## Path Planning Using Lazy PRM (2000)

Venue: | In IEEE Int. Conf. Robot. & Autom |

Citations: | 202 - 14 self |

### BibTeX

@INPROCEEDINGS{Bohlin00pathplanning,

author = {Robert Bohlin and Lydia E. Kavraki},

title = {Path Planning Using Lazy PRM},

booktitle = {In IEEE Int. Conf. Robot. & Autom},

year = {2000},

pages = {521--528}

}

### Years of Citing Articles

### OpenURL

### Abstract

This paper describes a new approach to probabilistic roadmap planners (PRMs). The overall theme of the algorithm, called Lazy PRM, is to minimize the number of collision checks performed during planning and hence minimize the running time of the planner. Our algorithm builds a roadmap in the configuration space, whose nodes are the user-defined initial and goal configurations and a number of randomly generated nodes. Neighboring nodes are connected by edges representing paths between the nodes. In contrast with PRMs, our planner initially assumes that all nodes and edges in the roadmap are collision-free, and searches the roadmap at hand for a shortest path between the initial and the goal node. The nodes and edges along the path are then checked for collision. If a collision with the obstacles occurs, the corresponding nodes and edges are removed from the roadmap. Our planner either finds a new shortest path, or first updates the roadmap with new nodes and edges, and then searches for a shortest path. The above process is repeated until a collision-free path is returned.

### Citations

2171 |
Robot Motion Planning
- Latombe
- 1991
(Show Context)
Citation Context ..., the problem can be stated as follows: given an initial configuration qinit and a goal configuration qgoal in :c, find a continuous curve in :c connecting these points, or determine that none exists =-=[20]-=-. Path planning is becoming increasingly important for automated manufacturing and for mobile robots, but it has also found applications in computer animations, medical surgery, and molecular biology ... |

714 |
Applied multivariate statistical analysis
- Johnson, Wichern
- 1982
(Show Context)
Citation Context ...incesis diagonal, this simply means that each component qi, i = 1,..., d, of q is normally distributed with mean /i and variance i,i' To show (2), we use that (q-j )*-(q-j ) is distributed with d dof =-=[16]-=-. Thus, the event (q--J)'S-(q--j ) _ X() has probability 1 - a. Using (1) and (3), gives the confidence ellipsoid in (2). If we choose a = 0.05, then the parameter A controls the size of the 95% confi... |

509 |
The Complexity of Robot Motion Planning
- Canny
- 1987
(Show Context)
Citation Context ... proposed; see [10, 14, 20] for overviews. An algorithm is called complete if it always will find a solution or determine that none exists. However, due to the complexity of the path planning problem =-=[8]-=-, complete planners are far too slow to be useful in practice. Trading completeness for speed, probabilistic planners have been successfully applied to many problems in high-dimensional configuration ... |

456 | Randomized kinodynamic planning
- LaValle, Kuffner
(Show Context)
Citation Context ...ed component of the roadmap. See also [9] and [15] for related algorithms. The latter paper gives an adaptive scheme for adjusting the power of the local planner. Other methods, described in [13] and =-=[22]-=-, build two trees rooted at the initial and goal configurations respectively. As soon as the two trees intersect, a feasible path can be extracted. In [13], the trees are expanded by generating new no... |

353 |
Robot motion planning: A distributed representation approach
- Barraquand, Latombe
- 1989
(Show Context)
Citation Context ...eful in practice. Trading completeness for speed, probabilistic planners have been successfully applied to many problems in high-dimensional configuration spaces. The Randomized Path Planner (RPP) in =-=[4]-=- has successfully solved problems for robots with more than 60 degrees of freedom [19]. The planner uses a potential field as a guidance towards the goal, and random walks to escape local minima. Anot... |

236 | OBPRM: An obstacle-based PRM for 3D workspaces
- Amato, Bayazit, et al.
- 1998
(Show Context)
Citation Context ...ths, and are therefore inevitable. Solutions for the cases of cluttered configuration spaces and fast collision-checking are also proposed, but neither these cases nor the narrow passage problem (see =-=[2, 6, 12]-=-) are our main objectives. Our algorithm, called Lazy PRM, is described in detail in Section 3, and experimentally evaluated in Section 4 using a real industrial environment. 2 Probabilistic Technique... |

231 | Path planning in expansive configuration spaces
- Hsu, Latombe, et al.
- 1997
(Show Context)
Citation Context ...e connected component of the roadmap. See also [9] and [15] for related algorithms. The latter paper gives an adaptive scheme for adjusting the power of the local planner. Other methods, described in =-=[13]-=- and [22], build two trees rooted at the initial and goal configurations respectively. As soon as the two trees intersect, a feasible path can be extracted. In [13], the trees are expanded by generati... |

173 | On finding narrow passages with probabilistic roadmap planners
- Hsu, Kavraki, et al.
- 1998
(Show Context)
Citation Context ...ths, and are therefore inevitable. Solutions for the cases of cluttered configuration spaces and fast collision-checking are also proposed, but neither these cases nor the narrow passage problem (see =-=[2, 6, 12]-=-) are our main objectives. Our algorithm, called Lazy PRM, is described in detail in Section 3, and experimentally evaluated in Section 4 using a real industrial environment. 2 Probabilistic Technique... |

