Speaker
Description
The development of robust, precise and computationally efficient track reconstruction is important for CEPC offline reconstruction and its future high-level trigger (HLT) development, where tracking algorithms must balance reconstruction performance with the computational demands of large-scale event processing. In particular, charged-particle motion in magnetic fields and interactions with detector material introduce nonlinearities that can affect the estimation of track states and their uncertainties. Motivated by these considerations, this work investigates nonlinear effects within the ACTS-based tracking framework as a step towards the development and evaluation of Unscented Kalman Filter (UKF)-based tracking for CEPC.
As an initial study, we investigate the ACTS FreeToBoundCorrection (FTBC), which supplements the conventional linearised free-to-bound covariance transformation with a correction for nonlinear effects. We found that FTBC was not being applied in the Combinatorial Kalman Filter (CKF) propagation path used in our study and have enabled and systematically evaluated its use. While FTBC provides an additional correction for nonlinear effects not fully captured by the linearised transformation, the underlying filtering approach still relies on conventional linearisation. The UKF, in contrast, propagates sigma points through nonlinear transformations and reconstructs the state and covariance from their transformed distribution, allowing nonlinear effects to be incorporated directly into the estimation process. The FTBC study therefore establishes a baseline for determining whether UKF-based estimation can provide additional advantages.
Controlled simulation studies are used to examine the influence of FTBC on predicted and filtered track states, covariance transport and measurement association. The results show that small FTBC-induced changes in the propagated state and covariance can, under particular trajectory and incidence conditions, alter the source-link selected during combinatorial reconstruction. Selective suppression studies further demonstrate that a correction introduced at an upstream detector surface can influence a measurement-association decision several surfaces later. In a representative trajectory, suppressing the upstream correction removes the later association switch, whereas suppressing corrections at intermediate and switching surfaces does not. These observations demonstrate how small continuous changes in state and uncertainty propagation can interact with discrete decisions in the reconstruction chain.
The dependence of these effects on track momentum and incidence conditions is being investigated, providing a controlled baseline for studies in realistic CEPC detector configurations. Ongoing work will extend the framework towards UKF-based track fitting and reconstruction and compare it with conventional Kalman-filter-based estimation in terms of state accuracy, uncertainty propagation, robustness under nonlinear conditions, measurement association and computational requirements. Such studies will help assess the potential of nonlinear filtering methodologies for future CEPC offline reconstruction and HLT applications.
