In this presentation, we study estimation and inference for nonlinear regression models with integrated time series through quantile regression method. In the model, the regression function derivative is given by integrable function or asymptotically homogeneous function, respectively, which has been analyzed in Park and Phillips (1999, 2001) (PP) by nonlinear least squares. We first derive the asymptotic distributions of the nonlinear quantile regression (NQR) estimators. We then find that one of the estimators does not converge to mixed normal as in PP, so that a fully-modified type estimator is proposed in that case. In consequence, standard inference such as a Wald test for parameter restrictions becomes available. Finally, we observe from simulations that our NQR estimators are desirable when distributions of regression errors possess fat tails.