Efficient Enumeration of Ground Actions for Lifted Classical Planning
Title: Efficient Enumeration of Ground Actions for Lifted Classical Planning
DNr: LiU-compute-2022-12
Project Type: LiU Compute
Principal Investigator: Simon Ståhlberg <simon.stahlberg@liu.se>
Affiliation: Linköpings universitet
Duration: 2022-04-07 – 2023-05-01
Classification: 10201
Keywords:

Abstract

In planning, it is common to give a domain description and a problem description to a planner. The domain description contains, inter alia, action schemas; which has free variables that need to be assigned to objects in the problem description. The procedure of finding valid assignments is called the grounding problem. In classical planning most planners enumerate all ground actions as a preprocessing step before solving the problem. That is, if a state is reachable, then all ground actions needed to reach it must be generated. However, in some domains the number of ground actions is exponential in the input size, this is a clear bottleneck that the research field has begun to address in lifted classical planning: enumerate only ground actions valid in the current state. In this project we will evaluate a new algorithm for tackling this problem.