#### DMCA

## Towards Autonomous Topological Place Detection Using the Extended Voronoi Graph,” (2005)

### Cached

### Download Links

Venue: | Proceeding of IEEE ICRA, |

Citations: | 53 - 9 self |

### Citations

1295 |
The Image of the City
- Lynch
- 1960
(Show Context)
Citation Context ...ph in Section VI. We conclude with a summary of this work. II. DEFINING TOPOLOGICAL PLACES Topological navigation is a behavior that is used by a variety of different animal species, including humans =-=[5]-=-– [7]. Topological representations discretize the continuous world into a finite set of places connected by paths. This helps facilitate large-scale spatial reasoning, mainly due to the compactness of... |

470 | A robot exploration and mapping strategy based on a semantic hierarchy of spatial representations, Robotics and Autonomous Systems 8(1-2): 47–63
- Kuipers, Byun
- 1991
(Show Context)
Citation Context ...terion, a place should be defined at this location; however, no junction point exists. The appropriate control at this point is coastal navigation, keeping at least one obstacle in sight at all times =-=[21]-=-, [22]. Instead of defining paths parallel to the walls, the Voronoi graph defines a path that will quickly bring the robot to a location where no visible obstacles exist. Once this occurs, the Vorono... |

405 |
Sensor fusion in certainty grids for mobile robots
- Moravec
- 1988
(Show Context)
Citation Context ...ven simple environments [15]. A reliably detected set of well-separated but connected places leads to efficient Fig. 1. The generalized Voronoi graph of a global metrical map. Using an occupancy grid =-=[19]-=-, [20], the Voronoi graph can be drawn by using any non-free cells as obstacles. Junction points include intersections in corridor environments. map-building, while unreliable, non-deterministic place... |

262 |
The development of spatial representations of large-scale environments.
- Siegel, White
- 1975
(Show Context)
Citation Context ...using familiar highway intersections as markers for specific turn actions. It has long been known that humans use intersections as a basis for building spatial representations in unknown environments =-=[9]-=-—preferring first to build structural models of novel environments prior to detailed visual models [10]. Occasionally people may define places that are not at intersections. Dead-ends are one example ... |

223 | Topological simultaneous localization and mapping (SLAM): Toward exact localization without explicit localization
- Choset, Nagatani
- 2001
(Show Context)
Citation Context ... variation and sensor noise. III. PLACES AT VORONOI JUNCTIONS Choset has been a main proponent for the use of Voronoi graphs in topological map-building and navigation [16], [17]. Choset and Nagatani =-=[18]-=- define a generalized Voronoi graph (GVG) as a one-dimensional set of points equidistant to the n closest obstacles in n dimensions, where a preset threshold determines whether observations belong to ... |

194 | Inference of finite automata using homing sequences,
- Rivest, Schapire
- 1989
(Show Context)
Citation Context ...locations. Theoretically, if place detection is deterministic, topological map-building can be viewed as finding the minimal deterministic finite automaton that explains the experience of places [2], =-=[13]-=-, [14]. When place detection is not deterministic, topological map-building can be viewed as a partially observable Markov decision process, which needs a large amount of experience to accurately mode... |

164 |
Occupancy Grids: A Probabilistic Framework for Robot Perception and Navigation
- Elfes
- 1989
(Show Context)
Citation Context ...mple environments [15]. A reliably detected set of well-separated but connected places leads to efficient Fig. 1. The generalized Voronoi graph of a global metrical map. Using an occupancy grid [19], =-=[20]-=-, the Voronoi graph can be drawn by using any non-free cells as obstacles. Junction points include intersections in corridor environments. map-building, while unreliable, non-deterministic place detec... |

135 | Learning topological maps with weak local odometric information
- Shatkay, Kaelbling
- 1997
(Show Context)
Citation Context ...n is not deterministic, topological map-building can be viewed as a partially observable Markov decision process, which needs a large amount of experience to accurately model even simple environments =-=[15]-=-. A reliably detected set of well-separated but connected places leads to efficient Fig. 1. The generalized Voronoi graph of a global metrical map. Using an occupancy grid [19], [20], the Voronoi grap... |

