## Fast Replanning for Navigation in Unknown Terrain (2002)

### Cached

### Download Links

Citations: | 24 - 8 self |

### BibTeX

@MISC{Koenig02fastreplanning,

author = {Sven Koenig and et al.},

title = {Fast Replanning for Navigation in Unknown Terrain},

year = {2002}

}

### OpenURL

### Abstract

Mobile robots often operate in domains that are only incompletely known, for example, when they have to move from given start coordinates to given goal coordinates in unknown terrain. In this case, they need to be able to replan quickly as their knowledge of the terrain changes. Stentz’ Focussed Dynamic A * (D*) is a heuristic search method that repeatedly determines a shortest path from the current robot coordinates to the goal coordinates while the robot moves along the path. It is able to replan faster than planning from scratch since it modifies its previous search results locally. Consequently, it has been extensively used in mobile robotics. In this article, we introduce an alternative to D * that determines the same paths and thus moves the robot in the same way but is algorithmically different. D * Lite is simple, can be rigorously analyzed, extendible in multiple ways, and is at least as efficient as D*. We believe that our results will make D*-like replanning methods even more popular and enable robotics researchers to adapt them to additional applications.

### Citations

2215 |
Robot Motion Planning
- Latombe
- 1991
(Show Context)
Citation Context ...ent continue to hold in these grids, that might be more realistic models of outdoor terrain. VII. RELATED WORK Path planning for goal-directed navigation in known terrain has been studied extensively =-=[24]-=-. Path planning for robot navigation in unknown terrain has been studied less frequently. We are interested in those navigation methods that repeatedly determine a shortest path from the current robot... |

728 |
Heuristics: Intelligent Search Strategies for Computer Problem Solving
- Pearl
- 1984
(Show Context)
Citation Context ...since it combines the efficiency of heuristic and incremental searches, yet—different from real-time heuristic search methods [6]—still finds shortest paths. It achieves a large speedup over repeated =-=[7]-=- searches by modifying previous search results locally. has been extensively used on real robots. This includes indoor Nomad robots [8] as well as outdoor HMMWVs and the UGV Demo II vehicles as part o... |

436 | Real-time heuristic search - Korf - 1990 |

238 | The parti-game algorithm for variable resolution reinforcement learning in multidimensional state spaces
- Moore, Atkeson
- 1995
(Show Context)
Citation Context ...ots [13]. Finally, versions of have been used to implement greedy mapping [14], a particular mapping method [8], and the parti-game method [15], a particular reinforcement-learning method for control =-=[16]-=-. However, is complex and thus hard to understand, analyze, and extend. For example, while has been widely used as a black-box method, it has not been extended by other researchers. Building on our Li... |

211 | Optimal and Efficient Path Planning for Partially-Known Environments
- Stentz
- 1994
(Show Context)
Citation Context ...enever it detects that its current path is untraversable. The robot could use conventional graph-search methods but this is inefficient since most edge costs do not change between replanning episodes =-=[4]-=-. The most popular solution to this problem is Focussed Dynamic [5] since it combines the efficiency of heuristic and incremental searches, yet—different from real-time heuristic search methods [6]—st... |

121 | An incremental algorithm for a generalization of the shortest-path problem
- Ramalingam, Reps
- 1996
(Show Context)
Citation Context ... Fig. 1. Illustration of the navigation strategy. Gray cells are cells with unknown blockage status. III. LIFELONG PLANNING Lifelong Planning [17] generalizes both [7] and a version of DynamicSWSF-FP =-=[18]-=-. It is an incremental heuristic search method that repeatedly determines shortest Fig. 2. Simple example (part 1). Gray cells are cells with changed goal distances. paths between two given vertices a... |

110 | A complete navigation system for goal acquisition in unknown environments
- Stentz, Herbert
- 1995
(Show Context)
Citation Context ...ts locally. has been extensively used on real robots. This includes indoor Nomad robots [8] as well as outdoor HMMWVs and the UGV Demo II vehicles as part of the DARPA Unmanned Ground Vehicle program =-=[9]-=-. It is currently also being integrated into Mars Rover prototypes and tactical mobile robot prototypes for urban reconnaissance [10]–[12]. is also used as part of other software, including the GRAMMP... |

