The aim of this thesis is the modelling and robust control a special group of robots, namely the family of the spherical robots. The different kinematic and dynamic models of spherical robots and their properties are studied and presented. Different parameterizations of the orientation are used such as the Euler or Tait-Bryan angles and the quaternions. The exact linearizability of the nonlinear dynamic models is studied. Exact linearization based tracking controllers for the nonlinear systems are calculated and verified using simulation. A drawback of exact linearization is its apparent sensitivity to parameter uncertainty. Therefore a method to link exact state feedback linearization and robust controller synthesis techniques is presented. This method is based on the linearization of the dynamics obtained by exact linearization such that the nonlinear dynamical model is subject to parameter variations. This parameter variation is translated into a model set defined by an uncertainty structure which can be managed by standard synthesis techniques. This technique is applied and evaluated for the models of spherical robots.