SandersSchultes_Highway Hierarchies Hasten Exact shortest Path Queries.pdf

下载文档

7
0
约3.52万字
约 12页
2017-04-09 发布于江苏
举报
版权申诉
保障服务

SandersSchultes_Highway Hierarchies Hasten Exact shortest Path Queries.pdf

1、本文档共12页，可阅读全部内容。
2、原创力文档（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。
3、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。

Sanders

Highway Hierarchies Hasten Exact Shortest Path Queries Peter Sanders1, and Dominik Schultes1,2 1 Universita?t Karlsruhe (TH), 76128 Karlsruhe, Germany sanders@ira.uka.de 2 Universita?t des Saarlandes mail@dominik-schultes.de Abstract. We present a new speedup technique for route planning that exploits the hierarchy inherent in real world road networks. Our algo- rithm preprocesses the eight digit number of nodes needed for maps of the USA or Western Europe in a few hours using linear space. Shortest (i.e. fastest) path queries then take around eight milliseconds to pro- duce exact shortest paths. This is about 2 000 times faster than using Dijkstra’s algorithm. 1 Introduction Computing shortest paths in graphs (networks) with nonnegative edge weights is a classical problem of computer science. From a worst case perspective, the problem has largely been solved by Dijkstra in 1959 [1] who gave an algorithm that finds all shortest paths from a starting node s using at most m+n priority queue operations for a graph G = (V,E) with n nodes and m edges. However, motivated by important applications (e.g., in transportation net- works), there has recently been considerable interest in the problem of acceler- ating shortest path queries, i.e., the problem to find a shortest path between a source node s and a target node t. In this case, Dijkstra’s algorithm can stop as soon as the shortest path to t is found. A classical technique that gives a constant factor speedup is bidirectional search which simultaneously searches forward from s and backwards from t until the search frontiers meet. All further speedup techniques either need additional information (e.g., geometry information for goal directed search) or precomputa- tion. There is a trade-off between the time needed for precomputation, the space needed for storing the precomputed information, and the resulting query time. In Section 1 we review existing precomputation approaches, which have made significant progress, but