We show how an important class of nonlinearfeedback controllers can be designed using idealized abstractchemical reactions that can be implemented via DNA strand dis-placement (DSD) reactions. Exploitingchemical reaction networks(CRNs) as a programming language for the design of complexcircuits and networks, we show how a set of unimolecularand bimolecular reactions can be used to realize ultrasensitiveinput-output dynamics that produce a nonlinearquasi slidingmode(QSM) feedback controller. The kinetics of the requiredchemical reactions can then be implemented as enzyme-free,entropy-driven DNA reactions using DNA strand displacementvia Watson-Crick base pairing and branch migration. We demon-strate that the closed loop response of the nonlinear QSM con-troller outperforms a traditional linear controller by facilitatingmuch faster tracking response dynamics without introducingovershoots in the transient response. The resulting controller ishighly modular and is less affected by retroactivity effects thanstandard linear designs.
展开▼