Traffic radar uses MIMO technology to achieve the goal of obtaining high angular resolution with a small degree of freedom. However, due to factors such as highly integrated hardware processing technology and device aging, there are amplitude and phase errors between the channels of millimeter wave MIMO radar. Considering the calibration accuracy and traffic application scenarios, this paper proposes an error self-calibration method based on the minimum entropy of power (MPE). The error vector is randomly generated as the initial value of optimization, and the error vector corresponding to the minimum entropy is searched through an adaptive optimization algorithm, and each is calibrated in real time. Errors caused by influencing factors. The paper gives the derivation of the algorithm, the analysis results of the darkroom measured data and the actual scene simulation data, which proves the effectiveness and superiority of the proposed algorithm.