Abstract: Conventional interference alignment (IA) requires perfect channel state infromation and is sensitive to channel uncertainties. Based on the norm-bounded channel uncertainties model, this paper proposed a weight adjustable robust IA algorithm. The proposed algorithm uses the minimax value method to maximizes the minimum weighted sum power of the projection casted by both interference and desired signal in their corresponding subspace. The weighted factors are designed as the exponential function of SNR which can adjusted according to SNR, which can effectively overcome the influence of white noise of different SNR. Simulation results show that the proposed weight adjustable algorithm has a good robustness when channel uncertainties is norm-bounded, and outperforms the const weight algorithm at the low SNR region.