In passive RFID systems, how to reduce the collision among tags is an important issue at the medium access control layer. The Framed Slotted ALOHA and its variations are well-known anti-collision algorithms for RFID systems. However, when the Framed Slotted ALOHA is used, the system efficiency and the average time delay deteriorate rapidly when the total number of tags increases. On the other hand, the total number of slots in a frame can¡¦t be infinity. In this thesis, we first compare existing anti-collision protocols and then propose a novel algorithm based on the Enhanced Dynamic Framed Slotted ALOHA (EDFSA) and the Progressing Scanning (PS) algorithm. The proposed algorithm is called Dynamic Grouping (DG). The DG algorithm partitions the RFID tags according to the distances from tags to the reader in order to avoid using too many slots in a frame. Inparticular, the DG algorithm estimates the spatial distribution of tags based on previous scanning results and then adjusts the partition accordingly. Unlike PS algorithm, the DG algorithm is applicable when the RFID tags are uniformly distributed or normally distributed.