Loading...

Design and Implementation of a Planning-specific Sat Solver

Hamidian Shoormasti, Sina | 2018

483 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 52190 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Ghassem-Sani, Gholamreza
  7. Abstract:
  8. Automated planning is a branch of Artificial intelligence aimed at obtaining plans (i.e. sequences of actions) for solving complex problems or for governing the behavior of intelligent agents. Reduction of planning problems to satisfiability problems is one of the most successful approaches to automated planning. In this method first the planning problem with a preset length is coded to a satisfiability problem. Then, a general SAT solver is used for solving the obtained SAT formula. To find the plan with optimum length, plans with increasing coding lengths are evaluated by a SAT solver in a sequential manner. Since much of descriptive information of planning problems is lost in converting to SAT formula, general SAT solvers solve the formula without utilizing these information. In this thesis, it has been tried to provide utilization of these information. Some of the pieces of information that solvers ignore are the layered structure of SAT formulae and sequential solving of SAT formula. One of suggested methods in this M.Sc. project, provides the application of information derived from a layer of the SAT formula for being used in another layer. This information is usually added to the formula for pruning the search space. The presented method transfers these clauses to other layers of the formula and therefore prunning larger space. Also according to the sequential nature of solving formulae, clauses obtained from solving a formula can be used in other formulae. Ideas proposed in this thesis were implemented on the ITSAT planner, which has the capability of planning for classical and temporal planning problems. Empirical results confirm some positive effects of these methods in accelerating solution of SAT formulae
  9. Keywords:
  10. Satisfiability-based Planning ; Artificial Intelligence Planning ; Automatic Planning ; Satisfiability

 Digital Object List

 Bookmark

No TOC