The Internet of Things (IoT) increases the number of connected devices and supports the ever-growing complexity of applications. Owing to the constrained physical size, the IoT devices can significantly enhance the computational capacity by offloading computation-intensive tasks to the resource-rich edge servers deployed at the base station (BS) via wireless networks. However, how to achieve optimal resource scheduling remains a challenge due to stochastic task arrivals, time-varying wireless channels, and imperfect estimation of channel state information (CSI). In this article, by virtue of the Lyapunov optimization technique, we propose the toward optimal resource scheduling algorithm under imperfect CSI (TORS) to optimize resource scheduling in an IoT environment. A convex transmit power and subchannel allocation problem in TORS is formulated. This problem is then solved via the Lagrangian dual decomposition method. We derive analytical bounds for the time-averaged system throughput and queue backlog. We show that TORS can arbitrarily approach the optimal system throughput by simply tuning an introduced control parameter $\beta $ without prior knowledge of stochastic task arrivals and the CSI of wireless channels. Extensive simulation results confirm the theoretical analysis on the performance of TORS.

中文翻译：

---

不完善CSI下物联网的最优资源调度

物联网（IoT）可以增加连接设备的数量，并支持不断增长的应用程序复杂性。由于物理尺寸的限制，IoT设备可以通过无线网络将计算密集型任务卸载到部署在基站（BS）的资源丰富的边缘服务器上，从而显着提高计算能力。但是，由于随机任务的到来，时变无线信道以及信道状态信息（CSI）的估计不完善，如何实现最佳的资源调度仍然是一个挑战。在本文中，借助李雅普诺夫优化技术，我们提出了一种在不完全CSI（TORS）下向最优资源调度的算法，以优化物联网环境中的资源调度。提出了TORS中凸的发射功率和子信道分配问题。然后通过拉格朗日对偶分解法解决了这个问题。我们得出时间平均系统吞吐量和队列积压的分析界限。我们显示TORS可以通过简单地调整引入的控制参数$ \ beta $来任意接近最佳系统吞吐量，而无需事先了解随机任务的到来和无线信道的CSI。大量的仿真结果证实了TORS性能的理论分析。