Mobile Edge Computing: From Task Load Balancing to Real-World Mobile Sensing Applications

Publication Type:
Issue Date:
Full metadata record
With the rapid development of mobile computing technologies and the Internet of Things, there has been an increasing rise of capable and affordable edge devices that can provide in-proximity computing services for mobile users. Moreover, a massive amount of mobile edge computing (MEC) systems have been developed to enhance various aspects of people's daily life, including big mobile data, healthcare, intelligent transportation, connected vehicles, smart building control, indoor localization, and many others. Although MEC systems can provide mobile users with swift computing services and conserve devices' energy by processing their tasks, we confront significant research challenges in several perspectives, including resource management, task scheduling, service placement, application development, etc. For instance, computation offloading in MEC would significantly benefit mobile users and bring new challenges for service providers. Unbalance and inefficiency are the two challenging issues when making decisions on computation offloading among MEC servers. On the other hand, it is unprecedented to design and implement novel and practical applications for edge-assisted mobile computing and mobile sensing. The power of mobile edge computing has not been fully unleashed yet from theoretical and practical perspectives. In this thesis, to address the above challenges from both theoretical and practical perspectives, we present four research studies within the scope of MEC, including load balancing of computation task loading, fairness in workload scheduling, edge-assisted wireless sensing, and cross-domain learning for real-world edge sensing. The thesis consists of two major parts as follows. In the first part of this thesis, we investigate load balancing issues of computation offloading in MEC. First, we present a novel collaborative computation offloading mechanism for balanced mobile cloudlet networks. Then, a fairness-oriented task offloading scheme for IoT applications of MEC is further devised. The proposed computation offloading mechanisms incorporate algorithmic theories with the random mobility and opportunistic encounters of edge servers, thereby processing computation offloading for load balancing in a distributed manner. Through rigorous theoretical analyses and extensive simulations with real-world trace datasets, the proposed methods have demonstrated desirable results of significantly balanced computation offloading, showing great potential to be applied in practice. In the second part of this thesis, beyond theoretical perspectives, we further investigate two novel implementations with mobile edge computing, including edge-assisted wireless crowdsensing for outdoor RSS maps, and urban traffic prediction with cross-domain learning. We implement our ideas with the iMap system and the BuildSenSys system, and further demonstrate demos with real-world datasets to show the effectiveness of proposed applications. We believe that the above algorithms and applications hold great promise for future technological advancement in mobile edge computing.
Please use this identifier to cite or link to this item: