A recent problem in adaptive control has been the task of controlling a system that cannot be characterized by a single mathematical model, and instead undergoes changes within a certain family of models. One strategy for controlling such a system has been to use switching between a fixed number of controllers to compensate for changing plant dynamics. This paper considers the problem of controlling an unknown plant contained in a family of models together with the additional task of controlling a plant whose plant dynamics shift from one member of the family to another. In particular, this paper describes a new switching controller that uses observers to improve switching transients, while maintaining an acceptable level of robustness.