74 | Improved fast replanning for robot navigation in unknown terrain
- Koenig, Likhachev
(Show Context)
Citation Context ...rrowed from [18]) . The rhs-value of a vertex is based on the g-values of its predecessors and thus potentially better informed than them. It always satisfies the following relationship (Invariant 1) =-=[21]-=-: if otherwise. A vertex is called locally consistent if , otherwise it is called locally inconsistent. If all vertices are locally consistent then all of their g-values are equal to their respective ... |

68 | Incremental Algorithms for Minimal Length Paths - Ausiello, Italiano, et al. - 1991 |

63 | Fully dynamic algorithms for maintaining all-pairs shortest paths and transitive closure in digraphs
- King
- 1999
(Show Context)
Citation Context ...ange over time [40]. If arbitrary sequences of edge insertions, deletions, or weight changes are allowed, then the dynamic shortest path problems are called fully dynamic shortest path problems [19], =-=[41]-=-, [42]. An example of an incremental search method for fully dynamic shortest path problems is DynamicSWSF-FP [18], a variant of which has been applied to hierarchical motion planning [43]. Ref. [44] ... |

54 | On the computational complexity of dynamic graph problems
- Ramalingam, Reps
- 1996
(Show Context)
Citation Context ... dynamic shortest path problems, that is, path problems where shortest paths between a given start and goal vertex have to be determined repeatedly as the topology of a graph or its edge costs change =-=[27]-=-. Examples include [28]–[39]. They often differ in their assumptions, for example, whether they solve single-source or all-pairs shortest path problems, which performance measure they use, when they u... |

51 | Fully dynamic algorithms for maintaining shortest paths trees - Frigioni, Marchetti-Spaccamela, et al. |

47 | Distributed robotic mapping of extreme environments - Thayer, Digney, et al. - 2000 |

45 | Updating distances in dynamic graphs - Even, Gazit - 1985 |

44 | GRAMMPS: A generalized mission planner for multiple mobile robots in unstructured environments
- Bummit, Stentz
- 1988
(Show Context)
Citation Context ...d into Mars Rover prototypes and tactical mobile robot prototypes for urban reconnaissance [10]–[12]. is also used as part of other software, including the GRAMMPS mission planner for multiple robots =-=[13]-=-. Finally, versions of have been used to implement greedy mapping [14], a particular mapping method [8], and the parti-game method [15], a particular reinforcement-learning method for control [16]. Ho... |

43 | On finding and updating spanning trees and shortest paths
- Spira, Pan
- 1975
(Show Context)
Citation Context ...ems, that is, path problems where shortest paths between a given start and goal vertex have to be determined repeatedly as the topology of a graph or its edge costs change [27]. Examples include [28]–=-=[39]-=-. They often differ in their assumptions, for example, whether they solve single-source or all-pairs shortest path problems, which performance measure they use, when they update the shortest paths, wh... |

39 | 2002b. Incremental A
- Koenig, Likhachev
(Show Context)
Citation Context ...nd thus hard to understand, analyze, and extend. For example, while has been widely used as a black-box method, it has not been extended by other researchers. Building on our Lifelong Planning method =-=[17]-=-, we therefore present Lite, a novel replanning method that determines the same paths as and thus moves the robot in the same way but is algorithmically different. is an incremental version of that ha... |

36 | Fully dynamic all pairs shortest paths with real edge weights
- Demetrescu, Italiano
- 2001
(Show Context)
Citation Context ...ver time [40]. If arbitrary sequences of edge insertions, deletions, or weight changes are allowed, then the dynamic shortest path problems are called fully dynamic shortest path problems [19], [41], =-=[42]-=-. An example of an incremental search method for fully dynamic shortest path problems is DynamicSWSF-FP [18], a variant of which has been applied to hierarchical motion planning [43]. Ref. [44] demons... |

33 | Lifelong planning A - Koenig, Likhachev, et al. - 2004 |

33 | Finding paths and deleting edges in directed acyclic graphs - Italiano - 1988 |

31 | A Portable, Autonomous, Urban Reconnaissance Robot - Matthies, Xiong, et al. - 2000 |

28 | Greedy mapping of terrain - Koenig, Tovey, et al. - 2001 |

27 |
Sensor based planning and nonsmooth analysis
- Choset, Burdick
- 1994
(Show Context)
Citation Context ...its map and then repeats the procedure, until it eventually reaches the goal coordinates or all paths to them are untraversable. This navigation strategy is an example of sensor-based motion planning =-=[2]-=-, [3]. If we model the navigation problem as a graph-traversal problem on an eight-connected grid with edges that are Manuscript received June 17, 2002; revised June 24, 2003. This paper was recommend... |

