Chemotherapy Optimization using Moving Horizon Estimation based Nonlinear Model Predictive Control ?

摘要

A Moving Horizon Estimator (MHE) based Nonlinear Model Predictive Controller (NMPC) was designed for an impulsive minimal tumor growth model. The estimator computes the time-varying model parameters using mean square error with parameter deviation penalization and provides state estimations for the controller. The controller computes optimal doses for non-equidistant, fixed time instants while constraining the administered drug dose. Tuning of the MHE was based on experimental time series measurements, while for the NMPC a virtual population was generated. The robustness of the combined approach was tested in silico on a virtual population, where the simulation was tailored to a real experimental scenario.