## Particle Filters for Mobile Robot Localization (2001)

Citations: | 95 - 18 self |

### BibTeX

@MISC{Fox01particlefilters,

author = {Dieter Fox and Sebastian Thrun and Wolfram Burgard and Frank Dellaert},

title = {Particle Filters for Mobile Robot Localization},

year = {2001}

}

### Years of Citing Articles

### OpenURL

### Abstract

This article describes a family of methods, known as Monte Carlo localization (MCL) (Dellaert at al. 1999b, Fox et al. 1999b). The MCL algorithm is a particle filter combined with probabilistic models of robot perception and motion. Building on this, we will describe a variation of MCL which uses a different proposal distribution (a mixture distribution) that facilitates fast recovery from global localization failures. As we will see, this proposal distribution has a range of advantages over that used in standard MCL, but it comes at the price that it is more difficult to implement, and it requires an algorithm for sampling poses from sensor measurements, which might be difficult to obtain. Finally, we will present an extension of MCL to cooperative multi-robot localization of robots that can perceive each other during localization. All these approaches have been tested thoroughly in practice. Experimental results are provided to demonstrate their relative strengths and weaknesses in practical robot applications.

### Citations

666 | Robust Monte Carlo localization for mobile robots
- Thrun, Fox, et al.
- 2001
(Show Context)
Citation Context ...d-based representations, which previously was among the best known algorithms for the global localization problem. Similar results were obtained using a camera as the primary sensor for localization (=-=Dellaert et al. 1999-=-a). To test MCL under extreme circumstances, we evaluated it using data collected in a populated museum. During a twoweek exhibition, our robot Minerva (Figure 3) was employed as a tour-guide in the S... |

559 | Filtering via simulation: auxiliary particle filters - Pitt, Shepherd - 1999 |

519 |
An analysis of timedependent planning
- Dean, Boddy
- 1998
(Show Context)
Citation Context ... of all poses, particle lters are computationally e cient, since they focus their resources on regions in state space with high likelihood. Particle lters are also easily implementedasany-time lters (=-=Dean and Boddy 1988-=-, Zilberstein and Russell 1995), by dynamically adapting the number of samples based on the available computational resources. Finally, particle lters for localization are remarkably easy to implement... |

502 | Sequential Monte Carlo methods for dynamic systems
- Liu, Chen
- 1998
(Show Context)
Citation Context ...the perceptual model in navigating with a vision sensor. 3 MCL with Mixture Proposal Distributions 3.1 The Need For Better Sampling As noticed by several authors (Doucet 1998, Lenser and Veloso 2000, =-=Liu and Chen 1998-=-, Pitt and Shephard 1999), the basic particle lter performs poorly if the proposal distribution, which is used to generate samples, places too little samples in regions where the desired posterior Bel... |

432 | Carlo filter and smoother for non-Gaussian nonlinear state space models - Kitagawa, “Monte - 1996 |

307 | Markov Localization for Mobile Robots in Dynamic Environments
- Fox, Burgard, et al.
- 1999
(Show Context)
Citation Context ... localization are remarkably easy to implement, which also contributes to their popularity. This article describes a family of methods, known as Monte Carlo localization (MCL) (Dellaert at al. 1999b, =-=Fox et al. 1999-=-b). The MCL algorithm is a particle lter combined with probabilistic models of robot perception and motion. Building on this, we will describe a variation of MCL which uses a di erent proposal distrib... |

296 | Monte carlo localization: Efficient position estimation for mobile robots - Fox, Burgard, et al. - 1999 |

269 | Tractable inference for complex stochastic processes
- Boyen, Koller
- 1998
(Show Context)
Citation Context ...encies and instead represents the belief at time t by the product of its marginals Bel(xt) = NY i=1 Bel(x i t ) (4.1) Thus, our representation e ectively makes a (false) independence assumption| see (=-=Boyen and Koller 1998-=-) for an idea howtoovercome this independence assumption while still avoiding the exponential death of the full product space. When a robot detects another robot, the observation is folded into a robo... |

263 | Probabilistic robot navigation in partially observable environments
- Simmons, Koenig
- 1995
(Show Context)
Citation Context ...robot localization, Bayes lters are also known as Markov localization (Burgard, Fox, Hennig and Schmidt 1996, Fox at al. 1999a, Kaelbling et al. 1996, Koenig and Simmons 1996, Nourbakhsh et al. 1995, =-=Simmons and Koenig 1995-=-, Thrun 1998). To implement Markov localization, one needs to know three distributions: the initial belief Bel(x 0) (e.g., uniform), the next state probability p(xt j xt;1�ut;1) (called the motion mod... |

