Limited assignments: a new cutoff strategy for incomplete depth-first search

In this paper, we propose an extension of three incomplete depthfirst search techniques, namely depth-bounded backtrack search, credit search, and iterative broadening, towards producing incomplete solutions. We also propose a new cutoff strategy for incomplete depth-first search motivated by a human style of problem solving. This technique, called limited assignment number (LAN) search, is based on limiting the number of attempts tried to assign a value to the variable. A linear worstcase time complexity of LAN Search leads to promising stable time behavior in all accomplished experiments. The techniques are studied in the context of constraint satisfaction problems. Categories and Subject Descriptors I.2.8 [Artificial Intelligence]: Problem Solving, Control Methods, and Search – backtracking, graph and tree search strategies. General Terms Algorithms. Performance. Experimentation. Keywords Search. Constraint satisfaction.

• Type: Conference
• Year: 2005
• Where: SAC
• Authors: Roman Barták, Hana Rudová
• Keywords: Applied Computing, Depth-bounded Backtrack Search, Incomplete Depth-first Search, Incomplete Depthfirst Search, SAC 2005