cjameshuff
Critical Thinker
- Joined
- Feb 25, 2013
- Messages
- 370
So it uses a combination of assumed accelerations and velocities and actual measurements of same?
IIRC, it was one of the first major applications of Kalman filters.
It maintains an internal estimate of the system state which it uses as a basis for predicting the future state based on known dynamics of the system. It compares those predictions to actual measurements, correcting the estimate to be consistent with the actual state of the vehicle without relying entirely on either. If the dynamics model remains accurate, the result is a highly robust control system that converges to a good estimate of the system's state, but filters out spurious sensor inputs which are physically unlikely given recent history.
Reliability of different input sources can be taken into account, some sensor inputs having more importance relative to each other and to the internal estimate generated by all the previous sensor inputs. At the extreme, you might occasionally have a source trustworthy enough that you just overwrite the internal estimate with direct measurements.