223 | On sequential simulation-based methods for Bayesian filtering
- Doucet
- 1998
(Show Context)
Citation Context ... American History, which was used as the perceptual model in navigating with a vision sensor. 3 MCL with Mixture Proposal Distributions 3.1 The Need For Better Sampling As noticed by several authors (=-=Doucet 1998-=-, Lenser and Veloso 2000, Liu and Chen 1998, Pitt and Shephard 1999), the basic particle lter performs poorly if the proposal distribution, which is used to generate samples, places too little samples... |

192 |
An experiment in guidance and navigation of an autonomous robot vehicle
- Cox, Blanche
- 1991
(Show Context)
Citation Context ...us chapters in (Borenstein et al. 1996) and (Kortenkamp et al. 1998). Occasionally, it has been referred to as \the most fundamental problem to providing a mobile robot with autonomous capabilities" (=-=Cox 1991-=-). The mobile robot localization problem comes in di erent avors. The simplest localization problem|which has received by far the most attention in the literature|is position tracking. Here the initia... |

192 | A Probabilistic Approach to Collaborative Multirobotic
- Fox
(Show Context)
Citation Context ...aluated for the kidnapped robot problem in the Smithsonian museum. The middle curve re ects the performance of MCL with a small number of random samples added in the resampling step, as suggested in (=-=Fox et al. 2000-=-) as a means to recover from localization failures. The error rate is measured in percentage of time during which the robot lost track of its position. sense each other during localization. The abilit... |

189 |
Dynamic map building for an autonomous mobile robot
- Leonard, Durrant-Whyte, et al.
- 1992
(Show Context)
Citation Context ...problem is akey problem in mobile robotics, as it plays a fundamental role in various successful mobile robot systems� see e.g., (Cox and Wilfong 1990, Fukuda et al. 1993, Hinkel and Knieriemen 1988, =-=Leonard et al. 1992-=-, Rencken 1993, Simmons et al. 1997, Wei et al. 1994) and various chapters in (Borenstein et al. 1996) and (Kortenkamp et al. 1998). Occasionally, it has been referred to as \the most fundamental prob... |

189 |
Tools for Statistical Inference
- Tanner
- 1996
(Show Context)
Citation Context ...de ned in Equation (2.4), and uses the importance factors (2.6) to account for the di erence. It is well-known from the statistical literature (Doucet 1998, Pitt and Shephard 1999, Liu and Chen 1998, =-=Tanner 1993-=-) that the divergence between (2.5) and (2.4) determines the convergence speed. This di erence is accounted by the perceptual density p(yt j xt): If the sensors are entirely uninformative, this distri... |

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

170 |
DERVISH: An office-navigating robot
- Nourbakhsh, Powers, et al.
- 1995
(Show Context)
Citation Context ...n the context of mobile robot localization, Bayes lters are also known as Markov localization (Burgard, Fox, Hennig and Schmidt 1996, Fox at al. 1999a, Kaelbling et al. 1996, Koenig and Simmons 1996, =-=Nourbakhsh et al. 1995-=-, Simmons and Koenig 1995, Thrun 1998). To implement Markov localization, one needs to know three distributions: the initial belief Bel(x 0) (e.g., uniform), the next state probability p(xt j xt;1�ut;... |

140 | Sensor resetting localization for poorly modeled mobile robots
- Lenser, Veloso
- 2000
(Show Context)
Citation Context ...tory, which was used as the perceptual model in navigating with a vision sensor. 3 MCL with Mixture Proposal Distributions 3.1 The Need For Better Sampling As noticed by several authors (Doucet 1998, =-=Lenser and Veloso 2000-=-, Liu and Chen 1998, Pitt and Shephard 1999), the basic particle lter performs poorly if the proposal distribution, which is used to generate samples, places too little samples in regions where the de... |

133 | Navigating Mobile Robots: Systems and Techniques - Borenstein, Everett, et al. - 1996 |

122 | Bayesian landmark learning for mobile robot localization
- Thrun
- 1998
(Show Context)
Citation Context ... lters are also known as Markov localization (Burgard, Fox, Hennig and Schmidt 1996, Fox at al. 1999a, Kaelbling et al. 1996, Koenig and Simmons 1996, Nourbakhsh et al. 1995, Simmons and Koenig 1995, =-=Thrun 1998-=-). To implement Markov localization, one needs to know three distributions: the initial belief Bel(x 0) (e.g., uniform), the next state probability p(xt j xt;1�ut;1) (called the motion model), and the... |

