Deep Reinforcement Learning for the Design of Metamaterial Mechanisms with Functional Compliance Control

TL;DR

The paper tackles the design of metamaterial-based compliant mechanisms using a deep reinforcement learning framework that operates on a digitized, cell-based design domain evaluated by finite element analyses. By transforming the design problem into a Markov decision process and employing a dueling DQN, it learns placement policies that maximize a motion-focused reward, demonstrated on a door-latch and extended to a soft gripper with hinge-connectivity considerations. The RL-designed mechanisms show substantially improved horizontal compliance and reduced vertical deflection compared with human-guided designs, with experimental validation via additive manufacturing. This approach provides an efficient, scalable design methodology for programmable, highly compliant metamaterial mechanisms applicable to practical engineering tasks.

Abstract

Metamaterial mechanisms are micro-architectured compliant structures that operate through the elastic deformation of specially designed flexible members. This study develops an efficient design methodology for compliant mechanisms using deep reinforcement learning (RL). For this purpose, design domains are digitized into finite cells with various hinge connections, and finite element analyses (FEAs) are conducted to evaluate the deformation behaviors of the compliance mechanism with different cell combinations. The FEA data are learned through the RL method to obtain optimal compliant mechanisms for desired functional requirements. The RL algorithm is applied to the design of a compliant door-latch mechanism, exploring the effect of human guidance and tiling direction. The optimal result is achieved with minimal human guidance and inward tiling, resulting in a threefold increase in the predefined reward compared to human-designed mechanisms. The proposed approach is extended to the design of a soft gripper mechanism, where the effect of hinge connections is additionally considered. The optimal design under hinge penalization reveals remarkably enhanced compliance, and its performance is validated by experimental tests using an additively manufactured gripper. These findings demonstrate that RL-optimized designs outperform those developed with human insight, providing an efficient design methodology for cell-based compliant mechanisms in practical applications.

Figures (16)

• Figure 1: Design of a compliant door-latch mechanism: (a) basic design configuration (unit: mm), (b) selection of rigid elements, (c) square cells with different diagonal reinforcements, and (d) parallelogram cells with different diagonal reinforcements.
• Figure 2: Various FEA models: (a) square unit cell, (b) parallelogram unit cell, (c) three design cases for the door-latch structure (Designs 1-3).
• Figure 3: Experimental configuration for compliant door-latch structures: (a) schematics of the experimental setup, (b) assembled components for additively manufactured door-latch mechanism.
• Figure 4: Conversion of mechanism design problem into sequential decision-making problem: (a) conversion to a sequential decision-making problem by unfolding, (b) spiral tiling, (c) zigzag tiling.
• Figure 5: The neural network architectures: (a) DQN, (b) Dueling DQN.