Application-Aware Hierarchical Offloading for MEC-Enabled Autonomous Vehicle Architecture

Contemporary vehicular applications pose stringent latency and computation requirements for the autonomous vehicles (AVs). These requirements are hard to be met by the vehicles due to limited computation capabilities. One of the significant solutions is computation offloading in which delay-sensitive and complex applications are handed over to the network. However, computation offloading at the core network incurs excessive architecture-induced delay which is inefficient for applications with tight latency, data rate and computation requirements. Mobile Edge Computing (MEC) is one of the key enablers for 5G that offers computation resources at the edge of the network resulting in ultra-low latency, powerful computation, larger coverage area and context-awareness. European Telecommunication Standards Institute (ETSI) foresees vehicular communication as a use case for MEC. Therefore, we propose application-aware hierarchical offloading scheme (HOS) for MEC-enabled distributed AV architecture. The proposed architecture divides the network into three layers according to application requirements resulting in quick-response and efficient network that meets the application requirements. To decide the computation layer, each application is treated independently in accordance with the complexity, data rate and computation requirements. Thus, every application is handled at appropriate layer so as to meet its latency and computation requirements. Further, we also analyze the impact of task size on computation offloading decision. Finally, we compare our proposed architecture with local computation and distant placed mobile cloud computing (MCC) architecture.