For the nonlinear and uncertainty problems in the upper limb rehabilitation robot trajectory tracking control, combining the characteristic of repetitive execution of the rehabilitation robot system and the properties of iterative learning control, a nonlinear iterative learning control algorithm is proposed. This algorithm improves the commonly used linear robot dynamics control, so that the good control quality is obtained under the condition that the model information is not accurate and only the position feedback can be measured. The Lyapunov stability theory and LaSalle invariance principle are applied to prove the global asymptotic stability of the closed-loop system. The simulation results show that the proposed nonlinear iterative learning control has good control performance.