Bartender, I'll take a fuzzy Kalman please

Kalman filters seem to be the preferred solution to combine multiple, possibly conflicting, sources of sensor input. However, to properly apply one you have to know the noise characteristics of your different sensors.

Knowing the noise characteristics, and translating them to the proper weights in the Kalman filter is where the real magic comes in. Genetic algorithms are great at finding good combinations of parameters, and this article sounds like a promising combination of the two.

IngentaConnect A fuzzy Kalman filter optimized using a multi-objective genetic algorithm for enhanced autonomous underwater vehicle navigation