Intelligent Multi-Dimensional Resource Management in MEC-Assisted Vehicular Networks
MetadataShow full item record
Benefiting from advances in the automobile industry and wireless communication technologies, the vehicular network has been emerged as a key enabler of intelligent transportation services. Allowing real-time information exchanging between vehicle and everything, traffic safety and efficiency are significantly enhanced, and ubiquitous Internet access is enabled to support new data services and applications. However, with more and more services and applications, mobile data traffic generated by vehicles has been increasing and the issue on the overloaded computing task has been getting worse. Because of the limitation of spectrum and vehicles' on-board computing and caching resources, it is challenging to promote vehicular networking technologies to support the emerging services and applications, especially those requiring sensitive delay and diverse resources. To overcome these challenges, in this thesis, we propose a new vehicular network architecture and design efficient resource management schemes to support the emerging applications and services with different levels of quality-of-service (QoS) guarantee. Firstly, we propose a multi-access edge computing (MEC)-assisted vehicular network (MVNET) architecture that integrates the concepts of software-defined networking (SDN) and network function virtualization (NFV). With MEC, the interworking of multiple wireless access technologies can be realized to exploit the diversity gain over a wide range of radio spectrum, and at the same time, vehicle's computing/caching tasks can be offloaded to and processed by the MEC servers. By enabling NFV in MEC, different functions can be programmed on the server to support diversified vehicular applications, thus enhancing the server's flexibility. Moreover, by using SDN concepts in MEC, a unified control plane interface and global information can be provided, and by subsequently using this information, intelligent traffic steering and efficient resource management can be achieved. Secondly, under the proposed MVNET architecture, we propose a dynamic spectrum management framework to improve spectrum resource utilization while guaranteeing QoS requirements for different applications, in which, spectrum slicing, spectrum allocating, and transmit power controlling are jointly considered. Accordingly, three non-convex network utility maximization problems are formulated to slice spectrum among base stations (BSs), allocate spectrum among vehicles associated with the same BS, and control transmit powers of BSs, respectively. Via linear programming relaxation and first-order Taylor series approximation, these problems are transformed into tractable forms and then are jointly solved by a proposed alternate concave search algorithm. As a result, optimal spectrum slicing ratios among BSs, optimal BS-vehicle association patterns, optimal fractions of spectrum resources allocated to vehicles, and optimal transmit powers of BSs are obtained. Based on our simulation, a high aggregate network utility is achieved by the proposed spectrum management scheme compared with two existing schemes. Thirdly, we study the joint allocation of the spectrum, computing, and caching resources in MVNETs. To support different vehicular applications, we consider two typical MVNET architectures and formulate multi-dimensional resource optimization problems accordingly, which are usually with high computation complexity and overlong problem-solving time. Thus, we exploit reinforcement learning to transform the two formulated problems and solve them by leveraging the deep deterministic policy gradient (DDPG) and hierarchical learning architectures. Via off-line training, the network dynamics can be automatically learned and appropriate resource allocation decisions can be rapidly obtained to satisfy the QoS requirements of vehicular applications. From simulation results, the proposed resource management schemes can achieve high delay/QoS satisfaction ratios. Fourthly, we extend the proposed MVNET architecture to an unmanned aerial vehicle (UAV)-assisted MVNET and investigate multi-dimensional resource management for it. To efficiently provide on-demand resource access, the macro eNodeB and UAV, both mounted with MEC servers, cooperatively make association decisions and allocate proper amounts of resources to vehicles. Since there is no central controller, we formulate the resource allocation at the MEC servers as a distributive optimization problem to maximize the number of offloaded tasks while satisfying their heterogeneous QoS requirements, and then solve it with a multi-agent DDPG (MADDPG)-based method. Through centrally training the MADDPG model offline, the MEC servers, acting as learning agents, then can rapidly make vehicle association and resource allocation decisions during the online execution stage. From our simulation results, the MADDPG-based method can achieve a comparable convergence rate and higher delay/QoS satisfaction ratios than the benchmarks. In summary, we have proposed an MEC-assisted vehicular network architecture and investigated the spectrum slicing and allocation, and multi-dimensional resource allocation in the MEC- and/or UAV-assisted vehicular networks in this thesis. The proposed architecture and schemes should provide useful guidelines for future research in multi-dimensional resource management scheme designing and resource utilization enhancement in highly dynamic wireless networks with diversified data services and applications.
Cite this version of the work
Haixia Peng (2021). Intelligent Multi-Dimensional Resource Management in MEC-Assisted Vehicular Networks. UWSpace. http://hdl.handle.net/10012/16907