This paper presents a novel adaptive reference governor for robust tracking control in constrained linear systems with bounded disturbances. The proposed reference governor achieves a performance improvement over the existing reference governors by virtue of its added feature: adaptability. The design of such an adaptive reference governor involves a nonlinear non-deterministic polynomial time (NP)-hard optimization problem because the solution of the optimization problem must be searched for the infinite number of sequences of disturbance. The SDP relaxation method turns out to allow the nonlinear NP-hard problem to be recast into an SDP, which may be readily solved in polynomial time.