It is well known that channel estimation is important for overall performance of a VRVP-MQAM system. Therefore, the MMSE estimation based VRVP-MQAM scheme was studied and second order statistical characterization of the channel state information(CSI) imperfection over a Rayleigh flat-fading was derived. Then rate and power algorithms was proposed in two kinds of situations that are derived based on ρ, the correlation coefficient between the estimated SNR and its true value. Finally, the proposed VRVP-MQAM scheme was simulated with a given Rayleigh flat fading channel. The simulation results and their comparison with an alternative rate and power algorithm which exploits an ideal CSI assumption show preliminarily that: Compared with ideal CSI(ρ=1), the signal to noise ratio(SNR) of MMSE estimate error(ρ=0.9) is 1 to 3 dB; When average SNR is 20 dB, the average spectral efficient(ASE) of MMSE CSI is 0.3 bps/Hz lower than that of the ideal CSI. Thus the application of VRVP-MQAM scheme will greatly improve in the future.