Atomic clocks work on a standard frequency generated by the electron transitions in the atoms of the core material. A timescale is a reference frequency and phase measure generated by a set of atomic clocks. An ensemble algorithm combines the participating atomic clocks to form a “perfect” clock. The perfect clock is very stable and precise in terms of frequency and phase. There are many methods that exist to develop an ensemble for a timescale such as Kalman filter-based algorithms, inverse Allan variance-based algorithms, etc. A neural network-based realization of the ensemble algorithm for a timescale is discussed in this paper. The artificial neural network (ANN) model dynamically adapts the weights of the clocks to accommodate the behavioural changes in the clocks. This paper uses different types of M-sample deviations like overlapping Allan deviation and overlapping Hadamard deviation as the inputs to the model.
S. R. Nandita, Maharana, S., Dr. Rajathilagam B., T. Ganesh, S., and Krishnamoorthy, S., “An Artificial Neural Network Model for Timescale Atomic Clock Ensemble Algorithm”, MAPAN- Journal of Metrology Society of India, vol. 35, no. 4, pp. 547 - 554, 2020.