In case of the difficulty in modeling for exoskeleton robot, an adaptive fuzzy logic control was presented to directly approximate the equivalent control without estimating the unknown parameters in advance. And an adaptive law in controller was designed to adjust its parameters according to parameter changes of the control system. In order to decrease the steady-state error and prevent from the integrator windup, a new nonlinear saturation function derived from quasi-natural potential function was designed to improve the performance of traditional integral sliding mode control. When the error beyond a boundary layer, the integral action was restricted by adjusting factor to avoid the large overshoot and long adjustment time. When the error was small, the integral action was completed so as to reduce the steady-state error and improve the robustness. The stability of the proposed controller was proved by using Lyapunov method. Furthermore, the chattering of sliding mode control was alleviated by simplifying the fuzzy control instead of switching function without deteriorating the system reliability and robustness. Finally, without acknowledging the model parameters of hydraulic-driven exoskeleton robot system, experiments were implemented to demonstrate the robustness and effectiveness of the method. And the results showed that the control output can follow the reference position signal quickly and smoothly with anti-interference ability.