An Improved Q-Learning Algorithm Integrated into the Aloha Anti-Collision Protocol for Energy-Efficient RFID Systems

The collision in Aloha-based Radio Frequency Identification (RFID) systems is inevitable due to the random medium access nature of the Aloha protocol and the unknown number of tags within the reader’s coverage. Various Aloha anti-collision protocols have been proposed, and reducing collisions has always been the top priority. However, merely reducing collisions can increase the number of idle slots, the number of interrogation epochs, and bandwidth usage. This article proposes an approach to integrating Q-Learning into the Aloha anti-collision protocol, in which Interrogation Efficiency (IE), resulting in Energy Efficiency (EE), is the top priority. Two cases of fixed and dynamic frame sizes are considered. Experimental results show that the Q-Learning-integrated Aloha anti-collision protocols achieve the highest IE, in which the number of collision slots, idle slots, and interrogation epochs are reduced. The dynamic-frame Q-Learning-integrated Aloha anti-collision protocol achieves the best IE thanks to its ability to adjust frame size dynamically