DDPG-AmBC: Dynamic reflection control for batteryless IoT networks with ambient backscatter

Batteryless Internet of Things (IoT) networks enabled by Ambient Backscatter Communication (AmBC) represent a promising approach for large-scale and energy-sustainable deployments. Yet, dense and time-varying environments introduce major challenges, such as unpredictable fluctuations in incident RF power, strong interference, and unstable link quality. These non-stationary conditions make traditional fixed or reactive reflection-level control methods suboptimal. This work introduces DDPG-AmBC, a deep reinforcement learning (DRL) framework designed to perform continuous and adaptive reflection control for batteryless AmBC devices. The method employs the Deep Deterministic Policy Gradient (DDPG) algorithm to dynamically adjust the reflection coefficient ( Γ ) based on real-time observations of the wireless environment. This mechanism enables joint optimization of communication throughput and energy efficiency while preserving passive operation. The paper presents a complete system formulation, an expanded discussion of related studies, and comparative evaluations against fixed-reflection and competitive DRL-based baselines. The results show that DDPG-AmBC significantly improves throughput stability under varying interference conditions and supports enhanced energy-sustainable operation. Experiments conducted using OMNeT++ for network emulation and Google Colab for DRL training confirm the effectiveness of the proposed framework and its suitability for next-generation IoT scenarios.