CDOC: CoState Desensitized Optimal Control
Abstract
In this paper, costates are used to develop a framework that desensitizes the optimal cost. A general formulation for an optimal control problem with fixed final time is considered. The proposed scheme involves elevating the parameters of interest into states, and further augmenting the costate equations of the optimal control problem to the dynamical model. A running cost that penalizes the costates of the targeted parameters is then added to the original cost function. The solution obtained by minimizing the augmented cost yields a control which reduces the dispersion of the original cost with respect to parametric variations. The relationship between costates and the costtogo function, for any given control law, is established substantiating the approach. Numerical examples and MonteCarlo simulations that demonstrate the proposed scheme are discussed.
 Publication:

arXiv eprints
 Pub Date:
 September 2019
 arXiv:
 arXiv:1910.00046
 Bibcode:
 2019arXiv191000046R
 Keywords:

 Mathematics  Optimization and Control;
 Electrical Engineering and Systems Science  Systems and Control