Abstract: When a radio frequency identification (RFID) system identifies multiple tags, tag collisions will happen. The RFID system generally applies a tag anti-collision protocol to resolve the multi-tag collisions. To reduce identified time, this paper proposes a new adaptive dynamic framed Aloha (ADFA) for RFID tag collision arbitration. Based on dynamic framed Aloha protocol, ADFA adaptively allocates each identified tag a slot number. During the next reading round, the tags will be identified according to the slot number, which can reduce collision and idle slots when a reader repeatedly tags. In many RFID applications where a reader may repeatedly identify tags, such as supply chain operation, object tracking and locating, the proposed protocols can reduce time of re-identifying tags. Furthermore, to reduce more identified time, we improve ADFA protocol and propose a tag quantity estimate with low computational complexity and an optimal frame length. The tag estimate is based on Vogt method, and can reduce computational complexity by narrowing the search range of the tag quantity. And the optimal frame length scheme can achieve maximum throughput under the condition that the slot durations are different. The theoretical computation and simulation results both show that ADFA can reduce identified time when repeatedly reading tags, and the tag estimate in the improved ADFA can lower computational complexity. In addition, the optimal frame length in the improved ADFA can also advance system throughput.