Forecasting in a World where Uncertainty is the only Certainty

Accurate forecasting is an active area of research given its impact on life and property. Despite large uncertainties in current forecasts, the benefits that have resulted from better forecasting on paths of hurricanes, forecasting of Tsunami’s, are clearly manifested in the reduction of loss of life.

Kalman filtering is a technique to combine model predictions with actual measurements, to estimate the state of the system better than either use of the model or measurements alone. For example, to forecast the trajectory of an asteroid approaching the Earth, one can use a physics-based model to predict where it will be 10 days, 10 months, even 10 years from now. However, due to the inevitable small but cumulative errors in the dynamic model of the motion of the asteroid, the forecast may in the end prove dangerously misleading. On the other hand, if we were to be frightened off using a model at all and simply rely on current measurements, we give up our ability to use well-established physical principles to give an imperfect but reasoned prediction of the asteroid’s future course. The Kalman Filter paradigm blends prediction and measurement optimally, compensating for the errors in the model by recursively exploiting measurements to improve the quality of the estimates. This lecture will focus on some applications of the Kalman Filter to illustrate the potential applications of forecasting.

Speaker(s): Tarunraj Singh, Professor,

Agenda:

Dinner at 6pm with presentation to follow

Holiday Inn Buffalo Airport

4600 Genesee Street

Cheektowaga, NY 14225

Location:
Holiday Inn Buffalo Airport
4600 Genesee St
Cheektowaga, New York
14225