We consider a single-input–single-output minimum-phase nonlinear system with large parametric uncertainty. The system can be represented globally in the normal form and our goal is to find a dynamic output feedback control law to ensure that the output (practically) asymptotically tracks a bounded smooth reference signal. Earlier work used high-gain observers with saturation to derive adaptive as well as robust control laws for this problem. The adaptive control law requires the nonlinear functions to be linearly parameterized in the unknown parameters and could have unsatisfactory transient performance for a large parameter set. The robust control law is based on a worst-case design and could be overly conservative. High gain feedback is needed to implement both controllers in the case when the parameter set is large. As a result, the robust and adaptive controllers may perform poorly in the presence of unmodeled dynamics and measurement noise. In order to reduce the controller gain and improve performance we propose a new approach based on partitioning the set of uncertain parameters into smaller subsets. Robust control laws are designed for each subset and logic based switching is used to choose the appropriate control law. The switching rule uses an estimate of the derivative of a Lyapunov function, which is provided by a high-gain observer.