Optimal shape morphing control of 4D printed shape memory polymer based on reinforcement learning

4D printing technology, as a new generation of Additive Manufacturing methods, enables printed objects to further change their shapes or other properties upon external stimuli. One main category of 4D printing research is 4D printed thermal Shape Memory Polymer (SMP). Its morphing process has large time delay, is nonlinear time variant, and susceptible to unpredictable disturbances. Reaching an arbitrary position with high precision is an active research question. This paper applies the Reinforcement Learning (RL) method to develop an optimal control method to perform closed loop control of the SMP actuation. Precise and prompt shape morphing is achieved compared with previous control methods using a PI controller. The training efforts of RL are further reduced by simplifying the optimal control policy using the structural property of the prior trained results. Customized protective visors against COVID-19 are fabricated using the proposed control method.

Automated summary

This study develops an optimal control method using reinforcement learning to achieve closed-loop control of SMP actuation, resulting in more precise and prompt shape morphing compared to previous control methods.