An object-oriented extended Kalman filter tracking algorithm
Paper: 341
Session: A (talk)
Speaker: Brown, David, Ernest Orlando Lawrence Berkeley National Laboratory (LBL), Berkeley
Keywords: algorithms, C++
An object-oriented extended Kalman filter tracking algorithm
We present an object-oriented formulation of the extended
Kalman filter algorithm for track fitting and finding. Unlike previous
algorithms (Fruewirth, NIM A262(87)444 Boudreau, CHEP'94) this implementation makes
independent 'filtering' fits in both directions along the track, derriving
'smoothed' results at any point on the track from these. This approach is
mathematically and computationally
equivalent to 'filtering' away from the interaction point
and 'smoothing' back in, while having numerous advantages:
1. It allows a more efficient implementation of the track finding
operations of adding or subtracting hits from the track.
2. It treats both ends of the track with equal importance.
This accomodates the needs of PID systems commonly found outside the
tracking volume of many new HEP detectors.
3. It allows a log-likelihood treatment of non-gaussian hit errors and
multiple scattering uncertainty through a 'perturbative' extension to
the basic Kalman filter algorithm.
4. It is internally symmetric, resulting in fewer, simpler objects.
Another new feature is the separation of the 'parametric' and 'geometric' aspects
of a track into different objects. The parametric objects (Kalman Sites) are
formally independent of the geometric objects (Trajectories), resulting in a
more object oriented design. This algorithm will be used for track finding
and fitting with the BaBar detector.