Abstract: A B-spline neural network was trained to be a neural network chaotic controller to predict the time sequences of chaotic systems and obtain their perturbation signals for control of chaotic systems. The data around the fixed points of unstable periodic orbits were used as input samples to train the parametric perturbation model. The perturbation signals were embedded in a chaotic attractor on the inputs of parameter perturbation model. Numerical simulation on Henon mapping demonstrates that the proposed model is effective for nonlinear chaotic motion control.