Optical networks are evolving toward ultrawide bandwidth and autonomous operation. In this scenario, it is crucial to accurately model and control optical power evolutions (OPEs) through optical amplifiers (OAs), as they directly affect the signal-to-noise ratio and fiber nonlinearities. However, a fundamental contradiction arises between the complex physical phenomena in optical transmission and the required precision in network control. Traditional theoretical methods underperform due to ideal assumptions, while data-driven approaches entail exorbitant costs associated with acquiring massive amounts of data to achieve the desired level of accuracy. In this work, we propose a Bayesian inference framework (BIF) to construct the digital twin of OAs and control OPE in a data-efficient manner. Only the informative data are collected to balance the exploration and exploitation of the data space, thus enabling efficient autonomous-driving optical networks (ADONs). Simulations and experiments demonstrate that the BIF can reduce the data size for modeling erbium-doped fiber amplifiers by 80% and Raman amplifiers by 60%. Within 30 iterations, the optimal controlling performance can be achieved to realize target signal/gain profiles in links with different types of OAs. The results show that the BIF paves the way to accurately model and control OPE for future ADONs.

The emission wavelength of a laser is physically predetermined by the gain medium used. Consequently, arbitrary wavelength generation is a fundamental challenge in the science of light. Present solutions include optical parametric generation, requiring complex optical setups and spectrally sliced supercontinuum, taking advantage of a simpler fiber technology: a fixed-wavelength pump laser pulse is converted into a spectrally very broadband output, from which the required resulting wavelength is then optically filtered. Unfortunately, this process is associated with an inherently poor noise figure, which often precludes many realistic applications of such supercontinuum sources. Here, we show that by adding only one passive optical element—a tapered photonic crystal fiber—to a fixed-wavelength femtosecond laser, one can in a very simple manner resonantly convert the laser emission wavelength into an ultra-wide and continuous range of desired wavelengths, with very low inherent noise, and without mechanical realignment of the laser. This is achieved by exploiting the double interplay of nonlinearity and chirp in the laser source and chirp and phase matching in the tapered fiber. As a first demonstration of this simple and inexpensive technology, we present a femtosecond fiber laser continuously tunable across the entire red–green–blue spectral range.