P1.5
Similarities and Differences between Multi-sensor Data Fusion and Data Assimilation: Implications for Over-determined vs. Under-determined Systems
Andrew J. Annunzio, Penn State Univ., University Park, PA; and S. E. Haupt and G. Young
Both Data Assimilation and Multi-sensor data fusion provide frameworks for combining observations with forecasts to better estimate contaminant dispersion. In Multi-sensor data fusion the state of an entity is identified, while in data assimilation the state of a field variable is estimated. These two different frameworks give differing formulations for the problem. For an entity, multi-sensor data fusion uses association techniques to decipher which observations belong to an object via hypothesis generation, evaluation, and selection. In contrast, assimilation techniques estimate field variables by using interpolation to compensate for the lack of data. As a result, data assimilation is applied to an under-determined problem whereas multi-sensor data fusion is formulated for an over-determined problem. While the estimation algorithms are nearly identical, the problem formulation differs since an entity has more observations than unknowns while field variables typically have more unknowns than observations. Environmental noise further complicates the problem, making it difficult for the optimization techniques to converge to the appropriate solution. This poster examines these two different frameworks. It additionally shows how the resulting algorithms minimize the error when noise is added into both the over-determined and under-determined systems. Meteorological applications of the field vs. entity frameworks are included. This analysis illustrates the similarities and differences between data assimilation and multi-sensor data fusion and helps to bridge the gap between the two.
Poster Session 1, Poster Session P1
Monday, 21 January 2008, 2:30 PM-4:00 PM, Exhibit Hall B
Previous paper Next paper