PlanPilot: Efficient Navigation in Plan Space

Daniel GnadDaniel Gnad, Augusto B. CorreaAugusto B. Correa, Johannes K. FichteJohannes K. Fichte, David SpeckDavid Speck, Dominik RusovacDominik Rusovac, Sarah Alice GagglSarah Alice Gaggl, Markus HecherMarkus Hecher

Download

Daniel Gnad, Augusto B. Correa, Johannes K. Fichte, David Speck, Dominik Rusovac, Sarah Alice Gaggl, Markus Hecher
PlanPilot: Efficient Navigation in Plan Space
Workshop on Explainable Artificial Intelligence (XAI), August 2025

Details
Bibtex

KurzfassungAbstract
Many planning applications require not only a single solution but benefit substantially from having a set of possible plans from which users can select according to preferences. Surprisingly, planning research has primarily focused on quickly finding single plans for decades. Only recently have researchers started to investigate plan enumeration by top-k planning, offering more flexibility to the user. But simply enumerating the k best plans is
far from targeted due to the time-consuming nature of enumeration, likely feeding many similar
plans to the user, or forcing the user to define filters beforehand. In fact, in extensive search spaces, enumeration is hardly practical. We present an approach and a tool called PlanPilot to navigate solution spaces of planning tasks iteratively and interactively. We build on answer-set programming (ASP) to restrict the plan space. To that end, we employ facets, which are meaningful actions that appear in some, but not all plans. Enforcing or forbidding such facets allows for navigating even large plan spaces while ensuring desired properties quickly and step by step.
Forschungsgruppe:Research Group: Logische Programmierung und ArgumentationLogic Programming and Argumentation

@inproceedings{GCFSRGH2025,
  author    = {Daniel Gnad and Augusto B. Correa and Johannes K. Fichte and
               David Speck and Dominik Rusovac and Sarah Alice Gaggl and Markus
               Hecher},
  title     = {PlanPilot: Efficient Navigation in Plan Space},
  booktitle = {Workshop on Explainable Artificial Intelligence (XAI)},
  year      = {2025},
  month     = {August}
}