Enabling mechano-intelligence in adaptive structures utilizing physical computing

Abstract

With advances in new technologies, researchers are attempting to develop the next generation of adaptive structures with intelligence directly embedded within their mechanical domain, the so-called mechano-intelligence. These attempts will enable mechanical systems to perform intelligent tasks, such as sensing the environment, changing geometries, making decisions, and executing computation in an even more proactive and autonomous manner, as compared to traditional mechatronic systems. However, there is no systematic foundation in constructing and integrating different aspects of mechano-intelligence. To advance the state of art, this research proposes to enhance the mechano-intelligence in adaptive structures through a machine learning framework called Physical Reservoir Computing (PRC). We show that a tunable modular metastructure can learn from its own wave dynamics and adaptively tune its own band structure via PRC. In other words, the metastructure can sense different input waves, make decisions and output appropriate control commands to alter its own wave characteristics without digital signal processors and controllers, i.e., achieve autonomous and integrated mechano-intelligence. Overall, this research provides a novel method to achieve intelligent and adaptive vibration/wave control based on the concepts of physical computing and learning and forms the basis for multi-faceted functional-relevant mechano-intelligence to be embedded in future adaptive structures and material systems.