The use of Smooth Variable Structure Filter (SVSF) has been successfully overcoming the Localization problem. Generally, its performance depends on the knowledge of noise statistics for the process and measurement. Because this knowledge is not available, both are determined and kept to be constant for all iterations. However, this approach will lead SVSF to the divergence condition. Accordingly, a novel improvement, namely FISGA-IAE ASVSF, is proposed in this paper. This name represents the role of the Genetic Algorithm (GA) used to optimize the Fuzzy Inference System (FIS) that is initially applied for enhancing the adaptive SVSF. Unlike the traditional way, this strategy can recursively update the noise covariance of the process Q and measurement R. In detail, FIS supervises the adaptive SVSF to reduce the mismatch between the reference and estimated covariance of error innovation. To effectively arrange the membership function of FIS, the GA is adopted. Lastly, it is implemented to solve the localization problem of mobile robots in the synthetic simulation perception. By using the term RMSE, the comparatively presented results are analyzed. And the proposed method shows better performance in terms of accuracy and stability.