A robust precoding algorithm based on Signal to Leakage Noise Ratio(SLNR) is put forward in order to improve the performance of multiuser Multiple Input Multiple Output(MIMO) system in spatial and temporal correlated channel environment with channel error. Conditional mean and covariance of channel vector is derived considering spatial correlation and temporal, then the Channel State Information at the Transmitter(CSIT) and the Channel State Information at the Receiver(CSIR) are modeled. Based on these models, a robust SLNR precoding algorithm is derived when Channel State Information(CSI) at both ends is imperfect and not equivalent, and the corresponding receiving filters are also discussed. Simulation results show that the proposed robust SLNR precoding algorithm possesses robustness against channel error, spatial correlation and temporal correlation.