For traditional identification method, it is difficult to identify a system with measurements including both missing data and outliers. Thus, we propose a robust parameter estimation method to address these issues. First, a varying-interval-based sampling method is designed to address missing measurements by adjusting sampling intervals. Then, the outlier detection problem is converted into a matrix decomposition problem, in which the information matrix is refreshed by innovation-window. Subsequently, a varying-interval robust multi-innovation gradient-based iterative algorithm is derived, in which missing data and outliers can be handled simultaneously. Finally, the effectiveness of the proposed method is demonstrated through a numerical case.