1: \begin{abstract}

2: In this paper, we propose a fully distributed algorithm for frequency offsets estimation in decentralized systems.

3: With the proposed algorithm, each node estimates its frequency offsets by local computations and limited exchange of information with its direct neighbors.

4: Such algorithm does not require any centralized

5: information processing or knowledge of global network topology.

6: It is shown analytically that the proposed algorithm always converges to the optimal estimates regardless of network topology.

7: Simulation results demonstrate the fast convergence of the algorithm and

8: show that estimation mean-squared-error at each node touches the centralized Cram\'{e}r-Rao bound within a few

9: {iterations of message exchange}.

10: Therefore, the proposed method has low overhead and is scalable with network size.

11: \end{abstract}