121 | Using the condensation algorithm for robust, vision-based mobile robot localization
- Dellaert, Burgard, et al.
- 1999
(Show Context)
Citation Context ... localization are remarkably easy to implement, which also contributes to their popularity. This article describes a family of methods, known as Monte Carlo localization (MCL) (Dellaert at al. 1999b, =-=Fox et al. 1999-=-b). The MCL algorithm is a particle lter combined with probabilistic models of robot perception and motion. Building on this, we will describe a variation of MCL which uses a di erent proposal distrib... |

111 | Efficient memory-based learning for robot control. Doctoral dissertation
- Moore
- 1990
(Show Context)
Citation Context ...hen sampling from p(x (j) t these samples approximate p(x (j) t j ut;1�x (j) t;1) as described above. Obviously, j d 0:::t;1�ut;1). 2. Transform the resulting sample set into a kd-tree (Bentley 1980, =-=Moore 1990-=-). The tree generalizes samples to arbitrary poses x (j) t in pose space, which is necessary to calculate the desired importance factors. 3. Finally, sample x (i) t from our proposal distribution p(yt... |

110 |
Minerva: A second generation mobile tour-guide robot
- Thrun, Bennewitz, et al.
- 1999
(Show Context)
Citation Context ... in a populated museum. During a twoweek exhibition, our robot Minerva (Figure 3) was employed as a tour-guide in the Smithsonian's Museum of Natural History, during which ittraversed more than 44km (=-=Thrun et al. 1999-=-). To aid localization, Minerva is equipped with a camera pointed towards the ceiling. Using this camera, the brightnesss480 Fox, Thrun, Burgard & Dellaert Average estimation error [cm] 30 25 20 15 10... |

92 |
Acting under uncertainty: Discrete Bayesian models for mobile-robot navigation
- Kaelbling, Cassandra, et al.
- 1996
(Show Context)
Citation Context ...ere. 2.2 Models of Robot Motion and Perception In the context of mobile robot localization, Bayes lters are also known as Markov localization (Burgard, Fox, Hennig and Schmidt 1996, Fox at al. 1999a, =-=Kaelbling et al. 1996-=-, Koenig and Simmons 1996, Nourbakhsh et al. 1995, Simmons and Koenig 1995, Thrun 1998). To implement Markov localization, one needs to know three distributions: the initial belief Bel(x 0) (e.g., uni... |

89 | A layered architecture for office delivery robots - Simmons, Goodwin, et al. - 1997 |

81 | Efficient locally weighted polynomial regression predictions
- Moore, Schneider, et al.
- 1997
(Show Context)
Citation Context ...ets, and with probability onenotwo samples in these sets are the same. To solve this problem, our approach transforms sample sets into density functions using density trees (Koller and Fratkina 1998, =-=Moore et al. 1997-=-, Omohundro 1991). Density trees are continuations of sample sets which approximate the underlying density using a variable-resolution piecewise constant density. Figure 16 shows such a tree, which co... |

79 |
Autonomous Robot Vehicles
- Cox, Wilfong
- 1990
(Show Context)
Citation Context ...a robot's pose relative to a map of its environment. The localization problem is akey problem in mobile robotics, as it plays a fundamental role in various successful mobile robot systems� see e.g., (=-=Cox and Wilfong 1990-=-, Fukuda et al. 1993, Hinkel and Knieriemen 1988, Leonard et al. 1992, Rencken 1993, Simmons et al. 1997, Wei et al. 1994) and various chapters in (Borenstein et al. 1996) and (Kortenkamp et al. 1998)... |

67 |
Multidimensional divide and conquer
- Bentley
- 1980
(Show Context)
Citation Context ...(j) t;1) and then sampling from p(x (j) t these samples approximate p(x (j) t j ut;1�x (j) t;1) as described above. Obviously, j d 0:::t;1�ut;1). 2. Transform the resulting sample set into a kd-tree (=-=Bentley 1980-=-, Moore 1990). The tree generalizes samples to arbitrary poses x (j) t in pose space, which is necessary to calculate the desired importance factors. 3. Finally, sample x (i) t from our proposal distr... |

