计算机科学 ›› 2010, Vol. 37 ›› Issue (4): 192-.

• 人工智能 • 上一篇    下一篇

基于命题可满足性的经典最优规划方法

吕帅,刘磊,江鸿,魏唯   

  1. (吉林大学计算机科学与技术学院 长春130012)
  • 出版日期:2018-12-01 发布日期:2018-12-01
  • 基金资助:
    本文受国家自然科学基金(60603031,60773097,60873044)和教育部高等学校博十学科点专项科研基金(20060183044,20070183057)资助。

Classical Optimal Planning Method Based on Propositional Satisfiability

LU Shuai,LIU Lei,JIANG Hong,WEI Wei   

  • Online:2018-12-01 Published:2018-12-01

摘要: 基于Graphplan的编码方式是2006年国际规划竟赛中著名的最优规划系统SATPLAN2006采用的编码方式。首先给出与编码相关的概念与性质,在基于Graphplan的编码方式的基础上,设计一种新的编码方式:基于FA的编码方式,并从理论上证明该编码方式的有效性。设计并实现对应的规划系统FA-SP,利用国际规划竞赛选用的Benchmark问题予以测试。实验结果表明,与SATPLAN2006相比,FA-SP对于所测两类规划域编码规模均有所压缩,除个别问题外求解效率都有一定程度的提高;对于顺序规划域Bloc

关键词: 基于可满足性的规划,基于Graphplan的编码,编码,框架公理,规划系统

Abstract: Graphplan-based encoding is a novel encoding adopted by SATPLAN2006,which is the statcof-thcart optimal planner of International Planning Competition in 2006. It firstly defined the concepts and characters about encoding.Based on Graphplan-based encoding, it proposed a new encoding, called FA-based encoding, by reducing the mutex axioms of actions and appending frame axioms. It justifies the soundness and completeness of the corresponding proposed encoding. We designed and implemented the corresponding planner called FA-SP, and tested it with benchmarks adopted by International Planning Competition, respectively. Comparing with SATPLAN2006,the results validate th at for both all two domains tested, the scales of encodings of FA-SP are compacted and solver efficiencies of them arc improved, and it is notes to sec that for sequential planning problems Blocks World, the scales of encodings of FA-SP are compacted nearly 40% and solver efficiencies of them are improved nearly 2 times; for parallel planning problems Logistics, the scales of encodings of FA-SP arc compacted more than 75%. The solver efficiencies of FA-SP are improved several times, and the scales of them are enlarged less than 5% for overwhelming majority problems.

Key words: Planning as satisfiability, Graphplan-based encoding, Encoding, Frame axiom, Planner

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!