A Qualitative Physics Compiler

Predicting the behavior of physical systems is essential to both common sense and engineering tasks. It is made especially challenging by the lack of complete precise knowledge of the phenomena in the domain and the system being modelled. We present an implemented approach to automatically building and simulating qualitative models of physical systems. Imprecise knowledge of phenomenais expressed by qualitative representations of monotonic functions and variable values. Incomplete knowledge about the system is either inferred or alternative complete descriptions that will affect behavior are explored. The architecture and algorithms used support both effective implementation and formal analysis. The expressiveness of the modelling language and strength of the resulting predictions are demonstrated by substantial applications to complex systems.

• Type: Conference
• Year: 1994
• Where: AAAI
• Authors: Adam Farquhar
• Keywords: AAAI 1994, Alternative Complete Descriptions, Complete Precise Knowledge, Intelligent Agents, Physical Systems