BRITE Relaunch: AI-Based Modeling and Control of 3D/4D Printed Soft, Fiber-Reinforced Actuators

About This Grant

This BRITE Relaunch award supports research that enables the manufacturing, modeling, and control of fiber-reinforced actuators leveraging advances in additive manufacturing and artificial intelligence, thereby promoting the progress of science, and advancing prosperity and welfare. Fiber-reinforced actuators are common in soft robotics since they closely mimic biological muscles. One challenge with fiber-reinforced actuators is that significant variability can exist between actuators if manufacturing methods are not precisely controlled. Additionally, the actuators displace nonlinearly and are difficult to control. This project will solve this challenge by utilizing 3D/4D printing, in which 3D structures are printed with electronic capabilities that allow them to sense physical phenomena, to repeatedly fabricate, characterize, and test fiber-reinforced actuators. Data obtained will inform physics-informed artificial intelligence models, which will then be integrated with advanced control strategies to enable more precise control of soft robotic actuators. The methods developed could address a fundamental gap in soft robotics related to control for applications ranging from biomedical to space systems. In addition, a multifaceted approach to generating excitement about STEM disciplines is planned that includes K-12 outreach activities, undergraduate research and teaching experiences, and development of a freely available, online workshop curriculum. The field of soft robotics offers significant potential for advancing how robots interact with humans. Fiber-reinforced, pneumatic artificial muscle (PAM) actuators and sensors could serve as transformative technologies for various applications. Control of fiber-reinforced PAMs remains challenging due to inadequate mathematical models of the nonlinear dynamics associated with the actuators. Furthermore, although many designs have been presented in the literature, each PAM is slightly different and performs differently. As a result, empirical or numerical mathematical models derived from experimental data often do not adequately capture the nonlinear dynamics of the actuators and are not broadly applicable to a variety of actuator designs. The research aims to study how advances in 3D/4D printing, additive manufacturing, strain sensing fibers, and physics-informed artificial intelligence modeling can be used to develop robust mathematical models of PAMs. The use of 3D/4D printing will lead to more reproducible fabrication and subsequently a more robust dataset that characterizes actuator performance. The data will further be used to create a physics-informed neural network model, which will then be used to develop advanced control strategies. The outcome would be a novel, data-driven approach that integrates repeatable manufacturing with AI to transform modeling and control of fiber-reinforced actuators. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.