## A Continuous Query System for Dynamic Route Planning

Citations: | 5 - 0 self |

### BibTeX

@MISC{Malviya_acontinuous,

author = {Nirmesh Malviya and Samuel Madden and Arnab Bhattacharya},

title = {A Continuous Query System for Dynamic Route Planning},

year = {}

}

### OpenURL

### Abstract

Abstract—In this paper, we address the problem of answering continuous route planning queries over a road network, in the presence of updates to the delay (cost) estimates of links. A simple approach to this problem would be to recompute the best path for all queries on arrival of every delay update. However, such a naive approach scales poorly when there are many users who have requested routes in the system. Instead, we propose two new classes of approximate techniques – K-paths and proximity measures to substantially speed up processing of the set of designated routes specified by continuous route planning queries in the face of incoming traffic delay updates. Our techniques work through a combination of precomputation of likely good paths and by avoiding complete recalculations on every delay update, instead only sending the user new routes when delays change significantly. Based on an experimental evaluation with 7,000 drives from real taxi cabs, we found that the routes delivered by our techniques are within 5 % of the best shortest path and have run times an order of magnitude or less compared to a naive approach. I.

### Citations

980 |
A formal basis for the heuristic determination of minimum cost paths
- Hart, Nilsson, et al.
- 1968
(Show Context)
Citation Context ... able to scale our system to potentially hundreds of thousands of users (leading to hundreds of thousands of standing routing queries). For the same reason, hill climbing algorithms such as A* search =-=[5]-=- which employ heuristics for search pruning do not serve our purpose if they are merely used for complete recomputations on every update. Dynamic variants of A* search such as dynamic anytime A* [6] a... |

98 | Computing the shortest path: A* search meets graph theory
- GOLDBERG, HARRELSON
(Show Context)
Citation Context ...t edges to reduce nodes’s reaches during preprocessing and though useful for routing in large road networks, it overlooks real time delays and is thus not applicable to our problem. Landmark Indexing =-=[19]-=- and Transit Routing [20] speed up run time query performance by using precomputed distances between certain set of landmarks (or transit nodes) chosen according to sophisticated algorithms. As with h... |

71 | Anytime Dynamic A*: An Anytime, Replanning Algorithm
- Likhachev, Ferguson, et al.
- 2005
(Show Context)
Citation Context ...h [5] which employ heuristics for search pruning do not serve our purpose if they are merely used for complete recomputations on every update. Dynamic variants of A* search such as dynamic anytime A* =-=[6]-=- and life long planning A* [7] also do not help us due to scalability issues in using them at run time for a large set of registered routes. The performance of 1:n dynamic shortest path algorithms tha... |

68 | A new approach to dynamic all pairs shortest paths
- Demetrescu, Italiano
(Show Context)
Citation Context ...would need to set this bounding constant to a large value to take the variability in traffic conditions into account, making the solution unattractive. Demetrescu and Italiano’s dynamic APSP approach =-=[11]-=- uses combinatorial properties of graphs to guarantee O(n 2 log 3 n) amortized time per update but the O(mn) space requirement makes their solution unattractive (here m is the number of edges in the g... |

63 | Fully dynamic algorithms for maintaining all-pairs shortest paths and transitive closure in digraphs
- King
- 1999
(Show Context)
Citation Context ...)[2], making their use impractical for large road networks and given the high number and frequency of updates we need to process. For instance, King’s dynamic all-pairs shortest path (APSP) algorithm =-=[10]-=- supports approximate shortest path queries in an efficient O(n 2 log 2 n/ log log n) amortized time but is applicable only when all edge weights are integers bounded above by a small constant. We wou... |

59 | Reach for A ∗ : Efficient Point-to-Point Shortest Path Algorithms
- Goldberg, Kaplan, et al.
- 2006
(Show Context)
Citation Context ...oute Planner Delay DB Maps transportation planning systems is largely based on graph preprocessing under the assumption of static conditions to speed up query response times [17]. Reach based routing =-=[18]-=- improves query performance by adding shortcut edges to reduce nodes’s reaches during preprocessing and though useful for routing in large road networks, it overlooks real time delays and is thus not ... |

58 | CCAM : A Connectivity-Clustered Access Method for Aggregate Queries on Transportation Networks
- Shekhar, Liu
- 1995
(Show Context)
Citation Context ...egies [15] for hierarchical routing algorithms also offer a balance between increased space usage and reduced query time and are considered to be important in querying large spatial network databases =-=[16]-=-. But as discussed earlier, any scheme which merely returns precomputed results (without taking into account dynamic edge costs) defeats the purpose of our system. The current application of shortest ... |

55 | Exact and Approximate Distances in Graphs – A Survey
- Zwick
- 2001
(Show Context)
Citation Context ...Section III, and our techniques in Section IV and V. We experimentally evaluate our system in Section VI and conclude in Section VII. II. RELATED WORK Dijkstra’s algorithm has traditionally been used =-=[3]-=- for route planning in road networks. Though extremely efficient implementations of Dijkstra’s algorithm exist [4], even the tens of milliseconds of computation time they take per route translates int... |

