This paper investigates the problem of mobile robot localization in ultra high frequency-radio frequency identification(UHF-RFID), and a localization method of the mobile robot is presented based on an adaptive UKF bank to integrate the information of UHF-RFID and odometer for the mobile robot localization with an unknown initial pose. Firstly, by using the UHF-RFID, the initial estimate of the position of the mobile robot is obtained, and a set of estimates can be generated randomly according to the initial estimate. Then, taking the quantization error into consideration, an adaptive UKF is applied to update the estimate set, and accordingly, the estimate set is pruned and updated, which helps to improve the accuracy and stability of the filter. The results of the simulation show that the adaptive UKF bank has higher accuracy and faster convergence compared with the standard UKF.