Multi-mode Bessel–Gaussian orbital angular momentum holography based on actor–critic-assisted parameter optimization

With an increasing number of orbital angular momentum (OAM) modes, the sampling constant will change, limiting traditional OAM hologram methods. In order to achieve a high capacity and resolution, we present a novel, to our knowkledge, multi-mode Bessel–Gaussian (MBG) OAM holography method, based on integrated actor–critic reinforcement learning and convolutional feature extraction for multi-parameter optimization. Initially, MBG OAM-preserving holograms are generated by repeatedly superimposing Bessel–Gaussian (BG) phase patterns with varying combinations of encoding parameters, resulting in both selectable and multiplexed MBG OAM holograms. The selectable holograms are lighted with MBG beams featuring opposing parameter combinations, facilitating the reconstruction of the target image following a Fourier transform. Additionally, to tackle the robust interdependence among topological charge, axicon parameters, sampling constant, and iteration count, an actor–critic optimization framework utilizing convolutional neural networks and deep deterministic policy gradient (DDPG) is developed, with a reward function formulated using structural similarity (SSIM) and peak signal-to-noise ratio (PSNR) to facilitate adaptive exploration of high-quality parameter domains. Simulation results indicate that this method offers significant encoding flexibility for OAM holography via BG modes, hence facilitating the superposition of Gaussian point spots to enhance reconstructed image quality. The adaptable encoding and decoding framework has prospective uses in hologram switching and dynamic displays, while the implemented reinforcement learning optimization can efficiently limit the parameter search scope and lessen dependence on manual adjustments, offering a viable method for intelligent optical field management.