Constructing diagrams representing group motions
Certain domains, such as military activities and weather phenomena, are characterized by a large number of individual elements moving in a field, either in pursuit of an organized activity in groups at different levels of aggregation (military action), or subject to underlying physical forces that cluster different elements in different groups with common motion (weather). Reasoning about phenomena in such domains is often facilitated by abstracting the mass of data into diagrams of motions of groups, and overlaying them on diagrams that abstract static features into regions and curves. Constructing such diagrams of motion basically calls for clustering at different time instants and joining the centers of the clusters to produce lines of motion. However, because of incompleteness and noisiness of data, the best that can be done is to produce plausible hypotheses. We envision a multi-layered abductive inference approach in which hypotheses largely flow upwards from raw data to a diagram to be used by a problem solver, but there is also a top-down control that asks lower levels to supply alternatives if the original hypotheses are not deemed sufficiently coherent.
Lecture Notes in Artificial Intelligence (Subseries of Lecture Notes in Computer Science)
Banerjee, B., & Chandrasekaran, B. (2004). Constructing diagrams representing group motions. Lecture Notes in Artificial Intelligence (Subseries of Lecture Notes in Computer Science), 376-378. https://doi.org/10.1007/978-3-540-25931-2_43