Multisearch systems with hierarchical control: a tool for solving complex AI problems

The authors describe multisearch as a general purpose tool for development and complex AI (artificial intelligence) systems. The multisearch approach expands on abstraction of a search space to a combination of search spaces. Each search is represented as a search template, which is a simple recursive scheme that can call other search templates. Search templates are executed by the search interpreter, which is a common driving mechanism for all search templates. Runtime control is represented by a network of control states at different levels, including global control for the whole multisearch system, local control for each search, and local control for individual search nodes. Differentiating between various types of control allows for distinguishing the corresponding types of heuristics. Two computer systems that have been programmed with the use of the multisearch framework are described.