Feedback linearization technique is applied to design robust controllers for a class of nonlinear uncertain systems. Every system in the class is a series connection of finite number of nonlinear subsystems which are individually feedback linearizable and have significant but matched nonlinear functional uncertainties. The key difference between this work and existing robust results is that the matching conditions and the existence of feedback linearization transformation are only required individually for each subsystem. A constructive robust control design procedure is proposed for the unique control in the overall system. The procedure is conceptually simple and contains three steps. First, perform input-output feedback linearization for all individual subsystems. Second, design fictitious robust controllers for inputs of each subsystem as if the subsystem had an independent control. Finally, a mapping is developed which maps the individual fictitious controls recursively into the unique control of the overall system. Since neither the matching conditions nor feedback linearization are required for the overall system, this becomes the broadest class of nonlinear uncertain (or known) Systems for which a robust (regular) tracking control can be constructed using feedback linearization method. Global stability of the overall system is always guaranteed.