Search Strategies for Partial-Order Causal-Link Planning with Preferences

Search Strategies for Partial-Order Causal-Link Planning with Preferences. Bercher, P., Ginter, F., & Biundo, S. In 27th PuK Workshop ''Planen, Scheduling und Konfigurieren, Entwerfen'' (PuK 2013), pages 29–40, 2013.

Paper

Slides abstract bibtex

This paper studies how to solve classical planning problems with preferences by means of a partial-order causal-link (POCL) planning algorithm. Preferences are given by soft goals – optional goals which increase a plan's benefit if satisfied at the end of a plan. Thus, we aim at finding a plan with the best net-benefit, which is the difference of the achieved preferences' benefit minus the cost of all actions in the plan that achieves them. While many approaches compile soft goals away, we study how they can be addressed natively by a POCL planning system. We propose novel search and flaw selection strategies for that problem class and evaluate them empirically.

@Inproceedings{Bercher13POCLSoftGoals,
  author    = {Pascal Bercher and Fabian Ginter and Susanne Biundo},
  title     = {Search Strategies for Partial-Order Causal-Link Planning with Preferences},
  year      = {2013},
  pages     = {29--40},
  booktitle = {27th PuK Workshop ''Planen, Scheduling und Konfigurieren, Entwerfen'' ({PuK} 2013)},
  abstract  = {This paper studies how to solve classical planning problems with preferences by means of a partial-order causal-link (POCL) planning algorithm. Preferences are given by soft goals -- optional goals which increase a plan's benefit if satisfied at the end of a plan. Thus, we aim at finding a plan with the best \textit{net-benefit}, which is the difference of the achieved preferences' benefit minus the cost of all actions in the plan that achieves them. While many approaches compile soft goals away, we study how they can be addressed natively by a POCL planning system. We propose novel search and flaw selection strategies for that problem class and evaluate them empirically.},
  url_Paper  = {https://bercher.net/publications/2013/Bercher13POCLSoftGoals.pdf},
  url_Slides = {https://bercher.net/publications/2013/Bercher13POCLSoftGoalsSlides.pdf},
  keywords   = {workshop}
}

Downloads: 0