred to as “ the most fundamental problem to providing a mobile robot with autonomous capabilities” [8]. The mobile robot localization p... |

196 | Probabilistic algorithms and the interactive museum tour-guide robot minerva,” - Thrun, Beetz, et al. - 2000 |

185 | An experimental comparison of localization methods continued.
- Gutmann, Fox
- 2002
(Show Context)
Citation Context ...., the review [4]). The nature of small, incremental errors makes algorithms such as Kalman filters [28,37,47,68] applicable, which have been successfully applied in a range of fielded systems (e.g., =-=[27,44,42,63]-=-). Kalman filters estimate posterior distributions of robot poses conditioned on sensor data. Exploiting a range of restrictive assumptions—such as Gaussian-distributed noise and Gaussiandistributed i... |

185 | A mixed-state condensation tracker with automatic model-switching,”
- Isard, Blake
- 1998
(Show Context)
Citation Context ...robot (e.g. cruising, avoiding people, standing still), enabling one to select a different motion model for each mode. This idea has been explored with great success in the visual tracking literature =-=[22]-=-, and it might further improve localization performance. In addition, this would also allow us to generate symbolic descriptions of the robot’s behavior. Acknowledgments The authors would like to than... |

177 |
Dervish: An Office-Navigating Robot.
- Nourbakhsh
- 1998
(Show Context)
Citation Context ...much of the information acquired by a robot’s sensors. Markov localization algorithms, in contrast, represent beliefs by piecewise constant functions (histograms) over the space of all possible poses =-=[7,24,30,36,40,50,55,56,66,70]-=-. Just like Gaussian mixtures, piecewise constant functions are capable of representing complex, multi-modal representations. Some of these algorithms also rely on features [36,40,50,55,66,70], hence ... |

176 | Stochastic simulation algorithms for dynamic probabilistic networks.
- Kanazawa, Koller, et al.
- 1995
(Show Context)
Citation Context ... has also been proposed in the context of Bayes networks [57,29], in particular in the context of marginalization using Monte Carlo sampling. Under the name of “arc reversal,” Kanazawa and colleagues =-=[38]-=- have proposed an efficient sampling algorithm that jump-starts samples at Bayes network nodes whose value is known, then propagating those samples throughout the network to obtain an estimate of the ... |

164 |
Occupancy Grids: A Probabilistic Framework for Robot Perception and Navigation.
- Elfes
- 1989
(Show Context)
Citation Context ...nt, since it only involves a small number of random coin flips. Figure 14 illustrates our sampling algorithm in practice. Shown in Figure 14a is an example range scan along with an occupancy grid map =-=[19,54]-=- as described in Section 2.2. From this scan, our approach extracts the three features described 30(a) laser scan and map (b) tree for this scan (c) samples of poses generated from tree Fig. 14. Samp... |

158 | Sensor resetting localization for poorly modelled mobile robots.
- Lenser, Veloso
- 2000
(Show Context)
Citation Context ...lly, participle filters are surprisingly easy to implement, which makes them an attractive paradigm for mobile robot localization. Consequently, MCL has already been adopted by several research teams =-=[16,43]-=-, who have extended the basic paradigm in interesting new ways. However, there are pitfalls, too, arising from the stochastic nature of the approximation. Some of these pitfalls are obvious: For examp... |

137 | Using the condensation algorithm for robust, vision-based mobile robot localization
- Dellaert, Burgard, et al.
- 1999
(Show Context)
Citation Context ...e grids with 4cm resolution—which is infeasible given even our best computers. 2.7 Robot With Upward-Pointed Camera Similar results were obtained using a camera as the primary sensor for localization =-=[12]-=-. To test MCL under challenging real-world conditions, we evaluated it using data collected in a populated museum. During a two-week exhibition, our robot Minerva (Figure 8) was employed as a tour-gui... |

133 | Bayesian landmark learning for mobile robot navigation.
- Thrun
- 1998
(Show Context)
Citation Context |

124 |
Minerva: A second generation mobile tour-guide robot,”
- Thrun, Bennewitz, et al.
- 1999
(Show Context)
Citation Context ...rophic localization failures. Finally, all these problems are particularly hard in dynamic environments, e.g., if robots operate in the proximity of people who corrupt the robot’s sensor measurements =-=[5,71]-=-. The vast majority of existing algorithms address only the position tracking problem (see e.g., the review [4]). The nature of small, incremental errors makes algorithms such as Kalman filters [28,37... |