94 | Bootstrap learning for place recognition,” in
- Kuipers, Beeson
- 2002
(Show Context)
Citation Context ...have described (a) how places, along with paths and regions, can form topological maps [2], [3], and (b) how an agent can learn highly-accurate place recognition (global localization) from experience =-=[4]-=-. Here, we consider the more fundamental problem of place detection: how a finite set of discrete places is abstracted from experience in a continuous environment. Spatial knowledge is undoubtedly hie... |

92 | Prototypes, location, and associative networks (PLAN): Towards a unified theory of cognitive mapping. - Chown, Kaplan, et al. - 1995 |

83 | Towards a general theory of topological maps
- Remolina, Kuipers
- 2004
(Show Context)
Citation Context ... and by an IBM Faculty Research Award. This paper investigates the idea of defining places. In previous work, we have described (a) how places, along with paths and regions, can form topological maps =-=[2]-=-, [3], and (b) how an agent can learn highly-accurate place recognition (global localization) from experience [4]. Here, we consider the more fundamental problem of place detection: how a finite set o... |

65 | Local metrical and global topological maps in the hybrid spatial semantic hieararchy
- Kuipers, Modayil, et al.
- 2004
(Show Context)
Citation Context ...ory experience to low-level, symbolic spatial constructs, we are developing a foundation for the high-level common-sense knowledge necessary for an intelligent agent to act in environments with people and other animals. ∗Research of the Intelligent Robotics lab is supported in part by grants from the National Science Foundation (IIS-0413257), from the National Institutes of Health (EY016089), and by an IBM Faculty Research Award. This paper investigates the idea of defining places. In previous work, we have described (a) how places, along with paths and regions, can form topological maps [2], [3], and (b) how an agent can learn highly-accurate place recognition (global localization) from experience [4]. Here, we consider the more fundamental problem of place detection: how a finite set of discrete places is abstracted from experience in a continuous environment. Spatial knowledge is undoubtedly hierarchical, but we leave the issue of hierarchical spatial abstraction to future work, focusing instead on the detection of the smallest, atomic places necessary for low-level abstraction of experience to symbols. This paper is organized as follows. In Section II, we outline the two criteria ... |

60 | Sensor-based exploration: The hierarchical generalized Voronoi graph.
- CHOSET, BURDICK
- 2000
(Show Context)
Citation Context ...but more robust to environmental variation and sensor noise. III. PLACES AT VORONOI JUNCTIONS Choset has been a main proponent for the use of Voronoi graphs in topological map-building and navigation =-=[16]-=-, [17]. Choset and Nagatani [18] define a generalized Voronoi graph (GVG) as a one-dimensional set of points equidistant to the n closest obstacles in n dimensions, where a preset threshold determines... |

58 | Bootstrap learning for object discovery.
- Modayil, Kuipers
- 2004
(Show Context)
Citation Context ...elf completely enclosed by occupied cells in large-scale environments. Exceptions could be elevators or small rooms where doors get closed. In these instances, our LPM should detect the dynamic doors =-=[23]-=-, allowing the algorithm to ignore these as obstacles for the purposes of building the Voronoi graph. Thus, there is always some region of free cells that touch the edge of the occupancy grid or some ... |

52 | Coastal navigation mobile robot navigation with uncertainty in dynamic environments.
- Roy, Burgard, et al.
- 1999
(Show Context)
Citation Context ..., a place should be defined at this location; however, no junction point exists. The appropriate control at this point is coastal navigation, keeping at least one obstacle in sight at all times [21], =-=[22]-=-. Instead of defining paths parallel to the walls, the Voronoi graph defines a path that will quickly bring the robot to a location where no visible obstacles exist. Once this occurs, the Voronoi grap... |

45 |
Two strategies for learning a route in a driving simulator,”
- Aginsky, Harris, et al.
- 1997
(Show Context)
Citation Context ...tions or deadends do not yield any additional structural information. In fact, as people do start to remember landmarks, they are biased to first learn landmarks at critical path changes along routes =-=[11]-=-, [12]. 2) Places must be reliably detectable. Although this criterion seems simple, it is extremely important. Qualitatively interesting locations should be considered as possible places when they ex... |

33 | Coping with uncertainty in map learning.
- Basye, Dean, et al.
- 1989
(Show Context)
Citation Context ...ons. Theoretically, if place detection is deterministic, topological map-building can be viewed as finding the minimal deterministic finite automaton that explains the experience of places [2], [13], =-=[14]-=-. When place detection is not deterministic, topological map-building can be viewed as a partially observable Markov decision process, which needs a large amount of experience to accurately model even... |

