A neural network clock discipline algorithm for the RBIS clock synchronization protocol

A fundamental role in clock synchronization protocols is played by clock discipline algorithms, which achieve more accurate regulation of nodes clocks, by improving stability against timestamp errors, operating system latencies, and environmental phenomena like temperature variations. In this paper, the NN-CDA clock discipline algorithm, which relies on neural networks, was implemented and its performance assessed using experimental data acquired from a real testbed. Results highlight that NN-CDA offers many advantages over conventional approaches, like those relying on linear regression, the most important of which are higher robustness to temperature variations and better synchronization quality.