Reinforcement learning optimization of a 6 Li magneto-optical trap: DDPG versus SAC
- 8 September 2026
- journal article
- Published by Optica Publishing Group in Optics Express
- Vol. 34 (19) , 34983
Abstract
We report an experimental implementation of deep reinforcement learning (RL) for optimizing the loading stage of a 6 Li magneto-optical trap (MOT) in a high-dimensional continuous control space. An off-policy actor-critic agent observes in-situ fluorescence images at 111 ms intervals and updates seven experimental parameters over 18 decision steps. Learning is performed directly on the apparatus using a sparse terminal reward R = N × A extracted from an absorption image, where N is the atom number, and A is the peak optical depth. We benchmark Deep Deterministic Policy Gradient (DDPG) and Soft Actor-Critic (SAC) across three independent 300-episode training runs under identical alignment. Both algorithms surpass the human-optimized (HO) baseline, but the central finding concerns reliability rather than peak performance. In deterministic replay validation, the three SAC waveforms show reward coefficients of variation of 9–14%, and the minimum reward of each waveform exceeds the HO mean reward. The DDPG waveforms show up to three times the dispersion, and even the best-performing DDPG run contains a catastrophic near-zero shot. The SAC solutions are nearly time-independent and consistently converge to a blue-detuned repump laser with an elevated field gradient, reminiscent of compressed-MOT operating conditions. Our results show that off-policy actor-critic RL can autonomously optimize high-dimensional laser-cooling sequences in an operating ultracold-atom apparatus, and that SAC has a reproducibility advantage over DDPG in a real noisy environment.Keywords
Funding Information
- National Research Foundation of Korea (RS-2023-NR119928)
- National Research Foundation of Korea (RS-2023-00256050)
- National Research Foundation of Korea (RS-2025-02220735)
- National Research Foundation of Korea (RS-2026-25469704)
This publication has 31 references indexed in Scilit:
- Reinforcement learning in cold atom experimentsNature Communications, 2024
- Machine learning-enhanced optical tweezers for defect-free rearrangementCurrent Applied Physics, 2024
- Realizing a deep reinforcement learning agent for real-time quantum feedbackNature Communications, 2023
- Single-atom exploration of optimized nonequilibrium quantum thermodynamics by reinforcement learningCommunications Physics, 2023
- Magneto-optical trapping and sub-Doppler cooling of a polyatomic moleculeNature, 2022
- Multiparameter optimisation of a magneto-optical trap using deep learningNature Communications, 2018
- Fast machine-learning online optimization of ultra-cold-atom experimentsScientific Reports, 2016
- Human-level control through deep reinforcement learningNature, 2015
- A distributed, graphical user interface based, computer control system for atomic physics experimentsReview of Scientific Instruments, 2013
- Double-pass acousto-optic modulator systemReview of Scientific Instruments, 2005