Reinforcement-Learning-Based Residual Channel Resources Exploitation in Cross-technology Coexistence

In unlicensed spectrum bands, cross-technology coexistence (CTC) creates a unique spectrum utilization issue when wideband technologies, such as WiFi, using partially overlapping channels share frequency space with narrowband technologies, such as ZigBee. When a wideband device transmits on one channel, it often partially occupies spectrum that overlaps with its neighboring channels. The remaining portions of these partially occupied channels are too narrow for the wideband technology’s own transmissions, thereby becoming unusable residual channel resources (RCRs). This paper introduces a novel framework for exploiting these otherwise wasted spectrum resources by enabling narrowband technology transmissions within these RCRs. We address two fundamental challenges in RCR exploitation within CTC environments: 1) detecting RCR frequency ranges without complex computations or expensive specialized hardware, and 2) predicting the inherently random durations of RCRs before they become available. Our solution comprises a lightweight frequency range detection method that leverages packet preamble patterns and requires only standard receiver hardware, coupled with a reinforcement learning approach that predicts upcoming RCR durations by analyzing historical transmission patterns. Extensive simulations demonstrate that our integrated approach significantly enhances spectrum utilization efficiency in CTC, enabling substantial overall throughput improvements.