AbstractReaching a sensible compromise between computational tractability and degree of optimality still remains a significant challenge in robust model predictive control (MPC). Tube MPC offers an efficient approach which is based on pseudo–closed loop optimization but can thus be conservative. The degree of conservatism is reduced through the so–called disturbance affine control policy and it is the aim of this paper to propose a new methodology that deploys a more general policy thereby improving on system theoretic properties yet is no more computationally intensive. The work is based on a suitable parameterization of state and control tubes and is underpinned by guarantees of strong system theoretic properties.