Consider a stochastic control problem, $$v^C(0,x) = \mathbb{E} \Big[\int_0^\tau f(X_t,C_t) d t + (T-\tau)|X_\tau|\Big] $$ where $X_t$ is a weak solution to the SDE $$dX_t = C_t dB_t, \quad X_0 = x \in (-1,1).$$ $\tau:= \inf\{t: X_t \notin (-1,1)\} \wedge T$.

$C$ is a $\mathbb{R}_+$ valued progressively measurable stochastic process for some space $(\Omega, \mathbb{F}=(\mathcal{F}_t)_{t\ge 0}, \mathbb{P})$ for which the SDE above has a solution for some standard brownian motion $B_t$.

$f(\cdot,\cdot)$ is a measurable function. Also, $f(\cdot, c)$ is continuous for any $c$. I would like to claim the following:

Claim: Given any admissible control $C$, $\exists$ a Markovian control $C'$ (i.e. $C'_t = g(t,X_t)$ )(possibly on a different probability space) such that $v^{C'}(0,x) \ge v^C(0,x)$.

Does this sound true? There are various results in the vicinity of this that I have found in a survey by Borkar but, not being a practitioner myself, they are a bit hard to follow. Any help would be appreciated. Thanks.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.