Appearance Randomization is a simulation-based training strategy that perturbs visual attributes such as color, texture, and lighting in synthetic environments to improve the robustness of policies transferred from simulation to real-world settings.