27 | Fully dynamic output bounded single source shortest path problem - Frigoni, Marchetti-Spaccamela, et al. - 1996 |

26 | An on-line edge deletion problem - Even, Shiloach - 1981 |

23 | Experimental analysis of dynamic algorithms for the single source shortest paths problem
- Frigioni, Lo®reda, et al.
- 1998
(Show Context)
Citation Context ... [41], [42]. An example of an incremental search method for fully dynamic shortest path problems is DynamicSWSF-FP [18], a variant of which has been applied to hierarchical motion planning [43]. Ref. =-=[44]-=- demonstrates experimentally the benefits of incremental searches over complete ones for repetitive planning tasks. Incremental search methods are typically uninformed, however, including all incremen... |

22 | A Dynamization of the All Pairs Least Cost Path Problem - Rohnert - 1985 |

22 | Semi-dynamic algorithms for maintaining single source shortest paths trees
- Frigioni, Marchetti-Spaccamela, et al.
- 1998
(Show Context)
Citation Context ...erformance measure they use, when they update the shortest paths, which kinds of graph topology and edge costs they apply to, and how the graph topology and edge costs are allowed to change over time =-=[40]-=-. If arbitrary sequences of edge insertions, deletions, or weight changes are allowed, then the dynamic shortest path problems are called fully dynamic shortest path problems [19], [41], [42]. An exam... |

20 |
Efficient search and hierarchical motion planning by dynamically maintaining single-source shortest paths trees
- Barbehenn, Hutchinson
- 1995
(Show Context)
Citation Context ...blems [19], [41], [42]. An example of an incremental search method for fully dynamic shortest path problems is DynamicSWSF-FP [18], a variant of which has been applied to hierarchical motion planning =-=[43]-=-. Ref. [44] demonstrates experimentally the benefits of incremental searches over complete ones for repetitive planning tasks. Incremental search methods are typically uninformed, however, including a... |

17 | A fully dynamic approximation scheme for shortest paths in planar graphs. Algorithmica 22 - PN, Subramanian - 1998 |

17 |
Differential a*: An adaptive search method illustrated with robot path planning for moving obstacles and goals, and an uncertain environment
- Trovato
- 1990
(Show Context)
Citation Context ...cause the robot moves in the terrain. Some of these search methods first identify the perimeter of areas in which the previous movement decisions need to get updated and restart the search from there =-=[45]-=-, [46]. Other search methods discover these areas while updating previous movement decisions. To the best of our knowledge, the only search methods that fit this description are (Focussed) [4] and the... |

14 | Experiments with driving modes for urban robots - Hebert, McLachlan, et al. - 1999 |

14 | Map-Based Strategies for Robot Navigation in Unknown Environments. in AAAI Spring Symposium on Planning with Incomplete Information for Robot Problems
- Stentz
- 1996
(Show Context)
Citation Context ...than complete search even though each cell expansion takes longer, which confirms earlier experimental results [5]. In a second set of experiments, we used fractal terrain, similar to the one used in =-=[23]-=-. All edges existed but their cost varied from 5 to 14 according to traversal difficulty of the cell they were entering. The robot initially assumed that all edges were present with cost 5. The robot ... |

14 | Dynamic algorithms for shortest paths in planar graphs - Feuerstein, Marchetti-Spaccamela - 1993 |

12 |
Path planning and navigation of mobile robots in unknown environments
- Ersson, Hu
- 2001
(Show Context)
Citation Context ...died how results from previous searches can be used to speed up the current search. Some of the approaches to this problem are based on minimum cost flow problems solved by the network simplex method =-=[25]-=-. Other approaches are based on graph search problems solved with either massively parallel search methods [26] or incremental search methods. Incremental search methods typically solve dynamic shorte... |

12 |
A new solution for path planning in partially known or unknown environment for nonholonomic mobile robots
- Podsedkowski, Nowakowski, et al.
- 2001
(Show Context)
Citation Context ...the robot moves in the terrain. Some of these search methods first identify the perimeter of areas in which the previous movement decisions need to get updated and restart the search from there [45], =-=[46]-=-. Other search methods discover these areas while updating previous movement decisions. To the best of our knowledge, the only search methods that fit this description are (Focussed) [4] and the first... |

