| Popis: |
Neuromorphic computing is an interdisciplinary field that combines principles of computer engineering, electronics, and neuroscience, aiming to design hardware and software that can process information in a similar manner to biological brains, offering advantages in efficiency, adaptability, and cognitive capabilities. In this work, we propose a coevolutionary (or adaptive) mixed-feedback framework in which, mimicking a feedback control loop, a neuromorphic plant follows a predetermined desired rhythmic profile. This simple, yet efficient coevolutionary law adaptively minimizes the error between the responses of the reference and the plant through a node-to-node mapping. Such direct node correspondence ensures that each node in the plant is associated with a corresponding element in the reference, even in the presence of a discrepancy in the number of oscillators between the two networks. As a result, the controlled output of the plant effectively replicates the response of the reference in an orderly manner. Moreover, we demonstrate the effectiveness of our mixed-feedback control approach through several examples, including the amplitude and phase control when considering a plant and a reference with the same and different topologies on their associated networks, as well as the generation of complete synchrony by considering a single node reference. Additionally, we further test our control methodology by demonstrating its efficiency even in the presence of reference networks with a time-varying topology. Finally, we discuss future work for the development of similar coevolutionary laws that enable the control of networks which describe different dynamical systems, as well as higher-order topological dependencies. |
