4.II.29I

Optimization and Control
Part II, 2006

An investor has a (possibly negative) bank balance x(t)x(t) at time tt. For given positive x(0),T,μ,Ax(0), T, \mu, A and rr, he wishes to choose his spending rate u(t)0u(t) \geqslant 0 so as to maximize

Φ(u;μ)0Teβtlogu(t)dt+μeβTx(T),\Phi(u ; \mu) \equiv \int_{0}^{T} e^{-\beta t} \log u(t) d t+\mu e^{-\beta T} x(T),

where dx(t)/dt=A+rx(t)u(t)d x(t) / d t=A+r x(t)-u(t). Find the investor's optimal choice of control u(t)=u(t;μ)u(t)=u_{*}(t ; \mu).

Let x(t;μ)x_{*}(t ; \mu) denote the optimally-controlled bank balance. By considering next how x(T;μ)x_{*}(T ; \mu) depends on μ\mu, show that there is a unique positive μ\mu_{*} such that x(T;μ)=0x_{*}\left(T ; \mu_{*}\right)=0. If the original problem is modified by setting μ=0\mu=0, but requiring that x(T)0x(T) \geqslant 0, show that the optimal control for this modified problem is u(t)=u(t;μ)u(t)=u_{*}\left(t ; \mu_{*}\right).