All coronagraphs for exoplanet high-contrast imaging employ wavefront estimation and control systems to reject system aberrations and create sufficiently dark holes. Currently at Princeton and JPL, two control schemes are being employed: electric field conjugation (EFC) and stroke minimization. Both of them are based on a linearized state transition model and both have achieved great success over the past decade. In this paper, we develop a general framework for the control system and improve upon the robustness of wavefront controllers from the perspective of stochastic optimization. We start by analyzing the model uncertainty and reformulating the wavefront control as a stochastic optimization problem. Then, we apply this idea to a recently proposed linear constraint wavefront controller and develop a robust version. According to the numerical simulations, the robust linear controller reveals the potential for creating more uniform dark holes. In experiment, the robust linear controller achieves at least as good performance as Electric Field Conjugation (EFC) and stroke minimization.