In this article, we study the pricing and resource management in the Internet of Things (IoT) system with blockchain-as-a-service (BaaS) and mobile-edge computing (MEC). The BaaS model includes the cloud-based server to perform blockchain tasks and the set of peers to collect data from local IoT devices. The MEC model consists of the set of terrestrial and aerial base stations (BSs), i.e., unmanned aerial vehicles (UAVs), to forward the tasks of peers to the BaaS server. Each BS is also equipped with an MEC server to run some blockchain tasks. As the BSs can be privately owned or controlled by different operators, there is no information exchange among them. We show that the resource management and pricing in the BaaS-MEC system are modeled as a stochastic Stackelberg game with multiple leaders and incomplete information about actions of leaders/BSs and followers/peers. We formulate a novel hierarchical reinforcement learning (RL) algorithm for the decision makings of BSs and peers. We also develop an unsupervised hierarchical deep learning (HDL) algorithm that combines deep QQ -learning (DQL) for BSs with the Bayesian deep learning (BDL) for peers. We prove that the proposed algorithms converge to stable states in which the peers’ actions are the best responses to optimal actions of BSs.

中文翻译：

---

具有区块链即服务和支持无人机的移动边缘计算的物联网系统中的分布式动态资源管理和定价


在本文中，我们研究了具有区块链即服务 (BaaS) 和移动边缘计算 (MEC) 的物联网 (IoT) 系统中的定价和资源管理。 BaaS 模型包括用于执行区块链任务的基于云的服务器以及用于从本地 IoT 设备收集数据的对等点集。 MEC 模型由一组地面和空中基站（BS）（即无人机（UAV））组成，用于将节点的任务转发到 BaaS 服务器。每个BS还配备了MEC服务器来运行一些区块链任务。由于基站可以由不同的运营商私有或控制，因此它们之间不存在信息交换。我们表明，BaaS-MEC 系统中的资源管理和定价被建模为随机 Stackelberg 博弈，其中包含多个领导者以及有关领导者/BS 和追随者/同行行为的不完整信息。我们为 BS 和同行的决策制定了一种新颖的分层强化学习（RL）算法。我们还开发了一种无监督分层深度学习 (HDL) 算法，该算法将 BS 的深度 QQ 学习 (DQL) 与同行的贝叶斯深度学习 (BDL) 结合起来。我们证明所提出的算法收敛到稳定状态，其中对等点的动作是对 BS 最优动作的最佳响应。