The Acquisition of Human Planning Expertise
Abstract
This chapter describes DÆDALUS, a planning system designed with human behavior in mind. Unlike most recent work on learning and planning, DÆDALUS employs a combination of forward chaining and means ends search, represents knowledge in a probabilistic framework, stores both cases and abstractions, and learns through an incremental process of concept formation. DÆDALUS represents each operator in a similar manner, specifying its preconditions as a set of state descriptors and its effects as a set of differences. Like most planning systems, DÆDALUS must solve problems that involve transforming an initial state into a desired state through the application of operators. The system describes each state as a set of literals—predicates with arguments—and it describes each problem or subproblem as an initial state conjoined with a set of differences that must be eliminated. DÆDALUS represents each operator in a similar manner, specifying its preconditions as a set of state descriptors and its effects as a set of differences. The system operates within a problem-space framework, but when domain-specific knowledge is available, it uses this information to constrain search.
Cite
Text
Langley and Allen. "The Acquisition of Human Planning Expertise." International Conference on Machine Learning, 1991. doi:10.1016/B978-1-55860-200-7.50020-9Markdown
[Langley and Allen. "The Acquisition of Human Planning Expertise." International Conference on Machine Learning, 1991.](https://mlanthology.org/icml/1991/langley1991icml-acquisition/) doi:10.1016/B978-1-55860-200-7.50020-9BibTeX
@inproceedings{langley1991icml-acquisition,
title = {{The Acquisition of Human Planning Expertise}},
author = {Langley, Pat and Allen, John A.},
booktitle = {International Conference on Machine Learning},
year = {1991},
pages = {80-84},
doi = {10.1016/B978-1-55860-200-7.50020-9},
url = {https://mlanthology.org/icml/1991/langley1991icml-acquisition/}
}