No label defined (Q4352302)

The authors solve the problem of optimal decentralized control by using the classical Lagrange methodology applied within the maximum Pontryagin principle framework. The Lagrange multipliers introduced for information structure constraints given in the form of partial differential equations are defined as so-called Lagrange-Lyapunov functions. The system is described by a system of stochastic Ito equations and the performance index is an average value of a quadratic cost functional. The solution for the linear quadratic Gaussian case has the form of local linear feedback law and involves Riccati type equations. In the more general case the feedback control has a composite nature: the control depends on the state but also incorporates an information about the probability density defined by a solution of Fokker-Planck-Kolmogorov equation. For the entire collection see [Zbl 0852.00018].