Lyapunov stabilizability of controlled diffusions via a superoptimality principle for viscosity solutions (Q2492876): Difference between revisions

The author develops superoptimality principles for Hamilton-Jacobi-Bellman inequalities. More precisely, it is shown that every bounded lower semicontinuous viscosity supersolution \(V\) of the Hamilton-Jacobi-Inequality \[ \max_{\alpha\in A}\{-DV(x)f(x,\alpha) - \text{trace}[a(x,\alpha)D^2V(x)]\} \geq l(x) \] can be represented as \[ V(x) = \inf_{\alpha_{\cdot}}\sup_{t\geq 0}E_x\left[V(X_t^\alpha) + \int_0^t l(X_s^\alpha) \,ds\right], \] where \(X_t^\alpha\) is a solution of the controlled stochastic ODE \[ dx_t = f(X_t,\alpha_t)dt + \sigma(X_t,\alpha)dW_t \] and \(a=\sigma\sigma^T/2\). This result -- which is in fact interesting at its own right -- is then used in order to give sufficient Lyapunov function like conditions for different kinds of (asymptotic) stabilizability properties in probability (by open loop stochastic control processes) of \(X_t^\alpha\) in terms of Hamilton-Jacobi-Inequalities.