Bifurcation analysis of PID-controlled neuromuscular blockade in closed-loop anesthesia

•Nonlinear dynamics in closed-loop PID controlled anesthesia are considered.•Bifurcation analysis of two Wiener models for neuromuscular blockade is performed.•An Andronov–Hopf bifurcation and chaos are discovered and characterized in both models.•A reduced-order model is shown to predict well the closed-loop behaviors of a classical pharmacokinetic–pharmacodynamic one. The problem of PID-controlled neuromuscular blockade (NMB) in closed-loop anesthesia is considered. Contrary to the usual practice of designing PID-controllers for nonlinear systems on the basis of a linearized model and online tests, bifurcation analysis is utilized in this paper for that purpose. Two nonlinear Wiener models for the NMB are considered: a conventional pharmacokinetic/pharmacodynamic (PK/PD) model and a parsimonious model suitable for online parameter estimation. The models under a PID feedback are analyzed in order to discern the safe intervals of the controller parameters that are not subject to complex dynamical phenomena. The parsimony of the mathematical model is instrumental in minimizing the number of bifurcation parameters. The analyses show that the closed-loop systems undergo Andronov–Hopf bifurcation at a point in the model parameter space giving rise to nonlinear oscillations. For steeper, but still feasible slopes of the nonlinear function parameterizing the static nonlinearity of the Wiener models, deterministic chaos can arise in the closed loop for lower concentrations of the anesthetic drug. A model-based PID-controller tuning procedure is suggested that guarantees a certain settling time and robustness margin of the resulting loop with respect to the bifurcation. The tuning procedure is illustrated on mathematical models identified from patient data and the corresponding PID controllers.