Analogical representations of naive physics

• Published in: Artificial intelligence Vol. 38; no. 2; pp. 139 - 159
• Main Authors: Gardin, Francesco, Meltzer, Bernard
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.03.1989; Elsevier
• Subjects: Applied sciences; Artificial intelligence; Computer science; control theory; systems; Exact sciences and technology; Learning and adaptive systems; Reasoning; Knowledge representation; Analog model; Artificial intelligence; Modeling; Physics; Cognitive system
• ISSN: 0004-3702; 1872-7921
• DOI: 10.1016/0004-3702(89)90055-6

In classical quantitative and contemporary qualitative physics, objects, properties and relations are represented by names; in the light of deficiencies of such representations for the purposes of naive physics, experimental studies are presented of analogical models, which bear a structural similarity to what is being modeled. For this purpose, objects and substances are represented by pixel sets built up from base elements of pixel aggregates, the global physical behaviour resulting from message passing between adjacent base elements. These messages embody a small number of local constraints (four for strings and eight for liquids) derived from naive observation, such as material continuity and noncopenetrability. After presentation of the results obtained, limitations of the programs so far developed are discussed as well as the relationship of the approach to previous ones. Possible uses of analogical models are briefly discussed, an example being an implemented general “pouring program” for liquids. Serendipitous applications of the programs developed, for instance to robotics and economics, are briefly referred to.