Embedded Files
DRL-assisted Charging
References:
- A. Azarbahram, O. López, P. Popovski, S. Pandey, and M. Latva-aho, "Deep Reinforcement Learning for Multi-User RF Charging with Non-linear Energy Harvesters," IEEE Global Communications Conference (GLOBECOM), Cape Town, South Africa, 2024, pp. 3075-3080, doi: 10.1109/GLOBECOM52923.2024.10901606.
We examine dynamic wireless charging in a multi-antenna RF-WPT system serving multiple devices with nonlinear energy harvesters and time-varying power demands.
Instead of optimizing a fixed transmission pattern, the system must adapt its beamforming decisions across discrete time slots to satisfy energy requirements while minimizing the long-term transmit power.
The deep deterministic policy gradient (DDPG)-based control policy progressively learns efficient charging strategies over training episodes, leading to a steady reduction in average transmit power and energy outage events as learning converges. This confirms that reinforcement-learning-based adaptivity can successfully capture the temporal structure of device energy demands and propagate it into power-efficient scheduling decisions.
As shown in the above figures, both the average power consumption and the likelihood of energy shortage rise as the number of energy-harvesting devices K increases, reflecting the growing spatial competition for available transmit power and the challenge of meeting heterogeneous energy thresholds under multi-user interference. Despite this, the learning-based solution continues to converge to a stable policy, demonstrating robustness in larger-scale scenarios, while the lightweight beamforming heuristic provides a strong performance baseline with significantly lower complexity.
Overall, intelligent, demand-aware transmission scheduling can materially boost RF-WPT efficiency in dynamic IoT environments, while combining low-complexity beamforming with reinforcement learning offers a practical pathway for scalable closed-loop wireless charging control.
Sense-then-Charge (STC)
References:
- A. Azarbahram, O. López, R. Souza, P. Popovski, and M. Latva-aho, "Sense-then-Charge: Wireless Power Transfer to Unresponsive Devices with Unknown Location," IEEE Global Communications Conference (GLOBECOM) Workshops, 2025 (Accepted).
We investigate a dual-functional multi-antenna WPT system capable of sensing the environment and charging devices without relying on prior location or CSI knowledge.
The study focuses on unresponsive energy-harvesting devices and evaluates a sense-then-charge (STC) protocol in which a portion of each transmission period is used for active sensing to estimate device directions and channels via echo-based processing.
Numerical results show that allocating sensing time judiciously enables accurate LoS channel estimation, resulting in substantially improved beamforming efficiency during the subsequent charging phase. We identify an optimal balance between sensing and charging time that depends on system operating conditions, including the number of antennas and available transmit power.
As shown in the figures below, when the system is equipped with a sufficiently large receive array and adequate transmit power, the proposed STC framework approaches the performance of an idealized system with perfect CSI and significantly outperforms CSI-free broadcast-based benchmarks.
As the number of devices increases, or when the system operates under constrained antenna resources or low transmit power, the performance advantage narrows and the STC scheme becomes sensing-limited, highlighting the importance of array aperture and power budget in enabling reliable device discovery and efficient energy delivery.
Overall, we show that embedding sensing functionality into WPT systems can substantially enhance wireless charging efficiency in scenarios where user feedback and CSI acquisition are impractical, provided that the sensing resources are appropriately dimensioned relative to network scale and power constraints.
Page updated
Google Sites
Report abuse