120 | Efficient memory-based learning for robot control - Moore - 1990 |

115 | A comparison of position estimation techniques using occupancy grids.
- Schiele, Crowley
- 1994
(Show Context)
Citation Context ... The mobile robot localization problem comes in many different flavors [4,24]. The most simple localization problem—which has received by far the most attention in the literature—is position tracking =-=[4,75,64,74]-=-. Here the initial robot pose is known, and the problem is to compensate incremental errors in a robot’s odometry. Algorithms for position tracking often make restrictive assumptions on the Preprint s... |

101 | A Layered Architecture for Office Delivery Robots
- Simmons, Goodwin, et al.
- 1997
(Show Context)
Citation Context ...s pose (location, orientation) relative to its environment. The localization problem is a key problem in mobile robotics. it plays a pivotal role in various successful mobile robot systems (see e.g., =-=[10,25,31,45,59,65,74]-=- and various chapters in [4,41]). Occasionally, it has been referred to as “ the most fundamental problem to providing a mobile robot with autonomous capabilities” [8]. The mobile robot localization p... |

96 |
Modeling a dynamic environment using a Bayesian multiple hypothesis approach
- Cox, Leonard
- 1994
(Show Context)
Citation Context ...vercome by two related families of algorithms: localization with multi-hypothesis Kalman filters and Markov localization. Multi-hypothesis Kalman filters represent beliefs using mixtures of Gaussians =-=[9,34,60,61]-=-, thereby enabling them to pursue multiple, distinct hypotheses, each of which is represented by a separate Gaussian. However, this approach inherits from Kalman filters the Gaussian noise assumption.... |

96 |
Eds., Autonomous Robot Vehicles.
- Cox, Wilfong
- 1990
(Show Context)
Citation Context ...s pose (location, orientation) relative to its environment. The localization problem is a key problem in mobile robotics. it plays a pivotal role in various successful mobile robot systems (see e.g., =-=[10,25,31,45,59,65,74]-=- and various chapters in [4,41]). Occasionally, it has been referred to as “ the most fundamental problem to providing a mobile robot with autonomous capabilities” [8]. The mobile robot localization p... |

95 | Spatial learning for navigation in dynamic environment.
- Yamauchi, Beer
- 1996
(Show Context)
Citation Context ... The mobile robot localization problem comes in many different flavors [4,24]. The most simple localization problem—which has received by far the most attention in the literature—is position tracking =-=[4,75,64,74]-=-. Here the initial robot pose is known, and the problem is to compensate incremental errors in a robot’s odometry. Algorithms for position tracking often make restrictive assumptions on the Preprint s... |

94 |
Acting under uncertainty: Discrete bayesian models for mobile robot navigation,”
- Kaelbling, Cassandra, et al.
- 1996
(Show Context)
Citation Context |

88 |
Navigating Mobile Robots: Systems
- Borenstein, Everett, et al.
- 1996
(Show Context)
Citation Context ...environment. The localization problem is a key problem in mobile robotics. it plays a pivotal role in various successful mobile robot systems (see e.g., [10,25,31,45,59,65,74] and various chapters in =-=[4,41]-=-). Occasionally, it has been referred to as “ the most fundamental problem to providing a mobile robot with autonomous capabilities” [8]. The mobile robot localization problem comes in many different ... |

88 |
The EM algorithm and extensions. Wiley series in probability and statistics.
- McLachlan, Krishnan
- 1997
(Show Context)
Citation Context ... max-range measurements, which frequently occur when a range sensor fails to detect an object. The specific parameters of the density in Figure 3b have been estimated using an algorithm similar to EM =-=[15,52]-=-, which starts with a crude initial model and iteratively labels several million measurements collected in the Smithsonian museum, while refining the model. A smoothed version of these data is also sh... |

79 | Integrating global position estimation and position tracking for mobile robots: The dynamic markov localization approach,
- Burgard, Derr, et al.
- 1998
(Show Context)
Citation Context ...vier Preprint 28 February 2001size of the error and the shape of the robot’s uncertainty, required by a range of existing localization algorithms. More challenging is the global localization problem =-=[6,34,61]-=-, where a robot is not told its initial pose but instead has to determine it from scratch. The global localization problem is more difficult, since the error in the robot’s estimate cannot be assumed ... |

