Recently, a new design of a model-locked all-fibre Figure-8 laser employing a nonlinear amplifying loop mirror (NALM) with two active fibre segments and two independently controlled pump-power modules has been proposed and experimentally demonstrated. This laser layout combines the reliability and robustness of conventional Figure-8 lasers with the flexibility of nonlinear-polarisation-evolution (NPE) lasers, providing access to a variety of generation regimes with a relatively wide adjustment range of the pulse parameters. Moreover, it enables reliable and reproducible live electronic adjustment of the lasing regimes, which is practically impossible to do by adjusting fibre-based polarisation controllers in NPE lasers. The general issue of reaching a target mode-locked laser regime with a setup featuring many adjustable parameters can be intelligently addressed by using machine-learning techniques. Here, we apply predictive regression to find optimum operating regimes in the NALM laser that are accessible through independent control of the pump powers in the gain segments, Pp,1, Pp,2, and the laser output coupling ratio β. We use a piece-wise propagation model for generating data that characterises the laser. In the fibres, propagation follows a standard modified nonlinear-Schrodinger equation including gain saturation and spectral response for the active segments. The gain coefficient amplitude is dependent on the average signal and pump powers, the average power dynamics being described by standard rate equations. We have trained a gradient boosted tree algorithm on our dataset to identify high-energy, stable mode-locked solutions across the full variation range of the total pump level delivered to the active fibres, Pp,tot, the ratio Pp1/Pp,tot, and β (tens of thousands of points). The algorithm has quickly handled the whole parameter space. Our approach paves the way for alternative approaches to the optimisation of nonlinear cavity dynamics, and can be generalised to other complex systems and higher degrees of freedom.
We apply predictive regression to find optimum operating regimes in a recently proposed layout of a flexible Figure-8 laser having two independently pumped segments of active fibre in its bidirectional ring.
The interplay among the effects of dispersion, nonlinearity and gain/loss in optical fibres is a powerful tool to generate a broad range of pulse shapes with tuneable properties. Here we propose a method to optimise the systems parameters for a given pulse target. By reducing the system complexity and applying machine-learning strategies, we show that it is possible to efficiently identify the sets of parameters of interest. Two configurations are numerically investigated: pulse shaping in a passive normally dispersive fibre and pulse generation in a dual-pump nonlinear-amplifying-loop-mirror mode-locked fibre laser.
The interplay among the effects of dispersion, nonlinearity and gain/loss in optical fibre systems can be efficiently used to shape the pulses and manipulate and control the light dynamics and, hence, lead to different pulse-shaping regimes [1,2]. However, achieving a precise waveform with various prescribed characteristics is a complex issue that requires careful choice of the initial pulse conditions and system parameters. The general problem of optimisation towards a target operational regime in a complex multi-parameter space can be intelligently addressed by implementing machine-learning strategies. In this paper, we discuss a novel approach to the characterisation and optimisation of nonlinear shaping in fibre systems, which combines numerical simulations of the governing equations to identify the relevant parameters and the machine-learning method of neural networks (NNs) to make predictions across a larger range of the data domain. We illustrate this general method through application to two configurations.
In this article, a method is presented for generating a trajectory through predetermined waypoints, with jerk and duration as two conflicting objectives. The method uses a Seq2Seq neural network model to approximate Pareto efficient solutions. It trains on a set of random trajectories optimized by Sequential Quadratic Programming (SQP) with a novel initialization strategy. We consider an example pick-and-place task for a robot manipulator. Based on several metrics, we show that our model generalizes over diverse paths, outperforms a genetic algorithm, SQP with naive initialization, and scaled time-optimal methods. At the same time, our model features a negligible GPU-accelerated inference time of 5ms that demonstrates applicability of the approach for real-time control.
The interplay among the effects of dispersion, nonlinearity and gain/loss in optical fibers is a powerful tool to generate a broad range of pulse shapes with tunable properties. Here we propose a method to optimize the systems parameters for a given pulse target. By reducing the system complexity and applying machine-learning strategies, we show that it is possible to efficiently identify the sets of parameters of interest. Two configurations are numerically investigated: pulse shaping in a passive normally dispersive fiber and pulse generation in a dual-pump nonlinear-amplifying-loop-mirror mode-locked fiber laser.
We numerically characterise, in the three-dimensional space of adjustable cavity parameters, the performance of a recently reported layout of a flexible figure-8 laser having two independently pumped segments of active fibre in its bidirectional ring (Smirnov et al 2017 Opt. Lett. 42 1732–5). We show that this optimisation problem can be efficiently addressed by applying a regression model based on a neural-network algorithm.
'%3e%3cpath fill='%23012169' d='M0 0v30h60V0z'/%3e%3cpath stroke='%23fff' stroke-width='6' d='m0 0 60 30m0-30L0 30'/%3e%3cpath stroke='%23C8102E' stroke-width='4' d='m0 0 60 30m0-30L0 30' clip-path='url(%23b)'/%3e%3cpath stroke='%23fff' stroke-width='10' d='M30 0v30M0 15h60'/%3e%3cpath stroke='%23C8102E' stroke-width='6' d='M30 0v30M0 15h60'/%3e%3c/g%3e%3c/svg%3e)
