Agent-based interpretation of geospatial evidential data

We report on work developing and testing agent-based algorithms for marshaling available evidence, making inferences, and testing hypotheses using geospatially referenced data. In addition to developing algorithms, we also identified and obtained land cover and transportation data from public sources, selected COTS software for map preprocessing, studied the scaling behavior of the agent-based algorithms as a function of the number of map cells, and investigated the impact of inter-agent small-world communication networks and agent activation rules on processing speed and scaling. Our results demonstrate the viability of an agent-based "bottom-up" approach to solving problems that involve synthesizing data from multiple sources in a geospatial setting. One particularly interesting finding is the discovery of a synergistic combination of small-world networks and agent activation rules leading to emergent basins of attraction for information flow.