A novel fuzzy wavelet network based robust adaptive critic design was proposed for a class of non-affine nonlinear system with unknown control direction. The mean value theorem and Nussbaum function were used to handle with the problem of the nonlinear functions being implicitly functions with respect to the control input and the control direction being unknown. Two fuzzy wavelet networks (FWNs) with the same fuzzy basis functions were employed to implement the control element and the critic element, the weights, dilation and translation parameters of which were tuned online. In order to attenuate FWNs approximation errors, a robust term was designed by adaptive bounding technique. No a prior knowledge of the control direction and bound of uncertainty is needed. Moreover, the semi-globally uniformly ultimate boundedness of the closed-loop system was proved by Lyapunov theory. Simulation results demonstrate the effectiveness of the proposed design．
［1］ PROKHOROV D V, WUNSCH D C. Adaptive critic designs ［J］. IEEE Translations on Neural Networks,1997,8(5)：997-1007．
［2］ SI J,BARTO A G,POWELL W B,et al. Handbook of learning and approximate dynamic programming ［M］. New York: John Wiley & Sons, 2004,15-20．
［3］ KULJACA O, LEWIS F L. Adaptive critic design using nonlinear network structure ［J］. International Journal of Adaptive Control and Signal Progress,2003,17(6)：431-445．
［4］ KIM Y H, LEWIS F L. Reinforcement adaptive learning neuralnetbased friction compensation control for high speed and precision ［J］. IEEE Translations on Control System Technology, 2000, 8(1)：118-126．
［5］ WEI Qinglai,ZHANG Huaguang,CUI Lili. Databased optimal control for discretetime zerosum games of 2D systems using adaptive critic designs ［J］. ACTA Automatica Sinica,2009,35(6)：682-692．
［6］ ZHANG Huaguang,LUO Yanhong,LIU Derong. RBF neural networkbased nearoptimal control for a class of discretetime affine nonlinear systems with control constraints ［J］. IEEE Translations on Neural Networks,2009,20(9)：1490-1503．
［7］ YANG Q,VANCE J B, JAGANNATHAN S. Control of nonaffine nonlinear discretetime systems using reinforcementlearningbased linearly parameterized neural networks ［J］. IEEE Translations on Systems, Man, and Cybernetics, Part B: Cybernetics,2008,38(4)：994-1001．
［8］ ZHANG T,GE S S. Adaptive neural network tracking control of MIMO nonlinear systems with unknown deadzones and control directions ［J］. IEEE Translations on Neural Networks,2009,20(3)：483-497．
［9］ GE S S,ZHANG T. Neuralnetwork control of nonaffine nonlinear system with zero dynamics by state and output feedback ［J］. IEEE Translations on Neural Networks,2003,14(4)：900-918．
［10］ LIN C K.H∞Reinforcement learning control of robot manipulators using fuzzy wavelet networks ［J］. Fuzzy sets and Systems,2009,160(12)：1765-1786．
