Policy-transferred reinforcement learning enables sample-efficient inverse design of thermally active composites with programmable 4D shape transformation.
Lee et al. (Thu,) studied this question.