Analysis and computation of an optimality equation arising in an impulse control problem with discrete and costly observations (Q2332705)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Analysis and computation of an optimality equation arising in an impulse control problem with discrete and costly observations |
scientific article |
Statements
Analysis and computation of an optimality equation arising in an impulse control problem with discrete and costly observations (English)
0 references
5 November 2019
0 references
The authors treat a particular control problem for an Itô diffusion $(X(t))_{t\geq0}$ pieced together as follows: The process is observed at random times $0 = \tau_0 < \tau_1 < \tau_2\dots$ with $\forall k$ $\tau_{k+1} - \tau_k \geq m$, $m > 0$ fixed. At $\tau_k$ an amount $\eta_k \geq 0$ is taken out (``harvested'') leaving $X(\tau_k-)-\eta_k$ as starting point for the diffusion continuing up to the next harvest. The time $\tau_k$ and the harvest $\eta_{k-1}$ are measurable with respect to the $\sigma$-algebra generated by $(\tau_0,\eta_0),\dots,(\tau_{k-1},\eta_{k-1})$. The cost of each observation is $d > 0$, the cost of harvest $\eta$ is $c(\eta) = 1_{\eta > 0}(k_0+k_1\eta)$, and the disutility of the process between observation times is represented by $f(X(t))$ where $f$ is a Hölder function with exponent in $(0,1]$. The strategy $((\tau_0,\eta_0),\dots,(\tau_k,\eta_k))$ is defined to be optimal if it minimizes the expectation of the sum over k from $1$ to $\infty$ of $(d+c(\eta_k))\exp\{-\delta\tau_k\}$ plus the expectation of the sum over $k$ from $0$ to $\infty$ of the integral of $f(X(t))e^{-\delta t}$ over $[\tau_k ,\tau_k+1)$, where $\delta$ is a discount rate. The authors derive and analyze the corresponding optimality equation, prove that it has a unique solution, which can be approximated recursively, and show how to calculate the optimal strategy from it. Based on the recursion formula and a finite difference scheme they construct an algorithm and analyze it with respect to monotony, stability, and convergence to the solution. As a demonstrative example the numerical procedures is explicitly worked out for the management of a one-species population with $\mu X(t)(1-X(t))^\theta$ and $\sigma X(t)(1-X(t))^\theta$ as diffusion and drift coefficient, respectively, with specific growth rate $\mu > 0$, volatility $\sigma > 0$, and $\theta \geq 1$. As an example of ecological applications the authors suggest the population management of the waterfowl \textit{Phalacrocorax cabo} which on the one hand is a major predator of commercially and ecologically important fishery resources and on the other hand should be saved from extinction. Finally, an advanced problem subject to model ambiguity, as it may occur in reality, is formulated and numerically analyzed on the basis of the multiplier robust formalism [\textit{L. P. Hansen} and \textit{T. J Sargent}, ``Robust Control and Model Uncertainty'', Am. Econ. Rev. 91, No. 2, 60--66 (2001; \url{doi:10.1257/aer.91.2.60})].
0 references
discrete and costly observations
0 references
optimality equation
0 references
fixed point problem
0 references
finite difference scheme
0 references
ecological management
0 references
model ambiguity
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references