The problem of estimating average treatment effect is of fundamental importance when evaluating the effectiveness of medical treatments or social intervention policies. Most of the existing methods for estimating average treatment effect rely on some parametric assumptions on the propensity score model or outcome regression model one way or the other. In reality, both models are prone to misspecification, which can have undue influence on the estimated average treatment effect. We propose a new robust approach to estimating the average treatment effect based on observational data in the challenging situation when neither a plausible parametric out come model nor a reliable parametric propensity score model is available. Our estimator can be considered as a robust extension of the popular class of propensity score weighted estimators. The new approach has the advantage of being robust, flexible, data adaptive and it can handle many covariates simultaneously. Adapting a dimension reduction approach, we estimate the propensity score weights semi-parametrically by using a nonparametric link function to relate the treatment assignment indicator to a low-dimensional structure of the covariates which are formed typically by several linear combinations of the covariates. We develop a class of consistent estimators for the average treatment effect and studied their theoretical properties. We demonstrate the robust performance of the new estimators on simulated data and a real data example of analyzing the effect of maternal smoking on babies’ birth weight.