Abstract: The filtered-x least mean square (FxLMS) algorithm is a classical algorithm for active noise control, however, the incompatibility problem between the convergence speed and steady-state error needs to be solved. One of the solutions is to use the variable step size FxLMS algorithm. The existing variable step size models based on the nonlinear function of error signal are summarized and applied to the FxLMS algorithm to improve the performance of the algorithm. Three common signals are used as reference input signals to carry out simulation experiments and the performance of variable step size algorithms with different nonlinear functions is compared. The results show that the variable step size FxLMS algorithm can achieve faster convergence speed and lower steady-state error when the reference signals are Gaussian white noise and sine wave, and the optimal algorithms are different under different noise environments, furthermore, the variable step size algorithm cannot improve the algorithm performance when the reference is impulsive noise. This research provides a reference for the selection of algorithms in different application. The variable step size FxLMS algorithm has been applied to the active noise control of a certain vehicle, and the results show that the variable step size FxLMS can significantly improve the system performance in constant speed condition, but its effect in rapid acceleration condition is not prominent.