An intelligent target tracker that we developed for a joint Air Force-Navy project has a similar structure to perception and cognition. It has internal models of tracks, and tracking consists in finding a correspondence between the internal track-models and subsets of the sensory data about the world. This involves an iterative process of partitioning the world among models, while adapting the models to the data. In the past, approaches to the development of tracking algorithms were similar to rule-system or production systems, in which evaluation of the intermediate results was based on logical structures similar to preferences in Soar. This led to combinatorial explosion of computations. In our approach, evaluation of intermediate results is accomplished by fuzzy similarity measures, which play the role of emotional signals. The paper compares concepts of the tracker architecture to those of semiotics and psychology. This leads to a better understanding of the architectural concepts as well as identifies mathematical tools for semiotics.
展开▼