This paper investigates the problem of fault detection and isolation (FDI) for vehicle lateral dynamics. The system under consideration is a Linear Parameter Varying (LPV) model, where the scheduling parameter is related to the vehicle speed. The heart of the proposed approach relies in synthesizing robust residuals for some specified working speed. The robustness guaranties the validity between each working points. Synthesis of robust residuals is inspired from the well-known parity-space method but extended for uncertain systems. The final fault detector is reconstructed by switching from the different residuals according to the speed. An applicative illustration is presented to detect a sensor fault on a vehicle lateral dynamic system. The measurements have been provided by the MIPS laboratory in collaboration within the French INOVE project.