In practical applications, tower crane systems always exhibit double-pendulum effects, because of non-ignorable hook mass and large payload scale, which makes the model more complicated and most existing control methods inapplicable. Additionally, most available control methods for tower cranes need to linearize the original dynamics and require exact knowledge of system parameters, which may degrade the control performance significantly and make them sensitive to parametric uncertainties. To tackle these problems, a novel adaptive neural network controller is designed based on the original dynamics of double-pendulum tower crane systems without any linear processing. For this reason, the neural network structures are utilized to estimate the parametric uncertainties and external disturbances. Based on the estimated information, and adaptive controller is then designed. The stability of the overall closed-loop system is proved by Lyapunov techniques. Simulation results are illustrated to verify the superiority and effectiveness of the proposed control law.