Generative adversarial network-based inverse design of self-deploying soft kirigami composites for targeted shape transformation

Abstract

The design and development of morphing structures that transition from compact, transportable forms to stable, deployable configurations is crucial for advances in soft robotics, healthcare applications, and biomimetic systems. These structures often require customized functionalities and must self-deploy into precise target shapes. Therefore, the deformed shapes of such structures are usually prescribed and the parameters for their design are unknown. To obtain the fabrication parameters, the inverse problem needs to be solved, which quickly becomes quite challenging using conventional methods due to the high-dimensional nature of the inverse problem as well as the material and geometric nonlinearities. To overcome these challenges, we combine the best of the two worlds – physics and data – and present a data-driven approach for the inverse design of two-layered soft composites that utilize the principles of kirigami and strain mismatch to self-deploy into different three-dimensional shapes. At the center of our methodology is the generative adversarial network, designed to generate the necessary fabrication parameters. By using a pre-trained simulator network, we condition the generative model to generate feasible and accurate fabrication parameters that are used to make composites that deploy into the target shapes. Our findings demonstrate that the generative model is able to effectively predict kirigami patterns and pre-stretch values required to realize complex three-dimensional shapes from simple and diverse planar designs. By performing simulations and precise desktop experiments, we compare the target with deployed shapes and demonstrate the predictive capacity of the method.