Using direct driven in-wheel motor as the driving system of distributed driven electric vehicle，the power distribution of each driving wheel can be realized quickly and accurately. Aiming at the matter of current sensor fault and rotational speed sensor fault in distributed driven electric vehicle, the robust fault detection and location method is studied. Considering the unknown input and noise in motor model, decoupling the unknown input via the method of system order reduction, using Kalman filter (KF) to filter the white noise of the decoupled subsystem, the optimal unknown input observer (UIO) is designed to realize the system state estimation, and a robust residual generator is obtained. At the same time, the generalized likelihood ratio (GLR) method is used to evaluate the residual signals and determine the threshold value, a sensor fault location method is proposed. Bench experiment results show that the sensor fault diagnosis method based on optimal UIO realizes the sensor fault diagnosis and location of the direct driven in-wheel motor system.