Angle-of-Arrival estimation in localized hybrid array is very challenging due to the phase ambiguity problem. A differential beam searching (DBS) algorithm was proposed to solve the problem. However, it converges slowly and suffers from a zigzag effect where estimates may jump between correct and wrong estimates. In this paper, we first propose an improved DBS algorithm to mitigate the zigzag effect. We then analyze the error sources for general DBS algorithms, and propose a new adaptive searching and tracking (AST) algorithm. By using carefully designed updating functions and exit conditions, the proposed AST algorithm allows automatic switching between searching and tracking modes, and can significantly speed up estimation process and reduce estimation error. Numerical results are provided to verify the effectiveness of the proposed schemes.