Loading...

Improving Representation and Search in Concurrent Temporal Planning

Feyzbakhsh Rankouh, Masoud | 2016

451 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 48679 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Ghassem Sani, Gholamreza
  7. Abstract:
  8. Temporal planning is one of the branches of AI planning, which has attracted a considerable amount of attention in recent years due to the importance of temporal properties in real world planning problems. Like in many other AI fields, there exists an important trade-off in temporal planning between the variety of solvable problems and the efficiency of the planner. Most of the previously proposed temporal planners use heuristic state-space search method. This method, despite being quite efficient, is not complete for so-called “problems with required concurrency”. On the other hand, plan-space based and satisfiability based temporal planners, which can tackle problems with required concurrency, have not been competitive with state-space based planners from the speed point of view. The main goal of this thesis is to design an efficient temporal planner capable of solving problems with required concurrency. To reach this goal, we employ satisfiability based approach to produce concurrent plans. We introduce several methods for translating a given temporal planning problem to a satisfiability problem. These methods are based on the separation of causal and temporal reasoning phases of solving a temporal problem. We first abstract out the temporal information of the given problem and try to find a plan that is only causally valid. We then try to schedule the actions of this causally valid plan to find a temporally valid one. We show that the separation of causal and temporal reasoning enables us to employ compact encodings that are based on the -step and -step semantics of parallel plans. We also present two preprocessing methods for mutual exclusion relation extraction and action compression. We show how such information can be utilized to enhance the efficiency of searching for valid plans. All the methods introduced in this thesis are implemented as a temporal planner named ITSAT. Our empirical results show that not only ITSAT outperforms the state-of-the-art temporally expressive planners, it is also competitive with the fast temporal planners that cannot handle required concurrency
  9. Keywords:
  10. Temporal Planning ; Artificial Intelligence Planning ; Constraint Satisfaction ; Required Concurrency ; Satisfiability-based Planning

 Digital Object List

 Bookmark

No TOC