This paper deals with the modeling and the robust control of a nonlinear electrostatic microgripper dedicated to embedded microrobotics applications. We first propose a polynomial linear parameter varying model of the system, where the varying parameter is the mean position of the microgripper that is used for the linearization. The controller is then derived using a multimodel and scheduled observer-based control strategy. The structure and the order of the controller are defined a priori allowing the derivation of a robust low-order controller suitable for a real-time implementation in embedded on-chip environments. Results show that a very wide (several tens of micrometers) and fast positioning of the gripping arm can be achieved using the control strategy. A robustness analysis and experimental implementation results show the efficiency of the controller and the relevance of the theoretical approach.