Kalman filtering for multi-path network synchronization

The accuracy of time distribution over packet networks relies on propagation delay stability and two-way symmetry in the path between the master clock and a slave clock. Multipath time synchronization provides a possibility to safeguard accuracy by exploiting path diversity, while improving security and fault-tolerance through inherent redundancy. In this paper we focus on the problem of combining time information received through multiple paths, for which we propose a Kalman filter-based algorithm. We show that the algorithm allows adaptive combination of time information, based on the associated measurement uncertainty and prove that the protocol and algorithm costs of redundancy can be repaid in terms of increased synchronization robustness.