Reinforcement learning-enabled robust phase control for OAM beam generation in coherent beam combining systems.

Area of Science:

• Optics and Photonics
• Laser Physics
• Artificial Intelligence in Science

Background:

• Orbital angular momentum (OAM) beams are crucial for advanced applications.
• High-power OAM generation faces challenges like mode competition and low damage thresholds.
• Coherent beam combination (CBC) offers a solution, but precise phase control is difficult.

Purpose of the Study:

• To develop an intelligent, label-free, and high-precision phase control framework for OAM beam generation in CBC systems.
• To leverage reinforcement learning (RL) for robust and efficient OAM beam synthesis.

Main Methods:

• Implementation of a reinforcement learning-based phase control framework.
• Incorporation of a physically informed reward function.
• Integration of a vector-quantized (VQ) module within actor and critic networks.
• Application to a 12-channel CBC system for OAM beam generation.

Main Results:

• Rapid and robust single-stage generation of ±1 and ±2 OAM beams.
• Achieved mode purities exceeding 0.99, even in noisy environments.
• Demonstrated superior control stability compared to the stochastic parallel gradient descent (SPGD) algorithm.
• Operated effectively at a significantly lower control frequency than SPGD.

Conclusions:

• The proposed RL framework enables advanced phase control for OAM beam generation in CBC systems.
• This approach offers a scalable and physically grounded paradigm for intelligent optical field modulation.
• The findings advance RL applications in coherent beam combination and optical systems.