Resource Access and Allocation Strategies in the Internet-of-Things

The Internet-of-Things (IoT) is envisioned as a transformative network paradigm to carry the massive interconnections among ubiquitous physical objects in real life. With the rapid advance in computing hardware industry and the ubiquity of wireless communication networks, the IoT has the potentials to greatly revolutionize the application scenarios and service types of our daily-used devices. However, with billions of devices anticipated to be connected, constructing the future IoT networks faces a series of significant challenges in diverse technical aspects. On one hand, due to the scarcity of spectrum resource, it is impractical to support the enormous volume of data transmission among the anticipated massive devices under current spectrum allocation scheme. On the other hand, because of the constrained power and computational resource at small-size mobile terminals, the traditional standalone operation of devices can hardly meet the increasingly stringent delay requirement for time-critical applications. To this end, a few emerging concepts in the recent literature, such as opportunistic spectrum access (OSA) and mobile edge computing (MEC), have been recognized as promising enablers for solving these resource-oriented issues in the IoT. In this thesis, we emphasis on endowing the IoT devices with intelligence, and propose several learning-based strategies for resource access and allocation in the communication/computing network design of IoT. Considering the scarce spectrum resource, we first design a novel learning-based spectrum sensing policy for IoT devices to achieve a fast spectrum access for unlicensed spectrum resource with a strong adaptability. We then propose a low-complexity task offloading strategy for a multi-layer mobile computing network, which enables IoT devices to effectively exploit the abundant computing resource deployed at network edges or cloud. In the last part, we developed an efficient strategy to provision different computing services for IoT devices, which enables an overall improvement in the quality of service (QoS) of smart applications. We concluded the thesis with a summary of important results and insights in our works, which is followed by an extensive discussion on other potential research directions in the scope of IoT for our future works