Physics-Informed Neural Network for Cross-Domain Predictive Control of Tapered Amplifier Thermal Stabilization

TL;DR

This work tackles thermally induced noise in tapered amplifiers that limits the sensitivity of quantum sensing with ultra_stable lasers. It introduces a physics_informed encoder_decoder GRU (PI_GRU) whose training enforces thermodynamic consistency via a lumped_parameter model, enabling accurate multi_step temperature predictions even when trained only on low_power data. The PI_GRU is integrated into a hierarchical parallel MPC that employs a GPU accelerated prediction layer and a parallel PSO optimization layer to achieve real_time, cross_domain thermal stabilization across multiple laser power regimes. Experimental validation demonstrates strong extrapolation to high_power conditions and substantial improvements in temperature stability, confirming the approach as a practical paradigm for robust cross_domain predictive control in nonlinear laser systems.

Abstract

Thermally induced laser noise poses a critical limitation to the sensitivity of quantum sensor arrays employing ultra-stable amplified lasers, primarily stemming from nonlinear gain-temperature coupling effects in tapered amplifiers (TAs). To address this challenge, we present a robust intelligent control strategy that synergistically integrates an encoder-decoder physics-informed gated recurrent unit (PI-GRU) network with a model predictive control (MPC) framework. Our methodology incorporates physical soft constraints into the neural network architecture, yielding a predictive model with enhanced physical consistency that demonstrates robust extrapolation capabilities beyond the training data distribution. Leveraging the PI-GRU model's accurate multi-step predictive performance, we implement a hierarchical parallel MPC architecture capable of real-time thermal instability compensation. This hybrid approach achieves cross-domain consistent thermal stabilization in TAs under diverse laser power operations. Remarkably, while trained exclusively on low-power operational data, our system demonstrates exceptional generalization, improving prediction accuracy by 58.2% and temperature stability by 69.1% in previously unseen high-power operating regimes, as experimentally validated. The novel synchronization of physics-informed neural networks with advanced MPC frameworks presented in this work establishes a groundbreaking paradigm for addressing robustness challenges in cross-domain predictive control applications, overcoming conventional modeling limitations.

Figures (7)

• Figure 1: (a) Schematic of the TA-based high-power semiconductor laser system. PBS: polarizing beam splitter, ${\lambda \mathord{\left/ {\newline} \right. \nulldelimiterspace} 2}$: half-wave plate. (b) Thermal management of the TA. (c) Experimental platform for TA temperature control. (d) Internal photograph of the TA controller. MCU: microcontroller unit.
• Figure 2: Framework of the proposed control strategy based on network predictive model
• Figure 3: Construction and training process of PI-GRU network
• Figure 4: Prediction results of GRU and PI-GRU across three laser power operations, with subfigures showing probability histograms of prediction errors. (a) GRU at $0.5\text{ W}$. (b) GRU at $1\text{ W}$. (c) GRU at $1.5\text{ W}$. (d) PI-GRU at $0.5\text{ W}$. (e) PI-GRU at $1\text{ W}$. (f) PI-GRU at $1.5\text{ W}$.
• Figure 5: Stepwise prediction errors of (a) GRU and (b) PI-GRU at three powers of $0.5\text{ W}$, $1\text{ W}$, and $1.5\text{ W}$.