The adaptive robust control scheme introduced by Veres and Sokolov (Automatica, 34, 723-730) provides optimal asymptotic performance under unmodelled dynamics, unknown disturbance bounds and unknown model orders that make the scheme desirable in practical applications. The amount of numerical computations associated with that scheme is large because of the possible large complexity of the appearing polyhedra. The geometrical details of the polyhedra need large memory space and their updating a lot of computation. This note gives a modification of the scheme so that polyhedron complexity is limited. Asymptotic performance of the scheme is the same as that of the original adaptive scheme in Veres and Sokolov (1998), transient performance can be worse. Simulation illustrates that practical implementation of the scheme is feasible.