Shack-Hartmann wavefront sensing (SHWS) is foundational to adaptive optics, yet its precision is often limited by artifactual aberrations when dealing with complex or non-ideal sources. This talk presents three advancements to address these limitations. First, I will discuss the challenge of wavefront sensing in the presence of sharp phase discontinuities (modeled via asymmetric bifocal optics), where traditional methods fail due to artifactual coma. I will show how our proposed dual-plane integration method significantly reduces the error. Second, I will describe the impact of volumetric scattering and multiple reflecting layers—analogous to laser guide star profiles or thick biological samples—which introduce centroid bias. We recently demonstrated a computationally efficient coordinate-transformation approach to model these effects. Finally, to meet the real-time demands of high-resolution adaptive optics, we implemented hybrid machine
learning strategies that achieve sub-millisecond reconstruction times. These techniques together offer a framework for improving SHWS accuracy and speed in challenging conditions.