Task Partitioning for Migration with Collaborative Edge Computing in Vehicular Networks

Abstract

Multi-access edge computing (MEC) is considered a promising technology to facilitate mission-critical vehicular applications, such as automatic driving, path-planning, and navigation. By offloading delay-sensitive or computation-intensive tasks from vehicles to MEC servers (MECSs), edge computing significantly enhances the computing capacity of vehicles with limited computing resources. However, MECSs may have uneven loads as vehicles are not evenly distributed across MEC systems and vehicles do not offload their tasks evenly. As a result, those offloaded tasks have high latency or be blocked. In addition, service interruption would happen frequently due to task migration caused by the high mobility. Due to the high mobility of vehicles and load dynamics at MECSs, computation tasks can migrate simultaneously to a particular MECS or migrate to a heavily congested MECS. Therefore, it is challenging to determine the migration decision, i.e., whether/where to migrate, among MECSs. In conventional methods, computation tasks are fully migrated to the MECS corresponding to the vehicle’s trajectory. By contrast, in this study, tasks are migrated partially or fully to other MECSs in the collaborative edge computing system. To reduce the task execution latency and improve the system throughput, the proposed method selects a MECS that optimizes load balancing among MECSs and partitions the task to migrate for the MECS. Through simulations, compared with the conventional methods, the proposed method can increase the satisfaction of quality of service (QoS) requirements and MEC system throughput by optimizing the load balancing and task partitioning.