Based on hysteresis functions, a kind of multi-scroll chaos systems constructing method is proposed, parameters of which can be adjusted. For a class of chaotic systems with complicated uncertainties, a kind of adaptive repetitive learning synchronization controller is presented by combining the adaptive neural network method and the repetitive learning scheme. The difficulty from periodic time-varying parametric uncertainties are overcomed by using the adaptive repetitive learning method scheme, while the function uncertainties are compensated by using the neural approximation technique. The robust learning term is designed to estimate the upper bounds of neural approximation error and the disturbance. A Lyapunov- like function is constructed to prove the convergence of synchronization errors. Simulation results show the effectiveness of the proposed adaptive repetitive learning synchronization scheme.