ML Anthology

Goal-Clobbering Avoidance in Non-Linear Planners

Rujith de Silva
AAAI 1994 pp. 1440
/aaai/1994/desilva1994aaai-goal/

Abstract

A central issue in non-linear planning is the ordering of operators so as to avoid undesirable interactions between their effects. The Modal Truth Criterion (Chapman 1987) states the conditions under which these interactions will occur. Non-linear planners use the Criterion, directly or indirectly, to promote or demote operators, or to co-designate variables, so as to avoid interactions. This abstract describes a method, called Goal Clobbering Avoidance (GCA), to avoid some interactions in a partially-ordered plan by promoting or demoting a sequence of operators, rather than individual operators. Effectively, it simultaneously applies the Modal Truth Criterion to all operators in the sequence, using pm-compiled information about the domain. GCA will be illustrated in the familiar Blocksworld domain, with the operators

PDF AAAI Semantic Scholar

Cite

Text

de Silva. "Goal-Clobbering Avoidance in Non-Linear Planners." AAAI Conference on Artificial Intelligence, 1994.

Markdown

[de Silva. "Goal-Clobbering Avoidance in Non-Linear Planners." AAAI Conference on Artificial Intelligence, 1994.](https://mlanthology.org/aaai/1994/desilva1994aaai-goal/)

BibTeX

@inproceedings{desilva1994aaai-goal,
  title     = {{Goal-Clobbering Avoidance in Non-Linear Planners}},
  author    = {de Silva, Rujith},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {1994},
  pages     = {1440},
  url       = {https://mlanthology.org/aaai/1994/desilva1994aaai-goal/}
}