Supersolutions and subsolutions techniques in a minimax optimal control problem with infinite horizon (Q1287062)

The infinite horizon problem with \(L^{\infty}\) cost functional is studied. The problem has dynamics \(dy/dt=g(y(t),\alpha(t))\) for \(t >0\) with \(y(0)=x \). The control \(\alpha\) is chosen to minimize the cost \(J(x,\alpha)=\text{ess sup}_{t \geq 0} f(y(t),\alpha(t))\). If we denote the associated value function by \(u(x)\), we see that \(u(x)\) may not be continuous, or even upper or lower semicontinuous. Formally, the Bellman equation should be \(\max\{H(x,u,Du),\min_a f(x,a)-u\}=0\) where \(H(x,r,p)=\min\{p\cdot g(x,a)| f(x,a) \leq r\}\), and this equation has no coercive term. Therefore, uniqueness is a serious problem and it seems that the problem may not be well posed. Examples of discontinuity are given in this paper and the main result is that the value function is the minimal supersolution of the equation. The equation is treated in integrated form throughout, i.e., through the use of the dynamic programming principle rather than via viscosity solutions.