UnmodeledAccelerations allows the addition of process noise to the covariance to account for unmodeled accelerations in the radial, intrack, and crosstrack directions, which may be due to a phenomenon such as outgassing. Adding process noise in this fashion keeps the covariance from being unrealistically optimistic, and the larger covariance allows the filter to capture measurements after long predicts. UnmodeledAccelerations is ignored in the simulation process. These attributes only apply in the filter and smoother.
Unmodeled acceleration process noise is added to the filtered covariance at each time update. The actual process noise that is added is computed by scaling the specified process noise based on the ratio of the time update span over the specified time interval.