Answers are grounded in this paper and cite their sources.
Stochastic control problems are considered in which a state process Xₜ and an observation process Yₜ are governed by Ito-sense stochastic differential equations driven by independent Brownian motions. The control Uₜ enters linearly in the dynamics of Xₜ. A "separated"control problem is introduced, in which the state at any time t is a measure Λₜ representing an unnormalized conditional distribution for Xₜ given Yₛ, Uₛ for s t. The method depends on introducing a pathwise version of Λₜ which depends continuously on observation and control trajectories Y, U. Existence of an optimal control is obtained in a suitable class, larger than the usual class of controls admissible in the strict sense that Uₜ is measurable on the σ-algebra Fₜ (Y) generated by observations Yₛ, s t. The dynamics of Λₜ are studied using a method of forward and backward partial differential equations. Under a suitable nondegeneracy condition, the measure Λₜ has a density $q(t,x)$ with respect to Lebesgue measure.
No takes yet. Share an insight, caveat, or question.
Fleming et al. (1982) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: