Abstract: An adaptive FNN control scheme for a class of unknown nonlinear systems is presented. The control architecture first employs two fixed structural FNNs to approximate plant uncertainties, then uses additional parallel adaptive FNNs to compensate for the modeling errors. A sliding mode terra is introduced to improve robustness. Based on the Lyapunov theory, the adaptation laws to tune on-line both the membership functions and weighting factors are deprived. Therefore, the closed-loop system is stable with the tracking error converging to zero.