Let $\theta \in \{0,1\}$ be an unknown state of the world. Let $P_0 := Prob(\theta = 1)$ at time $0$.

Let $G_t$ and $B_t$ be two Poisson processes with stochastic intensity $\lambda_g e_t \theta$ and $\lambda_b (1-e_t) (1-\theta)$. Let $\mathcal{F_t} := \sigma(G_s,B_s: s\in [0,t])$.

Let $P_t := Prob(\theta = 1 | \mathcal{F}_t)$.

The agent's control variable is $e_t$, which is $\mathcal{F}_t$ adapted.

Notice that at most one of $G_t$ or $B_t$ will generate any arrival. Moreover, if $G_t = 1 \Rightarrow P_t = 1$, $B_t = 1 \Rightarrow P_t = 0$. Otherwise, $$ dP_t = - [\lambda_g e_t - \lambda_b (1-e_t)] P_t (1-P_t) dt$$

Let $(p_1,p_2)$ be an interval such that $0 < p_1 < p_2 < 1$ and $s(p)$ be a continuous, bounded function on $[p_1,p_2]$.

\begin{align*} \tau := \inf \{t: P_t \notin(p_1,p_2)\} \end{align*}

The agent solves the following problem:

\begin{align*} V(p) := \sup_{e_t\in[0,1]} \mathbb{E} \left[ \int_0^\tau \exp\left(\int_0^t -s(P_u) du\right) \exp(-t) dt + e^{-\tau} V(P_\tau) \big| P_0 = p\right] \end{align*}

I would guess that for a smooth, well-behaved (Lipschitz) $s(p)$, $V$ would be $C^1$. But only argument to this effect that I can think of would need the viscosity approach. Is there a shorter way of doing this?



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.