Lattice reduction (LR) and successive interference cancellation (SIC) are two well-known techniques that can be used to improve detection performance over linear detectors for multiple-input multiple-output (MIMO) systems. However, the LR technique and the SIC technique usually need perfect knowledge on the channel at the receiver, and the use of these techniques with an erroneous channel matrix even worsens the detection performance compared to linear detectors. In this correspondence, we shall show how to modify these techniques to make them robust under imperfect channel estimation. Information needed for the proposed algorithm is moderate; the variance of channel estimation error is the only requirement. Furthermore, our algorithm is not sensitive to the error in the variance of channel estimation error.