Reasoning in dynamic systems: From raw data to temporal abstract information

Understanding the behaviour of systems requires dealing with the raw data obtained from variables that define the system's state and using high-level knowledge for deep processing tasks. Therefore, intermediate processes to reduce this cognitive gap are needed. Besides, when dealing with dynamic systems, in which state variables evolves with time, the way in which time is perceived plays a crucial role and affects the performance of deep processing tasks. Such an intermediate process, which helps overcome this cognitive gap between raw data and deep knowledge processing levels, while taking into account the time dimension, is called temporal abstraction (TA). The final objective of TA is to provide an explanation that describes the behaviour of dynamic systems arising from the temporal evolution of state variables and to locate the data in their correct temporal context. In this article we propose a novel method for TA, based on an abductive strategy and a temporal constraint model as an underlying formalism for time representation and reasoning. The abductive vision of this problem is suitable for a general case in which data are heterogeneous and where there is temporal imprecision.