48 |
Fully dynamic algorithms for maintaining shortest paths trees
- Frigioni, Marchetti-Spaccamela, et al.
(Show Context)
Citation Context ...e pre-calculation of any input, its running time is governed by the number of affected nodes which can be large as we need to maintain shortest paths from many source nodes. Frigioni’s DSSS algorithm =-=[9]-=- also requires a prohibitive shortest path tree computation. Dynamic all-pairs-shortest-path algorithms have a space and time overheads(per delay update) of at least O(n 2 ) and O(n 2 log n)[2], makin... |

35 | Experimental analysis of dynamic all pairs shortest path algorithms
- Demetrescu, Italiano
(Show Context)
Citation Context ...ntinuous routing queries is relatively small, it is unlikely to scale as the number of queries go up. Algorithms that are able to update shortest paths in the presence of link changes do exist (e.g., =-=[2]-=-), but they typically have a higher space or computation overhead than is acceptable for our setting (see Section II for a discussion). In this paper, we discuss several new techniques for maintenance... |

35 |
Bounded incremental computation
- Ramalingam
- 1993
(Show Context)
Citation Context ... in comparison to fromscratch recomputations of their static counterparts. However, they have other limitations that make them unusable in our setting. For example, Ramalingam’s incremental algorithm =-=[8]-=- for the dynamic single-source shortest (DSSS) path requires in-memory storage of a pre-calculated shortest-path-tree from every source node. In the setting we consider, this translates to a prohibiti... |

29 | Lifelong planning a
- Koenig, Likhachev, et al.
- 2004
(Show Context)
Citation Context ...for search pruning do not serve our purpose if they are merely used for complete recomputations on every update. Dynamic variants of A* search such as dynamic anytime A* [6] and life long planning A* =-=[7]-=- also do not help us due to scalability issues in using them at run time for a large set of registered routes. The performance of 1:n dynamic shortest path algorithms that update routes from a given s... |

28 | Materialization trade-offs in hierarchical shortest path algorithms, in:Symposium on Large Spatial Databases
- Shekhar, Fetterer, et al.
- 1997
(Show Context)
Citation Context ...s precomputation of paths from all source nodes only up to a certain distance or breaking the underlying graph into a set of several fragmented graphs and a boundary graph. Materialization strategies =-=[15]-=- for hierarchical routing algorithms also offer a balance between increased space usage and reduced query time and are considered to be important in querying large spatial network databases [16]. But ... |

27 | Fast Shortest Path Distance Estimation in Large Networks
- Potamias, Bonchi, et al.
- 2009
(Show Context)
Citation Context ...hitecture of our system. Route Planner Delay DB Maps transportation planning systems is largely based on graph preprocessing under the assumption of static conditions to speed up query response times =-=[17]-=-. Reach based routing [18] improves query performance by adding shortcut edges to reduce nodes’s reaches during preprocessing and though useful for routing in large road networks, it overlooks real ti... |

26 | Hierarchical optimization of optimal path finding for transportation applications
- Jing, Huang, et al.
- 1996
(Show Context)
Citation Context ...conditions are not taken into account. A precomputation based solution to APSP problem has a significant space overhead but this can be reduced by using approaches based on hierarchical decomposition =-=[14]-=- of the road network, such as precomputation of paths from all source nodes only up to a certain distance or breaking the underlying graph into a set of several fragmented graphs and a boundary graph.... |

20 | Adaptive fastest path computation on a road network: A traffic mining approach
- Gonzalez, Han, et al.
- 2007
(Show Context)
Citation Context ... queries registered with the system, and our proximity approach runs in worst case O(r(u + (m + n log n)) for a batch of delay updates of size u. Work on adaptive fastest path computation by Gonzalez =-=[12]-=- uses historical data to mine traffic patterns at different points of time in an attempt to take non-concrete parameters about different road segments into account. Only historical information is used... |

18 | Finding fastest paths on a road network with speed patterns
- Kanoulas, Du, et al.
- 2006
(Show Context)
Citation Context ...e in an attempt to take non-concrete parameters about different road segments into account. Only historical information is used in their algorithm with no real time traffic monitoring. Kanoulas et al =-=[13]-=- use historical data to get a different average speed for each hour of the day and uses that to compute the best route for a given day and time. Again, real-time traffic conditions are not taken into ... |

14 | Implementations of Dijkstra's Algorithm Based on Multi-Level Buckets
- Goldberg, Silverstein
- 1995
(Show Context)
Citation Context ...ude in Section VII. II. RELATED WORK Dijkstra’s algorithm has traditionally been used [3] for route planning in road networks. Though extremely efficient implementations of Dijkstra’s algorithm exist =-=[4]-=-, even the tens of milliseconds of computation time they take per route translates into an untenable solution for our problem given that our ultimate aim is to be able to scale our system to potential... |