Value-Gradient Iteration with Quadratic Approximate Value Functions

Alan Yang, Stephen Boyd

We propose a method for designing policies for convex stochastic control problems characterized by random linear dynamics and convex stage cost. We consider policies that employ qu…