New developments in the application of optimal control theory to therapeutic protocols (Q905822): Difference between revisions

The authors discuss an optimal control problem related to optimization of treatment protocols from the point of view of human immune system response. A dynamical system is represented by a model of immune response proposed by \textit{G. I. Marchuk} [Mathematical models in immunology. Transl. from the Russian. Translation Series in Mathematics and Engineering. York: Optimization Software, Inc. (1983; Zbl 0556.92006)] more than thirty years ago in the form of four state equations. Four control actions representing the effect of therapeutic agents are introduced additively into this model. The objective is to minimize a quadratic performance index containing both a terminal cost of chosen state variables and an integral cost of state and control variables. The authors use the Pontryagin maximum principle to find necessary conditions of optimality and propose an optimization algorithm based on the coordinate cyclic descent method [\textit{D. G. Luenberger}, Linear and nonlinear programming. 2nd ed. Boston, MA: Kluwer Academic Publishers (2003; Zbl 1134.90040)] to solve the TPBVP resulting from these conditions. Results of simulations are also presented based on parameter values from [\textit{R. F. Stengel} and \textit{R. Ghigliazza}, Math. Biosci. 191, No. 2, 123--142 (2004; Zbl 1051.92026)]. From the mathematical and numerical point of view the paper may be interesting, nevertheless its biomedical meaning is doubtful. The most important reservation is related to the quadratic cost functional which could not be considered as a measure of costs in the considered problem. Quadratic terms are usually introduced in optimization problems to measure costs of energy or variances of input or output variables in the controlled system. In the problem considered in this paper, none of these interpretations may be taken into account. Yet another unrealistic assumption is related to the additive form of the control variables in the model of the system. For example the effect of agent responsible for pathogen killing is multiplicative since its role is to increase mortality rate for the pathogen population. Other agents do not act additively either.