74 | Amos: Comparison of scan matching approaches for self-localization in indoor environments
- Gutmann, Schlegel
- 1996
(Show Context)
Citation Context ...[5,71]. The vast majority of existing algorithms address only the position tracking problem (see e.g., the review [4]). The nature of small, incremental errors makes algorithms such as Kalman filters =-=[28,37,47,68]-=- applicable, which have been successfully applied in a range of fielded systems (e.g., [27,44,42,63]). Kalman filters estimate posterior distributions of robot poses conditioned on sensor data. Exploi... |

70 |
Multidimensional divide and conquer,
- Bentley
- 1980
(Show Context)
Citation Context ...alculating importance factors, the ‘trick’ is to transform the samples 23c Wed 9G"H# " # for any sample c c c ™ ™ c c c œ œ Wed E C c c c Wed 9 " # % m " # c c ˜ into a kernel density tree (kd-tree) =-=[3,53]-=- that represents the predictive disd c¡ 9 " #$ " # tribution in closed form. Using this tree, we can now calculate Wed generated by the dual sampler. We now have all pieces together to define the seco... |

68 |
Error correction in mobile robot map learning,”
- Engelson, McDermott
- 1992
(Show Context)
Citation Context ...since the error in the robot’s estimate cannot be assumed to be small. Consequently, a robot should be able to handle multiple, distinct hypotheses. Even more difficult is the kidnapped robot problem =-=[20,24]-=-, in which a well-localized robot is tele-ported to some other place without being told. This problem differs from the global localization problem in that the robot might firmly believe itself to be s... |

68 | Monte Carlo localization with mixture proposal distribution.
- Thrun, Fox, et al.
- 2000
(Show Context)
Citation Context ...mate the correct density; which makes the interpretation of their results difficult. To overcome these problems, this article describes an extension of MCL closely related to [43], called Mixture-MCL =-=[72]-=-. Mixture-MCL addresses all these problems in a way that is mathematically motivated. The key idea is to modify the way samples are generated in MCL. Mixture-MCL combines regular MCL sampling with a “... |

66 | Active markov localization for mobile robots,” in Robotics and Autonomous Systems - Fox, Burgard, et al. - 1998 |

60 | Using learning for approximation in stochastic processes - Koller, Fratkina - 1998 |

59 |
editors, AI-based Mobile Robots: Case studies of successful robot systems,
- Kortenkamp, Bonasso, et al.
- 1998
(Show Context)
Citation Context ...environment. The localization problem is a key problem in mobile robotics. it plays a pivotal role in various successful mobile robot systems (see e.g., [10,25,31,45,59,65,74] and various chapters in =-=[4,41]-=-). Occasionally, it has been referred to as “ the most fundamental problem to providing a mobile robot with autonomous capabilities” [8]. The mobile robot localization problem comes in many different ... |

56 | Approximate reasoning using anytime algorithms
- Zilberstein, Russell
- 1995
(Show Context)
Citation Context ...ber of samples). The advantage of such an implementation is its adaptivity to the available computational resources. Such resource-adaptive algorithms are sometimes referred to as any-time algorithms =-=[11,76]-=-. They have the advantage that when ported to a different computer (e.g., a new, faster PC), they can exploit the additional computational power without any modification to the program code. An argume... |

55 | Navigating mobile robots. - Borenstein, Everett, et al. - 1996 |

53 | Keeping Track of Position and Orientation of Moving Indoor Systems by Correlation of Range-finder Scans.
- Weib, Wetzler, et al.
- 1994
(Show Context)
Citation Context |

49 | Passive distance learning for robot navigation
- Koenig, Simmons
- 1996
(Show Context)
Citation Context |

49 |
Concurrent localisation and map building for mobile robots using ultrasonic sensors”,
- Rencken
(Show Context)
Citation Context |

49 | Bayesian estimation and Kalman filtering: A unified framework for mobile robot localization,
- Roumeliotis, Bekey
- 2000
(Show Context)
Citation Context ...vier Preprint 28 February 2001size of the error and the shape of the robot’s uncertainty, required by a range of existing localization algorithms. More challenging is the global localization problem =-=[6,34,61]-=-, where a robot is not told its initial pose but instead has to determine it from scratch. The global localization problem is more difficult, since the error in the robot’s estimate cannot be assumed ... |

