Model-Based Control Can Improve the Performance of Artificial Cilia

Abstract

Artificial cilia are a prominent example of physical intelligence. Their mechanical properties are often designed so to achieve desired motions in response to very simple actuation patterns. Yet, variability in the mechanical properties are inherent in these systems. This may critically disrupt the input-output relation, resulting in a final behavior completely different from the desired one. In this Communication we investigate the possibility of designing a robotic brain that helps the cilium to maintain its physical intelligence. We achieve that by closing a model-based control loop which tracks the position of the end effector while compensating for drag forces. We propose experiments to characterize our model, and extensive simulations validating the results in different conditions. This work is intended as a proof of concept, which will be further expanded in future work.