Estimation of detector alignment parameters using the Kalman filter with annealing

Detector alignment is an essential step in the track reconstruction and analysis process. Alignment with tracks is the only possible strategy to estimate positions and orientations of components of a track detector with sufficiently high precision. We present a method for the estimation of alignment constants during track reconstruction in parallel with the estimation of the usual track parameters. The formalism is an extension of the standard Kalman filter and uses annealing in order to avoid suboptimal solutions. The algorithm has been implemented in the object-oriented reconstruction software of the CMS experiment and tested with a simulated test-beam-like setup of silicon detectors. We report results on the speed of convergence and the precision which can be achieved. (research notes from collaborations)