Polynomial optimization has been applied to multivariable control in the presence of stochastic inputs including coloured measurement noise. The method has been derived using the calculus of variations, which is thought to be a simpler formulation than completing squares. Both discrete-time and continuous-time systems are considered.