{"_id":"kLpDK4xKunNbpANSX","bibbaseid":"sievers-wehrle-helmert-ananalysisofmergestrategiesformergeandshrinkheuristics","downloads":0,"creationDate":"2016-03-09T03:04:31.949Z","title":"An Analysis of Merge Strategies for Merge-and-Shrink Heuristics","author_short":["Sievers, S.","Wehrle, M.","Helmert, M."],"year":null,"bibtype":"inproceedings","biburl":"icaps16.icaps-conference.org/papers.bib","bibdata":{"bibtype":"inproceedings","type":"inproceedings","track":"Main Track","title":"An Analysis of Merge Strategies for Merge-and-Shrink Heuristics","url":"http://www.aaai.org/ocs/index.php/ICAPS/ICAPS16/paper/view/13148","author":[{"firstnames":["Silvan"],"propositions":[],"lastnames":["Sievers"],"suffixes":[]},{"firstnames":["Martin"],"propositions":[],"lastnames":["Wehrle"],"suffixes":[]},{"firstnames":["Malte"],"propositions":[],"lastnames":["Helmert"],"suffixes":[]}],"abstract":"The merge-and-shrink framework provides a powerful basis for the computation of admissible planning heuristics. Recently, it has been shown theoretically that non-linear strategies are strictly more powerful than linear ones, and several non-linear strategies have been proposed and implemented. However, it is still unclear what makes up for a \"good\" merge strategy, and in particular, how much of the theoretical potential of non-linear strategies is actually exploited by the existing ones. In this paper, we address these questions on the accuracy of existing merge strategies. We provide an extensive experimental analysis, suggesting that there is still room for large improvements. Finally, we present a simple non-linear strategy that considerably improves over the efficiency of the existing merge strategies.","keywords":"Classical planning,Search techniques","bibtex":"@inproceedings {icaps16-118,\r\n track = {Main Track},\r\n title = {An Analysis of Merge Strategies for Merge-and-Shrink Heuristics},\r\n url = {http://www.aaai.org/ocs/index.php/ICAPS/ICAPS16/paper/view/13148},\r\n author = {Silvan Sievers and Martin Wehrle and Malte Helmert},\r\n abstract = {The merge-and-shrink framework provides a powerful basis for the\r\ncomputation of admissible planning heuristics. Recently, it has been\r\nshown theoretically that non-linear strategies are strictly more\r\npowerful than linear ones, and several non-linear strategies have\r\nbeen proposed and implemented. However, it is still unclear what\r\nmakes up for a \"good\" merge strategy, and in particular, how much\r\nof the theoretical potential of non-linear strategies is actually\r\nexploited by the existing ones. In this paper, we address these\r\nquestions on the accuracy of existing merge strategies. We provide\r\nan extensive experimental analysis, suggesting that there is still\r\nroom for large improvements. Finally, we present a simple non-linear\r\nstrategy that considerably improves over the efficiency of the\r\nexisting merge strategies.},\r\n keywords = {Classical planning,Search techniques}\r\n}\r\n\r\n","author_short":["Sievers, S.","Wehrle, M.","Helmert, M."],"key":"icaps16-118","id":"icaps16-118","bibbaseid":"sievers-wehrle-helmert-ananalysisofmergestrategiesformergeandshrinkheuristics","role":"author","urls":{"Paper":"http://www.aaai.org/ocs/index.php/ICAPS/ICAPS16/paper/view/13148"},"keyword":["Classical planning","Search techniques"],"metadata":{"authorlinks":{}},"downloads":0,"html":""},"search_terms":["analysis","merge","strategies","merge","shrink","heuristics","sievers","wehrle","helmert"],"keywords":["classical planning","search techniques"],"authorIDs":[],"dataSources":["iMkx859KiXcegwsin","EZtZjCTnxcdTTyeij"]}