We consider an IoT network with energy-harvesting capabilities. To extend the network lifetime, we propose a novel unmanned aerial vehicle (UAV)- enabled wireless power transfer (WPT) system, where UAVs move among IoT devices and act as data aggregators and wireless power providers. This paper addresses the decision-making problem since the limited buffer and energy resources constrain all nodes. Each IoT node must decide on whether to request a data transmission, to ask for a wireless energy transfer or to abstain and not take any action. When a UAV receives a request from an IoT device, either for data reception or wireless energy transmission, it has to accept or decline. In this paper, we aim to find a proper packet delivery and energy transfer policy according to the system state that maximizes the data transmission efficiency of the system. We first formulate the problem as a Markov Decision Process (MDP) to tackle the successive decision issues, to optimize a utility for each node upon a casual environment. As the MDP formalism achieves its limits when the interactions between different nodes are considered, we formulate the problem as a Graph-based MDP (GMDP). The transition functions and rewards are then decomposed into local functions, and a graph illustrates the dependency’ relations among the nodes. To obtain the optimal policy despite the system’s variations, Mean-Field Approximation (MFA) and Approximate linear-programming (ALP) algorithms were proposed to solve the GMDP problem.

我们考虑具有能量收集功能的物联网网络。为了延长网络寿命，我们提出了一种新型的启用了无人机（UAV）的无线电力传输（WPT）系统，其中，无人机在IoT设备之间移动，并充当数据聚合器和无线电力提供者。由于有限的缓冲区和能源限制了所有节点，因此本文解决了决策问题。每个物联网节点必须决定是请求数据传输，请求无线能量传输还是弃权而不采取任何行动。当无人机接收到来自物联网设备的数据接收或无线能量传输请求时，它必须接受或拒绝。在本文中，我们旨在根据系统状态找到合适的数据包传递和能量传输策略，以最大程度地提高系统的数据传输效率。我们首先将问题表述为马尔可夫决策过程（MDP），以解决连续的决策问题，以在偶然的环境中为每个节点优化实用程序。当考虑不同节点之间的交互时，由于MDP形式主义达到了其极限，因此我们将该问题公式化为基于图的MDP（GMDP）。然后将转换函数和奖励分解为局部函数，并用图形说明节点之间的依存关系。为了获得不受系统变化影响的最优策略，提出了均值场近似（MFA）和近似线性编程（ALP）算法来解决GMDP问题。当考虑不同节点之间的交互时，由于MDP形式主义达到了其极限，因此我们将该问题公式化为基于图的MDP（GMDP）。然后将转换函数和奖励分解为局部函数，并用图形说明节点之间的依存关系。为了获得不受系统变化影响的最优策略，提出了均场近似（MFA）和近似线性编程（ALP）算法来解决GMDP问题。当考虑不同节点之间的交互时，由于MDP形式主义达到了其极限，因此我们将该问题公式化为基于图的MDP（GMDP）。然后将转换函数和奖励分解为局部函数，并用图形说明节点之间的依存关系。为了获得不受系统变化影响的最优策略，提出了均值场近似（MFA）和近似线性编程（ALP）算法来解决GMDP问题。