Many future Internet of Things (IoT) applications are expected to rely heavily on reconfigurable intelligent surface (RIS)-aided unmanned aerial vehicles (UAVs). However, the endurance of such systems is constrained by the limited onboard energy, where frequent recharging or battery replacements are required. This consequently disrupts continuous operation and may be impractical in disaster scenarios. To address this challenge, we explore a dual energy harvesting (EH) framework that integrates time-switching (TS), power-splitting (PS), and element-splitting (ES) EH protocols for radio frequency energy, along with solar energy as a renewable source. First, we present the proposed system architecture and EH operating protocols, introducing the proposed hybrid ES-TS-PS EH strategy to extend UAV-mounted RIS endurance. Next, we outline key application scenarios and the associated design challenges. After that, a deep reinforcement learning-based framework is introduced to maximize the EH efficiency by jointly optimizing UAV trajectory, RIS phase shifts, and EH strategies. The framework considers dual EH, hardware impairments, and channel state information imperfections to reflect real-world deployment conditions. The optimization problem is formulated as a Markov decision process and solved using an enhanced deep deterministic policy gradient algorithm, incorporating clipped double Q-learning and softmax-based Q-value estimation for improved stability and efficiency. The results demonstrate significant performance gains compared to the considered baseline approaches. Finally, possible challenges and open research directions are presented, highlighting the transformative potential of energy-efficient UAV-mounted RIS networks for IoT systems.

翻译：许多未来的物联网应用预计将高度依赖于可重构智能表面辅助的无人机系统。然而，此类系统的续航能力受限于机载能量有限，需要频繁充电或更换电池，这会影响连续运行，且在灾害场景中可能不切实际。为应对这一挑战，我们探索了一种双能量收集框架，该框架集成了用于射频能量的时间切换、功率分配和单元分配能量收集协议，并将太阳能作为可再生能源。首先，我们介绍了所提出的系统架构与能量收集操作协议，提出了混合单元分配-时间切换-功率分配能量收集策略以延长无人机载可重构智能表面的续航时间。接着，我们概述了关键应用场景及相关设计挑战。随后，引入了一个基于深度强化学习的框架，通过联合优化无人机轨迹、可重构智能表面相位偏移和能量收集策略，以最大化能量收集效率。该框架考虑了双能量收集、硬件损伤和信道状态信息不完善性，以反映实际部署条件。优化问题被建模为马尔可夫决策过程，并采用改进的深度确定性策略梯度算法求解，该算法结合了裁剪双Q学习和基于Softmax的Q值估计，以提高稳定性和效率。结果表明，与所考虑的基线方法相比，该方法取得了显著的性能提升。最后，本文提出了可能的挑战和开放研究方向，强调了能效型无人机载可重构智能表面网络对物联网系统的变革潜力。