Reinforcement Learning for Scalable and Reliable Power Allocation in SDN-based Backscatter Heterogeneous Network

Backscatter heterogeneous networks are expected to usher a new era of massive connectivity of low-powered devices. With the integration of software-defined networking (SDN), such networks hold the promise to be a key enabling technology for massive Internet-of-things (IoT) due to myriad applications in industrial automation, healthcare, and logistics management. However, there are many aspects of SDN-based backscatter heterogeneous networks that need further development before practical realization. One of the challenging aspects is the high level of interference due to the reuse of spectral resources for backscatter communications. To partly address this issue, this article provides a reinforcement learning-based solution for effective interference management when backscatter tags coexist with other legacy devices in a heterogeneous network. Specifically, using reinforcement learning, the agents are trained to minimize the interference for macro-cell (legacy users) and small-cell (backscatter tags). Novel reward functions for both macro- and small-cells have been designed that help in controlling the transmission power levels of users. The results show that the proposed framework not only improves the performance of macro-cell users but also fulfills the quality of service requirements of backscatter tags by optimizing the long-term rewards.