48 |
Monte carlo techniques for prediction and filtering on nonlinear stochastic processes,
- Handschin
- 1970
(Show Context)
Citation Context ...e represent it by maintaining a set of samples that are randomly drawn from it. To update this density representation over time, we make use of Monte Carlo methods that were invented in the seventies =-=[6]-=-, and recently rediscovered independently in the target-tracking [7], statistical [8] and computer vision literature [9, 10]. By using a sampling-based representation we obtain a localization method t... |

46 | Active Global Localisation for a Mobile Robot Using Multiple Hypothesis Tracking.
- Jensfelt, Kristensen
- 1999
(Show Context)
Citation Context ...vier Preprint 28 February 2001size of the error and the shape of the robot’s uncertainty, required by a range of existing localization algorithms. More challenging is the global localization problem =-=[6,34,61]-=-, where a robot is not told its initial pose but instead has to determine it from scratch. The global localization problem is more difficult, since the error in the robot’s estimate cannot be assumed ... |

41 |
Comparison of scan matching approaches for self-localization in indoor environments
- Gutmann, Schlegel
- 1996
(Show Context)
Citation Context ...the density p(xkjZ k ) will remain Gaussian at all times. In this case, equations (1) and (2) can be evaluated in closed form, yielding the classical Kalman filter [12]. Kalmanfilter based techniques =-=[13, 14, 15]-=- have proven to be robust and accurate for keeping track of the robot’s position. Because of its concise representation (the mean and covariance matrix suffice to describe the entire density) it is al... |

41 | Collaborative multirobot localization,” in - Fox, Burgard, et al. - 1999 |

33 |
Environment perception with a laser radar in a fast moving robot
- Hinkel, Knieriemen
- 1988
(Show Context)
Citation Context |

30 |
A mobile robot that learns its place
- Oore, Hinton, et al.
- 1997
(Show Context)
Citation Context |

29 | Landmark-based autonomous navigation in sewerage pipes
- Hertzberg, Kirchner
- 1996
(Show Context)
Citation Context |

23 | Ewald von Puttkamer Keeping Track of Position and Orientation of Moving Indoor Systems by Correlation - Weiß, Wetzler - 1994 |

19 |
Navigation System based on Ceiling Landmark Recognition for Autonomous Mobile Robot",
- FUKUDA
- 1993
(Show Context)
Citation Context ...ric design principle of mobile robot software. 397 Related Work Mobile robot localization is a fundamental problem in mobile robotics, which has received considerable attention over the past decades =-=[4,10,25,31,41,45,59,65,74]-=-. As argued in the introduction of this article, the vast majority of work focuses on the position tracking problem, where errors are assumed to be small. Most approaches are incapable of recovering f... |

15 | A mobile robot that learns its place. Neural Computation 9(3):683–699 - Oore, Hinton, et al. - 1997 |

14 | Experiments in autonomous underground guidance
- Scheding, Nebot, et al.
- 1997
(Show Context)
Citation Context ...., the review [4]). The nature of small, incremental errors makes algorithms such as Kalman filters [28,37,47,68] applicable, which have been successfully applied in a range of fielded systems (e.g., =-=[27,44,42,63]-=-). Kalman filters estimate posterior distributions of robot poses conditioned on sensor data. Exploiting a range of restrictive assumptions—such as Gaussian-distributed noise and Gaussiandistributed i... |

13 | Combining computer graphics and computer vision for probabilistic self-localization. Internal report
- Denzler, Heigl, et al.
- 1999
(Show Context)
Citation Context ...lly, participle filters are surprisingly easy to implement, which makes them an attractive paradigm for mobile robot localization. Consequently, MCL has already been adopted by several research teams =-=[16,43]-=-, who have extended the basic paradigm in interesting new ways. However, there are pitfalls, too, arising from the stochastic nature of the approximation. Some of these pitfalls are obvious: For examp... |

9 |
Mobile robot self-localization using PDAB
- Reuter
(Show Context)
Citation Context ...vercome by two related families of algorithms: localization with multi-hypothesis Kalman filters and Markov localization. Multi-hypothesis Kalman filters represent beliefs using mixtures of Gaussians =-=[9,34,60,61]-=-, thereby enabling them to pursue multiple, distinct hypotheses, each of which is represented by a separate Gaussian. However, this approach inherits from Kalman filters the Gaussian noise assumption.... |