170 |
der Stappen, “The Gaussian sampling strategy for probabilistic roadmap planners
- Boor, Overmars, et al.
- 1999
(Show Context)
Citation Context ...ths, and are therefore inevitable. Solutions for the cases of cluttered configuration spaces and fast collision-checking are also proposed, but neither these cases nor the narrow passage problem (see =-=[2, 6, 12]-=-) are our main objectives. Our algorithm, called Lazy PRM, is described in detail in Section 3, and experimentally evaluated in Section 4 using a real industrial environment. 2 Probabilistic Technique... |

168 |
Gross motion planning. a survey
- Hwang, Ahuja
- 1992
(Show Context)
Citation Context ... using a real industrial environment. 2 Probabilistic Techniques The path planning problem has been extensively studied in the last two decades, and a number of different approaches are proposed; see =-=[10, 14, 20]-=- for overviews. An algorithm is called complete if it always will find a solution or determine that none exists. However, due to the complexity of the path planning problem [8], complete planners are ... |

163 | A fast algorithm for incremental distance calculation
- Lin, Canny
- 1991
(Show Context)
Citation Context ...eed to check the nodes and edges along the path for collision. The edges are discretised and checked with a certain resolution, so our algorithm only requires a collision checker for points in C; see =-=[7, 23, 28]-=-. The overall purpose of the Search, Check, and Remove steps of our algorithm (Figure 1), is roughly to identify and remove colliding nodes and edges of the roadmap until the shortest path between qin... |

159 | Maprm: A probabilistic roadmap planner with sampling on the medial axis of the free space
- Wilmarth, Amato, et al.
- 1998
(Show Context)
Citation Context ...point close to the first one. One of the points is added to the roadmap only if it is insand its pair is not. Another technique to increase the number of nodes in narrow passages of ' is presented in =-=[29]-=-; points are picked uniformly at random insand then retracted onto the medial axis of . A few methods using probabilistic roadmaps, do not divide the planning process into a learning phase and a query... |

153 | Randomized preprocessing of configuration space for fast path planning
- Kavraki, Latombe
- 1994
(Show Context)
Citation Context ...ues. In particular, the Probabilistic Roadmap Method [18] is described in detail, since the method forms the base of our solution. The idea behind the Probabilistic Roadmap Method (PRM), described in =-=[17, 18, 27]-=-, is to represent and capture the connectivity ofsby a random network, a roadmap, whose nodes and edges correspond to randomly selected configurations and path segments respectively. In a preprocessin... |

136 | A randomized roadmap method for path and manipulation planning
- Amato, Wu
- 1996
(Show Context)
Citation Context ...obstacles to a certain extent. Small neighborhoods around the configurations just in collision are then re-sampled in order to place nodes close to the boundary of . The Obstacle Based PRM (OBPRM) in =-=[2, 3]-=- determines configurations in collision to be origins of a number of rays. Binary search is then used along each ray to find points on the boundary of , where roadmap nodes are placed. In [6], the pla... |

133 | Planning motions with intentions
- KOGA, KONDO, et al.
- 1994
(Show Context)
Citation Context ...ccessfully applied to many problems in high-dimensional configuration spaces. The Randomized Path Planner (RPP) in [4] has successfully solved problems for robots with more than 60 degrees of freedom =-=[19]-=-. The planner uses a potential field as a guidance towards the goal, and random walks to escape local minima. Another interesting approach is presented in [24] - the Ariadne's clew algorithm. Consider... |

116 | A probabilistic learning approach to motion planning
- Overmars, Sv̌estka
- 1994
(Show Context)
Citation Context ...ues. In particular, the Probabilistic Roadmap Method [18] is described in detail, since the method forms the base of our solution. The idea behind the Probabilistic Roadmap Method (PRM), described in =-=[17, 18, 27]-=-, is to represent and capture the connectivity ofsby a random network, a roadmap, whose nodes and edges correspond to randomly selected configurations and path segments respectively. In a preprocessin... |

93 | Choosing good distance metrics and local planners for probabilistic roadmap methods
- Amato, Bayazit, et al.
- 2000
(Show Context)
Citation Context ...ely and use a relatively weak, but fast, local planner [18, 27]. The local planner may for instance only check the straight line between two nodes. Other local planners are discussed and evaluated in =-=[1]-=-. Often the learning phase of PRM has a node enhancement step in order to increase the connectivity of the roadmap by adding more nodes in difficult regions of . Different techniques are used to ident... |

80 | The Ariadne's Clew algorithm
- Mazer, Ahuactzin, et al.
- 1998
(Show Context)
Citation Context ...ots with more than 60 degrees of freedom [19]. The planner uses a potential field as a guidance towards the goal, and random walks to escape local minima. Another interesting approach is presented in =-=[24]-=- - the Ariadne's clew algorithm. Considering the initial configuration as a landmark, the planner incrementally builds a tree of feasible (i.e., collision-free) paths as follows. Genetic optimization ... |

