Optimal nonlinear feedback control of quasi-Hamiltonian systems (Q1974198)

The present paper proposes and investigates an innovative strategy for optimal nonlinear feedback control of stochastic systems, relying on the stochastic averaging method for quasi-Hamiltonian systems and on the stochastic dynamic programming principle. The main advantages of the proposed strategy are considered to be the following: (1) The external and/or parametric random excitations can be explicitly taken into account in the stochastic dynamic programming equation and in the optimal control force equations. (2) Both the response statistics of the uncontrolled and controlled systems and the statistics of the optimal control forces can be analytically predicted, once the statistics of random excitations are given. (3) The feedback control forces are made up of conservative parts and dissipative parts, this classification providing various and efficient physical implications and functions. The effectiveness of the proposed technique is illustrated on an example consisting of a controlled Duffing oscillator subject to Gaussian white noise or non-white real noise excitation.