Planning with natural actions in the situation calculus

  • Authors:
  • Fiora Pirri;Raymond Reiter

  • Affiliations:
  • Univ. degli Studi di Roma "La Sapienza", Roma, Italy;Univ. of Toronto, Toronto, Canada

  • Venue:
  • Logic-based artificial intelligence
  • Year:
  • 2000

Quantified Score

Hi-index 0.00

Visualization

Abstract

We describe the theory and implementation of a deductive planner in the situation calculus for domains with two kinds of actions: 1. "free will" actions on the part of agents with the ability to perform or withhold their actions, like choosing to pick up an object, or deciding to walk to some location. 2.Natural actions whose occurrence times are predictable in advance, in which case they must occur at those times unless something happens to prevent them, for example, objects moving under Newtonian laws, or trains arriving and departing in accordance with known schedules. The theoretical basis for our planner is an extension of the situation calculus to accommodate continuous time and natural actions. The planner itself is patterned after that proposed by (Bacchus and Kabanza, 1995; Bacchus and Kabanza, 2000); it is a forward reasoning planner that filters out partial plans using domain and problem-specific information supplied by the user. The planner is implemented in ECLPSE Prolog, and exploits that system's built-in linear constraint solver to do temporal reasoning. We illustrate the planner's workings on a space platform example that we fully axiomatize in the situation calculus.