This paper considers a new tracking control problem for a class of multivariate non-Gaussian system where the tracked target is a given joint probability density function.The control objective is to find an algorithm such that the entropy of tracking error is minimized.A function based model is established between the disturbance,state and error.A recursive optimization approach is presented such that the performance index is minimized.Finally,simulations are provided to demonstrate the effectiveness of the proposed minimum entropy control algorithm.