A Sound and Fast Goal Recognizer
Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence. Planning. Pages 1704-1709.
Abstract
Th e bulk of previous work on goal and plan recognition may be crudely stereotyped in one of two ways. "Neat" theories - rigorous, justified, but not yet practical. "Scruffy" systems - heuristic, domain specific, but practical. In contrast , we describe a goal recognition module that is provably sound and polynomial-time and that performs well in a real domain. Our goal recognizer observes actions executed by a human, and repeatedly prunes inconsistent actions and goals from a graph representation of the domain . We report on experiments on human subjects in the Unix domain that demonstrate our algorithm to be fast in practice. The average time to process an observed action with an initial set of 249 goal schemas and 22 action schemas was 1.4 cpu seconds on a SPARC-10.
@inproceedings{ijcai1995p219,
title = {A Sound and Fast Goal Recognizer},
author = {Lesh, Neal and Etzioni, Oren},
booktitle = {Proceedings of the Fourteenth International Joint Conference on
Artificial Intelligence, {IJCAI-95}},
publisher = {International Joint Conferences on Artificial Intelligence},
editor = {Chris S. Mellish},
pages = {1704--1709},
year = {1995},
note = {Planning},
url = {https://ijcai.org/proceedings/1995/219},
}
TY - CPAPER TI - A Sound and Fast Goal Recognizer AU - Lesh, Neal AU - Etzioni, Oren T2 - Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence, IJCAI-95 PB - International Joint Conferences on Artificial Intelligence PY - 1995 SP - 1704 EP - 1709 UR - https://ijcai.org/proceedings/1995/219 ER -Copied
Machine-readable versions of this paper, for tools and AI agents:
