Adaptive RFID Data Scheduling Using Proximal Policy Optimization for Reducing Data Processing Latency

Radio Frequency Identification (RFID) technology has become integral in various industries for automating processes and tracking items in real-time. However, efficiently offloading the large volume of data generated by RFID tags to readers poses significant challenges, particularly in dynamic environments where static or rule-based offloading methods fall short. This paper presents a novel approach for dynamically offloading data using deep reinforcement learning, specifically employing the Proximal Policy Optimization (PPO) algorithm. The proposed method utilizes a central controller equipped with the PPO model to make intelligent, real-time reader selection decisions based on environmental factors such as reader load, tag mobility, and network conditions. The reward function is designed to minimize data processing latency while maintaining balanced reader utilization, resulting in enhanced system efficiency. Extensive experiments demonstrate that the proposed PPO-based strategy significantly reduces average data processing latency by 30% and improves reader load balancing by 16% compared to conventional scheduling methods compared to conventional scheduling methods.