62 |
Active Markov localization for mobile robots. Robotics and Autonomous Systems
- Fox, Burgard, et al.
- 1998
(Show Context)
Citation Context ...age of particle lters over alternative representations, we are particularly interested in grid-based representations, which are at the core of an alternative family of Markov localization algorithms (=-=Fox et al. 1998-=-). The algorithm described in (Fox et al. 1998) relies on a ne-grained grid approximation of the belief Bel(), using otherwise identical sensor and motion models. Figure 5 plots the localization accur... |

61 |
Error correction in mobile robot map learning
- Engelson, McDermott
- 1992
(Show Context)
Citation Context ...ead has to determine it from scratch. The global localization problem is more di cult, since the robot's localization error can be arbitrarily large. Even more di cult is the kidnapped robot problem (=-=Engelson and McDermott 1992-=-), in which awell-localized robot is teleported to some other position without being told. This problem di ers from the global localization problem in that the robot might rmly believe to be somewhere... |

59 | Monte Carlo localization with mixture proposal distribution
- Thrun, Fox, et al.
(Show Context)
Citation Context ...e Proposal Distribution To alleviate this problem, one can use a di erent proposal distribution, one that samples according to the most recent sensor measurement yt (see also (Lenser and Veloso 2000, =-=Thrun et al. 2000-=-)). The key idea is to sample xt directly from a distribution that is proportional to the perceptual likelihood p(yt j xt): q := p(yt j xt) (yt) with (yt) = Z p(yt j xt) dxt (3.1) This new proposal di... |

58 |
AIbased Mobile Robots: Case studies of successful robot systems
- Kortenkamp, Bonasso, et al.
- 1998
(Show Context)
Citation Context ..., (Cox and Wilfong 1990, Fukuda et al. 1993, Hinkel and Knieriemen 1988, Leonard et al. 1992, Rencken 1993, Simmons et al. 1997, Wei et al. 1994) and various chapters in (Borenstein et al. 1996) and (=-=Kortenkamp et al. 1998-=-). Occasionally, it has been referred to as \the most fundamental problem to providing a mobile robot with autonomous capabilities" (Cox 1991). The mobile robot localization problem comes in di erent ... |

56 | Using learning for approximation in stochastic processes
- Koller, Fratkina
- 1998
(Show Context)
Citation Context ...es represented by sample sets, and with probability onenotwo samples in these sets are the same. To solve this problem, our approach transforms sample sets into density functions using density trees (=-=Koller and Fratkina 1998-=-, Moore et al. 1997, Omohundro 1991). Density trees are continuations of sample sets which approximate the underlying density using a variable-resolution piecewise constant density. Figure 16 shows su... |

53 | Keeping track of position and orientation of moving indoor systems by correlation of range-finder scans - Weiss, Wetzler, et al. - 1994 |

51 | Passive distance learning for robot navigation
- Koenig, Simmons
- 1996
(Show Context)
Citation Context ...t Motion and Perception In the context of mobile robot localization, Bayes lters are also known as Markov localization (Burgard, Fox, Hennig and Schmidt 1996, Fox at al. 1999a, Kaelbling et al. 1996, =-=Koenig and Simmons 1996-=-, Nourbakhsh et al. 1995, Simmons and Koenig 1995, Thrun 1998). To implement Markov localization, one needs to know three distributions: the initial belief Bel(x 0) (e.g., uniform), the next state pro... |

48 | Backward simulation in Bayesian networks - Fung, Favero - 1994 |

46 |
Navigating Mobile Robots
- Borenstein, Everett, et al.
- 1996
(Show Context)
Citation Context ...obile robot systems� see e.g., (Cox and Wilfong 1990, Fukuda et al. 1993, Hinkel and Knieriemen 1988, Leonard et al. 1992, Rencken 1993, Simmons et al. 1997, Wei et al. 1994) and various chapters in (=-=Borenstein et al. 1996-=-) and (Kortenkamp et al. 1998). Occasionally, it has been referred to as \the most fundamental problem to providing a mobile robot with autonomous capabilities" (Cox 1991). The mobile robot localizati... |

46 | Bumptrees for efficient function, constraint, and classification learning - Omohundro - 1991 |

43 |
Concurrent localisation and map building for mobile robots using ultrasonic sensors
- Rencken
- 1993
(Show Context)
Citation Context ...em in mobile robotics, as it plays a fundamental role in various successful mobile robot systems� see e.g., (Cox and Wilfong 1990, Fukuda et al. 1993, Hinkel and Knieriemen 1988, Leonard et al. 1992, =-=Rencken 1993-=-, Simmons et al. 1997, Wei et al. 1994) and various chapters in (Borenstein et al. 1996) and (Kortenkamp et al. 1998). Occasionally, it has been referred to as \the most fundamental problem to providi... |

