Linear regression is a well-known method to predict further values for dependent variable given observed independent variables. Sometimes, there is prior information about regression coefficients. Therefore, restricted regression takes into account prior information and combines it with the sample information. Unfortunately, restricted regression has same assumptions as a linear regression. In this paper, we propose robust estimation of restricted linear models. In the application part, it is demonstrated that the proposed methods are convenient method to give impressive results in the presence of outliers with stochastic or nonstochastic restriction.