Our object model is implemented by a Kalman Filter [8][5].
The quality of the object features (position and velocity) computed
from the scan data is very different for both sensors. For
example, the velocity information from the radar is more accurate
than that of the Lidar scanner, which must be computed
from measurements at different times. This was accounted for
by setting the values in the measurement covariance matrices
appropriately when updating the state vector.