Industry 4.0 describes future workpieces would be smart. They can control the individual stages of their production semiautonomously through negotiation with production resources. To this end, RFID is introduced into machining processes. The current applications of radio frequency identification mainly include two approaches: one is data-on-network approach which centrally stores workpiece-related data on network; another is data-on-tag approach, which stores the data on RFID tag. Data-on-network approach cannot adapt to flexible automated production lines, which require decentralized control system for flexible process and rapid response, and data-on-tag approach makes RFID tag store large amounts of data, which would result in slow data reading and writing problems. For solving this problem, a hybrid-data-on-tag approach is proposed. The on-tag data contains the basic machining information and the index of further machining information which is stored in the backend database. Based on the hybrid-data-on-tag approach, the control node model of decentralized control system and the corresponding cyber-physical systems architecture of flexible manufacturing shop floors are presented. Furthermore, a multi-agent-based decentralized control system is used to describe the communication of smart workpieces with production resources during their production procedures. Finally, a small-scale flexible automated production line is taken as an example to illustrate the utility of hybrid-data-on-tag approach.