Reference: Shahar, Y. & Molina, M. Knowledge-Based Spatiotemporal Abstraction. Knowledge Systems Laboratory, Medical Computer Science, March, 1996.
Abstract: We present a case study of reusing the same problem-solving method for the related tasks of temporal abstraction and linear-space abstraction. The method, known as knowledge-based temporal abstraction, was designed for abstraction of high-level concepts from time-stamped raw data, and was evaluated primarily in several clinical domains, such as monitoring of diabetes patients. This paper describes an application of this method to the domain of traffic control, in which the monitoring task requires linear reasoning along both space and time. First, we reused the method by mapping it to the spatial dimension to make abstractions about the highway state, given sensor measurements along each highway. Second, we reused the same method to make temporal abstractions about traffic behavior for the same space segments over time, by mapping it to the temporal dimension. Our results also show that the knowledge-based temporal-abstraction method can be generalized into a knowledge-based linear-abstraction method. This method solves tasks requiring abstraction of data along a linear distance measure. Both the spatial- and the temporal-abstraction methods are instances of that generalized method mapped to different dimensions. We show how a spatiotemporal-abstraction method can be assembled from these methods. The spatiotemporal-abstraction method can be applied to tasks requiring abstraction of time-oriented data over a linear, decomposable space, such as traffic over a set of highways.
Full paper available as ps.