Advanced Optimal Control Design for Nonlinear Systems including Impulsive Inputs with Applications to Automatic Cancer Treatment
Sakode, Chandrashekar M
MetadataShow full item record
The motivation of this research is to propose innovative nonlinear and optimal control design algorithms, which can be used in real life. The algorithms need to be computationally efficient, should deal with control constraints and should operate under state feedback. To show the efficacy of algorithms, automatic therapy for different cancer problems is chosen to be the ﬁeld of application. In this thesis, ﬁrst an advanced control design technique called ’optimal dynamic in-version’ has been successfully experimented with control constraints. The proposed approach has subsequently been shown to be quite effective in proposing automatic drug delivery schemes with simultaneous application of chemo and immunotherapy drugs for complete elimination of cancer cells in melanoma (a skin cancer) as well as glioma (a brain cancer). As per the current practice, the amount of drug dosages are generally given based on some apriori statistical study with a very small sample size, which in reality may either also lead to drug toxicity (due to excessive drug) or may become ineffective (due to insufficient drug) for a particular patient. Subject to the ﬁdelity of the mathematical model (which has been taken from published literature), it has been shown in this thesis that nonlinear control theory can be used for computation of drug dosages, which can then be used in a feedback strategy, thereby customizing the drug for the patient’s condition, to cure the disease successfully. Next, attention has been shifted to impulsive control of systems. Such impulsive con-trol systems appear in many other applications such as control of swings, control of spacecrafts and rockets using reaction control system, radiotherapy in cancer treatment and so on. Two impulsive control design philosophies are proposed in this thesis. In one approach, recently proposed model predictive static programming (MPSP) has been extended for impulsive control systems and has been named as impulsive-MPSP (I-MPSP). In other approach, another recent development, namely the Pseudospectral method has been utilized to consider both the magnitude of the control impulses as well as the time instants at which they are applied as the decision variables. It can be noted, that to the best of the knowledge of the author, the time instants of control application, being considered as decision variables is being proposed for the ﬁrst time in the nonlinear and optimal control framework. Both I-MPSP and Pseudospectral methods are computationally quite efficient and hence can be used for feedback control (I-MPSP happens to be computationally more efficient than the Pseudospectral method). Applicability of the proposed extensions have been shown by solving various benchmark problems such as (i) a scalar linear problem, (ii) Van der Pol’s oscillator problem and (iii) an inverted pendulum problem. Finally the applicability of the proposed I-MPSP strategy has been shown by solving challenging problems such as radiotherapy treatment of head and neck and adenocarcimona cancers. Radio-therapy model is considered with oxygen effect, in which radiosensitivity parameters are considered in different forms. Head and neck cancer is considered with constant radiosensitivity parameters and adenocarcinoma is considered with constant, linear, quadratic and saturation model of radiosensitivity parameters. Note that toxicity constraints on normal tissue, which are nonlinear control constraints, are also successfully incorporated in this control design.