Nonparametric Hammerstein Model Based Model Predictive Control for Heart Rate Regulation

This paper proposed a novel nonparametric modelbased model predictive control approach for the regulation ofheart rate during treadmill exercise. As the model structureof human cardiovascular system is often hard to determine,nonparametric modelling is a more realistic manner to describecomplex behaviours of cardiovascular system. This paperpresents a new nonparametric Hammerstein model identificationapproach for heart rate response modelling. Based on thepseudo-random binary sequence experiment data, we decouplethe identification of linear dynamic part and input nonlinearityof the Hammerstein system. Correlation analysis is applied toacquire step response of linear dynamic component. SupportVector Regression is adopted to obtain a nonparametric descriptionof the inverse of input static nonlinearity that is utilized toform an approximate linear model of the Hammerstein system.Based on the established model, a model predictive controllerunder predefined speed and acceleration constraints is designedto achieve safer treadmill exercise. Simulation results show thatthe proposed control algorithm can achieve optimal heart ratetracking performance under predefined constraints.