17 |
Navigation system based on ceiling landmark recognition for autonomous mobile robot
- Fukuda, Ito, et al.
- 1993
(Show Context)
Citation Context ...e to a map of its environment. The localization problem is akey problem in mobile robotics, as it plays a fundamental role in various successful mobile robot systems� see e.g., (Cox and Wilfong 1990, =-=Fukuda et al. 1993-=-, Hinkel and Knieriemen 1988, Leonard et al. 1992, Rencken 1993, Simmons et al. 1997, Wei et al. 1994) and various chapters in (Borenstein et al. 1996) and (Kortenkamp et al. 1998). Occasionally, it h... |

10 |
Bumptrees for E cient Function, Constraint, and Classi cation Learning
- Omohundro
- 1991
(Show Context)
Citation Context ...bility onenotwo samples in these sets are the same. To solve this problem, our approach transforms sample sets into density functions using density trees (Koller and Fratkina 1998, Moore et al. 1997, =-=Omohundro 1991-=-). Density trees are continuations of sample sets which approximate the underlying density using a variable-resolution piecewise constant density. Figure 16 shows such a tree, which corresponds to a r... |

8 | Mosaicing a large number of widely dispersed, noisy, and distorted images: A Bayesian approach
- Dellaert, Thorpe, et al.
- 1999
(Show Context)
Citation Context ...d-based representations, which previously was among the best known algorithms for the global localization problem. Similar results were obtained using a camera as the primary sensor for localization (=-=Dellaert et al. 1999-=-a). To test MCL under extreme circumstances, we evaluated it using data collected in a populated museum. During a twoweek exhibition, our robot Minerva (Figure 3) was employed as a tour-guide in the S... |

8 |
Monte Carlo lter and smoother for non-Gaussian nonlinear state space models
- Kitagawa
- 1996
(Show Context)
Citation Context ...pproximate) sample-based representation. Obviously, our algorithm constitutes just one possible implementation of the particle ltering idea� other sampling schemes exist that further reduce variance (=-=Kitagawa 1996-=-). Detailed convergence results can be found in Chapters 2 and 3 of this book. Further below, it will be convenient to notice that in this version of MCL, the proposal distribution for approximating B... |

7 |
A layered architecture for o ce delivery robots
- Simmons, Goodwin, et al.
- 1997
(Show Context)
Citation Context ...obotics, as it plays a fundamental role in various successful mobile robot systems� see e.g., (Cox and Wilfong 1990, Fukuda et al. 1993, Hinkel and Knieriemen 1988, Leonard et al. 1992, Rencken 1993, =-=Simmons et al. 1997-=-, Wei et al. 1994) and various chapters in (Borenstein et al. 1996) and (Kortenkamp et al. 1998). Occasionally, it has been referred to as \the most fundamental problem to providing a mobile robot wit... |

7 | Keeping track of position and orientation of moving indoor systems by correlation of range-finder scans - Wei��, Wetzler, et al. - 1994 |

5 |
Monte Carlo Localization: E cient Position Estimation for Mobile Robots
- Fox, Burgard, et al.
- 1999
(Show Context)
Citation Context ... localization are remarkably easy to implement, which also contributes to their popularity. This article describes a family of methods, known as Monte Carlo localization (MCL) (Dellaert at al. 1999b, =-=Fox et al. 1999-=-b). The MCL algorithm is a particle lter combined with probabilistic models of robot perception and motion. Building on this, we will describe a variation of MCL which uses a di erent proposal distrib... |

1 |
Particle Filters for Mobile Robot Localization 497
- Hinkel, Knieriemen
- 1988
(Show Context)
Citation Context ...vironment. The localization problem is akey problem in mobile robotics, as it plays a fundamental role in various successful mobile robot systems� see e.g., (Cox and Wilfong 1990, Fukuda et al. 1993, =-=Hinkel and Knieriemen 1988-=-, Leonard et al. 1992, Rencken 1993, Simmons et al. 1997, Wei et al. 1994) and various chapters in (Borenstein et al. 1996) and (Kortenkamp et al. 1998). Occasionally, it has been referred to as \the ... |

1 |
Filtering via simulation: auxiliary particle lter
- Fox, Burgard, et al.
- 1999
(Show Context)
Citation Context |