Pontryagin formalized optimal control problems as canonical systems, which are concepts from analytic mechanics. Recently, a canonical quantization method for optimal control problems was proposed. In this paper, we derive a sufficient condition that particular solutions to the canonical quantized optimal regulator problems converge to strict solutions to the original optimal control problems with infinite-time horizons. Further, we discuss possibility that the canonical quantization method yields useful approximate methods for nonlinear optimal regulator problems.

View full abstract

The Platen method is an explicit numerical method for stochastic differential equations. We derive an optimal Platen method for Stratonovich stochastic differential equations in this paper. We also study asymptotic stability of some Platen methods and present numerical results which demonstrate the convergence properties and stability properties of these methods.

View full abstract

In this paper, time history of temperature on concrete surface is observed by reinforcement bar heating based on the electromagnetic induction method, and we formulated reinforcement corrosion shape estimation problem so as to minimize difference between theoretical computed value and observed data. Design variable is coordinates on corrosion surface, and the adjoint variable method is applied to compute the gradient of performance function with respect to the coordinates. In this study, we could obtain the shape of reinforcement corrosion based on modification of thermal property and assumption that reinforcement corrosion shape is cylindrical shape.

View full abstract

The Arnoldi method was proposed to compute a few eigenpairs of large-scale generalized eigenvalue problems. For the iterative computation of eigenpairs, this method generates the basis of a subspace by solving linear systems. This leads to considerable computation time for the large-scale problems. In this paper, to reduce the computation time, we propose an Arnoldi (M, W, G) method based on the Arnoldi method.

View full abstract