A nonlinear robust adaptive sliding mode admittance controller is proposed for exoskeleton rehabilitation robots. The proposed controller has robustness against uncertainties of dynamic parameters using an adaptation law. Furthermore, an adaptive Sliding Mode Control (SMC) scheme is employed in the control law to provide robustness against disturbances (non-parametric uncertainties) with unknown bounds. For this purpose, another adaptation law is defined for the variation of the SMC gain. The proposed scheme is augmented with an admittance control method to provide the patient with compliance during interaction with the rehabilitation robot. The stability of the proposed controller and the tracking performance of the system are proven using the Lyapunov stability theorem. To verify the effectiveness of the proposed control method, some simulations are conducted for a nonlinear lower-limb exoskeleton robot interacting with a patient leg via some braces. Based on the obtained results, the controller is able to provide the patient with a flexibility and it appropriately responds to their non-compliant interaction torques. Moreover, the proposed controller significantly reduces the chattering of the input torques in comparison with a previous adaptive control method with a constant SMC gain, while it maintains a similar tracking performance. © 2018 Sharif University of Technology. All rights reserved.