Abstract: When there are a large number of tags in the recognition range of the reader, it is easy to cause the collision problem, which leads to the deterioration of the system throughput. To solve this problem, a grouped dynamic frame slotted ALOHA algorithm based on random numbers is proposed. Based on the original algorithm, the tags carried multiple random numbers to convert most idle time slots into successful time slots, which could reduce the time slots consumption and improve the system throughput rate. An improved corresponding relation between frame length and label number was proposed to improve the system stability. The simulation results show that the algorithm has good throughput and stability, and the throughput of the system increases from 36.8% to 55.3%.