22 | Honey bee cognition,” - Gould - 1990 |

21 | Sensorbased exploration: incremental construction of the hierarchical generalized Voronoi graph”,
- Choset, Walker, et al.
- 2000
(Show Context)
Citation Context ...re robust to environmental variation and sensor noise. III. PLACES AT VORONOI JUNCTIONS Choset has been a main proponent for the use of Voronoi graphs in topological map-building and navigation [16], =-=[17]-=-. Choset and Nagatani [18] define a generalized Voronoi graph (GVG) as a one-dimensional set of points equidistant to the n closest obstacles in n dimensions, where a preset threshold determines wheth... |

18 |
Metaphors We Live By. Chicago: The Univ.
- Lakoff, Johnson
- 1980
(Show Context)
Citation Context ...something is out of place.” The broad use of places is a specific example of a crucial point in the study of intelligent agents—spatial knowledge is a foundation for high-level common-sense knowledge =-=[1]-=-. We believe that by studying the problem of grounding continuous sensory experience to low-level, symbolic spatial constructs, we are developing a foundation for the high-level common-sense knowledge... |

17 | Positional entropy during pigeon homing II: navigational interpretation of bayesian latent state models,”
- Guilford, Roberts, et al.
- 2004
(Show Context)
Citation Context ...The majority of the time, this criterion defines places at intersections. The importance of path intersections in topological representations cannot be overstated. The recent work of Guilford et al. =-=[8]-=-, demonstrates that pigeons, long considered to use Earth’s magnetic field to fly long distances, follow man-made highway networks in wellknown environments, using familiar highway intersections as ma... |

11 |
Local metrical and global topological maps
- Kuipers, Modayil, et al.
- 2004
(Show Context)
Citation Context ...by an IBM Faculty Research Award. This paper investigates the idea of defining places. In previous work, we have described (a) how places, along with paths and regions, can form topological maps [2], =-=[3]-=-, and (b) how an agent can learn highly-accurate place recognition (global localization) from experience [4]. Here, we consider the more fundamental problem of place detection: how a finite set of dis... |

8 |
Way-marking behaviour: an aid to spatial navigation in the wood mouse (Apodemus sylvaticus),”
- Stopka, Macdonald
- 2003
(Show Context)
Citation Context ... Section VI. We conclude with a summary of this work. II. DEFINING TOPOLOGICAL PLACES Topological navigation is a behavior that is used by a variety of different animal species, including humans [5]– =-=[7]-=-. Topological representations discretize the continuous world into a finite set of places connected by paths. This helps facilitate large-scale spatial reasoning, mainly due to the compactness of the ... |

7 |
Positional entropy during pigeon homing II: navigational interpretation of bayesian latent state models
- Guilford, Roberts, et al.
- 2004
(Show Context)
Citation Context ...sThe majority of the time, this criterion defines places at intersections. The importance of path intersections in topological representations cannot be overstated. The recent work of Guilford et al. =-=[8]-=-, demonstrates that pigeons, long considered to use Earth’s magnetic field to fly long distances, follow man-made highway networks in wellknown environments, using familiar highway intersections as ma... |

3 |
The role of environmental features in route learning: Two exploratory studies of way finding,”
- Heft
- 1979
(Show Context)
Citation Context ...or deadends do not yield any additional structural information. In fact, as people do start to remember landmarks, they are biased to first learn landmarks at critical path changes along routes [11], =-=[12]-=-. 2) Places must be reliably detectable. Although this criterion seems simple, it is extremely important. Qualitatively interesting locations should be considered as possible places when they exist, b... |

1 |
Acquisition of object and structural landmark knowledge in virtual environment navigation,” in Object Perception and Memory,
- Kalia, Stankiewicz
- 2002
(Show Context)
Citation Context ... humans use intersections as a basis for building spatial representations in unknown environments [9]—preferring first to build structural models of novel environments prior to detailed visual models =-=[10]-=-. Occasionally people may define places that are not at intersections. Dead-ends are one example of important places that are not necessarily at an intersection of two or more paths. Less useful place... |