10 | A new shortest path updating algorithm - Goto, Sangiovanni-Vincentelli - 1978 |

9 | On the dynamic shortest path problem - Lin, Chang - 1990 |

8 | et al, “A portable, autonomous, urban reconnaissance robot - Matthies - 2000 |

5 |
Path planning strategies for point automaton moving amidst unknown obstacles of arbitrary shape
- Lumelsky, Stepanov
- 1987
(Show Context)
Citation Context ...ile robot has to move from its current coordinates to given goal coordinates. Robotics researchers have investigated various navigation strategies to solve it, including the well-known bug algorithms =-=[1]-=-. In this paper, we study the following navigation strategy: The robot always plans a shortest path from its current coordinates to the goal coordinates under the assumption that unknown terrain is tr... |

4 | Optimal route re-planning for mobile robots: A massively parallel incremental A* algorithm - Tao, Elssamadisy, et al. - 1997 |

3 | Incremental replanning for mapping
- Likhachev, Koenig
(Show Context)
Citation Context ...urban reconnaissance [10]–[12]. is also used as part of other software, including the GRAMMPS mission planner for multiple robots [13]. Finally, versions of have been used to implement greedy mapping =-=[14]-=-, a particular mapping method [8], and the parti-game method [15], a particular reinforcement-learning method for control [16]. However, is complex and thus hard to understand, analyze, and extend. Fo... |

2 | Semi-dynamic breadth-first search in digraphs - Franciosa, Frigioni, et al. - 2001 |

2 |
et al., “Distributed robotic mapping of extreme environments
- Thayer, Digney, et al.
- 2000
(Show Context)
Citation Context ... vehicles as part of the DARPA Unmanned Ground Vehicle program [9]. It is currently also being integrated into Mars Rover prototypes and tactical mobile robot prototypes for urban reconnaissance [10]–=-=[12]-=-. is also used as part of other software, including the GRAMMPS mission planner for multiple robots [13]. Finally, versions of have been used to implement greedy mapping [14], a particular mapping met... |

1 |
planning, part I: incremental construction of the generalized Voronoi graph
- “Sensor-based
- 1995
(Show Context)
Citation Context ...ap and then repeats the procedure, until it eventually reaches the goal coordinates or all paths to them are untraversable. This navigation strategy is an example of sensor-based motion planning [2], =-=[3]-=-. If we model the navigation problem as a graph-traversal problem on an eight-connected grid with edges that are Manuscript received June 17, 2002; revised June 24, 2003. This paper was recommended fo... |

1 |
focussed D algorithm for real-time replanning
- “The
- 1995
(Show Context)
Citation Context ... could use conventional graph-search methods but this is inefficient since most edge costs do not change between replanning episodes [4]. The most popular solution to this problem is Focussed Dynamic =-=[5]-=- since it combines the efficiency of heuristic and incremental searches, yet—different from real-time heuristic search methods [6]—still finds shortest paths. It achieves a large speedup over repeated... |

1 | fast replanning for robot navigation in unknown terrain - “Improved |

1 | et al., “Greedy mapping of terrain - Koenig |

1 |
et al., “Experiments with driving modes for urban robots
- Hebert
- 1999
(Show Context)
Citation Context ...mo II vehicles as part of the DARPA Unmanned Ground Vehicle program [9]. It is currently also being integrated into Mars Rover prototypes and tactical mobile robot prototypes for urban reconnaissance =-=[10]-=-–[12]. is also used as part of other software, including the GRAMMPS mission planner for multiple robots [13]. Finally, versions of have been used to implement greedy mapping [14], a particular mappin... |

1 |
et al., “Fully dynamic algorithms for maintaining shortest paths trees
- Frigioni
- 2000
(Show Context)
Citation Context ...ate only those356 IEEE TRANSACTIONS ON ROBOTICS, VOL. 21, NO. 3, JUNE 2005 start distances (that is, distances from the start vertex to a vertex) that have changed or have not been calculated before =-=[19]-=-, and a heuristic search tends to recalculate only those start distances that are relevant for recalculating a shortest path from the start vertex to the goal vertex [7]. thus recalculates only very f... |