8 |
Estimation and Tracking: Principles
- Bar-Shalom, Li
- 1993
(Show Context)
Citation Context ...acy is needed, but today’s best implementations yield somewhat inferior performance, as suggested by the comparison in Section 2.6. Localization algorithms based on the multi-hypothesis Kalman filter =-=[1,2]-=- represent beliefs using mixtures of Gaussians [9,34,60,61]. To calculate the covariance 40matrices of the individual Gaussian mixture components, the Kalman filtering approach linearizes the motion ... |

8 | Mosaicing a large number of widely dispersed, noisy, and distorted images: A Bayesian approach
- Dellaert, Thorpe, et al.
- 1999
(Show Context)
Citation Context ...ocalization [71]. The other data set contains image segments recorded with a camera pointed towards the museum’s ceiling, using a large-scale ceiling mosaic for cross-referencing the robot’s position =-=[14]-=-. In the past, these data have been used as benchmark, since localization in this crowded and feature4impoverished museum is a challenging problem. Our experiments suggest that our new MCL algorithm ... |

8 | Gordon (Eds.), Sequential Monte CarloMethods in Practice, - Doucet, Freitas, et al. - 2001 |

7 |
Humanoid robots
- Vukobratović, Borovać, et al.
- 2001
(Show Context)
Citation Context ...d in our implementation have been described in depth elsewhere [24]; therefore, we will only provide an informal account. The motion model, L * , is a probabilistic generalization of robot kinematics =-=[10,73]-=-. As noticed above, for a robot operating in the plane the poses and are three-dimensional variables. Each pose comprises a robot’s two-dimensional L Cartesian coordinates and its heading direction (o... |

5 | Tools for Statistical Inference. 3rd edn - Tanner - 1996 |

4 |
Towards autonomous excavation
- Le, Nguyen, et al.
- 1997
(Show Context)
Citation Context ...., the review [4]). The nature of small, incremental errors makes algorithms such as Kalman filters [28,37,47,68] applicable, which have been successfully applied in a range of fielded systems (e.g., =-=[27,44,42,63]-=-). Kalman filters estimate posterior distributions of robot poses conditioned on sensor data. Exploiting a range of restrictive assumptions—such as Gaussian-distributed noise and Gaussiandistributed i... |

4 | Robust mobile robot navigation using partially-observable semi-Markov decision processes. internal report
- Mahadevan, Khaleeli
- 1999
(Show Context)
Citation Context |

4 | Wilfong (Eds.), Autonomous Robot Vehicles - Cox, T - 1990 |

2 |
Bayes theorem and digital realisation for nonlinear filters
- Bucy
- 1969
(Show Context)
Citation Context ...er a grid of points. This involves discretizing the interesting part of the state space, and use it as the basis for an approximation of the density p(xkjZ k ), e.g. by a piece-wise constant function =-=[16]-=-. This idea forms the basis of our previously introduced grid-based Markov localization approach (see [5, 11]). Methods that use this type of representation are powerful, but suffer from the disadvant... |

2 |
tracing and graphics extensions
- Ray
- 1994
(Show Context)
Citation Context ...deal, noise-free sensor whose relative bearing X*W is . Since we assume that the robot is given a map of the environment such as the one shown in Figure Y * X*WG 3a, can be computed using ray tracing =-=[49]-=-. We assume that this “expected” Y * XMWA distance is a sufficient statistic for the W 8 probability , that is Z([ W Y M %X W 2 (10) W The exact W 8 density is shown in Figure 3b. This density is a mi... |

1 | A probabilisticexclusion principle for tracking multiple objects - MacCormick, Blake - 1999 |

1 | Estimation and Tracking: principles, Techniques, andSoftware. YBS - Tracking, Press - 1998 |

1 | Estimating the absolute positionof a mobile robot using position probability grids - Burgard, Fox, et al. - 1996 |

1 | Monte carlo localization for mobilerobots - Dellaert, Fox, et al. - 1999 |

1 | Collaborative multi-robot localization.Autonomous Robots - Fox, Burgard, et al. - 1998 |

1 | Markov Chain MonteCarlo in Practice. Chapman and Hall/CRC - Gilks, Richardson, et al. - 1996 |

1 | Towards autonomous excavation - Boget - 1998 |

1 | Dynamic map building for anautonomous mobile robot - Leonard, Durrant-Whyte, et al. - 1992 |