Aiming at the problem that the received signal quality drops sharply when transmit and receive beams have angular deviations in a mobile millimeter wave communication scenario， this paper proposes a beam tracking algorithm based on Adaptive Unscented Kalman Filter （AUKF）. Considering the applicability of low complexity for mobile scenarios， this algorithm designs an effective beam tracking scheme. In the analog beam forming architecture， a beam pair is trained to track an electromagnetic wave path to maintain an effective millimeter wave communication. By introducing an adaptive adjustment factor， the prediction and observation covariance matrices can be adjusted adaptive when there is an abnormal disturbance in the system， improving the estimation accuracy and the convergence speed. Simulation results show that the adaptive Unscented Kalman Filter algorithm in this paper significantly reduces beam tracking errors in mobile environments and has robust beam tracking capabilities.