Towards Continuous Process Supervision

Abstract

This paper provides a contribution to qualitative reasoning for continuous process supervision. Tools for formalizing empirical knowledge of continuous processes and a naive reasoning naturally integrating perception are proposed. A new paradigm emerges : maintenance of a single instantaneous and possibly partial understanding of a physical system during its evolution. This understanding in closed loop includes prediction and truth maintenance of whole histories of evolution at the data sampling rate. This technique relies on both deep and empirical knowledge, using mainly intuitive notions like order of magnitude of non-negligible quantity, gain, delay and duration. Our approach is compared to existing work in qualitative physics basically focused on qualitative calculus of classical physics formalization.