77 | A two level fuzzy PRM for manipulation planning
- Nielsen, Kavraki
- 2000
(Show Context)
Citation Context ...be feasible. The idea is to lazily evaluate the feasibility of the roadmap as planning queries are processed. Similar ideas about lazy evaluation have been developed concurrently but independently in =-=[25]-=- in a planner called Fuzzy PRM. In other words, let qinit, q9oal, and a number of uniformly distributed points form nodes in a roadmap, and connect by edges each pair of nodes being sufficiently close... |

62 | Motion planning: a journey of robots, molecules, digital actors, and other
- Latombe
- 1999
(Show Context)
Citation Context ...h planning is becoming increasingly important for automated manufacturing and for mobile robots, but it has also found applications in computer animations, medical surgery, and molecular biology (see =-=[21]-=-). Requirements for Single Query Path Planning Of particular interest are planners that with little preprocessing can answer single queries very quickly. Such planners can be used to re-plan paths in ... |

54 | A Random Approach to Motion Planning
- Overmars
- 1992
(Show Context)
Citation Context ... the medial axis of . A few methods using probabilistic roadmaps, do not divide the planning process into a learning phase and a query phase. Given an initial and a goal configuration, the planner in =-=[26]-=- inserts randomly distributed nodes in , one at a time, and connects them to the different components of the roadmap by a local planner. New nodes are inserted until the initial and goal configuration... |

44 | Probabilistic Roadmaps for Fast Path Planning in High-Dimensional Configuration Spaces - Kavraki, Svestka, et al. - 1996 |

43 |
Motion planning for many degrees of freedom — random reflections at c-space obstacles
- Horsch, Tolle
- 1994
(Show Context)
Citation Context ...lection can be based on a node's ratio of failed attempts by the local planner to find paths to other nodes [18]. Other techniques to increase the connectivity of the roadmap are described in [2] and =-=[11]-=-. PRM has been shown to work well in practice in highdimensional configuration spaces [17]. In particular, it is useful for multiple queries, since once an adequate roadmap has been created, queries c... |

26 |
Solving findpath by combination of goal-directed and randomized search
- Glavina
- 1990
(Show Context)
Citation Context ...m to the different components of the roadmap by a local planner. New nodes are inserted until the initial and goal configurations can be found in the same connected component of the roadmap. See also =-=[9]-=- and [15] for related algorithms. The latter paper gives an adaptive scheme for adjusting the power of the local planner. Other methods, described in [13] and [22], build two trees rooted at the initi... |

25 |
Interference detection between non-convex polyhedra revisited with a practical aim
- Thomas, Torras
- 1994
(Show Context)
Citation Context ...eed to check the nodes and edges along the path for collision. The edges are discretised and checked with a certain resolution, so our algorithm only requires a collision checker for points in C; see =-=[7, 23, 28]-=-. The overall purpose of the Search, Check, and Remove steps of our algorithm (Figure 1), is roughly to identify and remove colliding nodes and edges of the roadmap until the shortest path between qin... |

17 |
Practical motion planning in robotics
- Gupta, P
- 1998
(Show Context)
Citation Context ... using a real industrial environment. 2 Probabilistic Techniques The path planning problem has been extensively studied in the last two decades, and a number of different approaches are proposed; see =-=[10, 14, 20]-=- for overviews. An algorithm is called complete if it always will find a solution or determine that none exists. However, due to the complexity of the path planning problem [8], complete planners are ... |

12 | Path planning in expansive con spaces - Hsu, Latombe, et al. |

10 | A Two-level Search Algorithm for Motion Planning
- Isto
(Show Context)
Citation Context ... different components of the roadmap by a local planner. New nodes are inserted until the initial and goal configurations can be found in the same connected component of the roadmap. See also [9] and =-=[15]-=- for related algorithms. The latter paper gives an adaptive scheme for adjusting the power of the local planner. Other methods, described in [13] and [22], build two trees rooted at the initial and go... |

8 | Randomized preprocessing of con space for fast path planning - Kavraki, Latombe - 1994 |

5 |
Motion Planning for Industrial Robots. Licentiate thesis
- Bohlin
- 1999
(Show Context)
Citation Context ...odes in this order is that the probability of having the shortest feasible path via a particular node is higher if the node is close to either qinit or qgoal, SO we want to check end-nodes first; see =-=[5]-=- for a discussion. Checking Edges If all nodes along the path are in jc, we start checking the edges in a similar fashion; working from the outside in. However, to minimize the risk of doing unnecessa... |

3 |
Enhancing GJK: Computing minimum distance and penetration distanses between convex polyhedra
- Cameron
- 1997
(Show Context)
Citation Context ...eed to check the nodes and edges along the path for collision. The edges are discretised and checked with a certain resolution, so our algorithm only requires a collision checker for points in C; see =-=[7, 23, 28]-=-. The overall purpose of the Search, Check, and Remove steps of our algorithm (Figure 1), is roughly to identify and remove colliding nodes and edges of the roadmap until the shortest path between qin... |

2 | Solving by combination of goal-directed and randomized search - Glavina - 1990 |